Abstract
Our purpose was to identify clinical, radiological and physiological (CRP) determinants of survival and to develop a CRP scoring system that predicts survival in newly diagnosed cases of idiopathic pulmonary fibrosis (IPF). The study population consisted of 238 patients with biopsy confirmed usual interstitial pneumonia. For each patient, clinical manifestations, chest radiographs, and pulmonary physiology were prospectively assessed. We used Cox proportional-hazards models to assess the effect of these parameters on survival. The effects of age and smoking were included in the analysis. Survival was related to age, smoking status (longer in current smokers), clubbing, the extent of interstitial opacities and presence of pulmonary hypertension on the chest radiograph, reduced lung volume, and abnormal gas exchange during maximal exercise. A mathematical CRP score for predicting survival was derived from these parameters. We showed that this CRP score correlated with the extent and severity of the important histopathologic features of IPF, i.e., fibrosis, cellularity, the granulation/connective tissue deposition, and the total pathologic derangement. Using these models, clinicians are in a better position to provide prognostic information to patients with IPF and to improve the selection of the most appropriate patients for lung transplantation or other standard or novel therapeutic interventions.
Keywords: idiopathic pulmonary fibrosis; usual interstitial pneumonia; prospective studies; pulmonary function tests; smoking physiopathology; survival rate
Death occurs within 5 yr of diagnosis in the majority of patients with idiopathic pulmonary fibrosis (IPF). Physicians caring for such patients are frequently required to make complex and difficult decisions regarding whether or not to start, intensify, or stop treatment; or when to recommend referral of the patient for lung transplantation, which has been shown to lead to improved lung function and confers a survival benefit in these patients (1). These decisions would be made easier if accurate and objective measurements of the patient's current clinical status and risk of progression to death were available. Reported prognostic indicators of survival in patients with IPF have been inconsistent in different studies. This is because most studies have been based on a relatively small number of patients, and they have included patients with other conditions such as connective tissue disorders (2-4). Most importantly, in the majority of reports, the diagnosis of usual interstitial pneumonia (UIP), the histopathologic subset of idiopathic interstitial pneumonia found in IPF, was frequently not confirmed by assessment of surgical lung biopsy (5-7). This is an important distinction since response to treatment and outcome varies among the different histologic subsets of idiopathic interstitial pneumonia.
We previously described a composite clinical-radiologic-physiologic (CRP) scoring system to evaluate the clinical status of patients with IPF (8). This CRP scoring system was recommended as a quantitative tool for the serial assessment of clinical impairment in patients with IPF (8, 9). However, the value of this CRP scoring system for predicting survival was not assessed. In the present study, we prospectively analyzed the clinical course of 238 patients with IPF. Our goal was to determine whether the risk of death caused by respiratory failure could be predicted based on clinical, radiologic, and physiologic parameters obtained during initial entry into the study. Based upon the features that were shown to be the best determinants of survival, we have devised a scoring system that predicts survival in patients with IPF.
The study group consisted of 238 patients with IPF prospectively enrolled into a Specialized Center of Research Study at the National Jewish Medical and Research Center (NJC) between 1982 and 1996 (Table 1). The diagnosis of IPF was made based on established clinical and histologic criteria (5-7). Patients were excluded from the study if there was clinical evidence of a connective tissue disease, left ventricular failure, an occupational or environmental exposure that may result in interstitial lung disease, or a history of ingestion of a drug or an agent known to cause pulmonary fibrosis. At the initial visit to the NJC, all subjects underwent clinical, radiographic, and physiologic assessment before lung biopsy. Informed consent was obtained from each patient, and the Institutional Human Subject Review Committee approved the protocol.
| Characteristic | All Patients (n = 238) | Never Smokers (n = 84 ) | Former Smokers (n = 121) | Current Smokers (n = 33) | ||||
|---|---|---|---|---|---|---|---|---|
| Age, yr†,‡,‖ | 61.4 ± 10.5 | 62.3 ± 11.3 | 62.6 ± 9.1 | 54.7 ± 10.7 | ||||
| Sex‡,§ | ||||||||
| Male | 152 (64%) | 33 (39%) | 96 (79%) | 23 (70%) | ||||
| Female | 86 (36%) | 51 (61%) | 25 (21%) | 10 (30%) | ||||
| Race | ||||||||
| Caucasian | 211 (89%) | 71 (85%) | 113 (93%) | 27 (82%) | ||||
| Hispanic | 21 (9%) | 10 (12%) | 7 (6%) | 4 (12%) | ||||
| Black | 3 (1%) | 1 (1%) | 0 (0%) | 2 (6%) | ||||
| Other | 3 (1%) | 2 (2%) | 1 (1%) | 0 (0%) | ||||
| Duration of illness, mo¶ | 24 (12–46) | 24 (12–43) | 24 (12–48) | 24 (12–48) |
The study population consisted of 152 men and 86 women, with a mean age of 61.4 yr (range, 27 to 79 yr), and 211 (89%) were Caucasian (Table 1). Subjects were designated as current smokers if they had smoked cigarettes regularly within the previous year (n = 33), former smokers if they had not smoked cigarettes in the previous year but had smoked in the past (n = 121), and never smokers (n = 84). It can be seen in Table 1 that there are significant differences in age and sex among the smoking groups: current smokers were younger than never or former smokers; and among never smokers, the male/female ratio was 0.65:1.
A modified American Thoracic Society (ATS) questionnaire was used to collect demographic and medical information (8). The type and amount of exertion required to precipitate shortness of breath was assessed using a previously described dyspnea scale (8). The duration of illness was defined as the time from the onset of the disease, as determined from either the patient's recollection of the first appearance of cough throughout the day or of dyspnea walking up inclines. The median duration of illness was 2 yr (Table 1). There was no significant difference in duration of illness among the smoking groups. The median length of follow-up was 20 mo, with a maximum of 14.8 yr.
After entry into the study, most patients received treatment for their illness with prednisone alone, cyclophosphamide alone, or both in combination. The combination was usually given to patients with more severe disease that progressed during follow-up. Treatment appeared to have little or no impact on survival compared with no treatment (data not shown). Survival was assessed through July 31, 1998. Deaths were identified from follow-up with the patient's family or physician or by search of the national death registry. The cause of death was obtained by review of the hospital discharge information (and, when available, the autopsy reports) on all deaths. There were 155 deaths during the study period (65%): 125 died secondary to IPF, 19 died because of other causes. In 11 patients, in whom the cause of death was unknown, it was considered to be due to IPF. One hundred five patients (44%) were censored, including 79 who were alive at time of the analysis, 13 who had undergone lung transplantation, 12 who died of a cause other than IPF, and one who was lost to follow-up.
Assessment of the chest radiograph was carried out as previously described (12). The severity and profusion of parenchymal interstitial opacities, the extent of honeycomb changes, and the presence or absence of evidence of pulmonary hypertension were graded on standard posteroanterior chest radiographs.
Physiologic assessment included measurement of thoracic gas volume (Vtg) and TLC; FVC and FEV1; the volume-pressure relationship of the lungs, and the single-breath diffusing capacity for carbon monoxide (Dl CO). These methods have been previously described (10-16).
In 205 patients, the volume-pressure relationship of the lungs was measured in a body plethysmograph as previously described (15-17). The coefficient of elastic retraction was calculated by dividing maximal static transpulmonary pressure by TLC. The normal value in our laboratory is 3 to 8 cm H2O/L. The volume-pressure data were also subjected to an exponential curve fit where volume is expressed as a percentage of observed TLC, and the exponential K, a constant that is proportional to the total incremental compliance, was derived from the equation: V = A − Be-Kp, where V is the volume at a given pressure (p), A is the maximal theoretical volume at infinite pressure, and B is A minus the intercept of the fitted exponential on the volume axis (17).
Respiratory frequency, tidal volume, expired gas concentrations, heart rate, and blood pressure and arterial blood gas tensions were determined at rest and while exercising on an electrically braked bicycle ergometer during incremental work loads (maximal exercise testing), and during steady-state exercise at 50% of the maximum work load achieved during the incremental exercise testing. Arterial blood gases were determined with blood electrodes. The aaPO2 was calculated from the simplified alveolar air equation (18).
Because approximately half of the patients required supplemental oxygen during exercise, the steady-state exercise PaO2 and aaPO2 were corrected for Fi O2 , using the equation:
Dead space to tidal volume ratio (Vd/Vt) was calculated using the Bohr equation and corrected for the additional mechanical dead space imposed by the experimental apparatus: Vd = [(PaCO2 − Pe CO2 )/PaCO2 ] × Vt minus (mechanical dead space of apparatus).
Oxygen consumption (V˙o 2) and maximal work load achieved were expressed as percent of predicted reference values obtained from the age- and sex-adjusted equations of Jones and Campbell (19).
All patients underwent open thoracotomy or video-assisted thoracoscopic lung biopsy. In each patient, tissue was obtained from at least two sites, the upper and lower lobes of the same lung (when technically feasible). Findings consistent with usual interstitial pneumonia (UIP) were present in the patients with IPF (5-7). In light of the recent change in the definition of the pathologic features of the idiopathic interstitial pneumonia all of the biopsies were reviewed (5-7) and those patients who demonstrated the histopathologic patterns of desquamative interstitial pneumonia (DIP), nonspecific interstitial pneumonia (NSIP), respiratory bronchiolitis associated with interstitial lung disease (RBILD), lymphoid interstitial pneumonia (LIP), diffuse alveolar damage, pulmonary Langerhans' cell granulomatosis, hypersensitivity pneumonitis, sarcoidosis, or idiopathic bronchiolitis obliterans organizing pneumonia (idiopathic BOOP) were excluded from the study. For each patient, a semiquantitative assessment of inflammatory/exudative changes, fibrotic/reparative changes, and airway alterations, in addition to an overall assessment of cellularity and fibrosis was performed as previously described (20, 21).
The median and interquartile ranges (IQR) were calculated for each explanatory variable originally considered for the CRP model. Differences among smoking groups were assessed using the Kruskal-Wallis test, followed by Wilcoxon's pairwise rank-sum tests if the Kruskall-Wallis test was significant. The Kaplan-Meier method (22) was used to produce estimates and plots for the patient cohort as a whole, and stratified by sex, age, and smoking status. Survival time was calculated as the number of months from the patients' initial visits until their death or time of censoring. Patients were censored if they: (1) were still alive at the last contact, (2) had received a single lung transplant (n = 11), double lung transplant (n = 1), or a heart and lung transplant (n = 1) or, (3) died from a cause other than IPF. When the cause of death was unknown (n = 11), it was considered due to IPF. The log-rank test was used to compare survival time between groups. The final score met the assumption of proportional hazards assumed by the Cox model.
Univariate and age- and smoking-adjusted analysis. The effect of each potential explanatory variable on the hazard function was considered in a univariate analysis using Cox proportional hazards regression. Because the univariate analysis indicated that age and smoking status (never, former, or current smoker) were predictive of survival, Cox proportional hazards regression was used to model the effect of the explanatory variables on the hazard function, adjusting for the effects of age and smoking status. Hazard ratios, the relative amount of risk associated with a specified increase in the explanatory variable, are reported for the age- and smoking-adjusted analyses.
Influential points, outliers that had a large effect on the fit of the regression model, were identified through exploration of plots and by using likelihood displacement (LD) and lmax statistics (23). Points that were considered influential observations (n = 16) were excluded from the univariate analysis and subsequent analyses. Although this represents 8% of the patients without missing data, according to the lmax statistic these patients accounted for 85% of the influence on the parameter estimates (data not shown).
Multivariable analysis. Variables with a p value of 0.25 or lower in the age- and smoking-adjusted analyses were considered for inclusion in the multivariable model. In order to avoid multicollinearity, Pearson's correlation was performed to detect variables that were highly correlated. If the Pearson's correlation coefficient was 0.60 or greater, the variable that was most significant in the age- and smoking-adjusted analysis was entered into the multivariable model. A forward elimination process was used to develop a preliminary multivariable model. All measured variables (including those that were eliminated because of high correlation with other variables in the model) were substituted into this preliminary multivariable model in order to develop a final model. Multivariable influential points were identified through LD and lmax statistics (24), and eliminated if necessary. Two models were developed using this process: one with complete data, and one for patients who did not complete exercise and lung mechanics testing.
Development of a clinical-radiologic-physiologic (CRP) scoring system. To develop a reproducible, quantifiable means for assessing the clinical status and potential prognosis of patients with IPF we computed a composite CRP scoring system using the models derived from the multivariable analysis. Values for each variable in a model were graded into two to nine levels of severity, based on the range of the data for each parameter. These categories were weighted so that all variables had equivalent minimum and maximum scores. The model was rerun with the grouped variables, and Akaike's Information Criteria (AIC) (24) was used to compare the model with the categorical variables to the model with the continuous variables. A final score, with a maximum of 100 points, based on the p value and relative hazard ratio of each categorized variable, was developed. The total score was tested for nonproportional hazards.
A second abbreviated CRP scoring system was developed in an analogous manner to that described above for the complete model; however, this abbreviated model did not include exercise or lung mechanics testing. This analysis led to an abbreviated scoring system that contained a subset of the score developed in the complete model. For this reason, the scoring system derived from the abbreviated model was compared with the scoring system derived from the complete model (using AIC). It was determined that the scoring system derived from the complete CRP model was superior to the abbreviated CRP scoring system.
Scores from the two models were compared using AIC on the subset of patients (n = 148) in whom data were available to derive scores for both models. Because the abbreviated CRP model was a subset of the first model, a likelihood ratio test was used to compare the two models. Pearson's correlation was used to examine the relationship between the newly developed complete and abbreviated CRP scores, the original published CRP score, and the semiquantitative evaluation of the lung histopathology. In addition, a Cox regression model was used to determine if absence of lung mechanics or exercise testing was significantly related to the risk of death.
All data analyses were performed using the SAS statistical package (SAS Institute Inc., Cary, NC). Unless otherwise noted, all tests were two-sided and performed at the 0.05 significance level.
The clinical, radiographic, and physiologic measurements, obtained at the time of the initial visit are shown in Tables 2 and 3, for the group as a whole, and relative to smoking status.
| Variable | All Patients | Never Smokers | Former Smokers | Current Smokers | ||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| (n) | Med* | IQR† | (n) | Med | IQR | (n) | Med | IQR | (n) | Med | IQR | |||||||||||||
| Clinical | ||||||||||||||||||||||||
| Dyspnea | 238 | 10.0 | 6.0–16.0 | 84 | 12.0 | 8.0–16.0 | 121 | 12.0 | 6.0–16.0 | 33 | 10.0 | 6.0–12.0 | ||||||||||||
| Clubbing, Yes/No‡,§ | 65/173 (27%/73%) | 13/71 (15%/85%) | 40/81 (33%/67%) | 12/21 (36%/64%) | ||||||||||||||||||||
| Inspiratory crackles | ||||||||||||||||||||||||
| None or Few/ Mod or Many‡,‖ | 46/192 (19%/81%) | 13/71 (15%/85%) | 19/102 (16%/84%) | 14/19 (42%/58%) | ||||||||||||||||||||
| Radiographic abnormalities | ||||||||||||||||||||||||
| Profusion, 0–18§,‖ | 238 | 17.0 | 16.0–17.0 | 84 | 16.7 | 16.0–17.0 | 121 | 17.0 | 16.3–17.3 | 33 | 16.7 | 16.0–17.0 | ||||||||||||
| Honeycomb change, 0–6§,‖ | 238 | 2.0 | 0.0–4.0 | 84 | 0.0 | 0.0–2.0 | 121 | 2.0 | 0.0–4.0 | 33 | 0.0 | 0.0–3.0 | ||||||||||||
| Pulmonary hypertension, Yes/No‡,§ | 47/191 (20%/80%) | 9/75 (11%/89%) | 29/92 (24%/76%) | 9/24 (27%/73%) | ||||||||||||||||||||
| Pulmonary function | ||||||||||||||||||||||||
| SVC, % pred‡,§,‖ | 222 | 66.0 | 53.0–79.0 | 80 | 58.5 | 46.5–72.5 | 116 | 69.0 | 56.5–78.0 | 26 | 83.5 | 74.0–90.0 | ||||||||||||
| Vtg, % pred‡,‖ | 234 | 66.0 | 57.0–77.0 | 84 | 64.5 | 55.5–73.5 | 118 | 65.0 | 56.0–75.0 | 32 | 86.0 | 68.5–96.0 | ||||||||||||
| RV, % pred§,‖ | 235 | 80.0 | 61.0–109.0 | 83 | 89.0 | 63.0–114.0 | 120 | 72.0 | 58.0–97.5 | 32 | 93.0 | 78.5–125.5 | ||||||||||||
| TLC, % pred‡,‖ | 237 | 71.0 | 62.0–83.0 | 84 | 69.0 | 59.5–77.0 | 120 | 69.5 | 60.0–83.0 | 33 | 89.0 | 77.0–96.0 | ||||||||||||
| FVC, % pred‡,§,‖ | 237 | 61.0 | 48.0–74.0 | 84 | 51.5 | 41.5–66.0 | 120 | 63.0 | 50.5–73.5 | 33 | 79.0 | 69.0–89.0 | ||||||||||||
| FEV1, % pred‡,§,‖ | 237 | 71.0 | 58.0–84.0 | 84 | 63.0 | 51.5–79.5 | 120 | 71.5 | 59.5–84.0 | 33 | 80.0 | 72.0–89.0 | ||||||||||||
| FEV1/FVC ratio‡,§,‖ | 228 | 83.0 | 76.5–89.0 | 82 | 87.0 | 81.0–92.0 | 116 | 82.5 | 76.0–87.0 | 30 | 73.0 | 68.0–80.0 | ||||||||||||
| Lung mechanics | ||||||||||||||||||||||||
| log of “K Value”‡,‖ | 201 | −2.4 | −2.8–−2.1 | 70 | −2.5 | −2.9–−2.2 | 101 | −2.4 | −2.8–−2.1 | 30 | −2.1 | −2.3–−1.9 | ||||||||||||
| Coefficient of retraction, cm H2O/L‡,‖ | 205 | 12.3 | 8.6–16.1 | 72 | 14.4 | 10.3–17.6 | 102 | 12.2 | 9.2–17.0 | 31 | 6.6 | 5.4–11.1 | ||||||||||||
| Upstream resistance, cm H2O/L/s | 184 | 3.0 | 2.0–4.5 | 62 | 3.2 | 2.4–5.0 | 94 | 3.3 | 2.0–4.5 | 28 | 2.5 | 2.0–3.3 | ||||||||||||
| Variable | All Patients | Never Smokers | Former Smokers | Current Smokers | ||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| (n) | Med* | IQR† | (n) | Med | IQR | (n) | Med | IQR | (n) | Med | IQR | |||||||||||||
| Diffusion | ||||||||||||||||||||||||
| Dl co, % pred | 233 | 49.0 | 36.0–60.0 | 82 | 52.0 | 40.0–61.0 | 119 | 46.0 | 33.0–59.0 | 32 | 50.0 | 39.0–61.5 | ||||||||||||
| Dl co/Va, % pred‡,§,‖ | 233 | 81.0 | 62.0–95.0 | 81 | 90.0 | 77.0–101.0 | 119 | 81.0 | 56.0–94.0 | 33 | 63.0 | 47.0–76.0 | ||||||||||||
| Resting | ||||||||||||||||||||||||
| Respiratory rate/min | 151 | 21.1 | 17.1–25.1 | 60 | 21.2 | 17.3–25.7 | 70 | 19.2 | 16.4–24.1 | 21 | 22.1 | 19.1–25.2 | ||||||||||||
| Vt btps,ml§ | 151 | 659 | 565–820 | 60 | 593 | 518–717 | 70 | 696 | 596–899 | 21 | 664 | 596–746 | ||||||||||||
| V˙ e btps, L/min | 151 | 13.7 | 11.7–16.2 | 60 | 12.9 | 10.9–15.5 | 70 | 14.3 | 11.8–16.9 | 21 | 15.3 | 12.4–16.4 | ||||||||||||
| Vd/Vt, % | 150 | 40.2 | 32.3–46.4 | 60 | 38.1 | 29.2–46.1 | 69 | 41.0 | 32.6–46.4 | 21 | 41.2 | 34.3–50.0 | ||||||||||||
| O2 saturation | 230 | 91.0 | 86.0–93.0 | 81 | 92.0 | 88.0–93.0 | 116 | 91.0 | 85.0–93.0 | 33 | 90.5 | 86.0–93.0 | ||||||||||||
| PaO2, mm Hg | 231 | 63.0 | 52.0–70.0 | 82 | 65.0 | 55.0–70.0 | 116 | 60.0 | 50.5–69.0 | 33 | 64.0 | 52.0–71.0 | ||||||||||||
| aaPo 2, mm Hg§ | 231 | 17.2 | 11.0–26.7 | 82 | 14.1 | 10.0–24.7 | 116 | 20.8 | 13.0–29.4 | 33 | 16.8 | 11.0–25.1 | ||||||||||||
| Maximal exercise | ||||||||||||||||||||||||
| Work, watts‡,§ | 217 | 72.0 | 46.1–105.5 | 78 | 58.9 | 36.5–90.9 | 108 | 75.9 | 50.1–106.3 | 31 | 90.4 | 55.0–106.5 | ||||||||||||
| Work load, % pred | 207 | 50.4 | 36.4–71.3 | 74 | 58.3 | 39.1–76.4 | 105 | 47.4 | 34.1–65.6 | 28 | 50.8 | 41.9–68.0 | ||||||||||||
| Respiratory rate/min§ | 208 | 41.2 | 33.2–50.5 | 75 | 46.0 | 35.3–53.3 | 105 | 39.6 | 30.7–47.3 | 28 | 38.6 | 29.9–49.6 | ||||||||||||
| Vt btps, ml‡,§ | 208 | 1257 | 912–1706 | 75 | 1000 | 767–1477 | 105 | 1295 | 1026–1708 | 28 | 1536 | 1222–1843 | ||||||||||||
| V˙ e btps, L/min | 208 | 55.0 | 41.4–75.4 | 75 | 53.0 | 35.7–71.8 | 105 | 56.9 | 42.5–75.2 | 28 | 60.0 | 45.8–84.2 | ||||||||||||
| Vd/Vt, % | 207 | 38.4 | 31.9–45.7 | 75 | 36.3 | 30.6–45.2 | 105 | 40.2 | 33.8–47.2 | 27 | 37.2 | 31.7–45.2 | ||||||||||||
| PaO2 , mm Hg§,‖ | 200 | 48.7 | 42.0–56.8 | 72 | 50.7 | 42.8–58.5 | 105 | 46.1 | 39.1–54.9 | 23 | 53.3 | 46.9–56.8 | ||||||||||||
| aaPo 2, mm Hg§,‖ | 199 | 40.5 | 32.7–50.8 | 72 | 37.5 | 31.7–45.8 | 104 | 45.1 | 34.7–53.9 | 23 | 39.1 | 30.8–44.8 | ||||||||||||
| Steady-state exercise | ||||||||||||||||||||||||
| Respiratory rate/min§ | 196 | 31.1 | 25.5–38.1 | 67 | 34.1 | 29.1–41.2 | 104 | 29.4 | 24.0–35.9 | 25 | 30.4 | 24.8–38.2 | ||||||||||||
| Vt btps, ml‡,§,‖ | 196 | 1012 | 763–1382 | 67 | 856 | 677–1200 | 104 | 1012 | 813–1344 | 25 | 1502 | 1128–1708 | ||||||||||||
| V˙ e btps, L/min‡,‖ | 196 | 37.6 | 29.5–46.5 | 67 | 36.2 | 28.0–44.4 | 104 | 36.8 | 29.9–45.8 | 25 | 43.9 | 39.4–53.6 | ||||||||||||
| Vd/Vt, % | 196 | 33.9 | 28.2–40.9 | 67 | 32.3 | 27.5–37.9 | 104 | 34.0 | 28.1–43.4 | 25 | 35.5 | 31.7–40.9 | ||||||||||||
| PaO2 , mm Hg§,‖ | 191 | 54.4 | 46.5–60.1 | 65 | 56.2 | 52.5–60.7 | 103 | 50.3 | 42.2–59.0 | 23 | 55.8 | 51.5–63.5 | ||||||||||||
| aaPo 2, mm Hg§,‖ | 190 | 35.4 | 28.0–46.6 | 64 | 33.0 | 25.2–40.2 | 103 | 39.7 | 29.3–52.1 | 23 | 34.0 | 28.1–41.9 | ||||||||||||
The dyspnea grade was similar for all smoking classes. Moderate or many crackles were present in 192 subjects (81%) and finger clubbing was seen in 65 (27%) subjects. Never smokers had significantly less frequent finger clubbing than did former or current smokers. There was a higher proportion of current smokers with none or few crackles heard on chest examination than never and former smokers.
The profusion of parenchymal interstitial opacities on the chest radiograph were similar among the subjects stratified by smoking status, with former smokers exhibiting the greatest amount of profusion. The extent of honeycombing was significantly greater in the former smokers than in the never and current smokers. The number of patients with evidence of pulmonary hypertension on the chest radiograph was significantly less in never smokers than in former or current smokers.
Lung volumes were significantly lower among never smokers than among former (SVC) or current (SVC, Vtg, and TLC) smokers. Residual volume (RV) was significantly higher in never smokers than in former smokers. In addition, former smokers had significantly lower SVC, Vtg, RV, and TLC than did current smokers.
In never smokers, the FVC and FEV1 were significantly lower and the FEV1/FVC ratio was significantly higher than in former smokers. Among current smokers, the FVC and FEV1 were significantly higher and FEV1/FVC ratio was significantly lower when compared with never and former smokers.
The median coefficient of elastic retraction was significantly higher in never and former smokers than in current smokers, the median for the latter was in the normal range (i.e., 3 to 8 cm H2O/l). In addition, there was a significant difference in calculated “K values” (natural log) with current smokers having higher values than never and former smokers.
The diffusing capacity was reduced in all groups (Table 3). Among current smokers, the Dl CO/VA was lower than in never and former smokers and was lower in former smokers than in never smokers.
The median respiratory rate at rest was increased; however, there were no significant differences among the three groups. The Vd/Vt was increased in all groups. Among former smokers, the resting PaO2 was lower (p = NS for both comparisons) and, the aaPO2 was higher compared with never (p < 0.05) and current smokers (p = NS). O2 saturation was the same in all groups. The median Vd/Vt for the entire study group was abnormally elevated but was similar among the three subgroups.
The exercise physiology is summarized in Table 3. As a group, the patients demonstrated a limitation in exercise tolerance with a decreased maximal work load (median, 50.4% expressed as percentage of predicted), elevated Vd/Vt, and abnormal gas exchange (decreased PaO2 and elevated aaPo 2). There were several significant differences during maximal exercise testing among the three groups: (1) never smokers were able to carry out exercise at a significantly lower work load (watts achieved) than were current and former smokers; however, the groups were similar in maximal work load achieved as a percentage of the predicted; (2) the respiratory rate was higher and tidal volume (Vt) was lower in never smokers when compared with former and current smokers; (3) the PaO2 and the aaPo 2 were significantly worse in former smokers than in never or current smokers. As expected, these differences persisted during steady-state exercise testing.
The Kaplan-Meier plot of survival probability from the reported time of onset of disease and from the initial visit at NJC for all patients can be seen in Figure 1.
Fig. 1. Kaplan-Meier plot of survival probability from the time of the onset of symptoms (median survival was 80.8 mo; 95% CI: 65.5–89.3) compared with the time from initial visit (median survival was 35.2 mo; 95% CI: 23.2–48.5) (n = 238).
[More] [Minimize]There was no difference in median survival (p = 0.15) between men (30.0 mo, 95% CI: 19.1–44.3) and women (39.3 mo, 95% CI: 27.2–77.8) (Figure 2).
Fig. 2. Kaplan-Meier plot of survival probability from the time of the initial visit stratified by sex (Female: n = 86, Male: n = 152).
[More] [Minimize]The effect of age at the time of the initial visit on survival time is shown in Figure 3. There was a significant difference in survival among the age groups (p < 0.0001). The effect of smoking status on survival is illustrated in Figure 4.
Fig. 3. Kaplan-Meier plot of survival probability from the time of the initial visit stratified by age group (younger than 50 yr, n = 36; 50–60 yr, n = 48; 60–70 yr, n = 108; older than 70 yr, n = 46). The median survival time was: younger than 50 yr of age, 116.4 mo; 50–60 yr of age, 62.8 mo; 60 to 70 yr of age, 27.2 mo; 70 yr of age or older, 14.6 mo.
[More] [Minimize]
Fig. 4. Kaplan-Meier plot of survival probability from the time of the initial visit stratified by smoking group (Current: n = 33, Former: n = 121, Never: n = 84 ). There was a significant difference (p = 0.0003) between the three smoking groups. The median survival time in the current smokers was 116.4 mo and it was 25.3 mo in former smokers and 27.2 mo in never smokers.
[More] [Minimize]The hazard ratios of the variables tested individually, and when adjusted for age and smoking are presented in Table 4. The degree of dyspnea; presence of finger clubbing; the extent of profusion and honeycombing, as well as evidence of pulmonary hypertension on chest radiograph; the severity of impairment of lung volume, spirometry, lung mechanics, Dl CO, and gas exchange while at rest and during exercise were all significant (p < 0.05) predictors of survival.
| Variable | Magnitude of the Change | Hazard Ratio* | 95% Confidence Intervals | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Lower | Upper | p Value† | ||||||||
| Clinical | ||||||||||
| Dyspnea | Δ4 (0–20 scale) | 1.23 | 1.06 | 1.43 | 0.0064 | |||||
| Clubbing | Y/N | 2.53 | 1.68 | 3.80 | < 0.0001 | |||||
| Crackles | None or few/ moderate or many | 1.39 | 0.87 | 2.22 | 0.1679 | |||||
| Radiographic | ||||||||||
| Profusion | Δ2 (0–18) | 2.48 | 1.66 | 3.71 | < 0.0001 | |||||
| Honeycombing | Δ1 (0–6) | 1.25 | 1.15 | 1.37 | < 0.0001 | |||||
| Pulmonary hypertension | Y/N | 1.96 | 1.29 | 2.99 | 0.0017 | |||||
| Pulmonary function | ||||||||||
| SVC, % pred | Δ10 | 0.75 | 0.66 | 0.85 | < 0.0001 | |||||
| Vtg, *% pred | Δ10 | 0.79 | 0.69 | 0.89 | 0.0001 | |||||
| RV, % pred | Δ10 | 0.91 | 0.87 | 0.96 | 0.0008 | |||||
| TLC, % pred | Δ10 | 0.70 | 0.62 | 0.79 | < 0.0001 | |||||
| FVC, % pred | Δ10 | 0.73 | 0.65 | 0.83 | < 0.0001 | |||||
| FEV1, % pred | Δ10 | 0.81 | 0.72 | 0.90 | 0.0001 | |||||
| FEV1/FVC | Δ5 | 1.17 | 1.04 | 1.32 | 0.0082 | |||||
| Lung mechanics | ||||||||||
| “K Value”, (log) | Δ0.10 | 0.95 | 0.92 | 0.98 | 0.0007 | |||||
| Coefficient of retraction | Δ0.5 | 1.04 | 1.03 | 1.05 | < 0.0001 | |||||
| cm H2O/L | ||||||||||
| Diffusion | ||||||||||
| Dl CO, % pred | Δ10 | 0.77 | 0.68 | 0.86 | < 0.0001 | |||||
| Dl CO/VA, % pred | Δ10 | 0.94 | 0.86 | 1.02 | 0.1294 | |||||
| Resting | ||||||||||
| O2 saturation | Δ5 | 0.69 | 0.60 | 0.80 | < 0.0001 | |||||
| PaO2 , mm Hg | Δ5 | 0.85 | 0.78 | 0.92 | 0.0001 | |||||
| aaPo 2, mm Hg | Δ4 | 1.16 | 1.08 | 1.24 | < 0.0001 | |||||
| Respiratory rate | Δ5 | 0.84 | 0.68 | 1.04 | 0.1037 | |||||
| Vd/Vt, % | Δ5 | 1.07 | 0.97 | 1.19 | 0.1763 | |||||
| Maximal exercise | ||||||||||
| Work | Δ20 watts | 0.89 | 0.81 | 0.98 | 0.0192 | |||||
| Work, % pred | Δ10 | 0.92 | 0.85 | 1.00 | 0.0383 | |||||
| Vt btps, ml | Δ100 | 0.96 | 0.92 | 1.00 | 0.0336 | |||||
| Vd/Vt, % | Δ5 | 1.11 | 1.01 | 1.22 | 0.0228 | |||||
| PaO2 , mm Hg | Δ5 | 0.74 | 0.67 | 0.82 | < 0.0001 | |||||
| aaPo 2, mm Hg | Δ4 | 1.20 | 1.12 | 1.28 | < 0.0001 | |||||
| Steady-state exercise | ||||||||||
| Work | Δ20 watts | 0.69 | 0.54 | 0.89 | 0.0041 | |||||
| Work, % pred | Δ10 | 0.78 | 0.65 | 0.93 | 0.0057 | |||||
| Vt btps, ml | Δ100 | 0.92 | 0.87 | 0.97 | 0.0038 | |||||
| Vd/Vt, % | Δ5 | 1.15 | 1.04 | 1.26 | 0.0046 | |||||
| PaO2 , mm Hg | Δ5 | 0.77 | 0.69 | 0.86 | < 0.0001 | |||||
| aaPo 2, mm Hg | Δ4 | 1.15 | 1.08 | 1.23 | < 0.0001 | |||||
| Respiratory rate | Δ5 | 1.07 | 0.98 | 1.16 | 0.1273 | |||||
| V˙ e/V˙ o 2 | Δ4 | 1.06 | 1.01 | 1.11 | 0.0196 | |||||
Multivariable analysis was used to identify those manifestations that were significantly related to survival while controlling for the other factors in the model. The complete model was developed from the subset of patients (n = 183) in whom there were data for all variables. The variables considered in the complete model are shown in Table 4. Many of the variables considered for inclusion in the multivariable analysis were highly correlated with each other (r ⩾ 0.60). Thus, in order to avoid multicollinearity, the variable that was most significant in the age- and smoking-adjusted univariate analysis was used as a surrogate for the correlated variables. Each variable was entered into the multivariable complete model in the following order in descending order of significance: PaO2 , TLC, FEV1, Work, Rest O2 sat, Vd/Vt, and Dl CO/Va (Table 5). This analysis resulted in the complete model that included: age, smoking history; clubbing; extent of profusion of interstitial opacities, and presence or absence of pulmonary hypertension on the chest radiograph; % predicted TLC; and PaO2 at the end of maximal exercise (Table 6).
| Variable Considered | Correlated with | Correlation Coefficient | ||
|---|---|---|---|---|
| FEV1, % pred | FVC, % predicted | 0.83 | ||
| SVC, % pred | 0.76 | |||
| Vt btps, ml at maximal exercise | 0.61 | |||
| TLC, % pred | SVC, % pred | 0.76 | ||
| FVC, % pred | 0.75 | |||
| Vtg, % pred | 0.73 | |||
| RV, % pred | 0.63 | |||
| Dl co/Va,% pred | Dl CO,% pred | 0.72 | ||
| Resting O2 saturation | Resting PaO2 , mm Hg | 0.91 | ||
| Resting aaPo 2, mm Hg | −0.83 | |||
| PaO2 , mm Hg at maximal exercise | aaPo 2, mm Hg at maximal exercise | −0.89 | ||
| PaO2 , mm Hg at 50% of maximal exercise | 0.85 | |||
| aaPo 2, mm Hg at 50% of maximal exercise | −0.83 | |||
| Resting PaO2 , mm Hg | 0.68 | |||
| Resting aaPo 2, mm Hg | −0.65 | |||
| Dl CO, % pred | 0.64 | |||
| Work, % pred at maximal exercise | Work, % pred at 50% of maximal exercise | 0.92 | ||
| Work, watts at maximal exercise | 0.62 | |||
| Work, watts at 50% of maximal exercise | 0.63 | |||
| Vd/Vt at maximal exercise | Vd/Vt at 50% of maximal exercise | 0.79 |
| Age Category | Score | Clinical | Radiographic | Physiologic | Total | |||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Smoking Status | Clubbing | Profusion | Pulmonary Hypertension | TLC (% pred ) | PaO2 at Maximal Exercise | |||||||||||||||||||||||
| Category | Score | Category | Score | Category | Score | Category | Score | Category | Score | Category | Score | |||||||||||||||||
| < 40 | 0 | Current | 0 | No | 0 | < 15 | 0 | No | 0 | ⩾ 80 | 0 | ⩾ 65 | 0 | |||||||||||||||
| 40–44.9 | 3.2 | Former | 10.2 | Yes | 10.7 | 15–17.49 | 9.15 | Yes | 10.3 | 70–79.9 | 2.75 | 60–64.9 | 1.5 | |||||||||||||||
| 45–49.9 | 6.4 | Never | 13.6 | ⩾ 17.5 | 18.30 | 60–69.9 | 5.50 | 55–59.9 | 3.0 | |||||||||||||||||||
| 50–54.9 | 9.6 | 50–59.9 | 8.25 | 50–54.9 | 4.5 | |||||||||||||||||||||||
| 55–59.9 | 12.8 | < 50 | 11.00 | 45–49.9 | 6.0 | |||||||||||||||||||||||
| 60–64.9 | 16.0 | 40–44.9 | 7.5 | |||||||||||||||||||||||||
| 65–69.9 | 19.2 | 35–39.9 | 9.0 | |||||||||||||||||||||||||
| 70–74.9 | 22.4 | < 35 | 10.5 | |||||||||||||||||||||||||
| ⩾ 75 | 25.6 | |||||||||||||||||||||||||||
| Maximum score possible | 25.6 | 13.6 | 10.7 | 18.3 | 10.3 | 11.0 | 10.5 | 100 | ||||||||||||||||||||
Because lung mechanics and exercise testing are not routinely performed in many medical centers, we developed a second abbreviated model that excluded the pulmonary mechanics and exercise data in 228 patients who had all other parameters assessed. This analysis resulted in an abbreviated model with identical variables, with the exclusion of PaO2 during maximal exercise. The levels of severity of disturbance of each variable used in the abbreviated model and the corresponding assigned points are shown in Table 6. In both models the score rises as the patient ages, develops clubbing, worsening profusion of opacities or evidence of pulmonary hypertension on chest radiograph, or physiologic impairment worsens. The maximal possible CRP score being 100 points in the complete model, and 89.5 points in the abbreviated model.
We were able to compare the ability to predict survival time of the previously reported CRP score (8) with the complete CRP score in 82 patients, and with the abbreviated score in 91 patients. The variables in the original CRP score included: rest aaPo 2; calculation of an exercise score; lung function (FVC, FEV1, and Vtg); dyspnea; chest radiographic extent of interstitial opacities; honeycomb change and presence or absence of pulmonary hypertension; and Dl CO/Va. Although both the complete and the abbreviated model correlated significantly with the original CRP score (r = 0.48, p < 0.0001; and r = 0.46, p < 0.0001; respectively), AIC indicated that the complete CRP model was superior to the abbreviated model in predicting survival, and that both were superior to the original CRP score. In addition, the likelihood ratio test showed that the complete CRP scoring system was superior (p = 0.03) to both the abbreviated and our previously reported CRP scoring system.
We have recently demonstrated that analysis and quantification of the specific histopathologic features found in UIP is useful in predicting the prognosis of patients with IPF (25). Specifically, survival was longer in subjects with lesser degrees of granulation/connective tissue deposition (fibroblastic foci) (25). We sought to determine whether the complete or original CRP scoring systems could be used as an indication of the severity of the underlying histopathology. We assessed the relationship between the pathologic alterations using a semi-quantitative scoring system (17, 21) and the CRP scores, in those patients (n = 77) in whom the semiquantitative analysis of pathology was performed. As can be seen in Table 7, there was no significant relationship between the original CRP score and any pathologic feature. On the other hand, there was a significant relationship between the complete CRP scoring system and the fibrosis factor (r = 0.25, p = 0.046), the cellularity factor (r = 0.25, p = 0.045), the granulation/connective tissue factor (r = 0.43, p < 0.001), and the total pathology score (r = 0.37, p = 0.003). The abbreviated CRP scoring system correlated with the fibrosis factor (r = 0.26, p = 0.021), the granulation/connective tissue factor (r = 0.49, p < 0.001), and the total pathology score (r = 0.38, p < 0.001).
| Pathology Factor Score | Original CRP Score (n = 31) | Abbreviated Score (n = 77) | Complete Score (n = 63) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| r | p Value | r | p Value | r | p Value | |||||||
| Fibrosis factor | 0.02 | 0.914 | 0.26 | 0.021 | 0.25 | 0.046 | ||||||
| Cellularity factor | 0.05 | 0.771 | 0.22 | 0.059 | 0.25 | 0.045 | ||||||
| Granulation/Connective tissue factor | 0.16 | 0.389 | 0.49 | < 0.001 | 0.43 | < 0.001 | ||||||
| Desquamation factor | 0.13 | 0.491 | 0.10 | 0.369 | 0.22 | 0.085 | ||||||
| Total Pathology | 0.08 | 0.664 | 0.38 | < 0.001 | 0.37 | 0.003 | ||||||
In the present study, we prospectively evaluated 238 patients with biopsy-proven UIP. Our purpose was to define the clinical, radiologic, and physiologic determinants of survival. We sought to devise a new clinical-radiologic-physiologic scoring system, based on these determinants, that could predict survival in newly diagnosed cases of IPF.
In agreement with other studies, our study confirms that IPF is predominantly a disease of elderly men, with a mean age at presentation of 61 yr and a male-to-female ratio of 1.77:1. Survival is markedly reduced, the median survival being 81 mo from the estimated onset of the illness and 35 mo from the time of the initial visit (Figure 1). We found that survival was significantly related to age at presentation, presence or absence of finger clubbing, cigarette smoking history, profusion of interstitial opacities and evidence of pulmonary hypertension on the chest radiograph, reduced lung volume, and gas exchange abnormalities with exercise. In contradistinction to previous reports (26), we found that dyspnea, sex, and diffusing capacity were not independent predictors of survival in the multivariate analyses.
Using the identified independent determinants of survival, we developed a mathematical clinical-radiologic-physiologic scoring system for prediction of survival time in patients with IPF. This new complete CRP score included the following parameters: age, smoking history; clubbing; extent of profusion of interstitial opacities and presence or absence of pulmonary hypertension on the chest radiograph; % predicted TLC; and PaO2 at the end of maximal exercise. Further, we demonstrated that this complete CRP score correlated with the extent and severity of the important histopathologic features of IPF: fibrosis, cellularity, the granulation/connective tissue, and the total pathologic derangement.
Our data show that survival is extended in patients with IPF who are cigarette smokers at the time of their initial evaluation (25) when compared with former smokers or never smokers. Several studies have shown a high percentage of ever smokers among persons with IPF (3, 9, 26-28) and that cigarette smoking is an independent risk factor for the development of IPF (29). However, being a cigarette smoker has only recently been shown to be associated with improved survival in this disease (25). Of interest, in a case control study of IPF (29), when current and former smokers were analyzed separately, the hazard ratio for developing IPF was not increased in current smokers.
The explanations for the longer survival in patients with IPF who are smoking cigarettes at the time of their initial presentation are unclear. It is possible that smokers seek attention earlier because of smoking-related symptoms or, conversely, they may be better able to tolerate their symptoms and seek attention when their disease has progressed enough to cause cessation of smoking. This study has demonstrated differences in several features between current smokers and former smokers or never smokers. Current smokers compared with never smokers had: (1) less crackles on chest examination and more frequent digital clubbing; (2) more evidence of pulmonary hypertension on chest radiograph; (3) less reduction in lung volumes and FEV1; (4) a lower coefficient of elastic retraction; as well as (5) the capability of higher exercise work loads, which was associated with better gas exchange. These data are consistent with our previous demonstration of differences in pulmonary function (17), specifically a shift upward and to the left of the volume-pressure curve of the lungs in smokers with IPF (30). It is possible that the higher lung volume and lower FEV1/ FVC ratio and lower coefficient of elastic retraction in current smokers reflect concomitant emphysema. The effect of these two disease processes may markedly influence lung function and alter the prognosis of patients with IPF.
Interestingly, cigarette smoking alters the histopathologic appearance of IPF, and this may partially explain the longer survival in current smokers (25). Current smokers showed lesser degrees of overall cellularity but greater extent and severity of alveolar space cellularity—likely reflecting increased inflammation secondary to ongoing accumulation of macrophages as a result of smoking. The fibrotic changes are similar among the smoking groups. However, the extent and severity of granulation/connective tissue scores was less in current smokers. There are no differences in the pathology factor scores between never smokers and former smokers. Limited data exist regarding the effect of cigarette smoke on fibroblast function; however, our findings are interesting in light of data that suggest that cigarette smoke inhibits lung fibroblast proliferation as well as chemotaxis, and may impair lung repair after lung injury (31).
In our previous study (8), we proposed a CRP scoring system in which points were arbitrarily assigned to various features thought to be useful in monitoring the course of IPF. In descending order of magnitude these were gas exchange at rest and during exercise; ventilatory function; dyspnea; chest radiographic extent of interstitial opacities, honeycomb change and presence or absence of pulmonary hypertension; and diffusing capacity. Despite the arbitrary apportionment of points, the original scoring system reflected the extent and severity of the disease process, as judged by a rough assessment of histopathologic changes. No attempt was made to correlate the original CRP score with survival.
In the present study, we used hierarchical multivariable analysis of clinical, radiologic, and extensive physiologic variables to develop a model that would allow clinicians to make more precise prognostic estimations about patients with IPF. The complete CRP scoring system was derived in 183 patients with IPF, and based on: age, smoking history, finger clubbing, the extent of profusion of interstitial opacities and evidence of pulmonary hypertension on the chest radiograph, the percent predicted TLC, and the PaO2 during maximal exercise. A second abbreviated CRP scoring system, determined in 228 patients with data that excluded the PaO2 during maximal exercise, was shown to be inferior by AIC analysis and a likelihood ratio test in predicting survival. Both new scoring systems were better at predicting survival time compared with our original CRP score. The predicted survival curves for patients with a given CRP score using the complete model or the abbreviated model are shown in Figures 5 and 6.
Fig. 5. Predicted survival curves for patients with a given CRP score using the complete model. We used the data generated to predict survival curves for patients with IPF and a CRP score at 10-point intervals as shown. For example, the calculated 5-yr survival at CRP scores of 20, 40, 60, and 80 were 89%, 53%, 4%, and < 1%, respectively.
[More] [Minimize]
Fig. 6. Predicted survival curves for patients with a given CRP score using the abbreviated model. We used the data generated to predict survival curves for patients with IPF and a CRP score at 10- point intervals as shown. For example, the calculated 5-yr survival at CRP scores of 20, 40, 60, and 80 were 85%, 41%, 1%, and < 1%, respectively.
[More] [Minimize]It is of interest to compare the clinical, radiologic, and physiologic features that made up the previous CRP score with those of the newly derived scoring system.
Clinical features. In the original CRP score, dyspnea was assigned a significant proportion of the maximum score (20%) (8). The severity of dyspnea was a significant factor in predicting survival in the univariate analysis, but it was not an independent predictor in the multivariate analysis. Neither age nor smoking status was considered in the original model. However, the new data show both to be highly significant factors in survival in IPF, age composing as much as 25.6% of the score, and there was a difference in survival times based on smoking status.
Radiologic features. The radiographic features (profusion of interstitial opacities, honeycomb change, and presence or absence of pulmonary hypertension) were assigned only 10% of the previous maximum total score. This assignment was based on the view that although profusion of lung opacities may be quantified visually, using a modification of the ILO criteria for pneumoconiosis (8), the technique was time-consuming, and had not been proven to be clinically useful (32). The present study has shown that profusion of lung opacities and the presence or absence of pulmonary hypertension deserve considerably more weight, i.e., 28.6% of the maximum in predicting survival. Unfortunately, HRCT scans were not obtained in the early years of this study; therefore, they were not included in the present analysis. However, given current understanding of the superiority of HRCT over the chest radiograph for assessing the changes in IPF, it is likely that the addition of a quantitative assessment of HRCT scans to both current models would result in a superior scoring system.
Physiologic features. The original CRP scoring system allotted 25% of the maximum score to FVC, FEV1, and Vtg. Many of these parameters were excluded from the current model because of their interdependence with other parameters (see Table 4). The complete model assigns as much as 11% for TLC in its calculation. Gas exchange alterations were weighed heavily in the original CRP score, i.e., 45% of the maximum of points, including as much as 5% for the Dl co/Va, 10% for the resting aaPo 2, and 30% for measures of exercise gas exchange. Resting gas exchange was not important in the current models; however, exercise PaO2 was significantly predictive of survival and accounted for as much as 10.5% of the maximum score in the complete model.
Finally, there are several possible limitations of this study. First, all patients were required to have a surgical lung biopsy; therefore, the study did not include patients who were too ill or considered at high risk for anesthesia or postoperative complications. However, we were very aggressive in obtaining biopsies in our patients and enrolled many patients older than 70 yr of age (n = 46). Subjects referred to our institution might represent a subset of patients with IPF with more aggressive disease. This seems unlikely given that our study population was similar to that of previous studies and the survival was also comparable to recent series (5, 33, 34). Also, only 158 of the 238 patients underwent assessment of both lung mechanics and exercise response from which the complete model was derived, and it is possible that this subset of patients on whom the complete scoring system was developed was different from the larger patient population. However, several models run with indicators of lung mechanics and exercise testing demonstrated that their absence was not indicative of increased risk of death (data not shown). In addition, in constructing the model there were several patients who were highly influential in the analysis. Because our goal was to develop a score for the general population of persons with IPF, and we did not want a few “outliers” to drive the model. We chose to focus on the middle of the distribution where most of the patients are, rather than the ends. We used a bootstrapping approach to validate our final model (data not shown). Although it is clear that the model performs better for the patients who were not highly influential, when these “outliers” are included in the model, it still performs well. Also, we performed a sensitivity analyses to look at the effect of censoring transplanted patients and patients who did not die of ILD (data not shown). The final model was run assuming that all of the transplanted patients died at the time of transplant. Because this model is not nested within the model assuming that the transplanted patients were censored, these models could not be compared directly. However, the likelihood ratio test and the Wald tests for the individual variables in the model remained significant when the transplanted patients and the patients who did not die of ILD were treated as if they had died at the time of transplant or had died of IPF. In fact, the model that treated the transplanted patients as having died had p values for all variables that was the same or lower than our original model. Similar results were found for the sensitivity analysis treating the subset of patients who did not die of ILD. In fact, the group of patients who died of other causes than ILD had a higher proportion of current smokers and lived longer than those patients who died of ILD. Therefore, by censoring this group we were being conservative. This group includes patients who lived long enough to die of causes other than ILD. Importantly, this model must be validated in another cohort of patients with IPF.
In summary, we have provided data derived from a large prospectively enrolled cohort showing which clinical, radiologic, and physiologic features of IPF influence survival. This analysis derived a complete CRP model that can be used to estimate the survival time in a patient with IPF. The model includes the following parameters: age, smoking history; clubbing; extent of profusion of interstitial opacities, and presence or absence of pulmonary hypertension on the chest radiograph; % predicted TLC; and PaO2 at the end of maximal exercise. This complete model was superior to a second abbreviated model and to our originally reported CRP in predicting survival in IPF. Nevertheless, the abbreviated grading system would likely have more general applicability since most physicians do not have access to measurements of cardiopulmonary exercise testing.
Applying these models, clinicians will be in a better position to provide prognostic information to patients with IPF and to improve the selection of the most appropriate patients for lung transplantation or other standard or novel therapeutic interventions. Importantly, we have identified several factors that influence survival that must be taken into account for use in any future therapeutic trial.
The writers thank Drs. L.C. Watters, T.L. Dunn, A. Shen, and R.L. Mortenson for their role in enrolling patients; Alma (Dolly) Kervitsky, S. Arlene Niccoli, Martin Wallace, Trudy McDermott, and Janet Henson for their technical assistance; Drs. Thomas V. Colby, James A. Waldron, Jr., Andrew Flint, and William Thurlbeck; Becki Bucher Bartelson for her assistance with the data analysis; William Kastner and David Ikle for the design and maintenance of the computerized data for the ILD study; the referring physicians; and a special thanks to the patients for allowing the investigators to participate in their care.
Supported by a Specialized Center of Research (SCOR) Grant No. HL-27353 from the National Heart, Lung and Blood Institute.
| 1. | Hosenpud JD, Bennett LE, Keck BM, Edwards EB, Novick RJEffect of diagnosis on survival benefit of lung transplantation for end-stage lung disease [see comments]. Lancet351909519982427 |
| 2. | Turner-Warwick M, Burrows B, Johnson ACryptogenic fibrosing alveolitis: response to corticosteroid treatment and its effect on survival. Thorax351980593599 |
| 3. | Turner-Warwick M, Burrows B, Johnson ACryptogenic fibrosing alveolitis: clinical features and their influence on survival. Thorax351980171180 |
| 4. | Tukiainen P, Taskinen E, Holsti P, Korhola O, Valle MPrognosis of cryptogenic fibrosing alveolitis. Thorax381983349355 |
| 5. | Bjoraker JA, Ryu JH, Edwin MK, Myers JL, Tazelaar HD, Schroeder DR, Offord KPPrognostic significance of histopathologic subsets in idiopathic pulmonary fibrosis. Am J Respir Crit Care Med1571998199203 |
| 6. | Katzenstein ALA, Myers JLIdiopathic pulmonary fibrosis. Clinical relevance of pathologic classification. Am J Respir Crit Care Med157199813011315 |
| 7. | American Thoracic Society. Idiopathic pulmonary fibrosis: diagnosis and treatment. International consensus statement. Am J Respir Crit Care Med 2000;161:646–664. |
| 8. | Watters LC, King TE, Schwarz MI, Waldron JA, Stanford RE, Cherniack RMA clinical, radiographic, and physiologic scoring system for the longitudinal assessment of patients with idiopathic pulmonary fibrosis. Am Rev Respir Dis133198697103 |
| 9. | Watters LC, Schwarz MI, Cherniack RM, Waldron JA, Dunn TL, Stanford RE, King Jr TE. Idiopathic pulmonary fibrosis: pretreatment bronchoalveolar lavage cellular constituents and their relationships with lung histopathology and clinical response to therapy. Am Rev Respir Dis 1987; 135:696–704. |
| 10. | Dubois AB, Botelho SY, Comroe JHA new method for measuring airway resistance in man using a body plethysmograph: values in normal subjects and in patients with respiratory disease. J Clin Invest351956327335 |
| 11. | Goldman HI, Becklake MRRespiratory function tests. Normal value at medium altitude and the prediction of normal results. Am Rev Tuberc791959457467 |
| 12. | Morris JF, Koski A, Johnson LCSpirometric standard for healthy nonsmoking adult. Am Rev Respir Dis10319715767 |
| 13. | Ogilvie C, Foster R, Blakemore W, Morgan JA standardized breathholding technique for the clinical measurement of diffusing capacity of the lung for carbon monoxide. J Clin Invest361957117 |
| 14. | Crapo RO, Morris AHStandardized single breath values for carbon monoxide diffusing capacity. Am Rev Respir Dis1231981185189 |
| 15. | Gibson GJ, Pride NBLung distensibility. The static pressure-volume curve of the lungs and its use in clinical assessment. Br J Dis Chest701976143183 |
| 16. | Thompson MJ, Colebatch HJDecreased pulmonary distensibility in fibrosing alveolitis and its relation to decreased lung volume. Thorax441989725731 |
| 17. | Cherniack RM, Colby TV, Flint A, Thurlbeck WM, Waldron JA Jr, Ackerson L, Schwarz MI, King Jr TE. Correlation of structure and function in idiopathic pulmonary fibrosis. Am J Respir Crit Care Med 1995;151:1180– 1188. |
| 18. | Riley RL, Cournand AIdeal alveolar air and the analysis of ventilation-perfusion relationships in the lungs. J Appl Physiol11949825847 |
| 19. | Jones NL, Campbell EJM. Clinical exercise testing. Philadelphia: W. B. Saunders Co; 1982, p 249. |
| 20. | Cherniack RM, Colby TV, Flint A, Thurlbeck WM, Waldron JA, King Jr TE, and BAL Cooperative Group Steering Committee. Quantitative assessment of lung pathology in idiopathic pulmonary fibrosis. Am Rev Respir Dis 1991;144:892–900. |
| 21. | Hyde DM, King TE Jr, McDermott T, Waldron Fr JA, Colby TV, Thurlbeck WM, Flint A, Ackerson L, Cherniack RM. Idiopathic pulmonary fibrosis: the quantitative assessment of lung pathology. Comparison of a semiquantitative versus a morphometric histopathologic scoring system. Am Rev Respir Dis 1992;146:1042–1047. |
| 22. | Collett D. Modelling survival data in medical research. London: Chapman & Hall; 1994. |
| 23. | Pettitt AN, Bin Daud I. Case-weighted measures of influence for proportional hazards regression. Appl Stat 1989;38:51–67. |
| 24. | Akaike H. A new look at the statistical model identification. IEEE Trans Auto Control 1974;AC-19:716–723. |
| 25. | King Jr TE, Schwarz MI, Brown K, Tooze J, Colby TV, Waldron JA Jr, Flint A, Thurlbeck W, Cherniack RM. The relation between histopathological features and mortality in a prospective study of patients with idiopathic pulmonary fibrosis. Am J Respir Crit Care Med (in press) |
| 26. | Schwartz DA, Helmers RA, Galvin JR, Van Fossen DS, Frees KL, Dayton CS, Burmeister LF, Hunninghake GWDeterminants of survival in idiopathic pulmonary fibrosis. Am J Respir Crit Care Med1491994450454 |
| 27. | Carrington CB, Gaensler EA, Coutu RE, Fitzgerald MX, Gupta RGNatural history and treated course of usual and desquamative interstitial pneumonia. N Engl J Med2981978801809 |
| 28. | Johnston IDA, Prescott RJ, Chalmers JC, Rudd RMfor the Fibrosing Alveolitis Subcommittee of the Research Committee of the British Thoracic Society. British Thoracic Society study of cryptogenic fibrosing alveolitis: current presentation and initial management. Thorax5219973844 |
| 29. | Baumgartner KB, Samet J, Stidley CA, Colby TV, Waldron JAthe Collaborating Centers. Cigarette smoking: a risk factor for idiopathic pulmonary fibrosis. Am J Respir Crit Care Med1551997242248 |
| 30. | Hanley ME, King TE, Schwarz MI, Watters LC, Shen AS, Cherniack RMThe impact of smoking on mechanical properties of the lungs in idiopathic pulmonary fibrosis and sarcoidosis. Am Rev Respir Dis144199111021106 |
| 31. | Nakamura Y, Romberger DJ, Tate L, Ertl RF, Kawamoto M, Adachi Y, Mio T, Sisson JH, Spurzem JR, Rennard SICigarette smoke inhibits lung fibroblast proliferation and chemotaxis. Am J Respir Crit Care Med151199514971503 |
| 32. | Terriff BA, Kwan SY, Chan-Yeung MM, Mueller NLFibrosing alveolitis: chest radiology and CT as predictors of clinical and functional impairment at follow-up in 26 patients. Radiology1841992445449 |
| 33. | Daniil ZD, Gilchrist FC, Nicholson AG, Hansell DM, Harris J, Colby TV, du Bois RMA histologic pattern of nonspecific interstitial pneumonia is associated with a better prognosis than usual interstitial pneumonia in patients with cryptogenic fibrosing alveolitis. Am J Respir Crit Care Med1601999899905 |
| 34. | Nagai S, Kitaichi M, Itoh H, Nishimura K, Izumi T, Colby TVIdiopathic nonspecific interstitial pneumonia/fibrosis: comparison with idiopathic pulmonary fibrosis and BOOP. Eur Respir J12199810101019 |
