I. Obtaining Optimal-Quality Endobronchial Biopsies
II. Processing and Methods of Visualization
III. Variability and Quantification
IV. Endobronchial Biopsy versus Bronchoalveolar Lavage versus Induced Sputum
V. Overall Conclusions
The capacity to sample directly the conducting airway tissue of subjects with asthma or chronic obstructive pulmonary disease (COPD) for comparison with tissue obtained from normal, healthy individuals has generated hypotheses and advanced our understanding of pathogenesis. Such studies in adults are now being extended, with informed consent and ethical approval, to include children. In adults, approval to obtain endobronchial biopsies on more than one occasion has also allowed us to begin to test whether or not therapeutic intervention will significantly reduce airway inflammation or restore altered tissue structure (i.e., remodeling) to normal.
Endobronchial biopsy has been considered by some the standard by which to compare the validity and interpretation of results obtained by indirect, less invasive means, such as bronchoalveolar lavage (BAL) and induced sputum.
Asthma and COPD are recognized as inflammatory conditions of the airways, albeit with distinctive patterns of inflammation and tissue structure (1). A variety of endobronchial biopsy methodologies are being used that may compromise valid interpretation and confirmation of previously published findings. For example, in asthma, increases in connective tissue deposition are variably reported in association with severity of disease (2, 3), and controversy remains as to whether or not there is loss of surface epithelium in asthma (4, 5).
The quality of the biopsy obtained predetermines the quality of the staining and the reliability of the data generated. Several factors contribute to quality: the number of biopsies obtained and their location, size, and morphological integrity, which may often vary from research center to research center. There are numerous protocols for tissue fixation, processing for histology, visualization by immunostaining, in situ hybridization (ISH), and electron microscopy. Variability introduced by different methods of quantification is compounded by the variability that exists naturally between subjects, patient groups, and airway generations, as well as between individual biopsies and between sections of each biopsy from any one individual. Traditionally, qualitative examination of endobronchial biopsies, often with the aid of semiquantitative scores applied in a blinded manner, has led to the generation of hypotheses. Such observations have been strengthened scientifically by more rigorous use of quantitative techniques, and there are several available.
Investigators have also used data obtained from biopsies to test hypotheses, and treatment-related drug effects have been published (6–14). It is not clear, however, how many such clinical studies have failed to find treatment-related effects, whether because of inadequate patient numbers or because of insufficiently or inappropriately prepared or sampled tissue. Endobronchial biopsy samples are small and, moreover, the ‘signals’ obtained from them are also small and frequently may be lost within inherently large biological variability. Thus, the analyses used need to have the capacity to discriminate subtle yet clinically important differences between treatment groups.
Such analyses almost always benefit from the standardization of study design, avoidance of potential sources of bias, and attention to the validation of reliable endpoints, all part of well established principles in study design (15–28). Effective a priori planning followed by strict adherence, via quality control, to the specified study design and measurement protocols is essential. Critical components include identifying clear study aims and objectives, defining a hypothesis, identifying the primary and secondary objectives, assuring experimental standardization, and selecting valid endpoints. Such information will be needed to ensure that the sample size is adequate. To avoid controversy about study interpretation, a complete protocol should be in place before the study is initiated; the data analysis plan (see Appendix) also should be in place before the study is initiated, or at least before the study is unblinded.
It is crucial that standard procedures (detailed in a manual) for conducting each component of the study be defined and adhered to via regular evaluation, especially in multicenter trials, to reduce variability caused by procedural differences (29). Individual study sites should be trained and certified in the proper performance of these procedures. Before initiation of the study per se, pilot biopsies outside the study should be evaluated systematically from each center. Ideally, in multicenter studies, biopsy endpoints should be measured in a single central laboratory. Endpoints whose reproducibility and responsiveness are known should be chosen and their variance, effect size, and ultimately, sample size should be estimated. Frequently, pilot or validation studies are completed before designing the final study.
The goal of the international workshop reported here was to produce a document that would assist and guide investigators in making cross-comparisons of biopsy data. Our aim was to compare and discuss the methods currently used worldwide to obtain and analyze endobronchial biopsies in asthma and COPD and to work toward standardization of methodology.
Written documents were presented to and discussed by the entire group under the leadership of moderators and, based on the discussions, the written documents were revised. Four topics, having the following specific purposes, formed the core of the workshop:
Obtaining optimal-quality endobronchial biopsies: To discuss and recommend the best methods of safely obtaining good-quality endobronchial samples of adequate size for research in adults and also to consider the special issues related to endobronchial biopsy in children.
Processing of the biopsies and methods of visualization: To outline and record the advantages and disadvantages of different methods of processing tissue samples and of visualizing and expressing the biopsy parameters related to airway inflammation and remodeling.
Variability and quantification: To discuss the current practices of quantification to understand the limitations imposed on the interpretation of such data and to debate and clarify the use of alternative, less biased methods for the quantification of inflammatory and structural cells.
Advantages and limitations of endobronchial biopsy, BAL, and induced sputum: To highlight the distinct yet complementary data that can be gained from analyses of endobronchial biopsies, BAL, and induced sputum and to discuss how these data might relate to the clinical status of subjects.
Finally, not only can endobronchial biopsies provide information about treatment-induced changes in inflammation and remodeling, but they can also be used to isolate resident cells and to study their phenotype (30). These in vitro experimental models allow for the study of cell interactions under conditions as close as possible to their natural environment (31).
We hope that this report represents a step toward improving the reliability of testing hypotheses using airway tissue obtained by biopsy.
The report presented here relates primarily to research bronchoscopies aimed at obtaining endobronchial biopsies under direct vision. While we do comment briefly on transbronchial biopsies, these are not recommended routinely for research.
All research involving bronchoscopy must have approval by an institutional review board/ethics committee. Ethical considerations include bronchoscopy procedure, the appropriate use of samples, and the scientific validity of the study.
The subject must be fully informed about the study and of any potential risk and discomfort. This is often provided as an information sheet. In some countries, it is important to inform subjects that the samples will be archived. In both children and adults, when the bronchoscopy is being done for clinical reasons, but additional procedures are to be performed for research purposes, the ethical considerations described here also need to be followed.
Ethical concerns will vary locally and will relate both: (1) to the use of samples for purposes not originally defined in the consent form and (2) to the failure to use fully valuable archived tissue samples, a failure that would needlessly require additional volunteers to undertake bronchoscopy for the purposes of research.
Research bronchoscopy on volunteers is an invasive investigation that usually does not improve the volunteers' welfare directly. Thus, safety is of paramount importance (32). Although numerous reports suggest that the technique is well tolerated, even down to an FEV1 as low as 30% of predicted (33, 34), caution should be exercised in those with a low FEV1, particularly in subjects at risk of sudden severe bronchospasm. All research bronchoscopies should be indemnified.
Bronchoscopy needs to be performed by experienced bronchoscopists, preferably of specialist standing in appropriately equipped laboratories. Respiratory physicians in training can undertake the procedure under close supervision or after a period of training and accreditation. Research bronchoscopy is not to be used as part of the routine clinical training program (32, 35).
Continuous vigilance and monitoring throughout the procedure are essential. Subjects need to be premedicated with bronchodilators (unless specifically contraindicated by the study protocol). Atropine is often used to dry secretions and decrease cough. Premedication often incorporates an opiate, such as codeine, papaveretum, or fentanyl, which may be used to suppress the cough reflex. In adults, oxygen is routinely supplemented. The use of sedatives during bronchoscopy varies. In some centers, no sedation is used; in others, benzodiazepines, such as midazolam or diazepam, are used for sedation, sometimes in conjunction with a narcotic, such as fentanyl.
Before the bronchoscope is inserted, local anesthesia with lidocaine is applied in the nares and the oral pharynx to suppress sensation, cough, and the gag reflex unless the procedure is being performed under a general anesthetic. Lidocaine is a potentially toxic drug that can cause arrhythmias and seizures. There has been only one report of death that was thought to be due to lidocaine toxicity following a research bronchoscopy (36). Lidocaine may obscure recognition of laryngomalacia; consequently, the larynx needs to be examined before topical anesthesia is applied (37). The bronchoscope is passed nasally or orally, depending on local anatomy or the size of the bronchoscope. Local anesthetic is given via the bronchoscope at the vocal cords and throughout the tracheobronchial tree.
There are no set protocols for how lidocaine is applied (38–44); thus, comparison of the various studies of lidocaine levels is difficult. It may be given in a nebulized form (1,500 mg) prior to bronchoscopy (45), or as a gel or solution applied to the larynx and bronchi (1–4%). Premedication with opiates may reduce the dose of lidocaine required. Lidocaine is metabolized in the liver with a half-life of 100 to 120 minutes (39, 43). Some manufacturers recommend that the maximum dose be less than 300 mg and 4.5 mg or less per kilogram of body weight (46). Lidocaine, however, has been given safely at 9.3 mg/kg of body weight, and only occasionally exceeded the toxic level of 5 μg/ml (40, 41). Nevertheless, it is important to use as low a dose as possible and to keep a careful record of the lidocaine use.
During all research bronchoscopies, there needs to be an adequately trained physician and nurse or other assistant able to terminate the procedure if the clinical risk becomes unacceptable or the subject is distressed. In addition to the availability of full resuscitation facilities, benzodiazepine and opiate antagonists (flumazenil and naloxone) need to be at hand to reverse respiratory depression. Noting of clotting times and platelet counts is not recommended routinely prior to bronchoscopy, but should be undertaken if there are clinical features that raise concern (these measures should always be obtained prior to transbronchial biopsies).
BAL can provide beneficial information that complements the results of biopsy. However, BAL increases the duration of the procedure and, in a small proportion of individuals, results in cough, fever, or pleuritic pain after bronchoscopy. The exact frequency of these complications will vary according to the volume of BAL and the subject group, but frequencies of fever and pleurisy of between 2 and 5% have been reported. The lavage procedure is more likely than biopsy alone to induce arterial oxygen desaturation and bronchospasm. For these reasons, if both BAL and endobronchial biopsy are contemplated, it is preferable to take the biopsy samples first and then perform the BAL procedure in a different lung or lobe to avoid hematological contamination, especially if parameters measured in biopsies are the primary outcome variable. If, however, a BAL measurement is the primary outcome variable, this procedure should be performed first.
A physician or a specialist nurse is to see all subjects before discharge, and subjects need to be given a 24-hour contact telephone number. Also prior to discharge, subjects should be warned about the effects of topical anesthesia on swallowing and of sedation on driving
Although the best results are obtained using larger instruments, this goal needs to be balanced by the subject's comfort and safety. All aspects of the protocol and the processing of the biopsy samples are to be defined clearly before the study begins. Facilities for immediate handling and adequate storage are essential. It is important to set an order of priority in terms of sample processing to ensure that at least the primary outcome measures are obtained in case the sampling is incomplete.
Endobronchial biopsies taken under direct vision will, by definition, be from proximal airways, usually from subsegmental and segmental subcarinae and not from the lateral wall. This would usually be from the second to the fifth generation of airway branching. The choice of biopsy site needs to be defined by the protocol, especially in studies in which more than one sampling session over time is required. Because the consequences to the site of an endobronchial biopsy sample are not known, repeat samples (e.g., to assess the impact of intervention) should not be taken from the same anatomic site. Preintervention and postintervention biopsies may be randomized to each lung or come from the same airway generations in, for example, biopsies taken before and 24 hours after segmental allergen challenge.
Although no formal studies have been done, there is consensus among researchers that it is most difficult to obtain samples from normal volunteers and perhaps easiest to obtain them from subjects with asthma or COPD.
There is no systematically obtained evidence on the best way to take an endobronchial biopsy sample. Accumulated evidence suggests that reusable forceps should not be used in more than five bronchoscopies. Early experience with disposable forceps suggests a good yield. In principle, the forceps need to straddle the carina but not with excessive force, as this might increase artifactual damage. The forceps need to be closed firmly and held in place for a few seconds to allow them to grip before withdrawing. Biopsies performed in one lung usually yield eight samples. A minimum of two samples per processing procedure (e.g., embedding them in paraffin or resin or snap-freezing them) is needed to ensure at least one good sample for analysis. If multiple specimens are to be obtained, it is easiest to begin sampling in the most distal carina and subsequently to sample more proximally. This reduces the possibility of any bleeding obscuring the view. If poor samples are obtained at gross examination, consider changing the forceps. There have been no reports of excessive bleeding in otherwise healthy subjects. Therefore, use of epinephrine is not indicated. Alternatively, brush biopsy may be used to sample the endobronchial epithelium (47, 48).
It is important that an assistant be present at the bronchoscopy to check that the endobronchial biopsy sample is of an adequate size and quality. Rapid processing of samples and standard staining in the laboratory allow feedback as to the adequacy of the biopsy technique, which, in turn, allows for improvements to be made if needed during the course of the study.
All research involving bronchoscopy in children must have approval by an institutional review board/ethics committee. Ethical considerations include bronchoscopy procedure, the appropriate use of samples, and the scientific validity of the study. In young children, bronchoscopy cannot be performed solely for research purposes (49), but there is no ethical contraindication to performing an endobronchial biopsy if the child is undergoing a bronchoscopy for a clinically indicated purpose or is being intubated for another procedure, such as surgery, that permits the bronchoscope to be passed down the endotracheal tube. Indeed, research endobronchial biopsy in children has been well described and performed safely (49–53). If a research transbronchial biopsy is contemplated in a child, it is necessary to consider very carefully the increased risks of this procedure over direct endobronchial biopsy and the likelihood of obtaining worthwhile information before requesting approval for such studies (54).
A recent working group of the European Respiratory Society proposed that when general anesthesia or bronchoscopy is being performed for clinical purposes, it would be unethical not to consider taking an endobronchial biopsy and/or BAL sample after obtaining the appropriate informed consent (55). Even in children with severe asthma, bronchoscopy and airway biopsy may be performed safely under carefully controlled conditions (56, 57). Endobronchial biopsy (both forceps and brush) is a routine clinical procedure in, for example, bronchial tuberculosis in children. A series of several hundred samples taken at the time of a clinically indicated bronchoscopy in children aged 6 months to 15 years of age demonstrated that the procedure can be done without significant complication (53, 56, 58).
Infants and children may be examined under deep sedation or full general anesthesia (56, 59, 60). There are national and international guidelines for the safe administration of anesthesia and sedation (37, 61–66).
Pulmonary function should be monitored before and at 30-minute intervals after the procedure until it is within 10% of baseline (subjects are often too sleepy to give good efforts 30 minutes after the procedure). This recommendation cannot apply to children too young to have spirometry performed; in these children, pulse oximetry is usually relied upon. Intravenous atropine usually is administered to adults as part of the premedication to dry secretions and suppress laryngeal reflexes. Many anesthesiologists will choose not to give a premedication to children (61).
During the procedure, it is essential that the child be monitored by a person independent of the bronchoscopist, usually an anesthesiologist, someone who has taken the American Academy of Pediatrics Life Support course, or an equivalently trained pediatrician. The procedure needs to be terminated immediately if any decompensation in the child causes clinical concern to the independent subject advocate.
The largest bronchoscope that is safe should be used. This is generally a 4.9-mm endoscope (2.2-mm biopsy channel) in children at least 5 years of age and a 3.6-mm endoscope (1.2-mm biopsy channel) in younger children. The size of the bronchoscope used should be determined by the subject's age (bronchoscope diameter [mm] = 4 + age [years]/4) and the extent of reduction of airway caliber (airway area for ventilation = [(D − d)(D + d)/D2]%, where D is the diameter of the airway and d is the diameter of the bronchoscope). Thus, a 3.6-mm bronchoscope in a 5-mm airway occludes nearly half of the airway. Other considerations are similar to those in adults.
In view of their more advanced age and their potential for comorbid conditions, such as ischemic heart disease, all COPD subjects should have a prebronchoscopy ECG and, optimally, continuous ECG recording during bronchoscopy. Subjects known to have ischemic heart disease must be excluded from research bronchoscopies. In more advanced disease, where there is risk of CO2 retention, arterial blood gases should be measured and subjects with significant hypercapnia must be excluded (67).
Transbronchial biopsies should be performed only if the value of the information to be obtained outweighs the risk of the procedure as determined by the investigator and ethics committee. Because there is a lack of information about repeatability, the transbronchial technique cannot be recommended at this time for interventional (pharmacologic or challenge) studies. Transbronchial biopsies have, however, been taken for research (33, 68–70) and have provided information about inflammation in the lung parenchyma. In a series of more than 100 bronchoscopies, there was a report of one case of pneumothorax, which was treated conservatively.
The limited number of studies that have been done have used fluoroscopy to appropriately position the biopsy forceps. The routine use of fluoroscopy has not been shown to improve the safety of the procedure (71). The samples are generally obtained from the lateral segments of one lower lobe and are likely to contain both alveolar and peribronchial tissue. Samples can be taken from the standard subpleural position if one is interested in obtaining primarily parenchymal tissue or more proximally (two thirds to three quarters of the distance to the pleura) if the study is directed at smaller airways. Airway tissue is obtained in approximately 35 to 40% of samples. It is recommended that no more than four samples be taken for research purposes. A small-bore chest tube needs to be immediately available in case of pneumothorax. A postprocedure chest X-ray needs to be obtained. Subjects should be instructed to avoid the use of aspirin for 2 weeks prior to the procedure. Epinephrine should also be available in case of significant bleeding. In the opinion of the authors, this procedure can be performed for research purposes but only in centers with considerable research bronchoscopy experience. The safety of this procedure needs to be reviewed regularly. Pooling of data from different institutions would help to clarify the role of this procedure in research.
Considerable progress has been made in defining the best way to sample the airways, and this has revolutionized the concept of airway inflammation. However, continued audits, preferably using pooled data, will further the development of safer and more effective procedures. In the interests of safety and the ethical practice of this procedure, researchers are encouraged to report both positive and negative experiences with endobronchial biopsies. It is recommended that these reports be audited by panels of experts made up of representatives of relevant national and international societies.
Both the bronchoscopist and laboratory representatives need to participate directly in the initial planning. Prior to the study, pilot biopsies (i.e., samples not part of the study per se) should be taken to control for the adequacy of biopsy size and quality. The histology laboratory should ideally report back as soon as possible to the bronchoscopist on biopsy integrity and its viability for immunohistologic, molecular, and ultrastructural studies as appropriate to the study protocol. There may be logistical issues when there is a single central reporting laboratory in a study of multicenter/multicountry design.
For the purposes of research, an evaluable sample in adults should ideally have a subepithelial area, excluding crush artifact, cartilage, or blood clot, of at least 0.3 to 0.5 mm (72, 73). Samples of blood and mucus, scrapings of epithelium, and fragments of glands are unacceptable. If there are to be three distinct processing techniques, it is recommended that a minimum of six samples be taken from each subject, two samples for each processing technique. Endobronchial biopsy samples should not be placed in sterile phosphate-buffered saline as an interim holding procedure but must be fixed or snap-frozen immediately following their removal from the patient.
Methods used for tissue fixation and processing vary among researchers depending on study endpoints, local expertise, and experience (74–76). Whatever the method, there needs to be a standard operating protocol adhered to throughout the study. This is particularly important when comparing samples taken before and after challenge or therapeutic intervention. Fixation time needs to be long enough to preserve tissue morphology but not so long as to mask antigens, reduce accessibility, or destroy messenger RNA (mRNA)/DNA. Typically, fixation times vary between 4 and 24 hours, and for ISH, fixatives need to be free of RNAses and DNAses (77, 78).
For immunohistology, immediate snap-freezing of freshly obtained samples in melting isopentane (i.e., previously cooled in liquid nitrogen) and cryostat sectioning has the advantage of a rapid processing method (74–76, 79–81). An alternative is embedding the samples in optimal cutting temperature compound prior to freezing. Freezing enables all antigens of interest to be detected by immunohistochemistry (IHC). If ISH is to be performed on such snap-frozen samples, prefixation using freshly prepared 4% paraformaldehyde for 4 hours at room temperature and cryoprotection in 15% sucrose in phosphate-buffered saline is recommended. Fixation prior to freezing enhances morphology but can compromise IHC; for example, cell surface adhesion molecules and activation markers such as CD25 may be lost. Snap-frozen samples are usually cut at a thickness of 5 to 7 μm.
In comparison with samples embedded in paraffin or glycol methacrylate (GMA), frozen tissue provides relatively poor morphology because of its greater section thickness and has the potential disadvantage of ice crystal artifact. In the absence of prior fixation, there may be unwanted diffusion of antigens and proteins, and there is a need for low-temperature storage (−80°C in a freezer or storage in liquid nitrogen). In cut and stored sections, some antigens are lost quite rapidly. Also, because of the thickness of these sections, a smaller number can be prepared from each sample when compared with paraffin and GMA, both of which use thinner sections. Collaboration with centers worldwide requires the use of special thermal packaging for shipping of frozen biopsies, but samples cannot be guaranteed to remain frozen during transit and customs inspection.
Processing samples into paraffin wax (77), usually following fixation in 10% neutral-buffered formalin for IHC (77) or 4% paraformaldehyde or 10% neutral-buffered formalin for ISH (78), offers some advantages over snap-freezing (74–76, 82, 83). First, tissue architecture is better preserved. In addition, there is little or no diffusion of antigens, specialized storage facilities are not required, and facilities for processing are universally available. Finally, there are no difficulties of tissue preservation during transport of samples between laboratories, and both IHC and ISH procedures can be performed using paraffin-embedded material.
There are also some disadvantages to paraffin embedding. Antigen retrieval techniques may be required due to use of cross-linking fixatives that may mask selected antigens from detection by IHC; some antigens may thus not be retrievable. Tissue shrinkage occurs and is often ignored; this is usually of the order of 10%, but can be as much as 50%, depending on the temperature and the organic solvents used during the processing and embedding procedure.
A hydrophilic resin, GMA has recently been employed for IHC (84) and applied to numerous studies (70, 85, 86). When coupled with acetone fixation (with protease inhibitors), GMA offers some advantages over frozen and paraffin preparations. Acetone is a non–cross-linking fixative that, together with processing at a low temperature, ensures antigen preservation; all antigens demonstrable in frozen samples can also be seen in GMA but, compared with frozen samples, morphologic preservation is greatly improved. This is because sections are only 1 to 2 μm thick and provide essentially a single focal plane when viewed through the light microscope. Due to their thinness, adjacent sections can show more than one cut through a single cell, enabling the detection and “co-localization” of more than one epitope to a cell. Relatively large numbers of sections can be cut from a single sample, allowing for comparatively more markers to be investigated than with the other embedding techniques. As with paraffin embedding, GMA-embedded samples can be stored safely for long periods. However, there are disadvantages: GMA embedding is more time-consuming and less widely used than either frozen or paraffin preparations and is not yet suitable for studies of RNA/DNA and their visualization by ISH.
The methods used for sample fixation and embedding are summarized in Table 1
Antigen Preservation: IHC | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Embedding Medium | Recommended Fixation | Morphologic
Preservation
(Graded) | Ease
of Use | Cytoplasmic and
Nuclear Markers | Surface
Markers | Cytokines and
Chemokines | Adhesion
Molecules | ISH | |||
| Snap-frozen | Acetone or methanol | 3 | 3 | 5 | 5 | 5 | 5 | 3 | |||
| Fixed, then snap-frozen | 4% para-formaldehyde | 4 | 4 | 4 | 4 | 4 | 3 | 5 | |||
| Paraffin | 10% neutral buffer formalin or 4% para-formaldehyde | 5 | 5 | 4 | 4 | 3 | 1 | 5 | |||
| GMA | Acetone | 5 | 3 | 5 | 5 | 5 | 5 | 0 | |||
A dedicated laboratory and the support of an experienced immunohistochemist are needed. For batch staining of biopsy samples in clinical trials, an automatic staining machine is favored to ensure consistency and good quality of immunostaining over time. A high-quality microscope with photographic capability and access to integrated, interactive image analysis equipment is recommended. Various IHC methods have been reviewed and described in detail (77, 87, 88). Both monoclonal and polyclonal antibodies can be applied. The initial use of pilot tissue and positive controls known to contain the antigens of interest is recommended so that each antibody can be titrated by limiting dilution to provide an optimal working concentration (i.e., to obtain the best signal-to-noise ratio). Positive and negative controls should always be included with each subsequent run of tissue samples. Negative controls include omission of the primary antibody, using an isotype-matched Ig, or preabsorption of antigen prior to the IHC procedure. Although the methods of visualization for IHC are many and varied, most groups use horseradish peroxidase, alkaline phosphatase, or fluorescent two–three stage detection systems. If an enzymatic detection system is used, endogenous tissue enzymes must be blocked (89, 90). The following techniques are commonly available: streptavidin–biotin conjugates (91) (full horseradish peroxidase or alkaline phosphatase), alkaline phosphatase antialkaline phosphatase (92) (for monoclonal antibodies only), EnVision (horseradish peroxidase or alkaline phosphatase) (93), and fluorescence labeling. New techniques, such as immunopolymerase chain reaction (94) are now being validated for the detection of weakly expressed antigens. Detailed written standard operating protocols and standardized criteria for identifying and measuring the immunostained parameter of interest should be agreed upon and followed consistently, with appropriate quality control checks during the study.
ISH requires an investigator with extensive experience in the field and dedicated facilities. There is an absolute requirement to minimize RNAse contamination, which is ubiquitous, if riboprobes are to be used (95). If using radiolabeled probes, appropriate safety equipment and procedures are essential, including approval for radioactive work and a well-equipped darkroom. A microscope fitted for observation of fluorescence as well as dark- and bright-field viewing is recommended (90, 95, 96).
In order to visualize and localize mRNA, different types of probes are available, including DNA oligonucleotide probes and complementary RNA probes. The most commonly used RNA probes (97) offer the advantage of being more sensitive and specific for the detection of a particular mRNA. The probes can be labeled with radioisotope or with biotin, digoxigenin, or fluorescein isothiocyanate. Radiolabeled probes provide great sensitivity and should be used to detect mRNA encoding of actively synthesized proteins that are rapidly turned over and therefore have a relatively small amount of intracellular mRNA. However, radioactively labeled probes also have disadvantages in that they are difficult and can on occasion be dangerous to handle and have a short half-life. In addition, they require skill in tissue section autoradiography to visualize the tissue source of the radioactively labeled probes. Nonradiolabeled probes (98) are ideal for genes that are expressed at a high level (i.e., present in high copy number). They are safe to handle and can be stored for relatively long periods (i.e., 6 months). However, they are considered to be less sensitive than radiolabeled probes, and the required immunodetection procedure can result in nonspecific labeling, particularly of epithelia. The major advantage of nonradiolabeled probes is that the immunolabeled mRNA hybrids can be visualized by bright field microscopy without the need for autoradiography. Also, unlike the autoradiography of radiolabeled probes, which take a week or more to develop, nonradiolabeled probes can be viewed immediately after the IHC detection procedure. Double-labeling techniques in which the ISH procedure is followed by IHC can be used to identify the cell expressing a gene of particular interest (99). Appropriate controls are essential and need to include the complementary sense, an irrelevant sense, or an alternative irrelevant antisense probe, preferably of similar length to the antisense probe for the mRNA of interest (95). RNAse pretreatment (prior to hybridization) can be used in addition as a negative control. Positive controls must also be included when optimizing each probe.
An experienced investigator with knowledge of endobronchial biopsy, morphology, and ISH should interpret the results (100, 101). A number of inflammatory cells, including eosinophils and macrophages, can bind probes nonspecifically, especially if radioactive probes are used. In the case of radioactive probes, caution is to be exercised in interpreting the autoradiographic signal at the edge of the section and at the inner surface of blood vessels or airways (because of the “edge effect”) (102). There have been correlations measured between clinical pulmonary function and mRNA expression of inflammatory mediators (95, 103–105). The absence of a hybridization signal does not necessarily indicate the inability of cells to produce a specific protein. In the case of a transient protein product, mRNA can be expressed in low copy number and may be beyond the sensitivity of the ISH technique (106). Moreover, posttranscriptional and posttranslational events may affect gene expression. Small differences in the number of ISH-positive cells may have important functional and immunoregulatory implications.
Because it is limited by the wavelength of light, bright-field microscopy has a resolution of up to 0.25 μm and a maximum useful magnification of 2,000 ×. In contrast, transmission electron microscopy (TEM), which uses electrons rather than light, allows much greater resolution of biological samples, of the order of 1 to 2 nm, and magnifications greatly in excess of 200,000 ×. However, TEM requires an investment in expensive equipment and a dedicated laboratory with appropriately trained staff to process samples and to operate and maintain the equipment.
An advantage of TEM over immunologic methods is that tissue components and cells are all simultaneously visualized and may be quantified in one and the same section. In particular, it is possible to identify and assess the proportions of a variety of morphologically identifiable cellular phenotypes, including epithelial (4, 5) and endothelial cells (107, 108), distinct inflammatory cells (plasma cells, neutrophils, and macrophages), apoptotic cells, fibroblasts, and myofibroblasts (109, 110). Mast cell (111) and eosinophil granule subtypes (112) and degranulation (113) status can be assessed. Cell-to-cell contact and junctional complexes can be identified (114–117), particularly when TEM is used in combination with more specialized freeze-fracture techniques (115). Changes in the endothelium and in the composition of extracellular matrix (such as fragmentation of elastin (118) or dissolution of the reticular basement membrane) (119) can also be studied. In addition to freeze-fracture TEM, other, more specialized techniques include scanning electron microscopy, immunoelectron microscopy, and cryoultramicrotomy. These techniques have several unique advantages, but their specialized nature precludes their inclusion in this report.
For morphologic analysis by TEM, tissues are usually fixed first in a freshly prepared primary aldehyde fixative, such as buffered 0.5 to 3.0% glutaraldehyde or Karnovsky's fixative, to preserve the protein components. This is followed by secondary fixation, usually in 1% osmium tetroxide, to stabilize lipids. The slow penetration of these fixatives requires that the tissue samples be small (i.e., 2 mm), about the size of a bronchial biopsy. TEM requires the embedding of tissues in an electron-beam stable epoxy resin, such as Araldite or Epon, and the use of ultrathin sections (60–90 nm). To cut ultrathin sections requires specialized microtomes, and because of the hardness of the tissue block, the sections need to be cut with glass or diamond knives. To facilitate section support, the sections are usually picked up on copper mesh metallic grids. These can aid in quantification because the parts of the tissue section that are viewed lie between the grid spaces, providing a known area of tissue for the quantification. Mesh-free “slot grids,” Formvar-coated (Agar Scientific Ltd., Essex, UK) to support the specimen, can provide a large area of unobstructed view. Staining with heavy-metal salts, such as uranyl acetate or lead citrate, imparts electron density on selected tissue structures. Staining is performed either en bloc during the processing procedure or on the sections after they are cut. Epoxy resin-embedded sections of 1 μm thickness can also be cut, stained with toluidine blue, and viewed to advantage by light microscopy as is done with GMA resin.
Care in specimen handling is of primary importance, and the quality of processing, cutting, and staining must be evaluated first. Following examination of the “survey section” by light microscopy, the area of interest for TEM analysis needs to be selected and cut to the ultrathin level. The section, and particularly the area to be viewed and photographed, should be free of cutting artifacts (e.g., knife scores and holes) (120–122).
Thus, a variety of visualization techniques are available to the researcher: which to choose depends very much on the question or hypothesis being addressed.
Depending on the nature of the research question, semiquantitative or quantitative methods may be used. Indeed, there is evidence that a semiquantitative approach can provide results and interpretation similar to those derived from morphometric evaluation in, for example, assessments of lung fibrosis (123). Having said that, the basic equipment required for the objective application of quantitative methods is not extensive and has already been used in pulmonary medicine to quantify both epithelial and interstitial cells (123–129).
The standard practice of counting the number of cut cell profiles of interest in a tissue section using light microscopy and normalizing these counts to submucosal area or to length of the epithelial reticular basement membrane continues to be a popular quantitative approach. The methodologies involved in some of these procedures and their application to assess inflammation in bronchial biopsies have been summarized and discussed previously (130, 131).
The probability of visible cells being counted in a two-dimensional section is proportional to their size, density, and orientation, thus introducing a bias in favor of the larger cut profile (Figure 1)
Figure 1. Several cells intersected by a series of systematic uniform random planes represented by a set of lines. The number of times a cell is hit by the set of planes will depend on their height normal to the plane of sectioning. Although cells c and d have approximately the same dimensions, cell c is cut four times, while cell d is cut only once, purely as a consequence of orientation. Adapted from Howard CV, Reed MG. Unbiased stereology: three-dimensional measurement in microscopy. New York: Springer-Verlag; 1998, p. 71.
[More] [Minimize]The present authors recommend investigators consider the application of stereology, a less biased sampling strategy than that used thus far. Those interested are recommended to Howard and Reed's Unbiased Stereology: Three Dimensional Measurement in Microscopy (133), a well written, easy-to-read book that contains an excellent discussion of sampling and stereology-based morphometric techniques (see Chapter 5, pp 69–106). Stereology is the study of the three-dimensional properties of objects usually seen in two dimensions. Thus, by using stereology-based techniques, one can obtain quantitative three-dimensional information from measurements in the two dimensions of histology.
Because inflammation and remodeling in the airway in asthma and COPD are likely to be unevenly distributed in the airway mucosa throughout the bronchial tree, it is essential that biopsy sampling and the selection of biopsy sections and microscopic areas (fields) of tissue within biopsy sections be random. Random sampling eliminates bias and maximizes accuracy in the data generated. Randomness cannot be assumed; sampling strategies to ensure randomness need to be built into the measurement protocol.
Variability is introduced at the level of the subject, the biopsy (block), the biopsy section, the microscopic field of tissue, and the specific measurement used. Using the stereologic technique of “point counting,” the relative contributions of variance at these five levels have been estimated (134) and include 70% variability between subjects, 20% between blocks, 5% between sections, 3% between fields, and 2% between measurements.
It is important for investigators to acquire data in their own laboratories on the within- and between-observer variability for key outcomes and for the between-subject, between-biopsy, between-section, and between-field variability for these outcomes. Information about variability is essential for estimating required sample size for clinical studies involving bronchoscopy. For example, when Carroll and colleagues (135) examined airway tissues obtained postmortem from asthmatic persons, they demonstrated within- and between-observer coefficients of variation of approximately 1% for measurements of the dimensions of tissue structures and as high as 60% for counts of inflammatory cells. Additionally, there are examples in the pulmonary literature that demonstrate approximately 10 to 25% error (defined bias) in developing and injured airways using appropriate stereologic methods (125, 127). The variation found for counts of inflammatory cells and measures of reticular basement membrane thickness in endobronchial biopsies in subjects with asthma or COPD are on the order of 1 to 6%, and the variations between airway generations and subjects have been reported (72, 80, 136–139). High variability may preclude meaningful interpretation of the data in a study designed to test the effect of a treatment. This has led recently to a change in the current practice of analysis of one or in some cases two biopsy samples from which several sections and fields are quantified, to at least two biopsies per fixative, more subjects (as many as is practically and ethically possible), and relatively fewer sections and fields per biopsy. This change in approach to the quantitative analysis is captured in the saying “do more less well.”
There is relatively little bias in volumetric tissue analysis when “point counting” is used to establish volume fractions and the “disector” method is used to determine cell number (133) (see Number Estimation by the Disector Method, below). While not applicable to measurements of tissue volume or cell number, unbiased measurement of certain features that are orientation-sensitive, such as thickness and surface area, requires that the sections be random in their three-dimensional orientation (140, 141). As endobronchial biopsy tissue tends to curl on sampling, because of inherent tissue elasticity, it favors a random orientation of tissue structures included during embedding and in subsequent sectioning. This topic, referred to as isotropic randomization, can be complex and is reviewed clearly and simply in Unbiased Stereology (133).
To obtain a random sample, the authors recommend that a list of 10 to 15 potential biopsy sites located in the second, third, and fourth airway generations of the right or the left lung be generated before initiating the procedure. During bronchoscopy and moving in a distal (caudal) to proximal (rostral) direction, six to eight sites selected at random from this list need to be considered. Ideally, both lungs would be sampled equally, but potential complications secondary to endobronchial biopsy (bleeding and, rarely, pneumothorax) limit this possibility, and thus, only one lung should be biopsied during a bronchoscopy. If, for example, two fixation procedures are to be used and six samples are to be obtained, one sample from each of three generations would be placed in each fixative.
Several sections can be cut initially to determine orientation, followed by a new random start. In general, when there are layered structures (such as the mucosal layer of the airways) and the tissue sample is relatively large (not the case with endobronchial biopsies), unbiased estimates can be obtained by vertically sectioning (perpendicular to the surface epithelium and mucosa) the embedded mucosa and then randomly selecting the sections to be analyzed (133). Serial sections are cut and three to four unpaired sections are chosen randomly on a slide for each particular marker.
Given the small amount of tissue obtained by endobronchial biopsy, it is important to select the priority markers to be evaluated before sectioning the tissue. The distance between the sections needs to be greater than the largest element to be counted. With the newly recommended approach “do more less well,” an “adequate” biopsy section need have only one evaluable field of interest (a distinct difference from the technique used most widely at present, in which several fields are considered necessary for evaluation). Of course, if a biopsy has only one evaluable field, the remaining biopsies to be counted must be larger in order to yield a sufficient number of fields to complete the analysis for the subject.
Measurements in fields of interest require an invariant reference compartment that could be either the reticular basement membrane or the volume of the sample. Normalization to surface area of basement membrane (i.e., number of cells per surface area) is one recommended approach that has been used to document changes in epithelial and inflammatory cells in biopsies of airways (142).
The first step in selecting high-power fields for analysis is to decide on the tissue compartment of interest. Once an area has been chosen (e.g., epithelium or submucosa), the next step is to choose fields randomly. This can be done by dividing the area into quadrants and choosing fields randomly or by employing computerized sampling systems or random number or field generation.
Measures of density of cells, blood vessels, collagen, smooth muscle, or other tissue components can be determined by the simple technique of point counting. This determines the volume density (mm3/mm3) of objects in a reference volume (140), using the formula: Vv(O,R) = PO/PR, where Vv(O,R) denotes the volume of objects (O) contained in a reference volume (R), PO is the number of points from a grid falling onto the structure of interest, and PR is the number of points falling into the reference space. Vv is the volume density and corresponds to the volume proportion that the structure of interest occupies in the reference space. The beauty of this ratio of points is that it is proportional to volume (140). Therefore, the ratio of three-dimensional measurements can be obtained by assessing tissue elements in a two-dimensional plane, without taking any more (or even less) time than do currently used methods of counting cell area profiles.
A simple modification is used to determine surface area (e.g., of basement membrane). A grid with lines (as distinct from points) is superimposed on the tissue section, and the number of intersections of the test lines with the surface of interest (e.g., the reticular basement membrane) is recorded. The test grid can be provided in the form of an eyepiece reticule inserted into the microscope or by using an attached computerized system that can also facilitate the mathematical transformation of the intersection count to a surface area measurement (143). The subsequent determination of surface density (mm2/mm3) of objects in a reference volume is estimated therefore from both intersect and point counting (140) according to the formula: SV(O,R) = 2 × I(O)/L(R), where SV(O,R) denotes the surface area of an object (O) contained in a reference volume (R), I(O) is the number of intersections of the test line of length L in reference space (R) with the trace of the object (see Unbiased Stereology, Chapter 6, pp 107–123 in [133]) (140).
Determination of volume density is an elegant and simple way to avoid the criticisms of counts of cut cell profiles but does not directly provide an estimate of cell number. One simple approach to obtaining an estimate of cell number is to divide the calculated volume/density for any given cell type by an estimate of its overall cell size (144). Alternatively, cell number per volume can be more accurately calculated using the disector method of analysis, but this is more labor intensive and consumptive of tissue (see subsequent section).
Determination of the length of a tissue component may be required by the researcher in endobronchial biopsy tissue (e.g., measurement of the average length of a blood vessel in the submucosal compartment). The length density of a blood vessel can also be estimated using stereology and given as its length per unit volume of the reference space. In a section, it will be obtained by counting the number of times the vessel transects the section.
The number of cells can also be estimated and expressed per unit volume of tissue (i.e., the “reference space”) using serial sections and the disector method, which can only be outlined briefly here. The disector method relies on the principle that all cells will be counted with equal probability if the method is applied in one of two ways: use of either the physical disector or the optical disector. The physical disector is the most practical of the two for analysis of bronchial biopsies and consists of a pair of serial sections separated by a known distance; the distance chosen depends on the size of the smallest cell of interest. The principle is that a cell is counted only if it is seen in one section and not in the next (16, 133).
In conclusion, biopsy tissue can yield valuable quantitative information about both airway inflammation and remodeling. For interpretation of the data, we must assume that endobronchial biopsy of large airway subcarinae provides unbiased and representative samples of the entire bronchial tree, which currently is unproven and needs to be determined in larger lung samples obtained at surgery or postmortem. Animal studies, at least, indicate that the airway wall cellular characteristics are significantly different from those obtained from subcarinae (17). Stereology requires a common denominator reference space for counts of cells or measures of tissue structures to be expressed meaningfully. The surface area of the reticular basement membrane is one potentially useful reference, and simultaneous measurements of subepithelial and epithelial volume may also be useful as a way of normalizing count data (18, 19). In animal experiments, lung volume has been used as the reference space (145). Data from computer-assisted tomography has been used to provide a useful estimate of lung volume in humans (146).
Although area profile counts still remain acceptable, we recommend that investigators consider and try to apply less biased, stereology-based sampling methods, which are likely to be simpler, more discriminatory, and more powerful for hypothesis testing. In addition, the general incorporation of these stereologic methods will facilitate greater opportunities for metaanalyses of data generated from similarly designed studies worldwide (see Table 2
2-Dimensional: One Section per Biopsy or > Two-Step Section | 3-Dimensional: Multiple Biopsies and Multiple Sections per Biopsy | Semiquantitative Methods | Automated Methods | |
|---|---|---|---|---|
| Parameter | No. per area | Volume density (mm3/mm3) | Scoring system (1+, 2+, 3+) | No. per area |
| No. per length | Volume to surface | |||
| Advantages | Easily mastered | Uses established stereologic principles | Useful for pilot studies | Fast Consistency of staining |
| Volume-oriented measurement | ||||
| Can use technique to quantify cells and structures | ||||
| Easily mastered | ||||
| Disadvantages | May invite bias because larger
cells are overestimated, smaller
cells underestimated | “Do more less well”—requires more
subjects and more biopsies | Not quantitative | Systems may have difficulty
discerning cells; differences
based on color, not cell type |
Airway inflammation includes the presence of activated inflammatory cells, the release of cellular mediators, and the functional and structural consequences of such inflammation. Response to tissue injury, abnormal tissue repair, and restructuring, inappropriate to the maintenance of normal function, are referred to as “airway wall remodeling” (147, 148). Inflammation and remodeling may result from parallel or sequential pathways, and the techniques of endobronchial biopsy, BAL, and induced sputum may be used to sample complementary components of these pathways.
The potential for the use of less invasive alternatives to endobronchial biopsy for measuring inflammation and remodeling in asthma (149) and COPD (150) requires a thorough understanding of the techniques and of their advantages and disadvantages. Each technique samples a different compartment and a different facet of the inflammatory process, and thus the relationships between data obtained by these techniques can be expected to be, at best, moderate (151–153).
Endobronchial biopsy, including endobronchial brush biopsy, directly samples the resident cells. Biopsies alone maintain the spatial relationships of structural components that may be important to functional change. The airway wall is composed of many interrelated structural components such as epithelium, connective tissue, vessels, muscle, and mucus-secreting gland acini. Endobronchial biopsy can provide an assessment of the status of the airway mucosa, whereas BAL and induced sputum cannot. A biopsy preserves structural relationships that may be visualized by microscopy. Individual structural components can be designated for study and, if required, removed and studied in isolation using laser capture dissection (154). Inflammatory cell subtypes can be identified by immunostaining, and biomarkers of epithelial damage, such as mitosis and apoptosis, can be measured. Immunologic cascades can be studied using immunohistochemistry, ISH, and microarray analysis (for assessment of gene expression) (see III. Processing and Methods of Visualization). Tissue explants and cells derived from biopsies can also be preserved in culture, and resident cells, such as airway smooth muscle myocytes, have successfully been dissected from endobronchial biopsies from subjects with asthma, allowing measurements of the proliferative responses of these cells (155).
Endobronchial biopsy samples are limited in overall size to between 1 and 2 mm in diameter; they may not be representative of the entire conducting airways because they come mainly from relatively proximal airways and from subcarinae rather than the lateral walls. Thus, there is a bias inherent in the sampling procedure. Moreover, peribronchial tissue (deep to the mucosa) is not included in endobronchial samples. Depending on the operator's experience, the procedure itself may induce artifacts, especially due to forceps damage apparent as crush artifact and mechanical loss of epithelium (5). In general, the more normal the airway, the more likely is tissue damage. In repetitive studies, the biopsy procedure itself creates scar tissue and can initiate cellular processes that can proceed during the interval (seconds) between sampling and fixation. It should be remembered that biopsy provides only a snapshot in time of an ongoing disease process. Finally, most endobronchial biopsies have been performed in adults and not children, although pediatric data are now forthcoming (50, 57).
Samples obtained by BAL include luminal cells (predominantly macrophages) and soluble factors coming from a relatively large airway surface area, including the peripheral airways. With sequential lavage, it is possible to observe the results of allergen and placebo challenge in the same individual over time. BAL includes both mediators and cells released into the fluid phase that may be studied individually or in specific combinations in suspension or in culture. International guidelines and monographs on BAL are already available (156, 157).
There is no spatial, structural, or histologic information in BAL samples, and the cellular signals may be affected by the method of collection. The source of the cell sample is imprecise and may include cells that have migrated into the lumen or that were flushed from the airway wall. The volume of sample recovered may relate to disease severity, and the sources and the dilution factors of the solutes obtained are imprecise. The procedure itself or the subsequent handling and processing may introduce artifacts (e.g., lidocaine and bronchodilators used during bronchoscopy may affect cell activity). The removal, washing, centrifuging, and resuspension of mucus can activate cells and confound the cell and solute information. Finally, the BAL process may lead to transient fever within 24 hours after the lavage, occasionally associated with lung infiltrates (134).
Hypertonic saline-induced sputum is increasingly used as a minimally invasive procedure to obtain markers of airway inflammation, such as differential cell counts and solutes in asthma and COPD (158). Recently, detailed international recommendations on safety precautions and standardized induction, processing, and analysis of sputum have become available (158). Sputum can even be obtained during an acute exacerbation of the disease, and it can also be obtained (albeit less successfully) from children (159). The precision, accuracy, and responsiveness to challenges and therapy are well validated in asthma. Sputum can be repeatedly sampled, after sufficient time (i.e., 48 hours) has elapsed between inductions to allow cellular events caused by the procedure itself to subside. It can be studied using immunocytochemistry or ISH (158) and by a limited range of cell culture procedures. Many cytokines/mediators can now be measured by bio-, enzyme-, or immunoassays and real-time polymerase chain reaction in the fluid phase of the sputum sampled. Interestingly, induced sputum appears to be useful clinically in the management of asthma because therapy, guided by sputum eosinophil counts, has been shown to result in a considerable improvement of disease outcome (160).
Airway structural information is not available from sputum. The exact site of origin of the material and the dilution factor are largely unknown. Hypertonic saline inhalation is not without risk in subjects with poorly controlled asthma: isotonic saline can be a good alternative (161). About 10 to 20% of subjects are unable to produce an adequate sputum specimen (up to 50% in children), especially those who are clinically stable and well controlled with therapy. Sputum requires the use of reducing agents to process the samples, which may affect the results of subsequent procedures. Cytokines, growth factors, and mediators may be encrypted in mucus, and the cellular profile and staining patterns are more difficult to interpret than they are with BAL preparations.
We are just starting to appreciate the pros and cons of the available biopsy techniques. There is little doubt that bronchial biopsies, BAL, induced sputum, and also markers in exhaled breath (162) provide complementary biological information on the complex and dynamic features of inflammation. The distribution of such abnormalities within the airway wall and along the tracheobronchial tree determines the severity of airway narrowing (163, 164). Each method samples specific parts of the pathologic process. Eventually, the choice among the available methods for monitoring airway inflammation critically depends on the specific hypotheses and primary objectives of each individual study in conjunction with the burden that such measurements impose on the subject.
Given that the safety of the volunteer subject is of paramount importance, bronchoscopy has an invaluable place in airway research, even in children. The conduct of clinical trials should be based on well established principles of study design and include effective a priori planning followed by strict adherence to the specified study design and measurement concepts. Laboratory research personnel or senior representatives concerned with tissue preparation and analysis need to be involved during both study design and execution. The procedures of tissue sampling and processing of biopsies require meticulous attention to the high premium placed on each biopsy: this is the only technique that directly samples the resident cells and maintains the spatial relationships with structural components—the information is additional and complementary to that obtained by lavage and sputum. Standard operating protocols need to be agreed upon and adhered to and quality controls are needed. Minimizing the bias and variability introduced by the sampling procedure and the biology will improve accuracy and the capacity to test hypotheses. Counting area profiles of inflammatory cells, where the nucleus is seen in the plane of the section, remains an acceptable procedure. However, the application of stereology for quantifying cells and tissue structures minimizes bias. The last consideration may improve statistical power, better enabling investigators to discriminate and determine the potential for current and new treatments aimed at attenuating the inflammation and remodeling of diseases such as asthma and COPD.
Neil Barnes, London, United Kingdom
Andrew Bush, London United Kingdom
Ratko Djukanovic, Southampton, United Kingdom
Peter Howarth, Southampton, United Kingdom
Douglas Robinson, London, United Kingdom
John Warner, Southampton, United Kingdom
Qutayba Hamid, Montreal, Quebec, Canada
Peter Jeffery, London, United Kingdom
Annika Laitinen, Helsinki, Finland
Maurizio Vignola, Palermo, Italy
Susan J. Wilson, Southampton, United Kingdom
Peter H. Burri, Bern, Switzerland
Neil Carroll, Nedlands, Australia
John Fahy, San Francisco, California
Peter Gehr, Bern, Switzerland
James Hogg, Vancouver, British Columbia, Canada
Dallas Hyde, Davis, California
Peter Jeffery, London, United Kingdom
Monica Kraft, Denver, Colorado
Louis-Philippe Boulet, Quebec City, Quebec, Canada
Pascal Chanez, Montpellier, France
Kathleen Haley, Boston, Massachusetts
Stephen Holgate, Southampton, United Kingdom
Makoto Hoshino, Tokyo, Japan
Yutaka Nakamura, Madison, Wisconsin
Theodore Reiss, Rahway, New Jersey
Peter J. Sterk, Leiden, The Netherlands
E. Hayden Walters, Tasmania, Australia
Sally Wenzel, Denver, Colorado
Peter Jeffery, London, United Kingdom
Theodore Reiss, Rahway, New Jersey
Tri Tat, London, United Kingdom
Ji Zhang, Rahway, New Jersey
| 1. | Jeffery P, Laitinen A, Venge P. Biopsy markers of airway inflammation and remodelling. Respir Med 2000;94:S9–S15. |
| 2. | Chu HW, Halliday JL, Martin RJ, Leung DY, Szefler SJ, Wenzel SE. Collagen deposition in large airways may not differentiate severe asthma from milder forms of the disease. Am J Respir Crit Care Med 1998;158:1936–1944. |
| 3. | Minshall EM, Leung DY, Martin RJ, Song YL, Cameron L, Ernst P, Hamid Q. Eosinophil-associated TGF-β1 mRNA expression and airways fibrosis in bronchial asthma. Am J Respir Cell Mol Biol 1997;17:326–333. |
| 4. | Jeffery PK, Wardlaw AJ, Nelson FC, Collins JV, Kay AB. Bronchial biopsies in asthma: an ultrastructural quantificative study and correlation with hyperreactivity. Am Rev Respir Dis 1989;140:1745–1753. |
| 5. | Ordonez C, Ferrando R, Hyde DM, Wong HH, Fahy JV. Epithelial desquamation in asthma: artifact or pathology? Am J Respir Crit Care Med 2000;162:2324–2329. |
| 6. | Ward C, Pais M, Bish R, Reid D, Feltis B, Johns D, Walters EH. Airway inflammation, basement membrane thickening and bronchial hyperresponsiveness in asthma. Thorax 2002;57:309–316. |
| 7. | Gizycki MJ, Hattotuwa KL, Barnes N, Jeffery PK. Effects of fluticasone propionate on inflammatory cells in COPD: an ultrastructural examination of endobronchial biopsy tissue. Thorax 2002;57:799–803. |
| 8. | Faul JL, Demers EA, Burke CM, Poulter LW. Alterations in airway inflammation and lung function during corticosteroid therapy for atopic asthma. Chest 2002;121:1414–1420. |
| 9. | Jeffery PK, Venge P, Gizycki MJ, Egerod I, Dahl R, Faurschou P. Effects of salmeterol on mucosal inflammation in asthma: a placebo-controlled study. Eur Respir J 2002;20:1378–1385. |
| 10. | O'Sullivan S, Akveld M, Burke CM, Poulter LW. Effect of the addition of montelukast to inhaled fluticasone propionate on airway inflammation. Am J Respir Crit Care Med 2003;167:745–750. |
| 11. | O'Sullivan S, Cormican L, Murphy M, Poulter LW, Burke CM. Effects of varying doses of fluticasone propionate on the physiology and bronchial wall immunopathology in mild-to-moderate asthma. Chest 2002;122:1966–1972. |
| 12. | Wallin A, Sandstrom T, Cioppa GD, Holgate S, Wilson S. The effects of regular inhaled formoterol and budesonide on preformed Th-2 cytokines in mild asthmatics. Respir Med 2002;96:1021–1025. |
| 13. | Flood-Page PT, Menzies-Gow AN, Kay AB, Robinson DS. Eosinophil's role remains uncertain as anti-interleukin-5 only partially depletes numbers in asthmatic airway. Am J Respir Crit Care Med 2003;167: 199–204. |
| 14. | Verhoeven GT, Hegmans JP, Mulder PG, Bogaard JM, Hoogsteden HC, Prins JB. Effects of fluticasone propionate in COPD patients with bronchial hyperresponsiveness. Thorax 2002;57:694–700. |
| 15. | Feinstein A. The architecture of clinical research. Philadelphia: WB Saunders Company; 1985. |
| 16. | Geiser M, Cruz-Orive LM, Im Hof V, Gehr P. Assessment of particle retention and clearance in the intrapulmonary conducting airways of hamster lungs with the fractionator. J Microsc 1990;160:75–88. |
| 17. | Plopper CG, Weir A, Murphy A, Wong V, Fanucchi M, Evans M, Van Winkle L, Nishio S, Schelegle ES, Miller L, et al. Cyclic exposure to ozone disrupts postnatal growth of distal conducting airways in infant monkeys. Am J Respir Crit Care Med 2002;165:A507. |
| 18. | Hays SR, Woodruff PG, Khashayar R, Ferrando RE, Liu J, Fung P, Zhao CQ, Wong HH, Fahy JV. Allergen challenge causes inflammation but not goblet cell degranulation in asthmatic subjects. J Allergy Clin Immunol 2001;108:784–790. |
| 19. | Ordonez CL, Khashayar R, Wong HH, Ferrando RE, Wu R, Hyde DM, Hotchkiss JA, Zhang Y, Novikov A, Dolganov G, et al. Mild and moderate asthma is associated with airway goblet cell hyperplasia and abnormalities in mucin gene expression. Am J Respir Crit Care Med 2001;163:517–523. |
| 20. | Popper K. The logic of scientific discovery. New York: Harper & Row; 1934. |
| 21. | Fischer RA, Mackenzie WA. Studies in crop variation II: the manurial response to different potato varieties. J Agric Sci 1923;13:3–5. |
| 22. | Weyman J. On the problem of the most efficient tests of statistical hypotheses. Philos Trans R Soc Lond A. 1933;231:289–337. |
| 23. | Fischer R. Statistical methods for research workers, 13th ed. New York: Hafner; 1958. |
| 24. | Goodman SN. Toward evidence-based medical statistics 2: the Bayes factor. Ann Intern Med 1999;130:1005–1013. |
| 25. | Hulley S, Cummings S. Designing clinical research. Baltimore: Williams and Wilkins; 1988. |
| 26. | Streinber DL, Norman GR. Health measurement scales. New York: Oxford University Press; 1991. |
| 27. | Hays RD, Hadyn D. Responsiveness to change: an aspect of validity, not a separate dimension. Qual Life Res 1992;1:37–75. |
| 28. | Deyo R. Assessing the responsiveness of functional scales to clinical change: an analogy to diagnostic test performance. J Chronic Dis 1986;39:897–906. |
| 29. | Malmstrom K, Peszek I, Botto AI, Lu S, Enright PL, Reiss TF. Quality assurance of asthma clinical trials. Control Clin Trials 2002;23: 143–156. |
| 30. | Chakir J, Dube J, Laviolette M, Goulet F, Germain L, Auger F, Boulet L-P. Characterisation of airway fibroblasts in culture. Methods in molecular medicine: asthma. New York: Humana Press; 2000. |
| 31. | Chakir J, Page N, Hamid Q, Laviolette M, Boulet L, Rouabhia M. Bronchial mucosa produced by tissue engineering: a new tool to study cellular interactions in asthma. J Allergy Clin Immunol 2001;107: 36–40. |
| 32. | Jarjour NN, Peters SP, Djukanovic R, Calhoun WJ. Investigative use of bronchoscopy in asthma. Am J Respir Crit Care Med 1998;157: 692–697. |
| 33. | Wenzel SE, Szefler SJ, Leung DY, Sloan SI, Rex MD, Martin RJ. Bronchoscopic evaluation of severe asthma: persistent inflammation associated with high dose glucocorticoids. Am J Respir Crit Care Med 1997;156:737–743. |
| 34. | Vrugt B, Wilson S, Underwood J, Bron A, de Bruyn R, Bradding P, Holgate ST, Djukanovic R, Aalbers R. Mucosal inflammation in severe glucocorticoid-dependent asthma. Eur Respir J 1999;13:1245–1252. |
| 35. | Djukanovic R, Dahl R, Jarjour NN, Aalbers R. Safety of biopsies and bronchoalveolar lavage. Eur Respir J Suppl 1998;26:39S–41S. |
| 36. | Day RO, Chalmers DR, Williams KM, Campbell TJ. The death of a healthy volunteer in a human research project: implications for Australian clinical research. Med J Aust 1998;168:449–451. |
| 37. | Nielson DW, Klu P, Egger M. Topical lidocaine exaggerates laryngomalacia during flexible bronchoscopy. Am J Respir Crit Care Med 2000; 161:147–151. |
| 38. | Loukides S, Katsoulis K, Tsarpalis K, Panagou P, Kalogeropoulos N. Serum concentrations of lignocaine before, during and after fiberoptic bronchoscopy. Respiration (Herrlisheim) 2000;67:13–17. |
| 39. | Sanderson DR. Lignocaine for topical anesthesia in the fiberoptic bronchoscopy. Respiration (Herrlisheim) 2000;67:9–10. |
| 40. | Langmack EL, Martin RJ, Pak J, Kraft M. Serum lidocaine concentrations in asthmatics undergoing research bronchoscopy. Chest 2000; 117:1055–1060. |
| 41. | Efthimiou J, Higenbottam T, Holt D, Cochrane GM. Plasma concentrations of lignocaine during fibreoptic bronchoscopy. Thorax 1982;37: 68–71. |
| 42. | Patterson JR, Blaschke TF, Hunt KK, Meffin PJ. Lidocaine blood concentrations during fiberoptic bronchoscopy. Am Rev Respir Dis 1975; 112:53–57. |
| 43. | Gomez F, Barrueco M, Lanao JM, Vicente MT, Dominguez-Gil A. Serum lidocaine levels in patients undergoing fibrobronchoscopy. Ther Drug Monit 1983;5:201–203. |
| 44. | Sutherland AD, Santamaria JD, Nana A. Patient comfort and plasma lignocaine concentrations during fibreoptic bronchoscopy. Anaesth Intensive Care 1985;13:370–374. |
| 45. | Berger R, McConnell JW, Phillips B, Overman TL. Safety and efficacy of using high-dose topical and nebulized anesthesia to obtain endobronchial cultures. Chest 1989;95:299–303. |
| 46. | Lidocaine HC topical solution USP 4% [package insert]. Columbus, OH: Roxane Laboratories Inc.; October 1996. |
| 47. | Chanez P, Vignola AM. Bronchial brushing. Eur Respir J Suppl 1998;26: 26S–29S. |
| 48. | Riise GC, Andersson B, Ahlstedt S, Enander I, Soderberg M, Lowhagen O, Larsson S. Bronchial brush biopsies for the studies of epithelial inflammation in stable asthma and nonobstructive chronic bronchitis. Eur Respir J 1996;9:1665–1671. |
| 49. | Teig N, Philippou S, Wagner C, Beck T, Schauer U, Rieger C. Subepithelial basement membrane thickness in wheezy infants and development of asthma at an age of 5 years [abstract]. Eur Respir J 1999; 14:281s. |
| 50. | Pohunek P, Roche WR, Turzikova J, Kudrmann J, Warner JO. Eosinophilic inflammation in the bronchial mucosa of children with bronchial asthma [abstract]. Eur Respir J 1997;11:160s. |
| 51. | Durieu I, Peyrol S, Gindre D, Bellon G, Durand DV, Pacheco Y. Subepithelial fibrosis and degradation of the bronchial extracellular matrix in cystic fibrosis. Am J Respir Crit Care Med 1998;158:580–588. |
| 52. | Heino M, Juntunen-Backman K, Leijala M, Rapola J, Laitinen LA. Bronchial epithelial inflammation in children with chronic cough after early lower respiratory tract illness. Am Rev Respir Dis 1990;141: 428–432. |
| 53. | Bush A, Pohunek P. Brush biopsy and mucosal biopsy. Am J Respir Crit Care Med 2000;162(Suppl):S18–S27. |
| 54. | Whitehead B, Scott JP, Helms P, Malone M, Macrae D, Higenbottam TW, Smyth RL, Wallwork J, Elliott M, de Leval M. Technique and use of transbronchial biopsy in children and adolescents. Pediatr Pulmonol 1992;12:240–246. |
| 55. | Bush A. Guidelines for the ethical conduct of medical research involving children. Arch Dis Child 2000;83:370. |
| 56. | Payne DN, McKenzie SA, Stacey S, Misra D, Haxby E, Bush A. Safety and ethics of bronchoscopy and endobronchial biopsy in difficult asthma. Arch Dis Child 2001;84:423–426. |
| 57. | Payne DN, Adcock IM, Wilson NM, Oates I, Scallon M, Bush A. Relationship between exhaled nitric oxide and mucosal eosinophilic inflammation in children with difficult asthma, after treatment with oral prednisolone. Am J Respir Crit Care Med 2001;164:1376–1381. |
| 58. | Barbato A, Magarotto M, Crivellaro M, Novello AJ, Cracco A, de Blic J, Scheinmann P, Warner JO, Zach M. Use of the paediatric bronchoscope, flexible and rigid, in 51 European centres. Eur Respir J 1997;10:1761–1766. |
| 59. | Smyth AR, Bowhay AR, Heaf LJ, Smyth RL. The laryngeal mask in airway fibreoptic bronchoscopy. Arch Dis Child 1996;75:344–345. |
| 60. | Raine J, Warner JO. Fibreoptic bronchoscopy without general anaesthetic. Arch Dis Child 1991;66:481–484. |
| 61. | Shaw CA, Kelleher AA, Gill CP, Murdoch LJ, Stables RH, Black AE. Comparison of the incidence of complications at induction and emergence in infants receiving oral atropine versus no premedication. Br J Anaesth 2000;84:174–178. |
| 62. | Guidelines for the elective use of conscious sedation, deep sedation and general anesthesia in pediatric patients. Committee on drugs, Section on anesthesiology. Pediatrics 1985;76:317–321. |
| 63. | Malviya S, Voepel-Lewis T, Tait AR. Adverse events and risk factors associated with the sedation of children by nonanaesthesiologists. Anesth Analg 1997;85:1207–1213. |
| 64. | Cote CJ. Sedation for the paediatric patient: a review. Pediatr Clin North Am 1994;41:31–58. |
| 65. | Chopra V, Bovill JG, Spierdijk J. Accidents, near accidents and complications during anaesthesia: a retrospective analysis of a 10-year period in teaching hospital. Anaesthesia 1990;45:3–6. |
| 66. | Fan LL, Sparks LM, Fix FJ. Flexible fibreoptic endoscopy for airway problems in the paediatric intensive care unit. Chest 1988;93:556–560. |
| 67. | Hattotuwa K, Gamble EA, O'Shaughnessy T, Jeffery PK, Barnes NC. Safety of bronchoscopy, biopsy, and BAL in research patients with COPD. Chest 2002;122:1909–1912. |
| 68. | Kraft M, Martin RJ, Wilson S, Djukanovic R, Holgate ST. Lymphocyte and eosinophil influx into alveolar tissue in nocturnal asthma. Am J Respir Crit Care Med 1999;159:228–234. |
| 69. | Wenzel SE, Balzar S, Gibbs R, Chu HW. Small airway inflammation in refractory asthmatics and the relationship to physiologic abnormalities [abstract]. Am J Respir Crit Care Med 2001;163:A19. |
| 70. | Kraft M, Djukanovic R, Wilson S, Holgate ST, Martin RJ. Alveolar tissue inflammation in asthma. Am J Respir Crit Care Med 1996;154: 1505–1510. |
| 71. | Anders GT, Johnson JE, Biush BA, Matthews JI. Transbronchial biopsy without fluoroscopy: a seven-year perspective. Chest 1988;94:557–560. |
| 72. | Sullivan P, Stephens D, Ansari T, Costello J, Jeffery P. Variation in the measurements of basement membrane thickness and inflammatory cell number in bronchial biopsies. Eur Respir J 1998;12:811–815. |
| 73. | ten Hacken NH, Aleva RM, Oosterhoff Y, Smith M, Kraan J, Postma DS, Timens W. Submucosa 1.0 × 0.1 mm in size is sufficient to count inflammatory cell numbers in human airway biopsy specimens. Mod Pathol 1998;11:292–294. |
| 74. | Bancroft ID, Stevens A. Theory and practice of histological techniques. New York: Churchill Livingstone; 1996. |
| 75. | Culling CF, Allison RT, Barr WT. Cellular pathology technique. London: Butterworths; 1985. |
| 76. | Drury RA, Wasllington EA. Carleton's histological technique. New York: Oxford University Press; 1967. |
| 77. | Polak JM, Van Noorden S. Introduction to immunocytochemistry. Oxford: BIOS Scientific Publishers, Ltd; 1997. |
| 78. | Schwarzacher T, Heslop-Harrison P. Practical in situ hybridization. Oxford: BIOS Scientific Publishers, Ltd.; 2000. |
| 79. | Azzawi M, Bradley B, Jeffery PK, Frew AJ, Wardlaw AJ, Knowles G, Assoufi B, Collins JV, Durham SR, Kay AB. Identification of activated T lymphocytes and eosinophils in bronchial biopsies in stable atopic asthma. Am Rev Respir Dis 1990;142:1407–1413. |
| 80. | O'Shaughnessy TC, Ansari TW, Barnes NC, Jeffery PK. Inflammation in bronchial biopsies of subjects with chronic bronchitis: inverse relationship of CD8+ T lymphocytes with FEV1. Am J Respir Crit Care Med 1997;155:852–857. |
| 81. | Ying S, Durham SR, Corrigan CJ, Hamid Q, Kay AB. Phenotype of cells expressing mRNA for TH2-type (interleukin 4 and interleukin 5) and TH1-type (interleukin 2 and interferon-gamma) cytokines in bronchoalveolar lavage and bronchial biopsies from atopic asthmatic and normal control subjects. Am J Respir Cell Mol Biol 1995;12: 477–487. |
| 82. | Zhu J, Majumdar S, Qiu YS, Ansar T, Oliva A, Kips JC, Pauwels RA, DeRose V, Jeffery PK. IL-4 and IL-5 gene expression and inflammation in the mucus-secreting glands and subepithelial tissue of smokers with chronic bronchitis: lack of a relationship with CD8+ cells. Am J Respir Crit Care Med 2001;164:2220–2228. |
| 83. | Zhu J, Qiu YS, Majumdar S, Gamble E, Matin D, Turato G, Fabbri LM, Barnes NC, Saetta M, Jeffery PK. Exacerbations of bronchitis: bronchial eosinophilia and gene expression for interleukin-4, interleukin-5 and eosinophil chemoattractants. Am J Respir Crit Care Med 2001;164:109–116. |
| 84. | Britten KM, Howarth PH, Roche WR. Immunohistochemistry on resin sections: a comparison of resin embedding techniques for small mucosal biopsies. Biotech Histochem 1993;68:271–280. |
| 85. | Bradding P, Roberts JA, Britten KM, Montefort S, Djukanovic R, Mueller R, Heusser CH, Howarth PH, Holgate ST. Interleukin-4, -5 and -6 and tumour necrosis factor-alpha in normal and asthmatic airways: evidence for the human mast cell as a source of these cytokines. Am J Respir Cell Mol Biol 1994;10:471–480. |
| 86. | Wenzel SE, Schwartz LB, Langmack EL, Halliday JL, Trudeau JB, Gibbs RL, Chu HW. Evidence that severe asthma can be divided pathologically into two inflammatory subtypes with distinct physiologic and clinical characteristics. Am J Respir Crit Care Med 1999; 160:1001–1008. |
| 87. | Polak JM, Van Noorden S. Immunocytochemistry: practical applications in pathology and biology. Bristol, UK: Wright & Sons; 1983. |
| 88. | Bullock GR, Petrusz P. Techniques in immunocytochemistry, Vols 1 & 2. London: Academic Press; 1982. |
| 89. | Hamid Q. Immunocytochemistry. In: Kay AB, editor. Allergy and allergic diseases. London: Blackwell Scientific; 1997, p. 775–783. |
| 90. | Christodoulopoulos P, Muro S, Hamid Q. Molecular pathology of allergic diseases. Chem Immunol 2000;78:159–165. |
| 91. | Giaid A, Michel RP, Stewart DJ, Sheppard M, Corrin B, Hamid Q. Expression of endothelin-1 in lungs of patients with cryptogenic fibrosing alveolitis. Lancet 1993;341:1550–1554. |
| 92. | Bentley AM, Meng Q, Robinson DS, Hamid Q, Kay AB, Durham SR. Increases in activated T lymphocytes, eosinophils and cytokine mRNA expression for interleukin-5 and granulocyte/macrophage colony-stimulating factor in bronchial biopsies after allergen inhalation challenge in atopic asthmatics. Am J Respir Cell Mol Biol 1993;8: 35–42. |
| 93. | Sabattini E, Bisgaard K, Ascani S, Poggi S, Piccioli M, Ceccarelli C, Pieri F, Fraternali-Orcioni G, Pileri SA. The EnVision ++ system: a new immunohistochemical method for diagnostics and research. Critical comparison with the APAAP, ChemMate, CSA, LABC, and SABC techniques. J Clin Pathol 1998;51:506–511. |
| 94. | Cao Y, Kopplow K, Liu GY. In-situ immuno-PCR to detect antigens. Lancet 2000;356:1002–1003. |
| 95. | Hamid Q, Azzawi M, Ying S, Moqbel R, Wardlaw AJ, Corrigan CJ, Bradley B, Durham SR, Collins JV, Jeffery PK, et al. Expression of mRNA for interleukin-5 in mucosal bronchial biopsies from asthma. J Clin Invest 1991;87:1541–1546. |
| 96. | Hamid Q, Waharton J, Terenghi G, Hassall CJ, Aimi J, Taylor KM, Nakazato H, Dixon JE, Biurnstock G, Polak JM. Localization of atrial natriuretic peptide mRNA and immunoreactivity in the rat heart and human atrial appendage. Proc Natl Acad Sci USA 1987;84: 6760–6764. |
| 97. | Cox KH, DeLeon DV, Angerer LM, Angerer RC. Detection of mRNAs in sea urchin embryos by in situ hybridization using asymmetric RNA probes. Dev Biol 1984;101:485–502. |
| 98. | Giaid A, Hamid Q, Adams C, Trenghi G, Polak JM. Non-isotopic RNA probes: comparison between different labels and detection systems. Histochemistry 1989;93:191–196. |
| 99. | Cameron L, Hamid Q, Wright E, Nakamura Y, Christodoulopoulos P, Muro S, Frenkiel S, Lavigne F, Durham SR, Gould H. Local synthesis of epsilon germline gene transcripts, IL-4, and IL-13 in allergic nasal mucosa after ex vivo allergen exposure. J Allergy Clin Immunol 2000; 106:46–52. |
| 100. | Hamid Q. In situ hybridization. In: Kay A, editor. Allergy and allergic diseases. London: Blackwell Scientific; 1997, p. 766–774. |
| 101. | Nakamura Y, Christodoulopoulos P, Hamid Q. Localization of IL-4 and IL-5 mRNA by in situ hybridization in bronchial biopsies. In: Chung KF, Adcock I, editors. Methods in molecular medicine: asthma mechanisms and protocols; 2000, p. 211–225. |
| 102. | Ying S, Durham SR, Jacobson MR, Rak S, Masuyama K, Lowhagen O, Kay AB, Hamid QA. T lymphocytes and mast cells express messenger RNA for interleukin-4 in the nasal mucosa in allergen-induced rhinitis. Immunology 1994;82:200–206. |
| 103. | Robinson DS, Hamid Q, Ying S, Tsicopoulos A, Barkans J, Bentley AM, Corrigan C, Durham SR, Kay AB. Predominant TH2-like bronchoalveolar T-lymphocyte population in atopic asthma. N Engl J Med 1992;326:298–304. |
| 104. | Yasruel Z, Humbert M, Kotsimbos TC, Ploysongsang Y, Minshall E, Durham SR, Pfister R, Menz G, Tavernier J, Kay AB, et al. Membrane-bound and soluble α IL-5 receptor mRNA in the bronchial mucosa of atopic and nonatopic asthmatics. Am J Respir Crit Care Med 1997;155:1413–1418. |
| 105. | Minshall E, Chakir J, Laviolette M, Molet S, Zhu Z, Olivenstein R, Elias JA, Hamid Q. IL-11 expression is increased in severe asthma: association with epithelial cells and eosinophils. J Allergy Clin Immunol 2000;105:232–238. |
| 106. | Davenport AP, Nunez DJ. Quantification of hybridization signal. In situ hybridization: principles and practice. Oxford: Oxford University Press; 1990. |
| 107. | Kerr JF, Wyllie AH, Currie AR. Apoptosis: a basic biological phenomenon with wide-ranging implications in tissue kinetics. Br J Cancer 1972;26:239–257. |
| 108. | Erjefalt JS, Greiff L, Andersson M, Matsson E, Petersen H, Linden M, Ansari T, Jeffery PK, Persson CG. Allergen-induced eosinophil cytolysis is a primary mechanism for granule protein release in human upper airways. Am J Respir Crit Care Med 1999;160:304–312. |
| 109. | Brewster CE, Howarth PH, Djukanovic R, Wilson J, Holgate ST, Roche WR. Myofibroblasts and subepithelial fibrosis in bronchial asthma. Am J Respir Cell Mol Biol 1990;3:507–511. |
| 110. | Gizycki MJ, Adelroth E, Rogers AV, O'Brien PC, O'Byrne PM, Jeffery PK. Myofibroblast involvement in the allergen-induced late response in mild atopic asthma. Am J Respir Cell Mol Biol 1997;16:664–673. |
| 111. | Irani AM, Bradford TR, Kelpley CL, Schjechter NM, Schwartz LB. Detection of MCTT and MCTC types of human mast cells by immunohistochemistry using new monoclonal anti-tryptase and anti-chymase antibodies. J Histochem Cytochem 1989;37:1509–1515. |
| 112. | Persson CG, Erjefalt JS. Eosinophil lysis and free granules: an in vivo paradigm for cell activation and drug development. Trends Pharmacol Sci 1997;18:117–123. |
| 113. | Erjefalt JS, Andersson JS, Greiff L, Korsgren M, Gizycki MJ, Jeffery PK, Persson CG. Cytolysis and piecemeal degranulation as distinct modes of activation of airway mucosal eosinophils. J Allergy Clin Immunol 1988;102:286–294. |
| 114. | Montefort S, Roberts JA, Beasley R, Holgate ST, Roche WR. The site of disruption of the bronchial epithelium in asthmatic and nonasthmatic subjects. Thorax 1992;47:499–503. |
| 115. | Godfrey RW. Human airway epithelial tight junctions. Microsc Res Tech 1997;38:488–499. |
| 116. | Elia C, Bucca C, Rolla G, Scappaticci E, Cantino D. A freeze-fracture study of human bronchial epithelium in normal, bronchitic and asthmatic subjects. J Submicrosc Cytol Pathol 1988;20:509–517. |
| 117. | Godfrey RW, Severs NJ, Jeffery PK. Structural alterations of airway epithelial tight junctions in cystic fibrosis: comparison of transplant and post-mortem tissue. Am J Respir Cell Mol Biol 1993;9:148–156. |
| 118. | Bousquet J, Lacoste JY, Chanez P, Vic P, Godard P, Michel FB. Bronchial elastic fibers in normal subjects and asthmatic patients. Am J Respir Crit Care Med 1996;153:1648–1654. |
| 119. | Payne D, Rogers A, Adelroth E, Guntapaldi K, Bush A, Jeffery PK. Reticular basement membrane thickness in children with difficult asthma [abstract]. Am J Respir Crit Care Med 2001;163:A19. |
| 120. | Laitinen LA, Laitinen A, Haahtela T. A comparative study of the effects of an inhaled corticosteroid, budesonide, and a beta 2-agonist, terbutaline, on airway inflammation in newly diagnosed asthma: a randomized, double-blind, parallel-group controlled trial. J Allergy Clin Immunol 1992;90:32–42. |
| 121. | Bozzola JJ, Russell LD. Electron microscopy. Boston: Jones and Bartlett; 1992. |
| 122. | Hayat MA. Basic techniques for transmission electron microscopy. Orlando: Academic Press; 1986. |
| 123. | Hyde DM, King TE Jr, McDermott T, Waldron JA Jr, Colby TV, Thurlbeck WM, Flint WM, Ackerson L, Cherniack RM. Idiopathic pulmonary fibrosis. Quantitative assessment of lung pathology: comparison of a semiquantitative and a morphometric histopathologic scoring system. Am Rev Respir Dis 1992;146:1042–1047. |
| 124. | Downey GP, Worthen GS, Henson PM, Hyde DM. Neutrophil sequestration and migration in localized pulmonary inflammation: capillary localization and migration across the interalveolar septum. Am Rev Respir Dis 1993;147:168–176. |
| 125. | Hyde DM, Plopper CG, Kass PH, Alley JL. Estimation of cell numbers and volumes of bronchiolar epithelium during rabbit lung maturation. Am J Anat 1983;167:359–370. |
| 126. | Hyde DM, Plopper CG, Weir AJ, Murnane RD, Warren DL, Last JA, Pepelko WE. Peribronchiolar fibrosis in lungs of cats chronically exposed to diesel exhaust. Lab Invest 1985;52:195–206. |
| 127. | Moffatt RK, Hyde DM, Plopper CG, Tyler WS, Putney LF. Ozone-induced adaptive and reactive cellular changes in respiratory bronchioles of bonnet monkeys. Exp Lung Res 1987;12:57–74. |
| 128. | Vesely KR, Schelegle ES, Stovall MY, Harkema JR, Green JF, Hyde DM. Breathing pattern response and epithelial labeling in ozone-induced airway injury in neutrophil-depleted rats. Am J Respir Cell Mol Biol 1999;20:699–709. |
| 129. | Worthen GS, Henson PM, Rosengren S, Downey GP, Hyde DM. Neutrophils increase volume during migration in vivo and in vitro. Am J Respir Cell Mol Biol 1994;10:1–7. |
| 130. | Jeffery PK, Bousquet J. Methods for assessment of airways inflammation. Eur Respir J Suppl 1998;11:1S–58S. |
| 131. | DeBoer WI, Hiemstia PS, Sont JK, De Heer E, Rabe KF, Van Krieken JH, Sterk PJ. Image analysis and quantification in lung tissue. Clin Exp Allergy 2001;31:504–508. |
| 132. | Gundersen HJ. Stereology of arbitrary particles: a review of unbiased number and size estimators and the presentation of some new methods. J Microsc 1986;122:235–237. |
| 133. | Howard CV, Reed MG. Unbiased stereology: three-dimensional measurement in microscopy. New York: Springer-Verlag; 1998. |
| 134. | Gundersen HJ, Osterby R. Optimizing sampling efficiency of stereological studies in biology or “do more less well!” J Microsc 1981;121:65–73. |
| 135. | Carroll N, Lehmann E, Barrett J, Morton A, Cooke C, James A. Variability of airway structure and inflammation in normal subjects and in cases of nonfatal and fatal asthma. Pathol Res Pract 1996;192:238–248. |
| 136. | Sont JK, Willems LN, Bel EH, van Krieken JH, Vandenbroucke JP, Sterk PJ. Clinical control and histopathologic outcome of asthma when using airway hyperresponsiveness as an additional guide to long-term treatment. The AMPUL Study Group. Am J Respir Crit Care Med 1999;159:1043–1051. |
| 137. | Richmond I, Booth H, Ward C, Walters EH. Intrasubject variability in airway inflammation in biopsies in mild to moderate stable asthma. Am J Respir Crit Care Med 1996;153:899–903. |
| 138. | Jeffery PK, Godfrey RW, Adelroth E, Nelson F, Rogers A, Johansson SA. Effects of treatment on airway inflammation and thickening of basement membrane reticular collagen in asthma: a quantitative light and electron microscopic study. Am Rev Respir Dis 1992;145:890–899. |
| 139. | Azzawi M, Bradley B, Jeffery PK, Durham SR, Frew AH, Assoufi B, Collins J, Kay AB. Coefficients of variation on counts of cellular infiltrate in bronchial biopsy in normal and asthmatic subjects [abstract]. Am Rev Respir Dis 1990;141:A500. |
| 140. | Weibel ER. Stereological methods: practical methods for biologic morphometry. London: Academic Press; 1979. |
| 141. | Weibel ER. Sterological methods. Vol. 2. Theoretical foundations. London: Academic Press; 1980. |
| 142. | Fujinaka LE, Hyde DM, Plopper CG, Tyler WS, Dungworth DL, Lollini LO. Respiratory bronchiolitis following long-term ozone exposure in bonnet monkeys: a morphometric study. Exp Lung Res 1985;8: 167–190. |
| 143. | Baddeley AJ, Gundersen HJ, Cruz-Orive LM. Estimation of surface area from vertical sections. J Microsc 1986;142:259–276. |
| 144. | Retamales I, Elliott WM, Meshi B, Coxson HO, Pare PD, Sciurba FC, Rogers RM, Hayashi S, Hogg JC. Amplification of inflammation in emphysema and its association with latent adenoviral infection. Am J Respir Crit Care Med 2001;164:469–473. |
| 145. | Geiser M, Baumann M, Cruz-Orive LM, Im Hof V, Waber U, Gehr P. The effect of particle inhalation on macrophage number and phagocytic activity in the intrapulmonary conducting airways of hamsters. Am J Respir Cell Mol Biol 1994;10:594–603. |
| 146. | Coxson HO, Rogers RM, Whittall KP, D'Yachkova Y, Pare PD, Sciurba FC, Hogg JC. A quantification of the lung surface area in emphysema using computed tomography. Am J Respir Crit Care Med 1999;159: 851–856. |
| 147. | Jeffery PK. Inflammation and remodeling in the adult and child with asthma. Pediatr Pulmonol 2001;21:3–16. |
| 148. | Ward C, Johns DP, Bish R, Pais M, Reid DW, Ingram C, Walters EH. Reduced airway distensibility, fixed airflow limitation, and airway wall remodeling in asthma. Am J Respir Crit Care Med 2001;164: 1718–1721. |
| 149. | Silkoff PH. Noninvasive measurement of airway inflammation using exhaled nitric oxide and induced sputum: current status and future use. Clin Chest Med 2000;21:345–360. |
| 150. | Rutgers SR, Timens W, Kauffman HF, Postma DS. Markers of active airway inflammation and remodelling in chronic obstructive pulmonary disease. Clin Exp Allergy 2001;31:193–205. |
| 151. | Maestrelli P, Saetta M, Dei Stefano A, Calcagni PG, Turato G, Ruggieri MP, Roggeri A, Mapp CE, Fabbri LM. Comparison of leukocyte counts in sputum, bronchial biopsies, and bronchoalveolar lavage. Am J Respir Crit Care Med 1995;152:1926–1931. |
| 152. | Grootendorst DC, Sont JK, Willems LN, Kluin-Nelemans JC, Van Krieken JH, Veselic-Charvat M, Sterk PJ. Comparison of inflammatory cell counts in asthma: induced sputum vs. bronchoalveolar lavage and bronchial biopsies. Clin Exp Allergy 1997;27:769–779. |
| 153. | Rutgers SR, Timens W, Kaufmann HF, van der Mark TW, Koeter GH, Postma DS. Comparison of induced sputum with bronchial wash, bronchoalveolar lavage and bronchial biopsies in COPD. Eur Respir J 2000;15:109–115. |
| 154. | Brightling CE, Bradding P, Symon FA, Holgate ST, Wardlaw AJ, Pavord ID. Mast-cell infiltration of airway smooth muscle in asthma. N Engl J Med 2002;346:1699–1705. |
| 155. | Johnson PR, Roth M, Tamm M, Hughes M, Ge Q, King G, Burges JK, Black JL. Airway smooth muscle cell proliferation is increased in asthma. Am J Respir Crit Care Med 2001;164:474–477. |
| 156. | Walters EH, Gardiner PV. Bronchoalveolar lavage as a research tool. Thorax 1991;46:613–618. |
| 157. | Haslam PL, Baughman RP. Report of ERS Task Force: guidelines for measurement of acellular components and standardization of BAL. Eur Respir J 1999;14:245–248. |
| 158. | Djukanovic R, Sterk PJ, Fahy JV, Hargreave FE. Standardized methodology of sputum induction and processing. Eur Respir J 2002;20(Suppl 37):1s–55s. |
| 159. | Wilson NM, Bridge P, Spanevello A, Silvermann M. Induced sputum in children: feasibility, repeatability, and relation of findings to asthma severity. Thorax 2000;55:768–774. |
| 160. | Green RH, Brightling CE, McKenna S, Hargadon B, Parker D, Bradding P, Wardlaw AJ, Pavord ID. Asthma exacerbations and sputum eosinophil counts: a randomised controlled trial. Lancet 2002;360: 1715–1721. |
| 161. | Ten Brinke A, de Lange C, Zwinderman AH, Rabe KF, Sterk PJ, Bel EH. Sputum induction in severe asthma by a standardized protocol: predictors of excessive bronchoconstriction. Am J Respir Crit Care Med 2001;164:749–753. |
| 162. | Kharitonov SA, Barnes PJ. Exhaled markers of pulmonary disease. Am J Respir Crit Care Med 2001;163:1693–1772. |
| 163. | Macklem PT. A theoretical analysis of the effect of airway smooth muscle load on airway narrowing. Am J Respir Crit Care Med 1996; 153:83–89. |
| 164. | Haley KJ, Drazen JM. Inflammation and airway function in asthma: what you see is not necessarily what you get. Am J Respir Crit Care Med 1998;157:1–3. |
APPENDIX
Bland and Altman (1) offer an excellent tutorial on the use of appropriate statistical techniques for assessing measurement agreement between different observers or methods (uniformity) and measurement repeatability within the same observer or method (reproducibility). Data analysis can be divided into three parts: (1) examining the characteristics of individual variables (e.g., endpoints), (2) describing the association between these variables, and (3) testing these associations. These points will be illustrated using a hypothetical randomized double-blind, placebo-controlled clinical trial in which endobronchial biopsies were collected at baseline and after a course of antiinflammatory therapy. In this example, the study objective was to estimate the mean (or median) effect of active drug treatment on mast cells, identified by their ultrastructural characteristics and TEM. The outcome measure was mast cell density, determined as cell profiles (only those cut with their nucleus seen in the plane of the section were counted) expressed per 0.1 mm2 tissue. Potentially important confounding variables included age and baseline disease severity, determined as FEV1 percent predicted.
Variables can be either categoric or numeric. Categoric variables fit into discrete groups or categories. Typically, they are characterized by displaying the frequency with which each group occurs and its proportionate representation of the total. Categoric variables can be further classified as “nominal” or “ordinal.” Nominal variables are those with only two categories (e.g., treatment and age groups in Table 1
Variable | Placebo (n = 14) | Active (n = 16) | Difference | p Value |
|---|---|---|---|---|
| Age, n (%) | ||||
| ⩾ 65 years | 6 (43) | 8 (50) | – | 0.70|| |
| < 65 years | 8 (57) | 8 (50) | – | |
| FEV1 %pred, n (%) | 0.64** | |||
| < 30 | 3 (21) | 2 (13) | – | |
| 30–49 | 8 (57) | 12 (75) | – | |
| ⩾ 50 | 3 (22) | 2 (12) | – | |
| Mast cell density, cells/0.1 mm2 | ||||
| Median (95% CI)* | 80.0 (22.9 to 158.3) | 46.4 (24.4 to 77.8) | −5.2‡ (−80.0 to 25)¶ | 0.71†† |
| Geometric mean (95% CI)† | 55.6 (24.5 to 126.3) | 50.3 (29.3 to 86.3) | 0.90§ (0.36 to 2.25)† | 0.82† |
Numeric variables are those that take values within a given (valid) range. They can be subclassified as discrete if they can take only integer values (e.g., raw cell counts) or continuous if they can take any value (e.g., cell density). Numeric values are characterized by measures of central location, variability, and skew. For example, the left side of Figure 1
Figure 1. Distribution of cell density and log cell density at baseline (n = 30). Frequency histogram (left panel) shows cell density at baseline in the example study, indicating that the frequency distribution is skewed to the left (positive skew). Nonnormally distributed data are usually mathematically transformed in an attempt to normalize them. The log-transformed values (right panel) are more evenly distributed and approximate a normal distribution. 1-Dimensional scatter plot (top), frequency histogram (middle), summary statistics (bottom). 95% confidence interval (CI) calculated by *nonparametric binomial-based or **parametric t-distribution.
[More] [Minimize]As depicted in the Figure 1, baseline cell density is not normally distributed. Normally distributed data are preferred in statistical analyses because they permit easier identification of outlying data points, clearer differentiation between observations, and valid inferences when applying statistical methods that assume a normal distribution. For these reasons, nonnormally distributed data are usually mathematically transformed in an attempt to normalize them. The right side of Figure 1 depicts the log transformation of the cell density values. These log-transformed values are more evenly distributed and approximate a normal distribution.
The disadvantage in mathematically transforming data is that it alters the interpretation of a unit change in the transformed parameter. Trivial transformations of a variable (e.g., X to X + 50) do not adversely alter the meaning of the variable. However, more complex transformations, especially those that change the shape of the distribution, may leave the variable difficult to interpret. The ideal transformation is one that achieves the desired distribution while ensuring that estimates based on the transformed data remain interpretable.
For example, although 1/ and log(X) are both valid normalizations, the log transformation is preferable because it has multiplicative meaning (e.g., 0.69 increment in loge[X] always corresponds to a doubling of values on the original X scale). Log transformation of variables with zero values require a “trivial” modification (e.g., loge[X + 0.5]). If the proportion of zero values is large, the transformed data may look bimodal (two peaks) and will not be distributed normally. Results of analyses using transformed data should be back-transformed and presented in the original units (e.g., geometric mean cell density in cells/0.1 mm2, not mean log cell density [Figure 1, right panel]).
The next step in data analysis is assessing possible associations between two or more variables of interest. In the example study, association between variables is used to assess baseline characteristics, summarize study design, and estimate and characterize treatment effects.
Descriptive methods can be used to summarize the association between each treatment category and baseline variables, such as age, %FEV1, and cell density (Table 1). The difference in cell density between treatment groups can be estimated as either an absolute difference or a proportional difference (ratio). However, the estimated difference in median values is not calculated as the difference between medians. Hence, it does not correspond to 46.4 minus 80.0. Instead, it and its 95% CI are calculated based on an extension of the Mann–Whitney test, which does not require a normal distribution (2). With respect to the log-transformed data, the back-transformed difference between log cell densities is equivalent to the ratio of the geometric means (i.e., 50.3/55.6 = 0.90, with the 95% CI calculated using a t-test [Table 1]) (3).
Association between two numeric variables (e.g., pre- and posttreatment cell densities) are often summarized with a Pearson product-moment (if both variables are normally distributed) or a Spearman rank correlation coefficient (if both variables are not normally distributed) with the 95% CI calculated via the z-transform method (2). Values vary between −1 and 1, with 0 indicating no association. In the example study, the Spearman rank correlation between baseline and posttreatment mast cell densities separately for placebo and active groups are rs = +0.38 (95% CI, 0.03 to 0.65) and rs = −0.14 (−0.47 to 0.24), respectively.
Longitudinal plots, by linking data from the same subject, can be used to depict the relationship between baseline and posttreatment values (Figure 2
Figure 2. Longitudinal plot of mast cell densities. Longitudinal plots can depict the relationship between the baseline and posttreatment values. The y axis is plotted using the log-transformed data of the example study. Qualitative evaluation reveals similar baseline means for (A) the placebo and (B) the active groups, similar between-subject variability for the active group compared with the placebo group at both baseline and posttreatment, an increase in mean cell density compared with baseline with placebo, and a decrease in mean cell density compared with baseline with active treatment. Open squares with ‘X’ represent mean values; error bars represent 95% CI.
[More] [Minimize]Up to now, the analyses in this hypothetical example have focused on estimation (e.g., the study sample median is used as an estimate of the unknown true population median, and the 95% CI is used to quantify the uncertainty of this estimation). Assessing the truth concerning associations requires hypothesis testing. Hypothesis testing starts by specifying two mutually exclusive hypotheses referred to as the null hypothesis and the alternative hypothesis. The probability of obtaining either the observed or a more extreme result in the study is then calculated, assuming the null hypothesis to be true. This probability, which falls between 0 and 1, is referred to as the p value. By convention, probabilities of 0.05 or less (or in some cases 0.01 or less) are assumed to be statistically significant, (i.e., so unlikely to have occurred by chance as to cause rejection of the null hypothesis and acceptance of the alternate hypothesis). Of note, the cutoff chosen for this statistical significance also defines the probability of a Type I error, i.e., rejecting the null hypothesis when it is in fact true. In contrast, failure to reject a false null hypothesis constitutes a Type II error. The probability of a Type II error is related to the power of the analysis (power equals 1 minus the probability of a Type II error) and is a function of sample size. Usually, a study should be designed to provide a power of at least 80%.
Sample size calculations can be approached from either a hypothesis testing or an estimation point of view. For hypothesis testing, power calculations are done to control the probability of a Type II error, assuming the alternative hypothesis to be true. With increasing sample size, the probability of a Type II error decreases, but the probability of a Type I error does not vary from 5% (when the significance level is 5%) regardless of sample size. The estimation approach uses precision calculations to control the maximum width of 95% CI. Large studies have smaller intervals and thus a good chance of estimating the true average magnitude of differences. Estimates from small studies have wider intervals and are more uncertain, irrespective of whether the findings are significant. Sample size calculations are usually based on a t-test. As such, in assessing biopsy data, they should be performed using log-transformed data to satisfy assumptions of normality.
A common statistical problem in the analysis of biopsy data is the inflation of Type I error brought about by assessing outcomes for multiple cell types. Ideally, the Type I error for the whole analysis, termed the ‘experiment-wise error rate,’ should not exceed 5%. However, when hypothesis tests are performed for three separate outcomes, each with a 5% Type I error rate, the experiment-wise error rate is inflated to 15%. The Bonferroni correction (4, 5) is a traditional method for capping the experiment-wise error rate at 5%. However, it has been shown to be overly conservative (experiment-wise error rate < 5%) as the correlation between the various outcomes being assessed increases (6). Holm's sequential testing procedure (7) and O'Brien's global test (8) are alternative methods with better overall validity.
The χ2 test (9) (see Age, Table 1) is frequently used to test hypotheses involving associations between categoric variables. However, this test makes a large sample approximation that requires the expected frequency in each cell of the tabulation to be greater than 5 to be valid. When this is not satisfied, the usual alternative is the Fisher's exact test (see %FEV1, Table 1) (9).
Differences in mast cell density at baseline can be assessed using a t-test and assuming equal variances on the log-transformed normalized data, because it is plausible that all subjects were sampled from the same population at baseline (Table 1). Alternatively, differences in the nonnormally distributed median data can be assessed using a Mann–Whitney test (9). Although this test is distribution-free, meaning that it does not assume any particular shape for the distribution of the data, it does assume that the distributions of the two groups being compared are similar.
Two additional issues must be considered prior to hypothesis testing of the primary objective. Namely, what is the comparative measure of treatment effect, and is the study treatment the only (important) source of variation in mast cell density following treatment?
Analysis of cell data usually assumes a multiplicative treatment effect—meaning that, on average, the posttreatment cell density for a subject on active therapy is X% lower (or higher) than on placebo—and is estimated as a ratio between means (or medians). Additionally, in the example study, baseline cell density, %FEV1, and age were a priori confounding variables that could bias the point estimate of treatment effect (and the position of the 95% CI, which centers around this point estimate) if imbalances between groups were not adequately adjusted for. In the example study, the cell density imbalance between groups at baseline was small (Table 1) and the point estimates for the treatment effect are similar with (ratio = 0.56) or without (ratio = 0.50) adjusting for this baseline imbalance (Table 2
Comparison | Placebo Mean (SE) | Active Mean (SE) | Effect Estimate Mean (SE) | Ratio (95% CI) | p Value |
|---|---|---|---|---|---|
| Baseline | log (p0) = 4.01 (0.38) | log (a0) = 3.92 (0.25) | – | – | – |
| Posttreatment | log (p1) = 4.27 (0.35) | log (a1) = 3.58 (0.25) | log (a1/p1) = −0.69 (0.42) | 0.50 (0.21 to 1.19) | 0.11*/0.06† |
| Postbase change | log (p1/p0) = +0.25 (0.44) | log (a1/a0) = −0.34 (0.38) | −0.59 (0.57) | 0.56 (0.17 to 1.80) | 0.31*/0.29† |
Because the example study was randomized with important confounders well balanced between treatment groups, it is valid to maximize statistical power (and precision) by using the posttreatment-only analysis. This analysis demonstrates that, on average, active drug therapy reduced mast cell density by 50% relative to placebo (95% CI of ratio, 0.21 to 1.19), but that this reduction was not statistically significant (t-test, p = 0.11).
Data analysis should consist of both descriptive and statistical components. Thorough descriptions of the data are necessary before formal statistical tests are performed to understand the effects at work within the study data and to ensure validity of statistical comparisons.
| 1. | Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307–310. |
| 2. | Altman DG. Practical statistics for medical research. London, Chapman and Hall; 1991. |
| 3. | Altman DG, Gardner MJ, Machin D, Bryant TN. Statistics with confidence: confidence intervals and statistical guidelines. London, BMJ Books; 2000. |
| 4. | Campbell MJ, Machin D. Medical statistics—a commonsense approach. Chichester: John Wiley & Sons; 1993. |
| 5. | Hocberg Y, Tamhane A. Multiple comparison procedures. New York: Wiley; 1987. |
| 6. | Pocock SJ, Geller NL, Tsiatis AA. The analyses of multiple endpoints in clinical trials. Biometrics 1987;43:487–498. |
| 7. | Holm S. A simple sequentially rejective multiple test procedure. Scand J Stat 1979;6:65–70. |
| 8. | O'Brien PC. Procedures for comparing samples in multiple endpoints. Biometrics 1984;40:1079–1087. |
| 9. | Seigal S, Castallan NJ. Nonparametric statistics for the behavioural sciences. New York, McGraw-Hill; 1998. |
| 10. | Wei L, Zhang J. Analysis of data with imbalance in baseline outcome variable for randomized clinical trial. Drug Info J 2001;35:1201–1214. |
