Vital Capacity As a Predictor of Incident Type 2 Diabetes the Atherosclerosis Risk in Communities Study

Metabolic Syndrome/Insulin Resistance Syndrome/Pre-Diabetes ORIGINAL ARTICLE

Vital Capacity as a Predictor of Incident Type 2 Diabetes The Atherosclerosis Risk in Communities study

1 3 HSIN-CHIEH YEH, PHD JAMES S. PANKOW, PHD studies (11–14) have consistently shown 1,2 4 NARESH M. PUNJABI, MD, PHD BRUCE B. DUNCAN, MD, PHD that adults with diabetes have lower vital 2 1,2 NAE-YUH WANG, PHD FREDERICK L. BRANCATI, MD, MHS capacity than their nondiabetic counterparts, but such studies cannot establish temporal sequence. Prospective studies have tentatively identiﬁed lower vital capacity as a predictor of hyperinsulinemia OBJECTIVE — To test the hypothesis that lower vital capacity is cross-sectionally associated (3) and type 2 diabetes (4,5,25) but have with features of insulin resistance and is an independent predictor of incident type 2 diabetes. had limitations related to sample size and availability of diabetes-related data. We RESEARCH DESIGN AND METHODS — We conducted a prospective cohort study of vital capacity as a predictor of incident type 2 diabetes using 9-year follow-up data on 11,479 therefore analyzed longitudinal data from middle-aged adults without diabetes at baseline from the Atherosclerosis Risk in Communities the Atherosclerosis Risk in Communities (ARIC) Study. (ARIC) study, a biracial community- based study of 15,792 middle-aged RESULTS — Forced vital capacity (FVC) and forced expiratory volume in 1 s were measured adults, to test the hypothesis that lower at baseline using standard spirometry. Incident type 2 diabetes cases were ascertained during lung function, as indicated by lower vital follow-up. At baseline, low FVC (% predicted) was independently associated with indicators of capacity, is cross-sectionally associated the insulin resistance syndrome, including higher fasting levels of glucose, insulin, and triglyc- with features of insulin resistance and is erides; lower fasting HDL cholesterol; and higher systolic blood pressure. In prospective analy- an independent predictor of incident type ses, there were graded associations between low FVC (% predicted) and incidence of type 2 2 diabetes. diabetes in men and women. These associations persisted in multivariable analyses that adjusted for age, race, adiposity, smoking, physical activity, and ARIC center. Compared with individuals in the highest quartile of FVC (% predicted), the fully adjusted hazard ratio (95% CI) of diabetes RESEARCH DESIGN AND in individuals in the lowest quartile was 1.6 (1.3–2.0) in men and 1.7 (1.3–2.1) in women. These METHODS relationships were stronger in those who have never smoked. — The ARIC study is an on-going, prospective cohort study de- CONCLUSIONS — Lower vital capacity is an independent predictor of incident type 2 signed to assess subclinical and clinical diabetes. Pulmonary factors related to vital capacity deserve attention as possible risk factors for atherosclerosis in a cohort of adults aged insulin resistance and diabetes. 45–64 years, selected using probability sampling from the following four U.S. Diabetes Care 28:1472–1479, 2005 communities: Forsyth County, NC; Jack- son, MS; the northwest suburbs of Min- neapolis, MN; and Washington County, mpaired lung function has attracted effects of hypoxemia on glucose and insu- MD. By design, the Jackson site exclu- growing interest as a potentially novel lin regulation (2,6), lung-related inﬂam- sively recruited African Americans, I risk factor for glucose intolerance matory mediators and their effects on thereby accounting for 90% of African (1,2), insulin resistance (3), and type 2 insulin signaling (1,7,8), and adverse ear- Americans in the study. Most of the re- diabetes (4,5,25). Possible mechanisms ly-life exposures and their effects on or- maining African Americans came from for the hypothesized link include direct gan development (9,10). Cross-sectional the Forsyth County cohort. The sampling ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● procedure and methods used in the ARIC

1 2 study have been previously described From the Department of Epidemiology, Johns Hopkins University, Baltimore, Maryland; the Department (15). The current study was based on of Medicine, Johns Hopkins University, Baltimore, Maryland; the 3Division of Epidemiology, University of Minnesota, Minneapolis, Minnesota; and the 4Social Medicine Department, Federal University of Rio Grande 9-year follow-up data that included a do Sul, Porto Alegre, Brazil. baseline visit from 1987 through 1989 Address correspondence and reprint requests to Dr. Frederick L. Brancati, The Welch Center for Preven- and three follow-up clinic visits sched- tion, Epidemiology, and Clinical Research, Johns Hopkins Medical Institutions, 2024 E. Monument St., Suite uled 3 years apart. Of participants still 2-600, Baltimore, MD 21205. E-mail: [email protected]. Received for publication 7 October 2004 and accepted in revised form 16 March 2005. alive at the time of the scheduled visits, Abbreviations: ARIC, Atherosclerosis Risk in Communities; DBP, diastolic blood pressure; FEV1, forced response rates for the second, third, and expiratory volume in 1 s; FVC, forced vital capacity; SBP, systolic blood pressure. fourth follow-ups were 93, 86, and 81%, A table elsewhere in this issue shows conventional and Syste`me International (SI) units and conversion respectively. factors for many substances. For the current analysis, individuals © 2005 by the American Diabetes Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby were excluded based on the following cri- marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. teria: ethnicity other than black or white

1472 DIABETES CARE, VOLUME 28, NUMBER 6, JUNE 2005 Yeh and Associates

Table 1—Baseline characteristics of men and women according to FVC (% predicted) quartile

FVC (% predicted) quartile* I (low) (Ͻ 90.8) II (90.9–99.4) III (99.5–108.4) IV (high) (Ͼ108.4) P for trend† Men n 1,283 1,281 1,281 1,282 African American 26 19 16 14 Ͻ0.0001 Age at baseline (years) 55 Ϯ 5.6 54 Ϯ 5.7 54 Ϯ 5.6 53 Ϯ 5.7 Ͻ0.0001 Education Ͻ12 years 26 20 20 15 Ͻ0.0001 Parental history of diabetes 12 13 12 10 0.11 Smoking status Current 36 27 22 18 Ͻ0.0001 Former 42 45 46 45 — Never 22 28 32 37 — Pack-years‡ 36 Ϯ 22.9 31 Ϯ 23.8 28 Ϯ 20.3 24 Ϯ 18.4 Ͻ0.0001 Sport index 2.4 Ϯ 0.8 2.6 Ϯ 0.8 2.7 Ϯ 0.8 2.8 Ϯ 0.8 Ͻ0.0001 BMI (kg/m2) 28.1 Ϯ 4.6 27.5 Ϯ 4.1 26.9 Ϯ 3.6 26.6 Ϯ 3.3 Ͻ0.0001 Waist-to-hip ratio 0.96 Ϯ 0.06 0.96 Ϯ 0.05 0.96 Ϯ 0.05 0.94 Ϯ 0.05 Ͻ0.0001 Height (cm) 176 Ϯ 6.5 176 Ϯ 6.5 176 Ϯ 6.7 177 Ϯ 6.6 0.06 Waist (cm) 101 Ϯ 11.7 99 Ϯ 10.4 97 Ϯ 9.8 96 Ϯ 9.1 Ͻ0.0001 Metabolic syndrome§ 36 26 22 15 Ͻ0.0001 Fibrinogen (g/l) 3.1 Ϯ 0.7 3.0 Ϯ 0.6 2.9 Ϯ 0.6 2.8 Ϯ 0.6 Ͻ0.0001 White blood cell count (ϫ109/l) 6.4 Ϯ 1.9 6.2 Ϯ 1.8 6.1 Ϯ 2.6 5.9 Ϯ 1.7 Ͻ0.0001 FVC (l) 3.8 Ϯ 0.5 4.4 Ϯ 0.5 4.8 Ϯ 0.5 5.5 Ϯ 0.6 Ͻ0.0001 Ϯ Ϯ Ϯ Ϯ Ͻ FEV1 (l) 2.7 0.5 3.3 0.5 3.6 0.5 4.0 0.6 0.0001 Ϯ Ϯ Ϯ Ϯ FEV1/FVC 72.9 9.3 74.3 7.4 74.2 6.8 73.6 6.4 0.05

I (low) (Ͻ 95.7) II (95.8–105.3) III (105.4–114.6) IV (high) (Ͼ114.6) P for trend† Women n 1,588 1,588 1,588 1,588 African American 33 26 21 20 Ͻ0.0001 Age at baseline (years) 54 Ϯ 5.8 53 Ϯ 5.6 53 Ϯ 5.6 54 Ϯ 5.6 0.06 Education Ͻ12 years 26 19 17 15 Ͻ0.0001 Parental history of diabetes 12 13 14 13 0.71 Smoking status Current 34 22 19 15 Ͻ0.0001 Former 18 22 24 25 — Never 48 56 57 60 — Pack-years‡ 25.0 Ϯ 17.8 20.6 Ϯ 17.3 17.6 Ϯ 15.8 17.3 Ϯ 15.6 Ͻ0.0001 Sport index 2.2 Ϯ 0.7 2.3 Ϯ 0.7 2.4 Ϯ 0.8 2.5 Ϯ 0.8 Ͻ0.0001 BMI (kg/m2) 28.7 Ϯ 6.6 27.3 Ϯ 5.8 26.5 Ϯ 5.1 26.1 Ϯ 4.7 Ͻ0.0001 Waist-to-hip ratio 0.91 Ϯ 0.08 0.89 Ϯ 0.08 0.87 Ϯ 0.08 0.87 Ϯ 0.08 Ͻ0.0001 Height (cm) 163 Ϯ 5.8 163 Ϯ 5.9 162 Ϯ 6.0 162 Ϯ 6.1 Ͻ0.0001 Waist (cm) 98 Ϯ 16.3 94 Ϯ 14.9 91 Ϯ 13.4 90 Ϯ 12.8 Ͻ0.0001 Metabolic syndrome§ 35 27 21 16 Ͻ0.0001 Fibrinogen (g/l) 3.2 Ϯ 0.7 3.0 Ϯ 0.6 2.9 Ϯ 0.5 2.9 Ϯ 0.6 Ͻ0.0001 WBC count (ϫ109/l) 6.3 Ϯ 1.9 5.9 Ϯ 1.9 5.7 Ϯ 1.6 5.5 Ϯ 1.6 Ͻ0.0001 FVC (l) 2.7 Ϯ 0.4 3.2 Ϯ 0.4 3.5 Ϯ 0.4 3.8 Ϯ 0.5 Ͻ0.0001 Ϯ Ϯ Ϯ Ϯ Ͻ FEV1 (l) 2.0 0.4 2.4 0.3 2.6 0.3 2.9 0.4 0.0001 Ϯ Ϯ Ϯ Ϯ FEV1/FVC 76.0 8.2 76.6 6.1 76.2 5.6 75.1 5.8 0.46 Data are mean Ϯ SD or percent. *Quartiles are sex specific. †P values correspond to tests for linear trend across quartiles. ‡Pack-years of cigarette smoking in ever- smokers only. §Metabolic syndrome was defined based on guidelines published in the National Cholesterol Education Program Adult Treatment Panel III report (46). The clinical identification of the metabolic syndrome is based upon the presence of any three of the following traits: 1) abdominal obesity, defined as a waist circumference in men Ͼ102 cm (40 in) and in women Ͼ88 cm (35 in); 2) serum triglycerides 150 mg/dl (1.7 mmol/l); 3) serum HDL cholesterol Ͻ40 mg/dl (1 mmol/l) in men and Ͻ50 mg/dl (1.3 mmol/l) in women; 4) blood pressure 130/85 mmHg; and 5) fasting plasma glucose 110 mg/dl (6.1 mmol/l).

DIABETES CARE, VOLUME 28, NUMBER 6, JUNE 2005 1473 Vital capacity and incident diabetes

Table 2—Adjusted levels of selected features of the insulin resistance syndrome by sex and FVC (% predicted) quartile at baseline: the ARIC study

Fasting glucose Fasting insulin HDL Triglycerides SBP DBP FVC quartile* (mg/dl) (␮u/ml) HOMA-IR† (mg/dl) (mg/dl) (mmHg) (mmHg) Men I (low) 101.5 12.5 3.2 43.5 148 122 75 II 100.2 11.1 2.8 44.6 137 121 75 III 100.5 10.6 2.7 44.3 134 121 75 IV (high) 100.1 10.0 2.5 46.1 124 119 74 P value‡ 0.0004 Ͻ0.0001 Ͻ0.0001 Ͻ0.0001 Ͻ0.0001 Ͻ0.0001 0.12 Women I (low) 97.7 11.7 2.9 58.0 118 120 72 II 97.2 10.7 2.6 58.5 117 118 72 III 96.9 10.1 2.5 59.0 114 118 72 IV (high) 96.2 9.6 2.3 59.8 110 117 71 P value‡ Ͻ0.0001 Ͻ0.0001 Ͻ 0.0001 0.01 0.005 Ͻ0.0001 0.11 Adjusted simultaneously for age, race, waist circumference, pack-years of cigarette smoking, sport activity index, and ARIC center. HOMA-IR, homeostasis model assessment of insulin resistance. *Quartiles of FVC (% predicted) are sex speciﬁc. †HOMA-IR ϭ (fasting insulin ͓␮U/ml͔ϫfasting glucose ͓mmol/l͔/22.5). ‡P values correspond to tests for linear trend across quartiles.

(n ϭ 48), preexisting diabetes at baseline ogy was standardized across the four field follow-up. If participants did not know (n ϭ 1,867), missing data on spirometry centers. Quality control and reproducibil- their birth weights, they were asked to or diabetes status at baseline (n ϭ 192), ity were coordinated by a centralized pul- report birth weight categories: low (Ͻ5 no follow-up or with incomplete incident monary function reading center (Johns 1/2 lb), medium (5 1/2 to 9 lb), or high diabetes information (n ϭ 873), missing Hopkins School of Public Health, Balti- (Ͼ9 lb). data on relevant baseline covariates (n ϭ more, MD). 231), or self-reported asthma or chronic Data analysis ϭ lung diseases (n 1,102). The final study Incident type 2 diabetes Predicted FVC and FEV1 were calculated sample consisted of 11,479 adults with- Individuals were classified as having dia- by the ARIC Data Coordinating Center out diabetes at baseline. betes if any of the following criteria, using the equations developed by Crapo, adapted from 1997 American Diabetes Morris, and Gardner (20) in nonsmokers Spirometry Association criteria, were met: fasting glu- that included age, sex, height, and race. At the baseline visit, pulmonary function cose level of at least 7.0 mmol/l (126 mg/ Before analysis, crude data on FVC and was assessed using a water-sealed Collin dl) (90%), nonfasting glucose level of at FEV1 were divided by predicted FVC and Survey II volume displacement spirome- least 11.1 mmol/l (200 mg/l) (1%), cur- FEV1, respectively, to yield FVC (% pre- ter (Warren E. Collins, Braintree, MA) rent use of antidiabetic medication (1%), dicted) and FEV1 (% predicted). Because with a computer interface. Calibration or a positive response to the question lung capacity differs dramatically in men and analytic programs were installed on “Has a doctor ever told you that you had and women, all analyses were stratified by the computer to assist the operator in diabetes (sugar in the blood)?” (8%). Per- sex. Distributions of FVC (% predicted) daily calibration, spirometric testing, and sons classified as having diabetes at base- and FEV1 (% predicted) were approxi- analysis. Measurement of forced expira- line were excluded. mately normal and were categorized into tory volume in 1 s (FEV1) and forced vital sex-specific quartiles. Highest quartiles capacity (FVC) were performed based on Other baseline variables were used as reference groups. Means and recommendations from the Epidemiol- The definitions and methods used for frequencies of potential confounders as- ogy Standardization Project (16) and the other baseline measurements (age, race, sessed at the baseline visit were deter- American Thoracic Society (17). Partici- education level, cigarette smoking status mined for each quartile of FVC (% pants were asked to perform a maximum and pack-years, sport activity index, predicted) and FEV1 (% predicted). of five forced expiratory maneuvers to ob- blood pressure, parental history of diabe- ANOVA and ␹2 analysis were used to as- tain three acceptable spirograms, of tes, height, BMI, waist and hip circumfer- sess the statistical significance of the dif- which at least two were reproducible, de- ences, HDL cholesterol, triglycerides, ferences across quartiles. Ϯ fined as within 5% of the FEV1 and glucose, insulin, white blood cell count, To determine associations between FVC. Acceptability and reproducibility and fibrinogen) have been previously re- vital capacity and features of the insulin were also indicated by the computer pro- ported (7,18). C-reactive protein was as- resistance syndrome, a series of cross- gram and confirmed by the technician by sessed in a subgroup of 581 incident cases sectional analyses were performed using observing the volume-time spirograms. of type 2 diabetes and a comparison co- data at baseline among nondiabetic par- Tests were performed by trained and cer- hort of 572 noncases (19). Birth weight ticipants. The relation between quartiles tified pulmonary technicians. Methodol- was self-reported at the fourth scheduled of FVC (% predicted) and FEV1 (% pre-

1474 DIABETES CARE, VOLUME 28, NUMBER 6, JUNE 2005 Yeh and Associates

Figure 1—Nine-year adjusted hazard ratios (HR) for incident diabetes by FVC (% predicted) quartile, sex, and smoking status. All hazard ratios are simultaneously adjusted for age, race, pack-years of cigarette smoking (A and C), waist circumference, sport activity index, and ARIC center. FVC (% predicted) quartiles are categorized based on sex-specific values in total population. I indicates the lowest FVC quartile and IV the highest. Bars indicate 95% CIs. dicted) and fasting insulin, fasting glu- the risk of incident diabetes in lower three waist circumference, and waist-to-hip ra- cose, total triglycerides, HDL cholesterol, versus the highest quartile of FVC (% pre- tio; were more likely to have metabolic systolic blood pressure (SBP), and dia- dicted) and FEV1 (% predicted) after syndrome; and to have higher blood lev- stolic blood pressure (DBP) were assessed adjustment for confounding factors. els of white cell counts and fibrinogen. As by using linear regression models, after To minimize the possibility of residual expected, adults with low FVC (% pre- adjustment for confounding factors in- confounding secondary to smoking, mul- dicted) also tended to have low FVC and cluding age, race, pack-years of smoking, tivariable analyses were repeated in nev- FEV1. waist circumference, sport activity index er-smokers. Additional multivariable To determine whether a lower vital (scale one [lowest] to five [highest]), and analyses were performed to investigate capacity was associated at baseline with ARIC center. the roles of inflammatory markers and features of the insulin resistance syn- Incidence rates of diabetes were cal- low birth weight as potential confound- drome, cross-sectional analyses were per- culated for each quartile of FVC (% pre- ers. All tests of significance were two- formed (Table 2). In men and women, ␣ dicted) and FEV1 (% predicted) using a tailed, with an level of 0.05. All analyses lower FVC (% predicted) was indepen- person-years approach. For participants were performed using SAS (Cary, NC) dently associated with higher fasting glu- without diabetes, person-years were cal- statistical software package (version cose, insulin, and triglycerides; higher culated from baseline to the last visit date. 8.01). SBP; and lower HDL cholesterol. There Incident cases were assumed to have oc- was no association between FVC (% pre- curred at the midpoint between the last RESULTS dicted) and DBP. visit at which diabetes was found to be absent and the first visit at which diabetes Baseline characteristics and cross- Incident type 2 diabetes was found to be present. sectional associations During 9 years of follow-up, 673 men and Time to incident diabetes was as- Baseline characteristics for men and 673 women developed type 2 diabetes. In sessed by quartiles of FVC (% predicted) women are shown in Table 1 by quartiles both groups, there were graded inverse and FEV1 (% predicted) using survival of FVC (% predicted). In both groups, in- relationships between FVC (% predicted) analysis. Cox proportional hazards mod- dividuals with lower FVC (% predicted) and incidence rates of type 2 diabetes. In els were used for the multivariable analy- were significantly more likely to be Afri- men, diabetes incidence rose from 11.3 sis when the data met the proportionality can American, older, and less educated; to per 1,000 person-years in the highest assumption implicit in the models. Rela- have smoked more cigarettes; to be less FVC (% predicted) quartile to 28.3 per tive hazard ratios were used to compare physically active; to have higher BMI, 1,000 person-years in the lowest FVC (%

DIABETES CARE, VOLUME 28, NUMBER 6, JUNE 2005 1475 Vital capacity and incident diabetes

predicted) quartile, corresponding to an Subsidiary analyses FEV1 (% predicted) and incident age- and race-adjusted relative risk of 2.4 To investigate the relationship of FVC (% type 2 diabetes (95% CI 1.9–3.0). In women, diabetes predicted) to other emerging predictors When FEV1 (% predicted) was used as an incidence rose from 7.9 per 1,000 person- of diabetes risk, we conducted subsidiary indicator of lung function, similar pat- years in the highest FVC (% predicted) analyses in the full cohort (smokers and terns of results were obtained. For exam- quartile to 22.9 per 1,000 person-years in nonsmokers). First, to determine whether ple, in multivariable models adjusting for the lowest FVC (% predicted) quartile, FVC-related differences in inflammatory age, race, waist circumference, pack-years corresponding to an age- and race- markers might help explain the relation- of smoking, sport index, and ARIC center, adjusted relative risk of 2.5 (2.0–3.2). ship of lower FVC (% predicted) to dia- compared with their counterparts in the betes risk, we performed additional highest quartile of FEV1 (% predicted), analyses after introducing white cell men in the lowest quartile were 1.7-fold Multivariable analyses of FVC (% count and plasma fibrinogen concentra- (95%CI 1.3–2.1) as likely to develop type predicted) tion into multivariable models that 2 diabetes, and women in the lowest In multivariable analyses using the Cox already included age, race, waist circum- quartile were 1.5-fold (1.1–1.9) as likely. proportional hazards models, the signifi- ference, smoking, sport activity index, In contrast, there was no association be- cant graded inverse association between and ARIC center. In a case-cohort sub- tween the ratio of FEV1 to FVC and the FVC (% predicted) and incidence of dia- group of 1,153 participants, we addition- subsequent risk of incident diabetes in betes persisted in both men and women ally adjusted for C-reactive protein. In either men or women (data not shown). after adjustment for age, race, pack-years neither analysis did adjustment for in- of smoking, waist circumference, sport flammatory markers influence associa- CONCLUSIONS — In this prospec- activity index, and ARIC center (Figs. 1A tions between lower FVC and incident tive study of middle-aged men and and C). Compared with individuals in the type 2 diabetes. For example, in men in women without known lung disease, highest quartile of FVC (% predicted), the the case-cohort subgroup, the fully ad- lower vital capacity predicted the subse- adjusted hazard ratio (95% CI) of diabetes justed hazard ratios (95% CI) of incident quent development of type 2 diabetes. in individuals in the lowest quartile was diabetes in the third, second, and first The association was graded, was indepen- 1.6 (1.3–2.0) in men and 1.7 (1.3–2.1) in (lowest) quartile of FVC (% predicted) dent of a variety of potentially confound- women. Additional adjustment for pres- compared with those in the fourth (high- ing factors, was stronger in never est) quartile were 1.3 (0.7–2.7), 1.4 (0.6– smokers, and was similar in magnitude to ence of the metabolic syndrome attenu- 2.8), and 2.1 (1.1–3.9), respectively (P for well-established diabetes risk factors like ated the association slightly: compared trend ϭ 0.04, data not shown). physical inactivity and family history of with their counterparts in the highest Second, in a subset of individuals for diabetes (21,22). It also appeared to be quartile of FVC (% predicted), men and whom data on birth weight or birth specific to vital capacity, since the ratio of women in the lowest quartile of FVC (% weight categories were available (ϳ80%), FEV to FVC, an indicator of obstructive predicted) remained 1.4-fold (95% CI 1 we conducted analyses to determine A) airways disease, was not at all related to 1.1–1.8) and 1.5-fold (1.2–1.9) more the relationship of low birth weight with diabetes risk. Moreover, adults with lower likely to develop type 2 diabetes, respec- low FVC (% predicted) and B) the extent FVC (% predicted) had many features of tively. Adjustment for fasting glucose, ho- to which the FVC (% predicted)–diabetes insulin resistance at baseline, including meostasis model assessment of insulin association was explained by low birth higher blood levels of glucose, insulin, resistance, and SBP rather than metabolic weight. In multivariable analyses that ad- and triglycerides; lower HDL cholesterol; syndrome produced a similar attenua- justed for age and race, FVC (% pre- and higher SBP. tion: compared with their counterparts in dicted) was positively correlated with Our results are generally consistent the highest quartile of FVC (% predicted), birth weight in men (␤ coefficient ϭ 0.5% with previous prospective studies on lung men and women in the lowest quartile of per lb, P ϭ 0.007) but not in women (␤ function and the subsequent occurrence FVC (% predicted) were 1.3-fold (1.0– coefficient ϭϪ0.1% per lb, P ϭ 0.47). of diabetes and related conditions. In 1.7) and 1.4-fold (1.1–1.8) more likely to The relationship of FVC (% predicted) to 1,050 initially nondiabetic healthy male develop type 2 diabetes, respectively. incident diabetes was largely independent veterans in the Normative Aging Study, To minimize the possibility of resid- of low birth weight. In multivariable anal- Lazarus, Sparrow, and Weiss (3) found ual confounding due to cigarette smok- yses that included both FVC (% pre- that lower FVC, lower FEV1, and lower ing, analyses were repeated after dicted) and low birth weight, as well as maximal midexpiratory flow rate at base- restricting the sample to adults who never age, race, waist circumference, pack-years line predicted hyperinsulinemia and esti- smoked (Figs. 1B and D). In both men of smoking, sport index, and ARIC center, mated insulin resistance over 20 years of and women who never smoked, the asso- low FVC (% predicted) remained signifi- follow-up, independent of age, adiposity, ciations were stronger than in the full cantly associated with diabetes risk in a and smoking. In a Swedish study (23) of population: men and women in the low- graded fashion: in these models, men and 4,637 nondiabetic middle-aged men, est quartile of FVC (% predicted) had 1.9- women in the lowest quartiles of FVC (% baseline mean vital capacity was 10% fold (95% CI 1.2–3.0) and 1.9-fold (1.4– predicted) remained 1.7-fold (95% CI lower among 116 men who developed di- 2.6) risk of developing type 2 diabetes, 1.3–2.1) and 1.7-fold (1.3–2.2) more abetes during 6-years follow-up than respectively, compared with the individ- likely to develop type 2 diabetes, respec- those who did not develop diabetes. More uals in the highest quartiles of FVC (% tively, than their counterparts in the high- recently, Engstrom et al. (4) reported on predicted). est quartiles of FVC (% predicted). 382 initially nondiabetic Swedish men

1476 DIABETES CARE, VOLUME 28, NUMBER 6, JUNE 2005 Yeh and Associates born in 1914 who underwent spirometry thus making it a less plausible intermedi- sponse. In any case, our subsidiary anal- at age 55 years. In this cohort, low FVC ate in the putative causal pathway linking yses of white cell count or plasma and FEV1 predicted the presence of dia- low FVC with incident type 2 diabetes. fibrinogen suggested little or no explana- betes at follow-up 13 years later, indepen- Moreover, several lines of conflicting evi- tory role for these two general markers of dent of adiposity and weight gain. dence suggest that hypoxemia may have inflammation, although we lacked suffi- However, the use of urine test strips to no harmful effects in insulin sensitivity cient data on other more specific markers screen for diabetes at baseline raises con- (32–36). to exclude a role for inflammation conclu- cern about the possible inclusion of adults Another possible explanation is that sively. with preexisting diabetes into the incep- lower vital capacity and risk of diabetes Finally, decreased muscle strength tion cohort. Later on, Engstrom et al. (5) are both partially determined by adverse might be the link between impaired lung confirmed this finding in a larger cohort fetal or early-life conditions via effects on function and hyperinsulinemia (3). Ven- of Swedish men and women. This study organogenesis and metabolic pathway tilatory function is partially determined also found incidence of cardiovascular programming, which, through long- by respiratory muscle strength; low level diseases was significantly increased standing altered gene expression, would of skeletal muscle strength (as indicated among subjects with low FVC (% pre- favor resistance to insulin action. Prior by hand-grip strength) predicted higher dicted) who had developed insulin resis- studies suggest that low birth weight is levels of fasting insulin in the Normative tance. This observation suggested insulin associated in adults with reduced lung Aging Study (45). However, no data on resistance may be a mediator between low function (10), reduced ␤-cell function muscle strength were available in the FVC (% predicted) and cardiovascular (37), insulin resistance (38), and an in- ARIC study to test this hypothesis. disease, which has been noted previously creased incidence of type 2 diabetes (39). Strengths of this study include a com- (24). Finally, Ford and Mannino (25) re- Moreover, in adults, shorter leg length, an munity-based sampling method, standard- ported that FVC and FEV1 but not the indicator of adverse environmental influ- ized spirometric techniques, extensive data ratio of FEV1 to FVC were significantly ences on prepubertal development, is as- on potential confounders, and a large sam- and inversely associated with the inci- sociated with lower FVC and FEV1 (40) ple size that increased precision and permit- dence of diabetes in the National Health and with impaired glucose tolerance (41). ted multiple statistical adjustments. The and Nutrition Examination Survey Epide- Under this hypothesis, lower vital capac- study’s main limitation was lack of data on miologic Follow-Up Study. That study ity might either represent a marker of un- insulin sensitivity, oxygen saturation, vis- also found that restrictive lung disease derdevelopment without a direct role in ceral fat, and specific inflammatory mark- was significantly associated with the inci- diabetes pathophysiology or might be a ers, which precluded more detailed dence of diabetes but not obstructive lung mediator of the adverse metabolic effects investigation of causal pathways. Self- disease. Nevertheless, the above studies of underdevelopment. Indeed, we found reported birth weight data may also be sub- were limited to whites (4,5,23), used non- that low FVC (% predicted) was associ- ject to recall bias and misclassification. standard spirometry methods (5,25), or ated with low birth weight. However, the Finally, we could not establish temporal se- had a relatively small number of incident relation of FVC (% predicted) to incident quence for the cross-sectional relationship cases, which precluded extensive simul- diabetes was largely independent of low between insulin resistance and reduced taneous adjustment for potential con- birth weight, which suggests a distinct lung function that we observed at baseline. founders (4,5,23). pathway from low FVC to diabetes risk. The main implication of our study is One possible explanation for the link A third possible explanation involves that lower vital capacity of the lung de- between low vital capacity and diabetes some relationship of low vital capacity to serves attention as an emerging, novel risk risk is related to hypoxemia-induced in- emerging inflammatory precursors of in- factor for type 2 diabetes. Even if it turns sulin resistance. Experimental hypoxia sulin resistance and diabetes (7). Nuclear out not to lie within a causal pathway to produces insulin resistance and hyperin- factor interleukin-6, early growth re- diabetes, FVC might still be a useful risk sulinemia in animals (26,27). In humans, sponse-1, and hypoxia-inducible factor-1 predictor, and the FVC-diabetes link lower vital capacity can be associated with mediate inflammatory responses to could suggest explanations for other phe- lower oxygen saturation at rest with fur- chronic hypoxia in macrophages, pulmo- nomena, like the elevated risk of heart ther worsening with exertion (28). More- nary vascular endothelium, and smooth disease associated with low vital capacity. over, exposure to altitude hypoxia (2) or muscle (8). Cigarette smoking, recently On the other hand, if vital capacity does hypobaric hypoxia (6) can reduce insulin established as an independent predictor indeed lie in a casual pathway, then sensitivity and predispose to the develop- of type 2 diabetes (42,43), provokes an pathophysiologic derangements associ- ment of type 2 diabetes. These effects may inflammatory response (44) and is in- ated with lower FVC, like hypoxemia and be partially mediated by sympathetic versely associated with vital capacity. lung-related inflammation, might represent nerve activity: hypoxia increases sympa- However, in our study, the link between novel targets for interventions aimed at the thetic nerve activity (29,30), and in the lower vital capacity and diabetes risk was primary prevention of type 2 diabetes. ARIC study, indicators of sympathetic completely independent of cigarette ex- nerve activity (low heart rate variability posure and was stronger in never- and high heart rate) independently pre- smokers. Furthermore, reduced vital Acknowledgments— The ARIC study is sup- dict incident diabetes (31). However, the capacity is a common residual effect of ported by contracts from the National Heart, mild decrements lung function, as noted lower respiratory tract infections, includ- Lung, and Blood Institute (N01-HC-55015, in the current analyses, are not likely to be ing those in childhood and infancy (10), N01-HC-55016, N01-HC-55018, N01-HC- associated with significant hypoxemia, that might provoke an inflammatory re- 55019, N01-HC-55020, N01-HC-55021,

DIABETES CARE, VOLUME 28, NUMBER 6, JUNE 2005 1477 Vital capacity and incident diabetes and N01-HC-55022) and by the National lung function in a cross-sectional popula- 1996 Institute of Diabetes and Digestive and Kid- tion study. Eur Rrespir J 2:14–19, 1989 24. Kannel WB, Hubert H, Lew EA: Vital ca- ney Diseases Grant (5R01-DK56918-03), 12. Barrett-Connor E, Frette C: NIDDM, im- pacity as a predictor of cardiovascular dis- Bethesda, Maryland. paired glucose tolerance, and pulmonary ease: the Framingham study. Am Heart J The authors thank the staff and partici- function in older adults: the Rancho Ber- 105:311–315, 1983 pants in the ARIC study for their important nardo Study. Diabetes Care 19:1441– 25. Ford ES, Mannino DM: Prospective asso- contributions. 1444, 1996 ciation between lung function and the in- 13. Davis TME, Knuiman M, Kendall P, Vu H, cidence of diabetes: findings from the Davis WA: Reduced pulmonary function National Health and Nutrition Examina- References and its association in type 2 diabetes: the tion Survey Epidemiologic Follow-up 1. Hjalmarsen A, Aasebo U, Birkeland K, Fremantle Diabetes Study. Diabetes Res Study. Diabetes Care 27:2966–2970, Sager G, Jorde R: Impaired glucose toler- Clin Pract 50:153–159, 2000 2004 ance in patients with chronic hypoxic pul- 14. Walter RE, Beiser A, Givelber RJ, 26. Cheng N, Cai W, Jiang M, Wu S: Effect of monary disease. Diabetes Metab 22:37– O’Connor GT, Gottlieb DJ: Association hypoxia on blood glucose, hormones, and 42, 1996 between glycemic state and lung function: insulin receptor functions in newborn 2. Larsen JJ, Hansen JM, Olsen NV, Galbo H, the Framingham Heart Study. Am J Respir calves. Pediatr Res 41:852–856, 1997 Dela F: The effect of altitude hypoxia on Crit Care Med 167:911–916, 2003 27. Raff H, Bruder ED, Jankowski BM: The glucose homeostasis in men. J Physiol 504: 15. The ARIC Investigators: The Atheroscle- effect of hypoxia on plasma leptin and in- 241–249, 1997 rosis Risk in Communities (ARIC) Study: sulin in newborn and juvenile rats. Endo- 3. Lazarus R, Sparrow D, Weiss ST: Baseline design and objectives. Am J Epidemiol 129: crine 11:37–39, 1999 ventilatory function predicts the develop- 687–702, 1989 28. Pollard AJ, Barry PW, Mason NP, Collier ment of higher levels of fasting insulin and 16. Ferris BG Jr, Speizer FE, Bishop Y, Prang DJ, Pollard RC, Pollard PF, Martin I, fasting insulin resistance index: the Nor- G, Weener J: Spirometry for an epidemi- Fraser RS, Miller MR, Milledge JS: Hypox- mative Aging Study. Eur Respir J 12:641– ologic study: Deriving optimum summary ia, hypocapnia and spirometry at altitude. 645, 1998 statistics for each subject. Bull Europ Phys- Clin Sci (Lond) 92:593–598, 1997 4. Engstrom G, Janzon L: Risk of developing iopath Resp 14:146–166, 1978 29. Halliwill JR, Minson CT: Effect of hypoxia diabetes is inversely related to lung func- 17. Gardner RM, et al.: ATS statement: Snow- on arterial baroreflex control of heart rate tion: a population-based cohort study. bird workshop on standarization of spi- and muscle sympathetic nerve activity in Diabet Med 19:167–170, 2002 rometry. Am Rev Respir Dis 119:831–838, humans. J Appl Physiol 93:857–864, 2002 5. Engstrom G, Hedblad B, Nilsson P, 1979 30. Smith ML, Muenter NK: Effects of hypox- Wollmer P, Berglund G, Janzon L: Lung 18. Brancati FL, Kao WH, Folsom AR, ia on sympathetic neural control in hu- function, insulin resistance and incidence Watson RL, Szklo M: Incident type 2 dia- mans. Respir Physiol 121:163–171, 2000 of cardiovascular disease: a longitudinal betes mellitus in African American and 31. Carnethon MR, Golden SH, Folsom AR, cohort study. J Intern Med 253:574–581, white adults: the Atherosclerosis Risk in Haskell W, Liao D: A prospective investi- 2003 Communities Study. JAMA 283:2253– gation of autonomic nervous system func- 6. Braun B, Rock PB, Zamudio S, Wolfel GE, 2259, 2000 tion and the development of type 2 Mazzeo RS, Muza SR, Fulco CS, Moore 19. Duncan BB, Schmidt MI, Pankow JS, Bal- diabetes: the Atherosclerosis Risk in LG, Butterfield GE: Women at altitude: lantyne CM, Couper D, Vigo A, Hoo- Communities Study, 1987–1998. Circu- short-term exposure to hypoxia and/or al- geveen R, Folsom AR, Heiss G, the lation 107:2190–2195, 2003 pha(1)-adrenergic blockade reduces insu- Atherosclerosis Risk in Communities 32. Santos JL, Perez-Bravo F, Carrasco E, Cal- lin sensitivity. J Appl Physiol 91:623–631, Study: Low-grade systemic inflammation villan M, Albala C: Low prevalence of type 2001 and the development of type 2 diabetes: 2 diabetes despite a high average body 7. Schmidt MI, Duncan BB, Sharrett AR, the Atherosclerosis Risk in Communities mass index in the Aymara natives from Lindberg G, Savage PJ, Offenbacher S, Study. Diabetes 52:1799–1805, 2003 Chile. Nutrition 17:305–309, 2001 Azambuja MI, Tracy RP, Heiss G: Markers 20. Crapo RO, Morris AH, Gardner RM: Ref- 33. de Glisezinski I, Crampes F, Harant I, of inflammation and prediction of diabe- erence spirometric values using tech- Havlik P, Gardette B, Jammes Y, Souber- tes mellitus in adults (Atherosclerosis Risk niques and equipment that meet ATS bielle JC, Richalet JP, Riviere D: Decrease in Communities study): a cohort study. recommendations. Am Rev Respir Dis 123: of subcutaneous adipose tissue lipolysis Lancet 353:1649–1652, 1999 659–664, 1981 after exposure to hypoxia during a simu- 8. Semenza GL: Oxygen-regulated tran- 21. Manson JE, Rimm EB, Stampfer MJ, Cold- lated ascent of Mt Everest. Pflugers Arch scription factors and their role in pulmo- itz GA, Willett WC, Krolewski AS, Rosner 439:134–140, 1999 nary disease. Respir Res 1:159–162, 2000 B, Hennekens CH, Speizer FE: Physical 34. Lee WC, Chen JJ, Hunt DD, Hou CW, Lai 9. Hales CN, Barker DJ, Clark PM, Cox LJ, activity and incidence of non-insulin de- YC, Lin FC, Chen CY, Lin CH, Liao YH, Fall C, Osmond C, Winter PD: Fetal and pendent diabetes mellitus in women. Kuo CH: Effects of hiking at altitude on infant growth and impaired glucose toler- Lancet 338:774–778, 1991 body composition and insulin sensitivity ance at age 64. BMJ 303:1019–1022, 22. Shaten BJ, Smith GD, Kuller LH, and Nea- in recovering drug addicts. Prev Med 39: 1991 ton JD: Risk factors for the development 681–688, 2004 10. Barker DJ, Godfrey KM, Fall C, Osmond of type II diabetes among men enrolled in 35. Sauerwein HP, Schols AM: Glucose me- C, Winter PD, Shaheen SO: Relation of the usual care group of the Multiple Risk tabolism in chronic lung disease. Clin birth weight and childhood respiratory Factor Intervention Trial. Diabetes Care Nutr 21:367–371, 2002 infection to adult lung function and death 16:1331–1339, 1993 36. Krampl E, Kametas NA, Nowotny P, Ro- from chronic obstructive airways disease. 23. Eriksson KF, Lindgarde F: Poor physical den M, Nicolaides KH: Glucose metabo- BMJ 303:671–675, 1991 fitness, and impaired early insulin re- lism in pregnancy at high altitude. 11. Lange P, Groth S, Kastrup J, Mortensen J, sponse but late hyperinsulinaemia, as Diabetes Care 24:817–822, 2001 Appleyard M, Nyboe J, Jensen G, Schnohr predictors of NIDDM in middle-aged 37. Cook JT, Levy JC, Page RC, Shaw JA, Hat- P: Diabetes mellitus, plasma glucose and Swedish men. Diabetologia 39:573–579, tersley AT, Turner RC: Association of low

1478 DIABETES CARE, VOLUME 28, NUMBER 6, JUNE 2005 Yeh and Associates

birth weight with beta cell function in the in the Midspan Family Study. J Epidemiol 44. Zalokar JB, Richard JL, Claude JR: Leuko- adult ﬁrst degree relatives of non-insulin Community Health 57:141–146, 2003 cyte count, smoking, and myocardial independent diabetic subjects. BMJ 306: 41. Asao K, Kao WHL, Baptiste-Roberts K, farction. N Engl J Med 304:465–468, 1981 302–306, 1993 Bandeen-Roche K, Erlinger T, Brancati L: 45. Lazarus R, Sparrow D, Weiss ST: Hand- 38. Phillips DI, Barker DJ, Hales CN, Hirst S, Upper leg length and glucose tolerance in grip strength and insulin levels: cross-sec- Osmond C: Thinness at birth and insulin U.S. adults: NHANES III (Abstract). Cir- tional and prospective associations in the resistance in adult life. Diabetologia 37: culation 107:E7005, 2003 Normative Aging Study. Metabolism 46: 150–154, 1994 42. Wannamethee SG, Shaper AG, Perry IJ, 1266–1269, 1997 39. Forsen T, Eriksson J, Tuomilehto J, Reu- the British Regional Heart Study: Smok- 46. Expert Panel on Detection, Evaluation, nanen A, Osmond C, Barker D: The fetal ing as a modiﬁable risk factor for type 2 and Treatment of High Blood Cholesterol and childhood growth of persons who de- diabetes in middle-aged men. Diabetes in Adults: Executive Summary of the velop type 2 diabetes. Ann Intern Med 133: Care 24:1590–1595, 2001 Third Report of the National Cholesterol 176–182, 2000 43. Manson JE, Ajani UA, Liu S, Nathan DM, Education Program (NCEP) Expert Panel 40. GunnellD,WhitleyE,UptonMN,McCon- Hennekens CH: A prospective study of on Detection, Evaluation, and Treatment nachie A, Davey Smith G, Watt GC: Asso- cigarette smoking and the incidence of di- of High Blood Cholesterol in Adults ciations of height, leg length, and lung abetes mellitus among US male physi- (Adult Treatment Panel III). JAMA 285: function with cardiovascular risk factors cians. Am J Med 109:538–542, 2000 2486–2497, 2001

DIABETES CARE, VOLUME 28, NUMBER 6, JUNE 2005 1479