Abstract
We studied circadian variation in FVC, FEV1, PEF, TLC, VC, and RV between 9:00 a.m. and 9:00 p.m. and analyzed how this variation affected estimated longitudinal change. Data from 876 adults were obtained in a longitudinal survey of samples from two Dutch areas. Subjects participated in four surveys held at 3-yr intervals between 1975 and 1985. FVC, FEV1, PEF, and VC increased from 9:00 a.m. until noon and decreased afterwards. TLC was constant over the day, whereas RV decreased from 9:00 a.m. to noon. Average variation in FVC, FEV1 and PEF, expressed as percentages of average level, was 4.8% (200 ml), 2.8% (86 ml), and 3.1% (250 ml / s), respectively. These results are compatible with circadian changes in airway size. No differences in variability were found between men and women. Significantly larger changes between 9:00 a.m. and noon were found in young adults, smokers, and those with respiratory symptoms than in other subgroups. Adjustment for diurnal variation reduced, albeit slightly, residual standard deviations of estimated longitudinal declines. Average diurnal change was large relative to underlying longitudinal change. Its effect on longitudinal change within an individual can therefore be large depending on age, smoking habits, symptomatology, number of visits, time of measurement, and difference in time between measurements. However, when people are measured at random times during the day for at least three visits, or when measurements are made after 11:00 a.m., effects of diurnal variation in pulmonary function on estimated average longitudinal decline are minimal.
Airway function is one of the many biologic functions that exhibit variability over 24-h periods. This variability has been found to lie at the basis of the worsening of lung function at night in patients with nocturnal asthma and, to a lesser extent, with COPD (1-4). As nocturnal asthma is common and troublesome (1, 2, 5), circadian variation in airway function has been of considerable interest in respiratory medicine. It has been recognized that diurnal variation in airway caliber occurs in healthy subjects as well (6-8). In these subjects, a variation of about 4% of the average early morning level has been found in the FEV1 (6), and of about 8% of the average level in the peak expiratory flow (PEF) (7). This would mean a diurnal change in FEV1 of 140 ml for a healthy man 44 yr of age, 1.70 m tall, and with a FEV1 of 3,500 ml; the equivalent of many years of longitudinal decline (9, 10). In asthmatics, nocturnal wheezing is associated with a much larger diurnal variability, with a time course identical to that seen in healthy subjects. This has for instance been demonstrated in PEF, where, in asthmatics, the change over 24 h may even be greater than 50% of the largest daily recording (11).
Despite the generally acknowledged role of circadian variation in lung function, in follow-up studies and, least of all, in epidemiologic studies of the general population, this variation is frequently not accounted for by study design or by taking time of measurement into account. As far as we know, diurnal variation in pulmonary function has never been demonstrated or systematically investigated in large (longitudinal) studies of the natural history of asthma and COPD. Knowledge about diurnal variation in lung function mainly derives from clinical and occupational studies (12, 13). Studies of healthy subjects and of differences between subgroups from the general population are rare (14). The number of subjects involved in such studies and the number of measurements made within a subject over the day usually are small, whereas the recording of the time of day often lacks detail (“morning, noon, evening, bedtime”). Of the ventilatory function indices, circadian variation in PEF has been studied most often (14). Studies of FEV1 and FVC are particularly scarce (12, 13), and those of TLC and RV are unknown to us.
The purpose of our study was to estimate the magnitude of diurnal variability in lung function between 9:00 a.m. and 9:00 p.m. on normal survey days from a longitudinal study of two large populations in the Netherlands. We investigated diurnal variation in lung size (TLC) and its subdivisions (RV and VC), and in measurements of the dynamics of the forced expiration (PEF, FEV1). We also studied differences in diurnal variation related to age, smoking habits, and respiratory symptoms, and the influence of diurnal variation on estimates of longitudinal decline in different subgroups.
The data were obtained in field surveys conducted at 3-yr intervals between 1965 and 1990 in a rural area (Vlagtwedde) and in an industrial town (Vlaardingen) in the Netherlands. Each survey was completed in 1 wk in October. Details of aims, study design, and sample selection have been described previously (15, 16). The Vlagtwedde sample was a dynamic cohort. In every field survey all men and women born between 1921 and 1952 and living in a specified area were invited to take part. The Vlaardingen sample was a fixed cohort selected from the municipal register. In 1965 a random sample of men and women born between 1901 and 1925 was selected, and in 1969 an additional sample was selected from those born between 1930 and 1954.
Anthropometric and lung function measurements were obtained in a standardized way by trained operators. btps measurements of FEV1, FVC, and peak expiratory flow (PEF) were derived from maximal expiratory flow-volume (MEFV) curves, using pneumotachography as described previously (17). Measurements and procedures complied with recommendations of the European Community for Coal and Steel (18). Warming up of equipment started at 7:30 a.m. Calibration was performed approximately every hour from 9:00 a.m. until 9:00 p.m.
Measurements of TLC, residual volume (RV), and the largest value of either a slow expiratory or inspiratory vital capacity (VC) were obtained by the single-breath nitrogen technique extended with a forced rebreathing method, and converted to btps conditions. A detailed description of this method and its validity is given by Sterk and colleagues (19).
FEV1 and inspiratory vital capacity (IVC) measurements were also recorded using water-sealed spirometers. These measurements were made on every subject on every occasion since 1965. Equipment, measurements, and procedures have been described in previous reports (16). These measurements were used in this study only to check the quality of the other measurements.
Data relating to respiratory symptoms and smoking habits were obtained by trained interviewers using the MRC/ECCS questionnaire (20). For the present analyses, a subject was assumed to have had respiratory symptoms when one or more of the following conditions applied: cough and phlegm persisting for more than 3 mo each year, wheezing on the chest and dyspnea if experienced at all, and/or asthma, defined as ever having had an attack of shortness of breath with wheezing on the chest. In the analyses of smoking habits, current smokers were contrasted with those who never smoked or had stopped smoking at least 1 mo before a survey.
A schematic overview of the data selection is presented in Table 1. From 1972 until 1985 MEFV-curves (17) and nitrogen single-breath tests were collected in a large random subsample of all participants (about 80% over all surveys) not described in the base protocol (16). Results described in this report relate to this subsample.
| n | Remarks | Period | ||
|---|---|---|---|---|
| 12,916 | records from 2,322 men and 2,073 women with complete anthropometry, questionnaire, and MEFV-curve data obtained on 45 whole and 10 half days in five surveys at each of the two locations. Of these, | 1972–1985 | ||
| 9,889 | were records from 1,778 men and 1,587 women with known time of measurement and obtained on a whole survey day (of which there were 36). Of these, | 1975–1985 | ||
| 5,147 | were records obtained on 18 whole survey days during which accurate measurement time could be estimated (see text) and | 1975–1979 | ||
| 4,742 | records obtained on 18 whole survey days during which exact time of measurement was recorded. Of the 9,889 measurements were | 1981–1985 | ||
| 3,504 | records from 448 men and 428 women who visited all four surveys and whose measurements did not have large longitudinal irregularities (see text). Of these, | 1975–1985 | ||
| 1,616 | were records from 354 men and 327 women with a TLC measurement at one to four surveys. | 1975–1985 |
A total of 6,855 records from 2,322 men and 6,061 records from 2,073 women with complete MEFV-curve and questionnaire data, obtained during any of the five surveys at each location, were available (Table 1). However, only data from the eight surveys (four at each location) between 1975 and 1985 were analyzed because for these the times of day at which the measurements were made were recorded or could be accurately reconstructed (see below). Also, we used only data from subjects who attended all four surveys and who were consistent smokers or nonsmokers during the whole observation period.
Subjects were excluded from analysis if measurements derived from the MEFV-curve maneuver (FEV1 and FVC) showed large longitudinal irregularities accompanied by large discrepancies from corresponding measurements from the water-sealed spirometer (FEV1 and IVC) and, if available, from VC measurements obtained during the assessment of TLC and RV. This resulted in 1,792 measurements from 448 men and 1,712 from 428 women for both pneumotachographically and spirometrically derived indices (Table 1). In this selection only 854 records from 354 men and 762 records from 327 women also comprised TLC, RV, and VC measurements. The latter measurements, done within about 15 min of the MEFV-curve measurements, were much more time-consuming. Hence, for practical reasons only, they could not be obtained on all participants.
Of the women selected, 33% came from Vlaardingen, 32% were smokers, and 36% reported respiratory symptoms. Of the men selected, 41% came from Vlaardingen, 59% were smokers, and 37% reported symptoms. Average height was 164 cm in women and 177 cm in men. The age range was 20 to 73 yr; the range from 1st to 99th age percentile was 23 to 68 yr.
At all surveys, the date and the part of the day (morning, afternoon, evening) at which the questionnaire was taken were recorded (Table 1). From 1981 onwards the exact time of measurement was recorded. From 1975 onwards every participant at each survey was numbered consecutively when reporting at the registration desk (running number).
In panel A of Figure 1, the distribution of measurements over a survey day is shown, averaged over 36 whole survey days in the two areas of investigation combined. For the surveys at which the exact time of measurement was recorded, this distribution was almost the same for all complete survey days but differed slightly between the two areas. The same held for the distributions of measurements over the day based on recording of only the part of the day (morning, afternoon, evening), and in particular for those surveys without an exact time recording.
Fig. 1. Variability in measurement times on a day. (A) Distribution of measurements over a typical survey day. (B) Distribution of measurement times averaged per subject over the four surveys. (C ) Distribution of the average difference between measurement times per subject. (D) Distribution of the maximal difference in measurement times per subject.
[More] [Minimize]From the surveys at which exact measurement times were recorded, it was clear that the running number closely followed the time of day. Hence, by applying the distribution of measurements over the day from these surveys to the surveys at which exact times of measurement were not known, using the running numbers and the date and part of the day at which the measurements were made, we could estimate the time of measurement. Applying this same procedure to the running numbers of the surveys at which measurement times were known, this estimation turned out to be accurate to within 15 min. This justified its use in subsequent analyses.
Normally, completing the questionnaire occurred within at most an hour of the lung function measurements. Sometimes, however, subjects took their questionnaires and returned (much) later, as indicated by the information on the questionnaire. Such cases were excluded from the analyses, except for those surveys in which the exact time of measurement was recorded.
We used analysis of covariance (ANCOVA) for unbalanced repeated measures (21) to model relations between the lung function variables and time of day, and to control for effects of confounding variables. We tried to impose little a priori constraints on the shapes of the dependencies of the lung function values on time of day by modeling these with spline functions (22). Age and standing height were included as time-varying covariables, and indicator variables for symptoms, smoking, sex, and area of residence as fixed covariables in the ANCOVA models, to adjust for the very uneven distributions of these variables over the day. Interaction terms between the splines and the variables for symptoms, smoking, area of residence, sex, and age were added in order to study the relation between lung function indices and the time of day in different subgroups. In the models for the VC, RV, and TLC these interaction terms could not be estimated as too few observations were available. The significance of contributions of the splines and of the interaction terms was tested with the Wald test.
Using the estimated model parameters, the overall average diurnal change in lung function values and the average diurnal change in various subgroups were plotted, together with 90% confidence limits, for an average (reference) subject of 44.5 yr, a height of 170 cm, from either area of residence, and from either sex. These curves were given a zero intercept at 9:00 a.m. This adjustment makes it easy to read off diurnal change with respect to the value at 9:00 a.m. but makes it impossible to judge the significance of the differences in diurnal variation between subgroups by comparing the confidence intervals. Therefore, Wald test p values concerning these differences are given in the text. Statistical significance of peaks and troughs within the same curve, however, can be judged somewhat informally by inspecting whether the confidence intervals at 9:00 a.m. and at the peak or trough do or do not overlap. This procedure can be used to compare lung function values on the same curve at any pair of times. The α level of this test corresponds to about 0.05.
Estimates of individual annual longitudinal change in FEV1, FVC, and PEF were obtained from simple linear regressions of ventilatory function data on age for each subject separately. These regressions were done both before and after adjustment of the lung function values for diurnal variation. A categorization of the sample into those who were younger than the average age of 44.5 yr and those who were older was based on each individual's age halfway through the 12-yr measurement period.
Further details of the statistical procedures are given in the .
In Figure 1, the distribution of measurements over an average survey day is shown in panel A, and some characteristics of how subjects varied their time of visit over the four surveys are shown in panels B to D. In panel B it is shown that the distribution of measurement times averaged within subjects over the four surveys from each area of residence had a smaller range than that on a typical day but still covered most of the range between 9:00 a.m. and 9:00 p.m., with a median time at 2:21 p.m. Over the four surveys the average time difference between measurements made on two surveys (panel C) was rarely less than 1 h, with a median difference of 3 h 18 min. In some cases this average difference was more than 7 h, close to the theoretical maximum of 8 h.
The distribution of maximal time differences between any two measurements of a subject (panel D) had a median value of 5 h 57 min. So, often subjects came at different surveys at different times of the day, and few reported at approximately the same time. This implies that if lung function changes with the time of day, this can even show up in the average lung function of an individual over four surveys. Moreover, these changes can confound longitudinal change, as the difference in measurement times between surveys was appreciable on average (panel C) and was sometimes large (panel D).
The overall diurnal variation in the lung function variables with 90% confidence limits is shown in Figure 2. Almost every lung function index showed a significant rise (or fall in the case of RV) from 9:00 a.m. to around noon when a maximum value (a minimum in the RV) was reached (Table 2). The exception to the overall trend was the TLC, which hardly varied during the day. Variation in TLC was especially small when related to the average TLC levels in men and women (Table 2). TLC was also the only lung function index in which the relation to the time of day did not contribute significantly in the ANCOVA models (p = 0.121), whereas the contributions of time of day in the models for the other variables were all highly significant (p < 0.001; RV, p = 0.004).
Fig. 2. Diurnal change in lung function variables from 9:00 a.m. to 8:00 p.m. The plots refer to an average subject 44.5 yr of age with a standing height of 1.70 m from either area of residence, male or female, smoker or not, and with or without respiratory symptoms.
[More] [Minimize]| Age | Smoking | Symptoms | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Overall† | 30 yr | 60 yr | Never | Current | Absent | Present | ||||||||||
| FEV1 | Men | 3,273 | 3,674 | 2,794 | 3,332 | 3,201 | 3,436 | 3,085 | ||||||||
| Women | 2,903 | 3,304 | 2,429 | 2,767 | 2,831 | 3,071 | 2,720 | |||||||||
| Change | 86 | 120 | 50 | 57 | 121 | 49 | 152 | |||||||||
| Time | 12:01 | 12:03 | 11:58 | 12:01 | 12:01 | 11:45 | 12:14 | |||||||||
| FVC | Men | 4,260 | 4,666 | 3,753 | 4,331 | 4,179 | 4,161 | 4,267 | ||||||||
| Women | 3,725 | 4,126 | 3,213 | 3,796 | 3,644 | 3,626 | 3,732 | |||||||||
| Change | 200 | 152 | 251 | 121 | 294 | 150 | 287 | |||||||||
| Time | 12:13 | 12:11 | 12:15 | 12:05 | 12:08 | 12:11 | 12:16 | |||||||||
| PEF | Men | 8,986 | 9,295 | 8,317 | 9,177 | 8,769 | 9,406 | 8,260 | ||||||||
| Women | 7,271 | 7,580 | 6,602 | 7,457 | 7,054 | 7,691 | 6,545 | |||||||||
| Change | 250 | 294 | 203† | 120† | 441 | 191 | 361 | |||||||||
| Time | 12:16 | 12:24 | 12:07 | 14:16 | 12:00 | 11:57 | 12:46 | |||||||||
| Overall values for | VC | RV | TLC | |||||||||||||
| Men | 4,890 | 2,177 | 7,040 | |||||||||||||
| Women | 4,142 | 2,106 | 6,221 | |||||||||||||
| Change | 168 | −144 | 74‡ | |||||||||||||
| Time | 12:28 | 11:59 | 13:55 | |||||||||||||
The times at which the highest values (or lowest in the case of RV) were found were at or slightly past noon and did not differ much between those lung function indices in which a significant peak or trough value was found (they varied from 11:59 a.m. to 12:28 p.m.) (Table 2). The observed overall maximal change in FEV1 (the difference between the value at 9:00 a.m. and the maximum occurring at about noon) was 86 ml, which was 2.8% of the average FEV1 level during the day (Table 2 and Figure 2). The similarly observed overall maximal change in FVC was 200 ml (4.8% of the average level), that in PEF was 250 ml/s (3.1%), and that in VC was 168 ml (3.7%). The overall maximal decrease in RV came to 144 ml (7.1%). After the maximum was reached, FVC, VC, and RV did not change much, whereas the FEV1 and PEF decreased during the afternoon and early evening to levels close to those at 9:00 a.m.
Deviations from the average trends over the day in FEV1, FVC, and PEF shown in Figure 2 are shown in Figures 3-5 for young and old adults (panels A), for smokers and nonsmokers (panels B), and for those with or without respiratory symptoms (panels C). The 90% confidence intervals in these figures can be used as before to test differences between lung function levels in the same group at various times. In Table 2, the maximal (and minimal for the RV) diurnal changes in lung function values with respect to the levels at 9:00 a.m. and their estimated times of occurrence are given, depending on age, smoking habits, and respiratory symptoms. We had too few TLC, RV, and VC measurements available to find significant relations between diurnal variation in these variables and age, smoking, and symptoms.
Fig. 3. Diurnal change in FEV1 from 9:00 a.m. to 8:00 p.m. in several subgroups. For each group, deviations from the average curves in Figure 2 are shown for young (30 yr) and old (60 yr) adults (A), for those who smoked or who did not smoke (B), and for those who were symptom-free or who had one or more symptoms (C ).
[More] [Minimize]
Fig. 4. Diurnal change in FVC from 9:00 a.m. to 8:00 p.m. in several subgroups. See the legend of Figure 3 for further explanation.
[More] [Minimize]
Fig. 5. Diurnal change in PEF from 9:00 a.m. to 8:00 p.m. in several subgroups. See the legend of Figure 3 for further explanation.
[More] [Minimize]The maximal change in FEV1 was larger in young adults than in older adults (Figure 3A, and Table 2). FEV1 in young adults hardly decreased from noon until the end of the survey day, whereas in elderly adults it decreased considerably in the afternoon and early evening (Figure 4). In the FVC, the elderly subjects had the larger maximal change at noon, after which their FVC declined somewhat until at 8:00 p.m. it equaled that in young adults whose FVC had remained at a constant level after the initial rise (Figure 4A, and Table 2). The larger peak in FVC in elderly subjects contrasts markedly with the much lower peak in their FEV1, especially in view to the also much lower average FVC and FEV1 levels as compared with the younger subjects (panels A in Figures 3 and 4, and Table 2). The trends in PEF over the day in both age groups (Figure 5A) were similar to those in FEV1, but the effect of age on diurnal variation was small and the interaction term of age with time of day was not significant (p = 0.185).
The maximal changes in FEV1, FVC, and PEF were consistently higher in current smokers than in nonsmokers (panels B in Figures 3-5, and Table 2). In nonsmokers, diurnal variability in FEV1 was negligible, whereas it seemed not to exist at all in PEF. The FVC of nonsmokers, however, increased significantly during the morning. The interaction of smoking with time of day was significant in all three ANCOVA models for FEV1, FVC, and PEF (FEV1 and PEF: p < 0.001, FVC: p = 0.005).
The maximal diurnal changes in FEV1, FVC, and PEF were clearly larger in subjects with respiratory symptoms than in those without (panels C in Figures 3-5, and Table 2). In subjects without symptoms, the PEF appeared not to change during the day, and changes in FEV1 were small. The effect of respiratory symptoms on diurnal variation was significant in FEV1 (p < 0.037) and in FVC (p = 0.041), but not in PEF (p = 0.183). Nevertheless, the same trends were observed in PEF as in FEV1.
The maximal change in FEV1 varied, depending on age, smoking habits, and respiratory symptoms, from 49 to 152 ml (Table 2). In terms of the average FEV1 level during the day, this was from 1.6 to 5.6% (Table 2 and Figures 3-5). The maximal change in FVC varied between 121 and 294 ml (2.7 to 7.6%), and in PEF between 120 and 441 ml/s (1.3 and 6.0%).
We found highly significant differences between men and women in the levels of the lung function variables. The patterns of diurnal variation in FEV1, FVC, and PEF were, however, the same in both sexes as indicated by nonsignificant interactions between sex and the splines with which time of day was modeled (p values > 0.102). Level differences between the two areas of residence were small and marginally significant in RV only (p = 0.039, other p values > 0.148). Differences between the Vlaardingen and Vlagtwedde in the patterns of diurnal variation were irregular (data not shown) and only significant for FEV1 (p = 0.012).
The annual longitudinal change of FEV1, FVC, and PEF in eight subgroups over four surveys is compared before and after adjustment for their relation with time of day in Table 3. For each subgroup, the average values of the annual change and their standard deviations are shown before and after adjustment. The change in the standard deviations caused by adjustment is expressed as percentage of the standard deviations before adjustment.
| ΔℓFEV1 | ΔℓFVC | ΔℓPEF | ||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Subgroup | Age (n) | A (SD) | B (SD) | %SD | A (SD) | B (SD) | %SD | A (SD) | B (SD) | %SD | ||||||||||
| Nonsymptomatic, nonsmokers | ||||||||||||||||||||
| < 44.5 | 35.5 | −10.2 | −9.7 | 5.8 | 7.4 | 28.4 | 31.0 | |||||||||||||
| (193) | (19.3) | (19.2) | 0.5 | (27.6) | (27.6) | 0.0 | (77.6) | (77.6) | 0.0 | |||||||||||
| > 44.5 | 54.4 | −20.6 | −20.0 | −11.0 | −8.7 | −29.2 | −25.9 | |||||||||||||
| (211) | (20.6) | (20.2) | 1.7 | (30.8) | (30.8) | 0.0 | (80.3) | (80.3) | 0.0 | |||||||||||
| Nonsymptomatic, smokers | ||||||||||||||||||||
| < 44.5 | 35.0 | −15.6 | −15.7 | 6.5 | 8.0 | 7.3 | 9.3 | |||||||||||||
| (184) | (23.4) | (23.0) | 1.7 | (29.8) | (29.8) | 0.0 | (96.3) | (96.1) | 0.2 | |||||||||||
| > 44.5 | 52.8 | −33.1 | −31.9 | −20.1 | −17.6 | −64.4 | −60.3 | |||||||||||||
| (113) | (25.7) | (24.6) | 4.4 | (37.7) | (37.7) | 0.0 | (116.1) | (116.0) | 0.1 | |||||||||||
| Symptomatic, nonsmokers | ||||||||||||||||||||
| < 44.5 | 36.9 | −10.2 | −10.3 | −0.4 | 0.4 | −2.2 | −2.2 | |||||||||||||
| (21) | (22.9) | (22.8) | 0.7 | (28.3) | (27.7) | 2.0 | (91.8) | (88.9) | 3.1 | |||||||||||
| > 44.5 | 55.1 | −23.4 | −22.4 | −13.4 | −11.5 | −44.3 | −40.7 | |||||||||||||
| (49) | (19.6) | (18.9) | 3.6 | (31.6) | (31.3) | 0.8 | (81.9) | (81.9) | 0.0 | |||||||||||
| Symptomatic, smokers | ||||||||||||||||||||
| < 44.5 | 35.0 | −23.0 | −23.6 | 0.5 | 1.2 | −32.1 | −32.3 | |||||||||||||
| (46) | (22.1) | (21.5) | 3.0 | (34.2) | (33.1) | 3.4 | ( 95.2) | (94.1) | 1.2 | |||||||||||
| > 44.5 | 53.8 | −40.5 | −38.9 | −19.8 | −17.4 | −97.0 | −92.0 | |||||||||||||
| (59) | (29.5) | (27.5) | 6.9 | (42.0) | (40.2) | 4.2 | (117.7) | (114.8) | 2.5 | |||||||||||
Longitudinal decline was larger in older subjects, in smokers, and in those with respiratory symptoms. On average, there still was a significant increase in FVC and PEF in adults younger than 44.5 yr who did not have symptoms.
This overall picture was changed negligibly after adjustment for time of day. Moreover, the decreases in the standard deviations, and hence in the standard errors of the means, were only slight. The adjustment thus changed neither the qualitative description nor the statistical significance of differences between the subgroups. Nevertheless, the improvement in the standard deviations was clearly smallest in nonsmokers without symptoms (who had the smallest maximum changes), and was largest in symptomatic smokers (who had the largest change).
Subjects older than 44 yr of age participated more often during the morning than did younger subjects. During the morning, the changes in lung function were larger (Figures 2-5), and the effect of adjustment for time of day may have been more effective, resulting in a larger decrease in standard deviations among older than among younger adults (Table 3).
In this study, we estimated the daily variation in FVC, FEV1, PEF, TLC, VC, and RV from data obtained over a period of several years in a longitudinal survey of two Dutch populations and assessed how this variation influenced estimates of longitudinal change.
The splines used in the analysis of covariance to model the dependence of the lung function variables on the time of day yielded curve shapes that were difficult to obtain otherwise (for instance with more conventional polynomials). Age, sex, smoking habits, and symptomatology were strongly related to the levels of lung function variables. Control for their systematic relation with the time of day was therefore imperative for studying within-subject circadian variation in different subgroups. Many other factors associated with differences in lung function levels between subjects are approximately constant within a series of longitudinal observations on one subject. Because of this, unexplained level differences between subjects, which might have been caused by unknown and unquantified (health) differences were “absorbed” by the random subject-specific intercepts (see ). However, as shown in panel B of Figure 1, the within-subject average measurement times still covered most of the time range between 9:00 a.m. and 9:00 p.m. This meant that many subjects did not contribute to the same extent to all times of the day, which led to the type of confounding described in Results. Panel B in Figure 1 also illustrates that much information on the dependence of lung function values on the time of day was present in the variation between subjects as well. This was also exploited in the random effects ANCOVA models.
Differences between the two areas of residence were small and largely statistically insignificant. The irregular differences we did find were most likely not attributable to any systematic difference between the areas but to effects of differences in times of measurement of various subgroups that were not completely adjusted for by other confounding variables in the models such as age, height, symptoms, and smoking. Because area of residence was included in all models, effects of other variables on diurnal variation were adjusted for differences between the two sites anyway. Concentrating the analyses on the data of those subjects who attended all four surveys for which times of measurement were recorded or could be reconstructed furthermore ruled out differences that might have resulted from the different sampling strategies used in Vlaardingen and Vlagtwedde.
There were considerable differences between average levels of lung function variables in men and women, even after adjustment for standing height (see Figure 2 for a man and a woman 44.5 yr of age and 1.70 m tall), but their diurnal variation did not differ at all. Expressing the diurnal change in lung function as a percentage of the average level, therefore, suggests differences between men and women that do not exist in absolute terms. This ambiguity is also clear from the diurnal variations in RV and VC. These are clearly each other's mirror image in absolute terms, but not when the diurnal changes are expressed relative to their average levels (Table 2 and Figure 2).
TLC did not vary between 9:00 a.m. and 9:00 p.m. Therefore, the daily variation in the other lung volumes and ventilatory flows was not the result of changes in TLC. Furthermore, the finding of a constant TLC makes it unlikely that substantial circadian changes occur in lung elastic recoil. Our results therefore suggest, as do other studies, that in healthy subjects circadian changes in airway caliber are the main determinant of diurnal changes in ventilatory function. In healthy subjects, airway resistance increases at night (11, 23). Barnes (13) has argued that the most likely explanation for this is an increase in parasympathetic tone at night that leads to bronchoconstriction because this is the only rapidly reversible mechanism.
It is suggested in Figures 3-5 that partly different mechanisms may underlie the separate effects of age, smoking habits, and respiratory symptoms on diurnal variation. Prevalences of smoking and respiratory symptoms normally increase with age, whereas smoking itself is also associated with a higher prevalence of symptoms (9, 17). In Figures 3-5 the effects of age, smoking habits, and symptoms on the diurnal variation in lung function are shown, each independent of the average effects of the other two.
The increase in FVC from 9:00 a.m. until noon was larger in elderly subjects than in young ones, unlike the increases in FEV1 and PEF (Figures 3-5). This suggests that in older subjects airway caliber at low lung volumes varies more within a day than at high lung volumes, which may be related to age-related loss of lung elastic recoil (19, 24, 25). In addition, as demonstrated in a wide variety of physiologic variables in both human and animal models, the most consistently observed change in circadian variation with age is a decrease in amplitude (26). This agrees with our findings for FEV1 and PEF.
A much more pronounced peak in FVC, FEV1, and PEF at noon was found in smokers than in nonsmokers (Figures 3-5). As in other studies (17), a strong relation between current smoking and hypersecretion existed in our populations (9, 17). Mucus retention at night may amplify the diurnal variation, as even in healthy subjects mucociliary clearance may be diminished at night (27). Airway plugging may also occur in smokers with healthy lungs. Deplugging may open up lung compartments and thus contribute to VC, FEV1, and PEF. Also, inflammatory processes accompanying small airways disease observed in many smokers, even in those without respiratory symptoms, may be a significant factor that amplifies the normal daily rhythm.
Most research relating respiratory symptoms to diurnal variation has been done in nocturnal asthma. Such patients make up a small part of our symptomatic population. Recent work suggests a more severe inflammatory process, and hence smaller airway caliber, during the night (28). This mechanism may also be important in subjects with COPD. There is also evidence that the circadian rhythm modulates the response to allergen exposure (29). Other mechanisms, probably of lesser importance, are airway cooling, gastroesophageal reflux, and impaired mucociliary clearance (13).
The effect of adjustment for the dependency on the time of day on the estimated yearly longitudinal changes and their standard deviations was surprisingly small. This is especially so when the often substantial range of measurement times within subjects over the four surveys (panel D in Figure 1) and the magnitudes of the daily changes, that were often equivalent to many years of longitudinal change (9, 10) (Tables 2 and 3), are taken into consideration. However, it should be noted that most standard deviations of the estimated annual within-subject changes were 0.7 to about 4 times (not counting a number of very extreme values) larger than the corresponding averages, and thus were large to very large, even after adjustment. Also, these standard deviations not only reflect all kinds of measurement error and intrasubject variation but also differences between subjects. It is therefore not to be expected that a simple overall adjustment for measurement time will yield a large improvement. The absence of substantial changes in the estimated longitudinal declines themselves was attributable to the fact that people changed their time of visit randomly over surveys. Obviously, when a follow-up study constitutes only two measurements over a relatively short time interval and with one measurement made early in the morning and the other close to or after 12:00 noon, the times of measurement may completely determine the changes in lung function found. Also, in longitudinal studies of, for instance, acute effects of (occupational) exposure to noxious substances on ventilatory function, in which several measurements on the same subjects are made at different times of the day, the effects of exposure may become severely confounded with diurnal variation.
In this longitudinal study, we observed substantial diurnal changes in FVC, FEV1, PEF, VC, and RV that take place between 9:00 a.m. and about noon. We also found that the magnitudes of these changes varied considerably with age, smoking habits, and respiratory symptoms. Moreover, the average diurnal changes were large relative to underlying longitudinal change. Obviously, therefore, within an individual diurnal variability may cause large errors in estimates of longitudinal change, depending on the number of measurements and the times at which these are made. However, our results indicate that if a longitudinal survey comprises at least three measurement occasions and subjects visit at random times over the day, or alternatively, if all measurements are made after about 11:00 a.m., the actual measurement times will have little influence on the estimated average longitudinal change. Hence, whereas within an individual estimated longitudinal change may be fraught with relatively large errors, such studies provide valid estimates of longitudinal change at the group level.
The writers thank the populations of Vlaardingen and Vlagtwedde for their kind cooperation throughout the years, and Dr. J. Bogaard from the lung function department, Dijkzigt Academic Hospital, Rotterdam for critically reading the manuscript.
Supported by Grants 242 and 92.21 from the Netherlands Asthma Foundation and by Grant 7280-03-013 from the European Community for Coal and Steel.
| 1. | Turner-Warwick, M. 1984. The definition and recognition of nocturnal asthma (with discussion). In P. J. Barnes and J. Levy, editors. Nocturnal Asthma. Oxford University Press, Oxford. 3–9. |
| 2. | Turner-Warwick, M. 1988. The management of chronic asthma. In P. J. Barnes, I. W. Rodger, and N. C. Thomson, editors. Asthma: Basic Mechanisms and Clinical Management. Academic Press Limited, London. 721–732. |
| 3. | AnonymousAsthma at night (editorial). Lancet11983220222 |
| 4. | Bateman J. R. M., Clarke S. W.Sudden death in asthma. Thorax3419794044 |
| 5. | Storms W. W., Bodman S. F., Nathan R. A., Byer P.Nocturnal asthma symptoms may be more prevalent than we think. J. Asthma311994313318 |
| 6. | Guberan E., Williams M. K., Walford J., Smith M. M.Circadian variation of F.E.V. in shift workers. Br. J. Ind. Med.261969121125 |
| 7. | Hetzel M. R., Clark T. J. H.Comparison of normal and asthmatic circadian rhythms in peak expiratory flow rate. Thorax351980732738 |
| 8. | Kerr H. D.Diurnal variation of respiratory function independent of air quality: experience with an environmentally controlled exposure chamber for human subjects. Arch. Environ. Health261973144152 |
| 9. | van Pelt W., Borsboom G. J. J. M., Rijcken B., Schouten J. P., van Zomeren B. C., Quanjer P. H.Discrepancies between longitudinal and cross-sectional change in ventilatory function in 12 years of follow-up. Am. J. Respir. Crit. Care Med.149199412181226 |
| 10. | Burrows B., Lebowitz M. D., Camilli A. E., Knudson R. J.Longitudinal changes in forced expiratory volume in one second in adults: methodologic considerations and findings in healthy nonsmokers. Am. Rev. Respir. Dis.1331986974980 |
| 11. | Hetzel M. R.The pulmonary clock [Editorial]. Thorax361981481486 |
| 12. | Cinkotai F. F., Sharpe T. C., Gibbs A. C. C.Circadian rhythms in peak expiratory flow rate in workers exposed to cotton dust. Thorax391984759765 |
| 13. | Barnes, P. J. 1989. Circadian rhythms in the respiratory system. In J. Arendt, D. S. Minors, and J. M. Waterhouse, editors. Biological Rhythms in Clinical Practice. Wright, London. 71–82. |
| 14. | Quackenboss J. J., Lebowitz M. D., Krzyzanowski M.The normal range of diurnal changes in peak expiratory flow rates. Am. Rev. Respir. Dis.1431991323330 |
| 15. | van der Lende R., Kok T. J., Peset R., Reig, Quanjer P. H., Schouten J. P., Orie N. G. M.Decreases in VC and FEV1 with time: indicators for effects of smoking and air pollution. Bull. Eur. Physiopathol. Respir.171981775792 |
| 16. | van der Lende, R. 1969. Epidemiology of chronic non-specific lung disease. Thesis, Van Gorcum, Assen. |
| 17. | van Pelt W., Quanjer P. H., Wise M. E., van der Burg E., van der Lende R.Analysis of maximum expiratory flow volume curves using canonical correlation analysis. Methods of Information in Medicine24198591100 |
| 18. | Anonymous. 1983. Standardized lung function testing. Report working party ‘Standardization of Lung Function Tests', European Community for Coal and Steel, Luxembourg. Bull. Eur. Physiopathol. Respir. 19(Suppl. 5):1–95. |
| 19. | Sterk P. J., Quanjer P. H., van der Maas L. L. J., Wise M. E., van der Lende R.The validity of the single-breath nitrogen determination of residual volume. Bull. Eur. Physiopathol. Respir.161980195213 |
| 20. | van der Lende R., Orie N. G. M.The MRC/ECCS questionnaire on respiratory symptoms: use in epidemiology. Scand. J. Respir. Dis.531972218226 |
| 21. | Jennrich R. I., Schluchter M. D.Unbalanced repeated-measures models with structured covariance matrices. Biometrics421986805820 |
| 22. | Ramsay J. O.Monotone regression splines in action. Stat. Sci.31988425461 |
| 23. | Martin R. J.Nocturnal asthma: structure and function. Chest1071995158S161S |
| 24. | Colebatch H. J. H., Greaves I. A., Ng C. K. Y.Exponential analysis of elastic recoil and aging in healthy males and females. J. Appl. Physiol.471979683691 |
| 25. | Turner J. M., Mead J., Wohl M. E.Elasticity of human lungs in relation to age. J. Appl. Physiol.251968664671 |
| 26. | Richardson, G. S. 1990. Circadian rhythms and aging. In E. L. Schneider and J. W. Rowe, editors. Handbook of the Biology of Aging, 3rd ed. Academic Press, New York. 275–305. |
| 27. | Bateman J. R. M., Pavia D., Clarke S. W.The retention of lung secretions during the night in normal subjects. Clin. Sci. Mol. Med.551978523527 |
| 28. | Martin R. J., Cicutto L. C., Smith H. R., Ballard R. D., Szefler S. J.Airways inflammation in nocturnal asthma. Am. Rev. Respir. Dis.1431991351357 |
| 29. | Mohiuddin A. A., Martin R. J.Circadian basis of the late asthmatic response. Am. Rev. Respir. Dis.142199011531157 |
| 30. | Dixon, W. J., M. B. Brown, L. Engelman, and R. I. Jennrich, editors. 1988. BMDP Statistical Software Manual. University of California Press, Berkeley. |
| 31. | Hastie T., Tibshirani R.Generalized additive models. Stat. Sci.11986297318 |
