Fifteen-year Interval Spirometric Evaluation of the Oregon Predictive Equations

Fifteen-year Interval Spirometric Evaluation of the Oregon Predictive Equations* James F. Morris, M.D., F.C.C.P.;t Arthur Koski, Ed.D.;+ William P Temple, B.A.;§ Alan Claremont, Ph.D.,~ and David R. Thomas, Ph.D.11

The 1969 Oregon spirometric predictive equations were evaluated by retesting 199 of the 988 original sample population after 15 years. The 1969 data were used to test for sample bias between the retested and not-retested groups. There was no significant difference in mean values for age, height, or test results except for a five-year age difference in men. Regression analysis of residuals and the differences between calculated and predicted values of annual decrements of FVC, FEV l , and FEF25-75% on age revealed no

statistically significant age trend. Although residual means were statistically significant for FVC and FEV l for men and FVC and FEF25-75% for women, the differences between calculated and predicted annual decrements were significant only for women in FEF25-75%. Although group performance was accurately predicted for most tests, test SDs and SEMs demonstrated considerable individual variation. Lower limits of normality are suggested to assist in evaluating previously-tested patients.

predicted normal values are routinely used as reference standards in spirometric pulmonary function testing. 1 A considerable number of reference standards are available which were derived from cross-sectional studies of a variety of populations. Validity of a crosssectional approach has been questioned, however, and longitudinal studies are recommended to more accurately predict the changes in flows and volumes associated with aging in normal individuals.:" Predicted yearly decrements in vital capacity, FEV1 and forced expiratory flows are increasingly being used to determine respiratory impairment among individual patients and employees. Ideally, premorbid or preemployment test values should be available for comparison but, more often, prediction equations derived from cross-sectional testing are used. In 1969, we developed normal spirometric standards for healthy, nonsmoking Caucasian adults." A total of 988 residents of western Oregon performed forced expiratory flows and volumes. Regression equations for expired flows and volumes based on sex, age and height were derived. Cross-sectional analysis of the data indicated that ventilatory function declined progressively with advancing age. Identical testing was repeated in 1984 in a sample of the original group to

assess the predictive ability of the original regression equations with a 15-year interval between testing.

*From

the Pulmonary Disease Section, Veterans Administration Medical Center, Portland; and the College of Health and Physical Education and Department of Statistics, Oregon State University, Corvallis. tProfessor of Medicine, PVAMC. :j:Assistant Dean, Professor of Health, OSU. §Computer Specialist, PVAMC. ~Research Associate, OSU. IIProfessor of Statistics, OSU. Partially funded by NIH Biomedical Support Grant RR-07079. Manuscript received March 20; revision accepted May 26. Reprint requests: Dr. Morris, VA Medical Center, PO Box 1034, Portland, Oregon 97207

MATERIAL AND METHODS

A total of208 men and women from the original sample population were still residing in the Willamette River Valley of western Oregon and were available for retesting. No information was available concerning the whereabouts of the subjects who had relocated or died. Questionnaires similar to those in the original study were used to assess intercurrent cardiopulmonary diseases, cigarette smoking, and exposure to ambient and occupational air pollutants since 1969. Interviewers reviewed the responses with the subjects to check for accuracy and rejected nine individuals because of abnormal responses to those three categories. Standing height was measured without shoes, as before. The two Stead-Wells watersealed spirometers used in 1969 were again used for retesting and their accuracy periodically confirmed with a 3L calibration syringe. Tests were performed in the standing position with noseclips in place. A forced vital capacity (FVC) maneuver was repeated until three FVC values were obtained which varied no more than 100 ml or 5 percent, whichever was larger. FVC, FEV l , and FEF25-75% values were selected from the curve with the largest sum of FVC and FEV l , as determined in 1969. In addition, acceptable tests reflected maximum efforts, were begun without hesitation, lasted more than five sec, and were free from coughing. All gas volumes were corrected to body temperature. FEVl was calculated by both the method of Kory et al" (discarding the first 200 ml of FVC) and the method of back extrapolation of the FVC to time zero. 9,10 Standard t-tests and regression analyses were used for appropriate statistical analysis. 11 The two-sample t-test with a pooled estimate of variance was used for comparison of means for physical characteristics and pulmonary function measurements made in 1969 for the two subgroups: those retested in 1984 and those not retested (Table 1). Twenty-seven subjects were deleted from the 1969 study because of uncorrectable data errors. New multiple linear regression equations for pulmonary function measurements on the remaining 961 subjects were fitted to age and height. The 1984 age and height values for the retest subgroup were then substituted into the revised regression equations to produce residuals which are observed minus predicted pulmonary function values (Fig 1). Levels of statistical significance of CHEST / 92 / 1 / JANUARY, 1988

123

Table I-Comparison of 1969 Means of Physical Characteristics and Pulmonary Function Measurements of the Two Subgroups Retested and Not Retested in 1984 Women

Men Retested (n = 102)

Not retested (n = 400) 37.1 70.05 5.19 3.97 4.13

Age (yrs) Height (inches) FVC (L) FEV l (L) FEF25-75% (Usee)

42.0t 70.10 5.24 3.88 3.86

(0.75)* (0.15) (0.05) (0.04) (0.07)

(1.03) (0.28) (0.08) (0.06) (0.10)

Retested (n = 97)

Not retested (n =362) 42.6 64.1 3.46 2.65 3.14

41.8 64.5 3.61 2.75 3.14

(0.90) (0.12) (0.04) (0.03) (0.05)

(1.07) (0.24) (0.06) (0.05) (0.09)

*Standard errors of the mean. tp = 0.002 (all other values exceed the 0.05 significance level).

Table 2-Regression Analysis of Residuals of1984 Predicted and Observed Spirometric Values

Men FVC FEV l FEF25-75% Women FVC FEV l FEF25-75%

B2

SE(B2)

p

-0.0088 -0.0052 -0.0071

0.0056 0.0044 0.0087

0.12 0.23 0.42

-0.0079 -0.0013 0.0026

0.0052 0.0038 0.0078

0.13 0.74 0.74

pared with those of the remammg 762 subjects in the 1969 not retested subgroup to test possible sample bias. Values compared were spirometric results, height, and age for each sex (Table 3). No significant differences were found except for a mean age 4.9 years younger in the not retested men's subgroup (p = 0.0016). This age difference is attributable to a group of college students tested in 1969 but unavailable for retesting. Validity of the revised 1969 regression equations was examined by comparing predicted with observed 1984 retest values and analyzing the residuals (difference between predicted and observed values). These residuals were plotted against age and fitted to regression equations to determine any age effects not predicted by the 1969revised regression equations (Fig 1). Table 2 lists the slopes of the regression equations of the residuals on age, standard error of the slopes, and p-values for the slopes. The p-values show that the slopes are not statistically different from zero, which precludes significant unaccounted age or cohort effects in the predicted equations. Table 3 shows statistics for the residual values of the 1984 retest subgroup. Mean residual differences from zero are statistically significant only for FVC and FEVl for men and FVC and FEF25-75% for women. Despite statistical significance, clinical importance is considered minimal. Mean difference expressed as a percentage of 1969 sample mean value is 2.5 and 5.2 percent of FVC, and 2.5 and 1.1 percent for FEV l for men and women, respectively. The larger, negative discrepancy in the mean differences for FEF25-75%, particularly for women, may be explained both by the wider variation usually encountered in this test and the increased

B2 = slope of the regression equations of residuals on age; SE(B2)= the standard error of B2; V= the level of statistical significance of B2 (p>0.05 not significant) the slopes of the regression lines (Table 2) and of residual means (Table 3) were determined by the one-sample t-test. Annual decrements were calculated for each subject by the difference between the 1969 and 1984 test results divided by the intervening test periods. The difference between the calculated annual decrement (CAD)and the annual decrements predicted by the age coefficient of the regression equation corresponding to that test were analyzed by regression and t-tests. The study was approved by the Oregon State University Human Studies Committee. RESULTS

We reexamined the 1969 test values for 961 men and women after 27 deletions for erroneous value. The revised data were then fitted by age and height to multiple regression equations for men and women. Original and revised equations are listed in Table 4. Coefficients of multiple correlation and standard errors of estimate for the original and revised equations are listed in Table 5. The revised equations are similar to the original equations; therefore, the revised equations were used in the current study. The 1969 data for the 199 subjects in the 1984 retest subgroup were com-

Table 3-Mean Differences Between Observed and Predicted (Residuals) Spirometric Values in the Retest Subgroup Men

FVC (L) FEV l (L) FEF25-75% (Lisee)

Women

Mean (SEM)

p*

Mean (SEM)

p

0.13 (0.06) 0.10 (0.05) - 0.13 (0.09)

0.030 (2.5%)t 0.033 (2.5%) 0.175 (-3.1%)

0.20 (0.05) 0.08 (0.04) - 0.33 (0.07)

0.0001 (5.8%) 0.051 (2.8%) 0.0001 (-10.5%)

*Level of statistical significance of the mean (p<.05 is significant). tMean difference as the percent of the observed 1969 mean.

124

Fifteen-year Interval Spirometric Evaluation (Morris et al)

Table 4-Comparison of the Regression Equations for the 1969 Original and Revised Data

Men n Age (yrs) mean SD Height (inches) mean SD FVC (L) FEV1(L) FEF25-75% (Lis) Women n Age (yrs) mean SD Height (inches) mean SD FVC (L) FEV1(L) FEF25-75% (Us)

Original

Revised

517

502

37.9 14.3

38.1 14.3

70.1 2.9 0.148H - 0.025A - 4.241 0.092H - 0.032A -1.260 0.047H - 0.045A + 2.513

70.1 2.9 0.138H -0.027A -3.445 0.093H - O. 032A- 1.343 0.052H - 0.044A + 2.098

471

459

42.5 15.9

42.5 16.0

64.2 2.3 0.115H - 0.024A - 2.852 O. 089H - O. 025A- 1.932 O. 060H - O. 030A + 0.551

64.2 2.4 0.114H - 0.024A - 2.795 O. 085H - O. 025A-1. 692 0.055H - 0.031A + 0.919

observed FVC values in 1984. Calculated annual decrements (CAD) were fitted to regression equations with the initial test age as the independent variable to test for CAD age-dependency, similar to the treatment of residuals. Slopes were not significantly different from zero except for the women's FVC (p = .013). CAD data are summarized in Table 6. Differences between mean CAD values and the annual decrements predicted for each test (PAD) by the age coefficients of the corresponding regression equations were calculated and are also listed in Table 6. A remarkable agreement was found between CAD and PAD FEV1 for both men and women. Although other tests were not in the same close agreement as CAD and PAD, the only significant difference was found in the womens' FEF25-75% (p<.OOOI). Relative magnitudes of the 15-year accumulated CAD-PAD differences are demonstrated in Table 6 by expressing them as percentages of the 1984 mean test values. FVC and FEV1 can be standardized for body size by dividing each test value by the square of standing height, as recommended by Cole." Using Cole's formula as adapted by Burr et al," we also standardized FVC and FEV1 data for body size by substituting mean height for our male and female subjects. There was no significant difference in the results compared with test values not standardized for height.

H = height in inches; A = age in years

4

Men

FVC. (L.)

4

~

3

~

2

~

1

I

0

+' "C

a.. ~-1 >

en

~

4 3

~

2

+'

"C

f ~

30

40

50 60 70 80 Age (years)

rl<::"~.

-;;;;lS~~--;C•

~

3

~

2

•

.- .rf.;--

"C

~

1

a..

0

I T- -

;-2 en

~ xx~~ • -

x x

x

x

x x

•

x -----

50 60 70 80 Age (years)

90

•• :

1 _.--..'

a..

•-

~-1 >

30

40

50 60 70 80 Age (years)

."

•

••.....·.*:l·... .-- -

1;- x---,r "'-x-atoL.,.

x x

x

x

-4

90

FEV1. (L.)

4

~

3

~

2

+'

x

~~x ~~ .a::.C.=.J .. ~~-

Xx

~-1 > ~-2

-4

•

•

:.~~~-"r at "1~!Ji

0

I

en

'*

••

x

.---

en

40

2

~

FEF25-75%. (Lisee.)

_

x

g-3

"C Q)

30

40

50 60 70 80 Age (years)

90

FEF25-75%. (Lisee.)

....

x

•

" y .x • 0 -_x~:7~~-

c,

o.. I

~-1 > ~-2

en

8-3

8-3 30

;

Women 4 +'

•

3

~-2

-4

90

FVC. (L.)

~-1 >

-4

x !,r.-x

en

1 -

0

-:::x:--

x

g-3

B-3 -4

4

~

+'

x.

~-1 > ~-2

&i-2

FEV1. (L.)

8-3 30

40

50 60 70 80 Age (years)

90

-4

30

40

50 60 70 80 Age (years)

90

FIGURE 1. The dashed lines represent the regression lines plus or minus one standard error for the residuals (observed minus predicted values) of spirometric measurements on age.

CHEST I 92 I 1 I JANUARY, 1988

125

Table 5-Comparison of Multiple Correlation Coefficients, and Standard Errors ofEstimate for the 1969 Original and Revised Regression Equations FVC

FEF25-75%

FEV 1

~ ~ ~

Original Revised Original Revised Original Revised Men SEE Women SEE

0.65 0.74

0.69 0.65

0.73 0.55

0.75 0.52

0.53 1.12

0.54 1.08

0.71 0.52

0.70 0.54

0.73 0.47

0.79 0.38

0.56 0.89

0.57 0.78

r = coefficient of correlation between the observed and predicted 1969 values; SEE = standard error of estimate (root mean square error)

The differences between calculation of ~"'EV1 by the back extrapolation and the Kory techniques were mean values ofO.192±0.021 L for men and 0.191±0.015 L for women. These compare with a difference of 0.196 L for 20 normal subjects studied by Crapo et a}14 and 0.179 L for 100 patients including those with "significant obstructive impairment" reported by Smith and Caensler " Oregon predicted values for FEV1 can be corrected to conform to the back extrapolation technique recommended by the American Thoracic Society by the addition of192 In1 for both men and women, DISCUSSION

A major problem in any longterm study is the loss of subjects due to geographic relocation, illness, death, or failure to cooperate. The exact status of all our subjects not retested is uncertain, but the majority Table 6-Calculated Annual Decrements and Comparisons with Annual Decrements Predicted as Regression Equation Age Coefficients FVC

FEV 1

FEF25-75%

-- 0.0317 0.0295 -0.0266 - 0.0051 -0.0761 ,1.7380 -1.61

- 0.0314 0.0193 - 0.0319 0.0005 0.0077 3.3825 0.23

-0.0527 0.0464 -0.0444 -0.0083 -0.1239 3.0330 -4.08

-0.0220 0.0230 -0.0245 0.0025 0.0369 3.2638 1.13

-0.0262 0.0165 -0.0250 -0.0012 -0.0187 2.3237 -0.81

-0.0519 0.0475 -0.0313 -0.0206* -0.3084 2.2960 -13.43

(L)

Nlen CAD SD PAD Difference 15-yr AD 1984 test mean 15-yr AD/1984 mean % "Vomen CAD SD PAD Difference 15-yr AD 1984 test mean 15-yr AD/1984 mean (%)

(L)

(Lisee)

*p<.OOOI (all other p-values >.05). CAD = calculated annual decrement: PAD = predicted annual decrement; Difference = difference between calculated and predicted annual decrements; 15-yr AD = 15-year accumulated difference between calculated and predicted annual decrements; 15-yr AD/1984 mean (%) = 15-yr AD expressed as a percentage of the 1984 mean value. 126

moved out of the region. The survival effect of selective mortality over an extended period of time could result in a sample with relatively better ventilatory function" and possibly lower cross-sectionally predicted spirometric values compared with observed values." Theoretically, this survival effect could be less in an initial group of healthy nonsmokers compared with an age-matched group from the general population. Mortality rate has been shown by the Framingham study to be greater in those with initially decreased ventilatory function. 16 Various biases can affect longterm population studies. Selection bias is particularly important in the initial examination. Both the 1969 and 1984 subject samples consisted of volunteer nonsmokers from church membership lists who were screened by a medical history questionnaire. Age is the primary independent variable affecting ventilatory function in adults. Other variables to be considered include associated diseases (especially cardiopulmonary), ambient air quality, smoking habits, occupational exposure to noxious inhalants, level of physical fitness, changes in height and weight, variations in testing equipment and techniques, and medications. 17,18 In the present study, a health questionnaire was repeated after 15 years to detect intercurrent illnesses (particularly those affecting the cardiac and pulmonary systems), tobacco smoking, and medications affecting bronchomotor tone. Testing techniques similar to those used in 1969 included spirometers, methods, and time of day for testing. Standing height and weight were both measured, but only height was significant as a predictor. Lifetime height loss has been reported to be 2.9 em for men and 4.9 em for women." Our average loss tor 15 years was 0.6 em for both men and women. This represents 0.0157 inches per year, or the equivalent of a decrease of only about 2 ml per year in l?VC and FEV1 . Thus, the loss of height with age contributed only slightly to PAD. Included in the questionnaire were questions regarding the level and duration of nonoccupational regular aerobic activity. Classifying the subjects for three levels of aerobic physical activity showed no correlation with changes in ventilatory function. A learning effect resulting from farniliarity with testing techniques has been cited as a potential bias." With a IS-year interval between testing, any learning effect should be absent. Changes in ambient air quality should be considered in longitudinal studies of pulmonary function. 20 The Oregon Department of Environmental Quality provided air quality data for total suspended particulates (TSP) from the areas where the subjects resided. This is used as an indicator of ambient air quality. Sampling locations and techniques were similar over the testing period. Annual median values were below the National Ambient Air Quality Standard for TSp 21 Fifteen-year Interval Spirometric Evaluation (Morris et al)

and since 1976, there has been an improving trend in TSP in all Willamette Valley areas. The differences between observed and predicted individual test values and between calculated and predicted annual decrements were not attributable to changes in age, height, physical activity level, or deterioration in ambient air quality and they apparently resulted from other unidentified factors. The sample's test values closely followed the prediction equations, especially for FEV!? as demonstrated by small residuals and minor differences between calculated and predicted annual decrements. But the individuals varied considerably as shown by SEM of the residuals (Table 5) and SD of the annual decrements (Table 6). Serial testing of the same individual at more frequent intervals should more precisely reflect the effects of aging and other variables. Until there is a definitive longitudinal study outcome, we suggest use of appropriate regression' equations to predict normal levels for initial testing. For serial testing, we suggest using a 95 percent confidence limit for the calculated annual decrement to determine the lower limit of normal. This would represent a decrease of up to 70 ml for FVC, 55 ml for FEV l , and 130 ml/sec for FEF25-75%. These values are derived from the mean annual decrement plus one-tailed 95 percent confidence limit. We conclude from repeat testing 15 years later of a sample of healthy nonsmoking adults relatively free from air pollution and intercurrent diseases, that ventilatory function of individuals initially tested can be reliably predicted from an applicable cross-sectional study. To predict accurately the subsequent aging effect, more frequent serial testing is needed to explain the wide individual variation over time. REFERENCES 1 ACCP Scientific Section. Statement on spirometry: a report of the Section on Respiratory Pathophysiology. Chest 1983; 83:547-50 2 Rosenzweig DY, Arkins JA, Schrock LG. Ventilation studies on a normal population after a seven-year interval. Am Rev Respir Dis 1966; 194:74-78 3 Rowe JW. Clinical research on aging: strategies and directions. N

Engl J Med 1977; 297:1332-36 4 Beck OJ, Doyle CA, Schachter EN. A longitudinal study of respiratory health in a rural community. Am Rev Respir Dis 1982; 125:375-81 5 Glindmeyer HW, Diem JE, Jones RN, Weill H. Noncomparability of longitudinally and cross-sectionally determined annual change in spirometry. Am Rev Respir Dis 1982; 125:544-48 6 Dontas AS, Jacobs DR, Jr, Corcondilas A, Keys A, Hannan P Longitudinal vs cross-sectional vital capacity changes and affecting factors. J Geront 1984: 39:430-38 7 Morris JF, Koski A, Johnson LC. Spirometric standards for healthy nonsmoking adults. Am Rev Respir Dis 1971; 103:57-67 8 Kory RC, Callahan R, Boren HG, Syner JC. The Veterans Administration-Army cooperative study of pulmonary function. I. Clinical spirometry in normal men. Am J Med 1961; 30:243-58 9 Smith AA, Gaensler EA. Timing of forced expiratory volume in one second. Am Rev Respir Dis 1975; 112:882-85 10 Knudson RJ, Lebowitz ~1D, Slatin RC. The timing of the forced vital capacity. Am Rev Respir Dis 1979; 119:315-18 11 Neter J, Wasserman W, Kutner M. Applied linear regression models. Homewood, IL: Richard D Irwin Inc, 1983; 60-68 12 Cole TJ. Linear and proportional regression models in the prediction of ventilatory function. J Royal Statist Soc A 1975; 138:297-337 13 Burr ML, Phillips DM, Durst DN. Lung function in the elderly. Thorax 1985; 40:54-59 14 Crapo RO, Morris AH, Gardner RM. Reference spirometric values using techniques and equipment that meet ATS recommendations. Am Rev Respir Dis 1981; 123:659-64 15 Buist AS. Evaluation of lung function: concepts of normality. In: Simmons DH, ed. Current Pulmonology. New York: John Wiley & Sons, 1981; 141-65 16 Ashley F, Kannel WB, Sorlie PD, Masson R. Pulmonary function; relation to aging, cigarette habit, and mortality. The Framingham study. Ann Intern Med 1975; 82:739-45 17 Woolcock AJ, Colman MH, Blackburn CRB. Factors affecting normal values for ventilatory lung function. Am Rev Respir Dis 1972; 106:692-708 18 Morris JF: Spirometry in the evaluation of pulrnonary function. West J Med 1976; 125:110-18 19 Russell RM, McGandy RB, Jelliffe D. Reference weights. Practical considerations. Am J ~led 1984; 76:767-9 20 Dockery DW, Ware JH, Ferris BG Jr, Glicksberg DS, Fay ME, Spiro A III, Speizer FE. Distribution of forced expiratory volume in one second and forced vital capacity in healthy, white, adult never-smokers in six U.S. cities. Am Hev Respir Dis 1985; 131:511-20 21 US Environmental Protection Agency. National air quality and emissions trends report. Section 3. Research Triangle Park, NC: Office of Air Quality Planning and Standards, 1983

CHEST I 93 I 1 / JANUARY, 1988

127