SUMMARY Predictors of chronic obstructive pulmonary disease (COPD) have been identified in the prospective epidemiologic study of the population of ...
Risk of Chronic Obstructive Pulmonary Disease
Collaborative Assessment of the Validity of the Tecumseh Index of Risk1-3
MILLICENT W. HIGGINS, JACOB B. KELLER, J. RICHARD LANDIS, TERRI H. BEATY, BENJAMIN BURROWS, DAVID DEMETS, JOHN E. DIEM, IAN T. T. HIGGINS, EDWARD LAKATOS, MICHAEL D. LEBOWITZ, HAROLD MENKES, FRANK E. SPEIZER, IRA B. TAGER, and HANS WEILL
Introduction I n 1977, the National Heart, Lung and Blood Institute (NHLBI) issued a request for proposals (RFP) entitled "Development of an Index of Risk for Chronic Obstructive Lung Disease" (1). This was the first phase in a prevention and control program with the long-term goal of reducing the prevalence of smoking in those in the general population with greater than average risk of developing COPD. The RFP asked for "the design of a model for assessment of an individual's risk of chronic obstructive lung disease." The second phase of the program proposed to test the risk model, if one were developed, in selected small populations. The plan for the third and final phase was to "determine the effect of knowledge of personal risk on smokers enrolled in smoking cessation programs" (1). The index for appraising risk was to be based on existing epidemiologic data, and it was to provide a procedure for estimating personal risk of disability from COPD. An index of risk was developed at the University of Michigan using data collected in the Tecumseh Community Health Study. Initially, longitudinal data collected over a 15-yr period were used, and this index has been published (2). An index of risk of COPD developing within 10 yr was also prepared for the Tecumseh population (3). This was partly to see how well the 15-yr model predicted risk over a 10-yr interval, and partly to assess the contribution of additional, potential risk factors that had not been studied at the start of the 15-yr interval. The model and coefficients for 10- and 15-yr intervals were made available to investigators who had longitudinal epidemiologic studies supported by 380
SUMMARY Predictors of chronic obstructive pulmonary disease (COPD) have been identified in the prospective epidemiologic study of the population of Tecumseh, Michigan. Risk of developing COPD within 10 yr can be estimated from a profile that includes as risk factors age, sex, cigarette smoking habits, and forced expiratory volume in one second (FEV^. The index of risk placed 63% of male incidence cases and 64% of female incidence cases in the top 10% of the risk distribution and 81% of male and 86% of female COPD cases in the top 20% of the risk distribution for Tecumseh. The validity of the Tecumseh index of risk for other populations was determined in a collaborative investigation of data collected in longitudinal epidemiologic studies in Baltimore, Boston, Framingham, Louisiana, Staveley, and Tucson. The extent to which the risk model fitted these data sets was assessed by comparing predicted (or expected) onsets of COPD in each population with observations made in each study. The predictors of COPD identified in Tecumseh were shown to be the most important risk factors in the other populations as well. The goodnessof-fit of the index was satisfactory overall. In all populations, a high risk score was associated with an increased incidence of COPD, thus confirming the predictive ability of the risk index. The Tecumseh index of risk provides a practical method for developing a risk profile from answers to standard questions and simple tests of lung function. Risks are greatest for heavy cigarette smokers with reduced lung function, lower in smokers who stop smoking or reduce their cigarette consumption, and lowest of all in nonsmokers with above average lung function. Because cigarette smoking is the principal remediable risk factor, ability to identify high-risk men and women means that antismoking advice to the general public can be combined with a more vigorous approach to susceptible persons. AM REV RESPIR DIS 1984; 130:380-385
the NHLBI; the study populations were located in Baltimore, Boston, Framingham, Louisiana, Staveley (England), and Tucson. The purpose of this report is to present the Tecumseh models for calculating personal risk of developing COPD within 10 years and the results of using the models to predict onset of COPD in other populations. Methods For the purpose of these analyses, COPD was defined as obstructive airways disease manifested by a forced expiratory volume in one second (FEV,) less than 65% of the predicted value in combination with an FEV,/FVC ratio less than 80%. The validity of the diagnosis of COPD was confirmed in an independent clinical evaluation of patients and of healthy control subjects in Tecumseh (2).
(Received in original form November 4, 1983 and in revised form April 13, 1984)
1
From the University of Michigan, School of Public Health, Ann Arbor, Michigan; the Johns Hopkins University, School of Public Health, Baltimore, Maryland; the University of Arizona, College of Medicine, Tucson, Arizona; the Mathematics and Applied Statistics Branch, National Heart, Lung and Blood Institute; Tulane University, New Orleans, Louisiana; and Channing Laboratory, Brigham and Women's Hospital, Boston, Massachusetts. 2 Supported in part by Grants No. HL-18632, No. HL-14153, No. HL-14136, No. HL-15092, and Contracts No. NOl HR-82927 and No. NOl HR 8-2928 from the National Heart, Lung and Blood Institute. 3 Requests for reprints should be addressed to Millicent W. Higgins, M.D., the University of Michigan, School of Public Health, Department of Epidemiology, 109 Observatory Street, Ann Arbor, MI 48109.
381
RISK OF CHRONIC OBSTRUCTIVE LUNG DISEASE
Measures of ventilatory capacity were based on results of several maximal forced expirations following maximal inspirations. In Tecumseh, Wedge® spirometers (Med-Science Electronics, Burlington, MA) and recorders produced tracings of volume and flow on paper running at a speed of 25 mm/s. A minimum of 4 and a maximum of 6 tests were done to obtain 2 satisfactory tracings. The 2 tracings with the largest values of FVC plus FEVj were measured in accordance with the recommendations of the Snowbird Workshop on Standardization of Spirometry (4). Methods used in the other populations have been described (5-14). Predicted (or normal) values of FEV, were developed by regressing observed values of asymptomatic nonsmokers on age and height for each sex and each examination separately. Observed values were expressed as percent of predicted for each subject in each study population using denominators specific to each population, except that predicted values from Tecumseh were used for Staveley and Framingham. Thus, differences in measurement techniques among populations were partly controlled for, except in Staveley and Framingham. However, differences may exist because of variation in definition and selection of healthy subjects. Persons with an FEV! in the range of 65 to 69% of predicted were considered to be borderline abnormal, and they, as well as patients with COPD were excluded from baseline populations at risk of developing COPD. Persons with FEV, values in the borderline abnormal range at follow-up were included in the populations but not counted as cases of disease. Two subjects with a follow-up FEV, less than 65% of predicted and an FEV,/FVC ratio greater than 80% were excluded from the Tecumseh population as possible cases of restrictive lung disease. Predictors or risk factors for COPD were identified in the Tecumseh population by calculating age- and sex-specific incidence rates for those with and without specified characteristics and exposures, and according to specified levels of continuous variables such as lung function. Risk factors that were jointly significant were identified in stepwise discriminant and multiple logistic regression analyses. Various combinations of these risk factors were further evaluated in multiple logistic analyses. A person's risk of developing disease can be estimated from the multiple logistic model: P = [1 + exp(-a-IbiXi)]"1 where p is the probability of developing COPD, Xj is the value of the i- risk variable in the model, a is a constant term, and bi is the regression coefficient for the i- risk factor. The coefficients were obtained as maximal likelihood estimates (15), and the goodness of fit of each logistic model was assessed (16). The models selected for validation in other populations were those that fitted the Tecumseh data well and concentrated a high percentage of cases in the top levels of risk
TABLE 1 REGRESSION STATISTICS FOR PREDICTION OF FORCED EXPIRATORY VOLUME IN ONE SECOND FOR HEALTHY* NONSMOKERS: TECUMSEH POPULATION 25 TO 74 YEARS OF AGE, 1967-1969T Sex Male Female
n
a
b1
b2
R2
Syx
236 692
- 2.5870 - 0.9376
- 0.0294 - 0.0267
0.0445 0.0305
0.5350 0.5275
0.5292 0.4105
* Healthy population includes people without shortness of breath, cough, phlegm, asthma, or wheeze without colds, t Regression model: predicted FEV, = a + b, • age (yr) + b2 • height (cm).
using the fewest predictor variables. The investigators from each of the 6 validating studies assessed the extent to which the risk models fitted the data collected in their longitudinal studies. This was done by computing the risk probability for each subject, classifying individual subjects into categories of risk, and determining the expected number of incidence cases in each category by summing individual risk values for members of that category. Observed and expected numbers of incidence cases were then compared. Because some of the populations used to assess the validity of the Tecumseh index had been followed for only 4 to 6 yr, the assumption was made that the FEV, of each subject would continue to decline at the rate observed during the actual follow-up interval. Classification of subjects in these populations as cases or as noncases of COPD was based on linear extrapolation of FEV, values to a 10-yr interval. Framingham was the only population with a follow-up period long enough to assess the validity of the 15-yr index of risk. Results
The Tecumseh Index of Risk for COPD The Tecumseh index of risk for COPD within 10 yr was developed from observations on men and women 25 to 64 yr of age who were tested in the years 1967 to 1969 and retested approximately 10 yr later in 1978 and 1979. After excluding persons who already had definite or borderline COPD (FEV! less than 70% predicted) or inadequate spirograms, the population at risk consisted of 958 men and 1,159 women. Forty-three men and
28 women had a FEVj less than 65% predicted at follow-up; they are the cases of COPD on which the index of risk is based. Regression statistics for the baseline observations for healthy nonsmokers are given in table 1. They may be used to calculate normal values for other white populations, provided comparable measurement techniques are used. Age-, height-, and sex-specific normal values derived from these equations have been published (3). The risk factors included in the equation for calculating a person's risk of COPD are age, number of cigarettes smoked per day, FEV, percent predicted, and change in daily cigarette consumption during the follow-up interval. Multiple logistic regression coefficients for prediction of COPD within 10 yr are given in table 2. The odds ratios approximate the relative risks associated with a difference of 1 unit of measurement for each individual predictor, conditional on the others remaining fixed. For example, risk of COPD in men increases 2.5 times with every 10-yr increase in age, and over twice with every 1 pack per day increase in cigarette consumption. The odds ratio of 0.3 for FEVj percent predicted means that risk of COPD is reduced to one third with every 10% increase in FEVx percent predicted. A person's risk of developing disease was estimated by evaluating the equation given above at the observed values for
TABLE 2 TECUMSEH INDEX OF RISK: MULTIPLE LOGISTIC REGRESSION COEFFICIENTS FOR PREDICTION OF CHRONIC OBSTRUCTIVE PULMONARY DISEASE WITHIN 10 YEARS IN MEN AND WOMEN 25 TO 64 YEARS OF AGE Men
Risk Factors Age Cigarettes per day Change in cigarettes per dayt FEV„ % pred Constant (a)
bi
SE (bj)
0.08911 0.03890 0.02628 -0.11736 2.79280
0.02041 0.01169 0.01423 0.02011 2.05770
Women Odds Ratio* 2.4 2.2 1.7 0.3
bi
SE (bi)
0.07766 0.06875 0.04393 -0.11846 3.00410
0.02358 0.01777 0.02188 0.02239 2.28710
Odds Ratio* 2.2 4.0 2.4 0.3
Definition of abbreviations: bj = coefficients for the risk factors; SE = standard errors for the coefficients. * Units for odds ratios are: age, 10 years; cigarettes per day and change in cigarettes per day, 1 pack; FEV, % predicted, 10%. t Number of cigarettes per day at follow-up minus number at entry to the study.
382
HIGGINS, KELLER, LANDIS, ET AL
of cigarette smoking were higher in Boston, Louisiana, and Staveley than in other populations. There were small numbers of incidence cases in some of these populations, and the numbers of cases developing within 10 yr were estimated by linear extrapolation from observed rates of decline in FEVX for populations with follow-up of short duration, as described above (table 5). The distributions of observed and expected cases are also shown in table 5 by arbitrarily selected categories of risk. The goodness of fit of the index of risk can be judged from the distributions of observed and expected numbers. The main discrepancy is that expected cases exceed observed cases at high levels of risk, especially for Framingham and Staveley. However, observed cases cluster at the higher levels of risk in all studies. Discrepancies between observed and expected numbers are attributable to varying prevalences of COPD at follow-up, to the short duration of the observation period in some populations, to varying rates of mortality or loss to follow-up, and possibly to baseline levels of FEVj being too low in Framingham. Predictions of FEVX 10 yr in the future are likely to be less accurate when they are extrapolated from rates of decline observed over a short interval and when the initial levels of FEV are at the extremes of the range. Moreover, total observed and expected Validation of the Tecumseh Index of Risk numbers will not be equal in populations The numbers and characteristics of other than the population in which the men and women available to assess the model was developed. validity of the Tecumseh index of risk are The predictive ability of the model can shown in table 4. Comparison of base- be judged from information in table 6 line risk factor levels shows that in addi- where the frequencies of high-risk scores tion to having a longer follow-up, the are shown for the study populations and population of Framingham was older for individual subjects with COPD at and had lower mean FEVX values; rates follow-up. Risks of 0.1 or greater are con-
age, number of cigarettes smoked per day, and FEV! percent predicted, using the maximal likelihood estimates of the corresponding coefficients. Change in daily cigarette consumption during the interval was known for the Tecumseh population and was included in the equation. After computing the risk probability for each man and each woman, individual subjects were ranked and grouped into sex-specific deciles of risk. The extent to which the risk model fitted the Tecumseh data set is shown in table 3. There were approximately 96 men and 116 women in each decile. Agreement between the observed and expected numbers of cases within each decile was satisfactory, as indicated by the statistical test for goodness of fit. The discriminating ability of the model is indicated by the extent to which cases of disease were concentrated in the upper deciles. Among men, 63% of cases were in the top decile and 81% in the top quintile of risk; comparable figures for women were 64 and 86%. Incidence rates for the lowest quintile of risk were zero for both sexes, and for the highest quintile they were 18% for men and 10% for women (figure 1). Thus, the goodness of fit and the predictive power of the Tecumseh index of risk are very good for the population in which it was developed.
TABLE 3 TECUMSEH INDEX OF RISK: OBSERVED AND EXPECTED CASES OF CHRONIC OBSTRUCTIVE PULMONARY DISEASE WITHIN 10 YEARS BY DECILE OF RISK* Women
Men Cases
Cases
Observed
E xpected
Number in Decile
1 2 3 4 5 6 7 8 9 10
0 0 1 0 1 0 2 4 8 27
0.1 0.2 0.3 0.5 0.9 1.4 2.4 4.3 8.7 24.3
96 96 96 96 96 95 96 96 96 95
0 0 0 0 0 0 3 1 6 18
0.0 0.1 0.2 0.3 0.5 0.8 1.4 2.3 4.8 17.5
Total
43
43
958
28
28
Decile of Risk
k
2
Observed
Expected
Number in Decile 116 116 116 116 117 115 116 116 116 115 1,159
X (8 df) goodness of fit for the model for men was 4.53 and for women, 5.07; p > 0.10 (no significant lack of fit).
MEN (N=958)
T • T 1 2 3 4 5 QUINTILE OF RISK
WOMEN (N=1159)
1 2 3 4 5 QUINTILE OF RISK
Fig. 1. Incidence rates ( ± 1 SE) for COPD by Tecumseh index of risk score for men and women 25 to 64 yr of age in Tecumseh.
sidered high in these analyses; this cut point approximates the 90th percentile of the risk score distribution for the Tecumseh population. In Tecumseh, 65% of male cases and 43% of female cases had high risk scores. Larger proportions of male than of female cases had highrisk scores in each population; these proportions ranged from 60 to 78% for men (excluding Boston where there were only 2 cases) and from 43 to 70% for women (excluding Boston where there were only 5 cases). The frequency of highrisk scores in each population depends on the frequency and distribution of risk factors. The percentages with high-risk scores ranged from 13 in Tecumseh to 49 in Framingham for men, and from 6 in Tecumseh to 25 in Framingham for women, again excluding Boston. High-risk scores were more frequent in older populations and in those with larger proportions of cigarette smokers or lower mean FEVx values. In all populations, risk scores of 0.1 or more were associated with increased incidence of COPD. If we compare a risk score of 0.1 or greater to a positive test, and calculate its predictive value as the 10-yr incidence of COPD in persons who were test-positive at baseline, the predictive value was 23% for men and 17% for women in Tecumseh. Values for the other populations are shown in table 6. The relative risks of COPD associated with baseline risk scores of 0.1 or greater compared with scores of less than 0.1 are also shown in table 6. Relative risks were 12.8 for men and 11.3 for women in Tecumseh and from 2.2 to 17.1 in the other populations. Discussion Predictors of COPD identified in the population of Tecumseh also predicted onset of COPD in other populations in the United States and the United Kingdom. It is not surprising that these studies find that age, cigarette smoking, and re-
383
RISK OF CHRONIC OBSTRUCTIVE LUNG DISEASE TABLE 4 CHARACTERISTICS OF THE STUDY POPULATIONS Age (yr)
Men Tecumseh Baltimore Boston Louisiana Staveley Tucson Framingham Women Tecumseh Baltimore Boston Louisiana Staveley Tucson Framingham
FEV, (% pred)
Smokers
Duration of Follow-Up (mean in yr)
n
Mean
SD
(%)
Mean
SD
Dates of Examinations
958 700
41.9 44.3
9.6 12.4
45.5 40.8
96.6 97.8
12.7 13.2
1967-69; 1978-79 1971-75; 1976-81
10 4.7
97 309 502 377 1,088
42.3 41.0 40.5* 44.0 50.0
6.8 11.7 14.2 12.1 7.8
70.1 68.3 63.6 38.3 55.2
92.4 96.5 91.6 97.7 80.0
14.0 12.0 11.5 13.7 15.0
1975; 1980-81 1973-1980 1957; 1966 1972-75; 1977-81 1958-60; 1972-74
6 4 9 7.3 14
1,159 558
42.0 42.9
9.8 12.8
30.4 40.8
99.6 97.2
13.3 12.8
1967-69; 1978-79 1971-75; 1976-81
10 4.8
231 48
39.8 46.4
6.0 8.0
63.6 56.2
93.7 101.9
13.8 14.2
1975; 1980-81 1974-79
46.3 50.8
11.9 8.0
32.0 40.5
99.2 83.4
13.7 14.6
1972-75; 1977-81 1958-60; 1972-74
Type of Population
General Nonpatient control subjects & relatives General General & industrial General & industrial General General General Nonpatient control subjects & relatives General General
6 5
493 1,549
7.3 14
General General
" Subjects were 25-34 or 55-64 yr of age.
be inferred from the Tecumseh index of scribed in detail in previous publications risk. When the model is used to estimate (2, 3). Information on rate of decline of probability of developing disease in the FEVj in nonsmokers, cigarette smokers, future, the number of cigarettes a per- and ex-smokers, as reported by the U.S. son will be smoking at the end of the 10- Public Health Service (17) and by Fletchyr period must be estimated; several er and coworkers (19), is consistent with different estimates varying from no cig- these findings. arettes to more than current consumpSeveral other known or suspected tion provide a range of probabilities of predictors of disease have been investigatdeveloping disease, and demonstrate the ed and found to add little to prediction beneficial reduction in risk associated when these major risk factors are taken with stopping or reducing cigarette smok- into account. Other potential risk indiing. Use of the index to calculate risk for cators that were assessed during develindividual patients was illustrated or de- opment of the Tecumseh index of risk
duced FEVi are the most important risk factors for COPD. Morbidity and mortality rates for COPD are consistently higher at older ages and in men, and epidemiologic observations have established firmly that cigarette smoking is the most important cause of COPD, including chronic bronchitis and emphysema (17,18). Changes in exposure to cigarette smoke during the follow-up interval influenced incidence of COPD; the beneficial effect of stopping or reducing cigarette smoking and the harmful effect of starting or increasing use of cigarettes can
TABLE 5 OBSERVED AND EXPECTED NUMBERS OF CASES OF CHRONIC OBSTRUCTIVE PULMONARY DISEASE IN SPECIFIED CATEGORIES OF RISK Categories of Risk 0-0.03
Framinghamt Women Tecumseh Baltimore* Boston Louisiana* Tucson* Framinghamt
0.05--0.10
0.10-0.20
0.20-0.30
0.30 +
Total
Number in Population
E
O
E
O
E
O
E
O
E
O
E
O
E
4 9 0 4 3 1
5.3 3.6 0.6 1.5 2.5 1.9
2 7 0 1 1 2
3.2 2.3 0.6 .8 1.2 0.8
9 10 0 1 4 1
7.0 4.2 0.7 2.4 3.5 3.3
12 14 2 4 2 1
9.4 7.8 0.7 4.3 5.9 3.9
8 10 0 5 4 4
8.6 8.1 0.0 7.6 6.2 3.8
8 25 0
9.5 22.7 0.0
11 5
20.7 7.3
43 75 2 15 25 14
43.0 48.7 2.6 16.6 40.0 21.0
958 700 97 309 502 377
7
3.8
4
5.2
12
12.5
19
33.0
17
32.4
44
80.7
103
167.6
1,088
5 7 2 0 2
6.1 2.6 1.3 .2 2.3
4 3 2 0 1
2.9 1.2 0.5 .2 1.9
7 5 0 1 0
4.9 3.6 1.5 .3 3.1
6 7 1 0 1
6.2 6.7 2.3 .2 4.1
4 5 0 1 3
3.7 4.6 0.5 1.2 3.0
2 8 0
4.2 12.7 2.8
3
8.4
28 35 5 2 10
28.0 31.4 8.9 2.1 22.8
1,159 558 231 48 493
16
7.4
3
6.4
14
17.0
20
26.4
11
19.5
41
62.0
105
138.7
1,549
0 Men Tecumseh Baltimore* Boston Louisiana* Staveley Tucson*
0.03-0.05
* Observed cases based on linear extrapolation of FEV, to 10 yr of follow-up. t Expected cases based on index of risk within 15 yr.
384
HIGGINS, KELLER, LANDIS, ET AL
TABLE 6 FREQUENCIES, PREDICTIVE VALUES, AND RELATIVE RISKS OF HIGH-RISK SCORES (>0.1)
Number in Population Men Tecumseh Baltimore Boston Louisiana Staveley Tucson Framinghamt Women Tecumseh Baltimore Boston Louisiana Tucson Framinghamt
Number of Cases
Population with Risk ^0.1
Cases with Risk >0A
Predictive Value of Risk ^0.1
(%)
(%)
(%)
Relative Risk*
958 700 97 309 502 377
43 75 2 15 25 14
12.7 19.0 6.2 17.2 23.5 15.9
65.1 65.3 (100.0) 60.0 68.0 71.4
23.0 36.8 33.3 17.0 14.4 16.7
12.8 8.0
1,088
103
48.9
111
15.0
3.6
1,159 558 231 48 493
28 35 5 2 10
6.2 16.3 10.4 10.4 12.0
42.9 57.1 20.0 50.0 70.0
16.7 22.0 4.2 20.0 11.9
11.3 6.8 2.2 8.6 17.1
1,549
105
25.2
68.6
18.5
6.5
7.2 6.9 13.2
* Relative risk = incidence of COPD with risk ^ 0.1/incidence of COPD with risk < 0.1. t Based on index of risk within 15 yr.
include respiratory symptoms (cough, phlegm, wheeze, and shortness of breath), respiratory diseases (chronic bronchitis, asthma, and hay fever), histories of upper and lower respiratory infections (including bronchitis, pneumonia, and pleurisy), other measures of lung function (FVC, Vmax50, Vmax25, the nitrogen index of uneven ventilation, and bronchial reactivity to isoproterenol inhalation), skin sensitivity to allergens, education, occupation, alcohol consumption, respiratory disease in first degree relatives, and 12 genetic markers (2, 20,21). Possible risk factors that were not evaluated in the Tecumseh population include alphaj-antitrypsin phenotype and trypsin inhibitory capacity, and exposure to air pollution and specific occupational hazards. The latter 2 exposures were present among some of the populations involved in the validation study, but available evidence does not suggest that inclusion of these factors improves prediction to an extent that is of practical importance for general populations (22). However, this is not necessarily true for all occupational groups, even though adverse effects of hazards to which a person has been exposed are presumably reflected in the FEV\ measured at entry to the study. The Tecumseh index of risk identifies high-risk persons by combining information on a few simple factors into a composite risk score that ranks them correctly according to their level of risk. Absolute
levels of risk and incidence rates vary in different populations with characteristics of the population such as age, sex, frequency and amount of smoking, and distribution of lung function (tables 5 and 6). They are likely to vary also with changes in the composition of cigarettes and with the frequency and severity of other hazardous exposures. New onsets of disease defined in terms of impaired lung function can only be ascertained in survivors, whereas those most severely affected are likely to die. Members of the Tecumseh population who died had higher risk scores initially than did members who survived (2). There is also substantial evidence that reductions in FEVi and other measures of ventilatory lung function are associated with increased morbidity and mortality (23-25). Thus, estimates of risk based only on survivors are underestimates, and the extent of the underestimation will be greater for populations with higher death rates. Estimates of risk are at best average probabilities of developing disease for groups of people. Risk scores for individual subjects are useful in that they indicate advisability of preventive measures rather than certainty of adverse outcomes. There is some evidence that risk of disease can be reduced by modifying risk factors, and results of recent antismoking efforts in men at high risk for coronary heart disease are particularly encouraging with respect to getting smokers to quit (26-29). Whether knowl-
edge of increased risk of COPD will also be associated with success at quitting smoking remains to be determined. References 1. Development of an index of risk for chronic obstructive lung disease. RFP-NHLBI-HR-78-1, 1977. 2. Higgins MW, Keller JB, Becker M, et al. An index of risk for obstructive airways disease. Am Rev Respir Dis 1982; 125:144-51. 3. Higgins MW, Keller JB. Estimating your patient's risk of COPD. J Respir Dis 1983; 4:97-108. 4. American Thoracic Society. Statement of the Snowbird Workshop on Standardization of Spirometry. Am Rev Respir Dis 1979; 119:831-8. 5. Cohen BH, Ball WC, Brashears S, et al. Risk factors in chronic obstructive pulmonary disease (COPD). Am J Epidemiol 1977; 105:223-32. 6. Tager IB, Weiss ST, Rosner B, Speizer FE. Effect of parental cigarette smoking on the pulmonary function of children. Am J Epidemiol 1979; 110:15-26. 7. Lakatos E, DeMets DL, Kannel WB, Sorlie P, MacNamara P. Influence of cigarette smoking on lung function and COPD incidence. The Framingham Study. J Chronic Dis (submitted). 8. Weill H, Ziskind MM, Waggenstack C, Rosseiter CE. Lung function consequences of dust exposure in asbestos cement manufacturing plants. Arch Environ Health 1975; 30:88-97. 9. Diem JE, Jones RN, Hendrick DJ, et al. Fiveyear longitudinal study of workers employed in a new toluene diisocyanate manufacturing plant. Am Rev Respir Dis 1982; 126:420-8. 10. Glindmeyer HW, Diem JE, Jones RN, Weill H. Noncomparability of longitudinally and crosssectionally determined annual change in spirometry. Am Rev Respir Dis 1982; 125:544-8. 11. Jones RN, Hughes J, Hammad YY, et al. Respiratory health in cotton seed crushing mills. Chest 1981; 79(Suppl:30-3). 12. Higgins ITT, Cochrane AL, Gilson GC, Wood CH. Population studies of chronic respiratory disease: a comparison of miners, foundry workers and others in Staveley, Derbyshire. Br J Ind Med 1959; 16:255. 13. Higgins ITT, Gilson JC, Ferris BG, Waters ME, Campbell H, Higgins MW. Chronic respiratory disease in an industrial town: a nine-year follow-up study. Preliminary report. Am J Public Health 1968; 58:1667-76. 14. Knudson RJ, Lebowitz MD, Holberg CJ, Burrows B. Changes in the normal maximal expiratory flow-volume curve with growth and aging. Am Rev Respir Dis 1983; 127:725-34. 15. Walker SH, Duncan DB. Estimation of the probability of an event as a function of several independent variables. Biometrika 1967; 167-79. 16. Lemeshow S, Hosmer DW. A review of goodness of fit statistics for use in the development of logistic regressions models. Am J Epidemiol 1981; 116:92-106. 17. U.S. Public Health Service. Smoking and health. A report of the Surgeon General. Washington, D.C.: U.S. Government Printing Office, 1979. DHEW Publication No. (PHS)79-50066.) 18. N.C.H.S. Monthly Vital Statistics report, Vol. 31, No. 6 Supplement, September 30, 1982. 19. Fletcher CM, Peto R, Tinker CM, Speizer FE.
385
RISK OF CHRONIC OBSTRUCTIVE LUNG DISEASE
The natural history of chronic bronchitis and emphysema. Oxford, England: Oxford University Press, 1976. 20. Higgins MW, Keller JB, Howatt W, Weg J, Landis R. Bronchial reactivity, skin test responses and obstructive airways disease. Am Rev Respir Dis 1981; 123:80. 21. Higgins MW. Epidemiology of COPD. Chest. Supplement of the 26th Annual Aspen Lung Conference. June 8, 1982. Chest 1984; 85:3S-8S. 22. Higgins ITT, Higgins MW, Keller JB. Trial of prediction of obstructive airways disease. Trial of the Tecumseh index of risk in the population of Staveley, England. Am Rev Respir Dis 1982; 125:146.
23. Higgins MW, Keller JB. Predictors of mortality in the adult population of Tecumseh: respiratory symptoms, chronic respiratory disease, and ventilatory lung function. Arch Envir Hlth 1970; 21:418-24. 24. Beaty TH, Cohen BG, Newill CA, Menkes HA, Diamond EL, Chen CJ. Impaired pulmonary function as a risk factor for mortality. Am J Epidemiol 1982; 116:102-33. 25. Peto R, Speizer FE, Cochrane AL, et al. The relevance in adults of air-flow obstruction, but not of mucus hypersecretion to mortality from chronic lung disease. Am Rev Respir Dis 1983; 128:491-500.
26. Multiple risk factor intervention trial. Risk factor changes and mortality results. JAMA 1982; 248:1465-77. 27. Office of Population Censuses and Surveys. Trends in respiratory mortality, 1951-1975. London: Her Majesty's Stationery Office, 1981 (Series DH1, no. 7). 28. Hjermann I, Velve Byre K, Holme I, et al. Effect of diet and smoking intervention on the incidence of coronary heart disease. Report from the Oslo study group of a randomized trial in healthy men. Lancet 1981; 2:1303-10. 29. Kuller L, Meilahn E, Townsend M, et al. Control of cigarette smoking from a medical perspective. Annu Rev Public Health 1982; 3:153-78.