INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

An analysis of longitudinal data from the Normative Aging Study showed an inverse relationship between rate of decline of FEV₁ and plasma cortisol concentration in smoking, ex-smoking, and nonsmoking men (1). In men in the lower third of the cortisol range the rate of decline of FEV₁ was on average eight times greater than in those in the upper third, but despite this striking difference baseline FEV₁ did not differ significantly between cortisol tertiles.

Therapeutic doses of systemically administered corticosteriods may stabilize airway function in COPD (2), perhaps by suppressing inflammatory changes that lead to damage (3), but studies of basal plasma cortisol in COPD have shown no significant differences from normal (4).

We have reexamined the association between endogenous corticosteroid status and lung disease in a cross-sectional and longitudinal study of a predominantly smoking population of both sexes, excluding subjects with known obstructive lung disease. We have looked at the relationship of aspects of lung structure and function to serum cortisol concentration and urinary cortisol excretion.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Seventy-four current cigarette smokers (5 or more cigarettes/d) aged 35 to 65 yr of age (mean age, 51.3 ± 7.5 yr; M 38, F 36) and 20 lifetime nonsmoking control subjects (mean age, 50.7 ± 8.1 yr; M 7, F 13) were recruited after publicity in local media. All subjects were white. Exclusion criteria were a FEV₁ < 1.5 L, history of asthma, bronchodilator or corticosteroid medication, and use of other tobacco products. The study was approved by the North Tees Local Research Ethics Committee. Subjects were recruited after reading a description of the study and possible risks. Consent was written. Subjects were informed of their results and encouraged to stop smoking.

Lung Function

Forced spirometry and CO transfer were measured seated using automated apparatus (Model TTUSA; PK Morgan, Chatham, Kent, UK). Expiratory flow-volume curves were recorded from a computerized dry rolling-seal spirometer. The forced expiratory maneuver was repeated until duplicate estimates of FVC and FEV₁ were within 5% of each other; we accepted the highest value. CO transfer was measured by the single-breath method (8) using a 9-s breathhold time. Duplicate measurements were accepted where estimates of TL_CO and effective alveolar volume (VA) were within 5%; CO transfer coefficient (KCO) was derived (= TL_CO/VA). Hemoglobin was measured, and in all cases was within the normal range.

Predicted lung function values used were from Roberts and colleagues (9) based on a similar white, urban, nonsmoking population.

Lung Structure

High resolution CT (HRCT) scanning was performed using an IGE Sytec 3000i CT scanner. Three 1-mm cuts from the upper, middle, and lower zones of the right lung were taken at TLC. Images were analyzed independently by two radiologists using the criteria of Remy-Jardin and colleagues (10).

Serum Cortisol

Measurements of serum cortisol were made (three measurements in 70, two in 15, and one in nine) on separate days. Subjects attended the laboratory between 8:15 and 9:00 A.M. after an overnight fast and an abstinence from smoking of at least 8 h. Subjects lay supine for 30 min before venepuncture. Blood samples were centrifuged immediately, and serum was frozen at 20° C until determination of cortisol concentration. Serum cortisol was measured by enzyme-immunological assay using a Boehringer Mannheim ES 300 immunoassay analyser (Boehringer Mannheim Immunodiagnostics, Lewes, Sussex, UK).

24-Hour Urinary Cortisol

Eighty-six subjects (19 nonsmokers, 67 smokers) successfully collected their urine for 24 h. Urinary volume was measured and the concentration of cortisol ascertained by means of a cortisol-¹²⁵I radioimmunoassay (Orion Diagnostica, Espoo, Finland).

Longitudinal Study

Fifty-seven subjects (M/F = 26 / 31; 39 smokers, 18 nonsmokers) attended for repeat measurement of lung function after a mean interval of 520 ± 69 d.

Statistics

The data were analyzed using a statistical package (SPSS Inc., Chicago, IL). Unadjusted data from smokers and nonsmokers were compared using an unpaired t test. Analysis of covariance and multiple linear regression were used to examine the relationships of lung functional and structural variables to cortisol. A p value < 0.05 was accepted as significant. In line with this, for the lower limit of normality of FEV₁ and KCO, we adopted (mean, 1.645 SD). Values below this will occur by chance on 5% of occasions in a normal population (9).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Cross-Sectional Study

Lung function and HRCT findings. The lung function variables FEV₁ and KCO were significantly reduced in the smoking group (Table 1). In 10 smokers FEV₁ was abnormally low, one of whom also had a low KCO; KCO was low in a further nine smokers. HRCT scanning demonstrated mild emphysema (involving < 25% of the lung parenchyma) in 18 smokers, severe emphysema (involving > 75% of parenchyma) in one. Six emphysematous smokers had a low KCO. Lung function and structure was normal in the nonsmoking volunteers.

Effect of smoking on cortisol. Mean serum cortisol, adjusted for age and sex, and urinary cortisol, adjusted for age, sex, and body weight did not differ significantly between smoking and nonsmoking subjects (Table 1).

Cortisol levels with lung damage. There were no significant differences in adjusted serum cortisol concentration between subjects with and without low FEV₁, low KCO or HRCT emphysema (Table 2). Serum cortisol tended to be higher in subjects with a low FEV₁ (p = 0.09). Adjusted urinary cortisol was 19% higher in subjects with HRCT emphysema (p = 0.05); otherwise there were no significant differences of urinary cortisol between groups, but there were tendencies for urinary cortisol to be higher in subjects with low KCO (p = 0.09). The mean coefficient of variation of repeat serum cortisol measurements was 12.8%.

Correlations of cortisol with FEV₁ and KCO. In multiple linear regression (with age, sex, and smoking status) log urinary cortisol/body weight showed a significant negative correlation with KCO (p = 0.000) (Table 3). The correlation was similarly significant with urinary cortisol and log urinary cortisol. Otherwise neither lung function variable showed any significant relationship to serum or urinary cortisol.

FEV₁ and KCO by tertile of serum and urinary cortisol. There were no significant differences in lung function between tertiles of serum cortisol (Table 4). However, a significant difference was seen in KCO between tertiles of urinary cortisol/ body weight, KCO being 21% lower in the highest tertile than in the lowest.

Longitudinal Study

Decline of lung function and cortisol. On multiple linear regression (Table 5) there were no significant relationships between decline of FEV₁ or KCO, and either serum or urinary cortisol. There were no significant differences in rate of decline of FEV₁ (adjusted for height, age, sex, smoking status, and initial FEV₁) or of rate of decline of KCO (adjusted for age, sex, smoking status, and initial KCO) between tertiles of serum or urinary cortisol (Table 6).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We have failed to demonstrate any relationship between serum cortisol and lung function. Specifically, we have not confirmed the previous observation (1) that FEV₁ declines more rapidly in subjects with low normal serum cortisol, nor could we demonstrate an effect of cortisol on baseline FEV₁. Our findings are therefore in line with studies showing normal plasma cortisol concentrations in COPD (4).

We have to consider why our study and that of Sparrow and colleagues (1) have come to such different conclusions. Volunteers for both studies were obtained by media advertisement, and may therefore not be typical of their populations. Our advertisements stated that we were studying the effects of smoking on the lung, and this may have encouraged participation of subjects who were concerned about their respiratory health, and it is therefore possible that our study population was enriched with smokers with incipient COPD. It is unclear whether selection bias of a comparable nature applied to Sparrow's volunteers. The study of Sparrow and colleagues was on smoking (25%), ex-smoking (42%), and never-smoking (33%) men, excluding subjects with chronic diseases, including bronchitis and asthma. Sparrow and colleagues (1) were also at pains to establish the normality of volunteers' endocrine and gonadal systems. We studied smokers (79%) and never-smokers (21%) of both sexes. We excluded ex-smokers and subjects with a diagnosis of asthma or COPD or receiving medication for these conditions, and subjects with a FEV₁ below 1.5 L. We made no selection on the basis of endocrine status. Mean age of subjects in the two studies were similar, and FEV₁ values were almost identical.

It is not stated whether any of the men in the study of Sparrow and colleagues had abnormally low lung function results. In our population, despite exclusion criteria, we found 10 smokers with an abnormally low FEV₁. Plasma cortisol tended to be higher in those with reduced FEV₁ (p = 0.09), the reverse of what would be expected from the data of Sparrow and colleagues.

Although Sparrow and colleagues (1) describe their study as prospective, it is part of a wider study on gonadal function performed 18 yr prior to publication (11). They do not state whether they set out with the hypothesis that serum cortisol is a factor determining rate of FEV₁ decline or whether the relationship emerged fortuitously from their analysis. This is very relevant to the statistical interpretation of their finding.

Sparrow and colleagues suggest that the rate of decline of FEV₁ in subjects in the lowest tertile of serum cortisol concentration is 71.6 ml/yr faster than in subjects in the highest tertile. They measured this over a 4.7-yr period, but clearly such a difference would have to operate for longer to produce significant airflow obstruction. If we make the modest assumption that this difference has been maintained in the study group for 10 yr we would expect to see a 715-ml difference between tertiles in the cross-sectional study. With the standard deviation of adjusted FEV₁ (700 ml) we have a 90% chance of detecting the difference at the 1% level of significance with 31 subjects in each tertile. Our observed difference was 0 ml. The likelihood of a Type II error is low. By the same token Sparrow and colleagues should have observed a significant difference between tertiles in their cross-sectional study. Their observed difference was 240 ml (p = 0.08) but some of this arose because their cross-sectional data were gleaned from both first and second lung function measurements (depending on which measurement was temporally closest to the cortisol measurement). Thus, their cross-sectional data include measurements affected by the rapid change observed in the longitudinal study. Clearly their cross-sectional data are not compatible with a substantial difference in rate of decline being maintained for more than a few years.

By comparison with the previous study (1) our longitudinal study lacks statistical power because we were not able to extend follow-up as long as we would have wished. The short interval between measurements has meant a larger standard deviation of the adjusted change in FEV₁ compared with the study of Sparrow and colleagues. We calculate that there was an 85% chance of finding a difference of the magnitude that Sparrow and colleagues found at the 0.05 level of significance. Our observed difference between tertiles was 21 ml/yr (higher in the lowest cortisol tertile). The difference was nonsignificant, but clearly there is a real possibility of a Type II error.

Urinary cortisol gives a better indication of the biologically active free cortisol (12) despite intersubject differences in the renal tubular handling of cortisol. We were unable to show any significant relationship between cortisol excretion and FEV₁ in either cross-sectional or longitudinal parts of the study. Unexpectedly, we have demonstrated an inverse relationship between urinary cortisol excretion and KCO; low KCO is associated with the highest tertile of urinary cortisol excretion. Although highly statistically significant (p = 0.001), the finding needs to be treated with caution since we had no prior hypothesis in this area. We assume that emphysema was the main cause of reduced KCO in this population; urinary cortisol excretion seemed to be significantly elevated in smokers with HRCT emphysema (p = 0.05), but a higher level of significance is required since we made multiple statistical comparisons of cortisol with smoking, lung function, and structure. With one exception, smokers with emphysema were asymptomatic, and it is therefore unlikely that cortisol excretion was increased by the stress of illness. Because there was a trend for urinary cortisol to be lower in smokers, it is unlikely that the relationship with emphysema is due to smoking.

No coherent pattern has emerged from the literature on the effect of smoking on serum or urinary cortisol. We found no significant differences in serum or urinary cortisol between smokers and nonsmokers. Others have found serum/plasma cortisol to be elevated (13) or lowered (16) or unchanged (1) in smokers, whereas urinary cortisol was found to be higher in smokers (17). Smoking induces oxidative mechanisms responsible for cortisol metabolism, but excretion of 6--hydroxycortisol was the same in smokers and nonsmokers (18).

In conclusion, we have failed to confirm the findings of Sparrow and colleagues (1), but we have demonstrated a tendency for urinary cortisol excretion to be higher in smokers with emphysema and in subjects with reduced KCO.