A prospective study of weight change and systemic inflammation over 9 y 1 3

Transcription

1 A prospective study of weight change and systemic inflammation over 9 y 1 3 Andrew W Fogarty, Caoimhe Glancy, Stuart Jones, Sarah A Lewis, Tricia M McKeever, and John R Britton ABSTRACT Background: An increase in weight is a risk factor for cardiovascular disease and cancer. This increased risk may be mediated by inflammation, but no long-term data are available on the effect of weight gain on systemic inflammation. Objective: We tested the hypothesis that weight gain is associated with an increase in systemic inflammation during a 9-y period. Design: In 1991 data on body weight and a blood sample were collected from a random sample of 2425 randomly selected adults from a community-based cohort in Nottingham, United Kingdom. In 2000, these measures were repeated in 1301 of these participants. The main outcome measure was change in systemic inflammation as measured by serum C-reactive protein (CRP) from the 1222 participants who provided paired samples. Results: The mean change in weight from 1991 to 2000 was 2.9 kg (95% CI: 2.6, 3.2 kg). The geometric mean of CRP in 1991 was 1.22 mg/l (95% CI: 0.03, mg/l), and it increased to 1.76 mg/l (95% CI: 0.09, 62.0 mg/l) in 2000 (P 0.001). A linear association was observed between increase in weight and serum CRP, with a 1-kg increment in weight being associated with an additional increase in CRP of 0.09 mg/l (95% CI: 0.02, 0.16 mg/l) during this time period. Conclusion: During a 9-y period, an increase in weight is associated with an increase in systemic inflammation. This provides a mechanism that may explain some of the previously reported association of weight gain with an increased risk of both cancer and cardiovascular disease. Am J Clin Nutr 2008;87:30 5. KEY WORDS INTRODUCTION Inflammation, obesity, weight Inflammation has an important role in the pathogenesis of several chronic diseases, including cardiovascular disease, diabetes, and cancer. Obesity is also an important risk factor for many of these conditions, but the mechanisms underlying these associations and the role of systemic inflammation in them are not fully understood. However, it is now well established that C-reactive protein (CRP), a nonspecific marker of systemic inflammation, is elevated in overweight and obese adults (1), suggesting that the link with disease may be mediated at least in part by an effect of obesity on systemic inflammation. This is an important area because the prevalence of obesity is increasing rapidly in many countries (2, 3). However, although a recent review concluded that weight loss may be an effective nonpharmacologic strategy for lowering CRP (4), it is uncertain whether an increase in body weight increases systematic inflammation during a long-term period. To our knowledge, there are no large prospective studies of the association between change in weight and its relation to changes in systemic inflammation in samples representative of a general adult population. We have therefore used data from a longitudinal community-based study to determine whether a change in body weight is associated with a change in systemic inflammation, as measured by CRP, in a large prospective population-based study of adults aged y based in the United Kingdom during a 9-y period. SUBJECTS AND METHODS Study population The participants were drawn from a previously reported cohort of adults aged y who lived in a Local Authority Area in Nottingham, United Kingdom, in 1991, and participated in a study of the relation between diet and respiratory disease described elsewhere (5 7). In summary, 7016 adults aged y were identified by systematic sampling from a random point in the local electoral register in Of those adults, 2633 (representing 48 59% of the initial sample subsequently ascertained to be alive and resident in the study area at the time of the study) participated, and all surviving persons were invited to participate in a follow-up study in 2000 (8). Both studies were approved by the Nottingham City Hospital Ethics Committee. Data collection on anthropometry and serum CRP In 1991 and 2000 participants completed a computeradministered lifestyle questionnaire that included questions on smoking and a semiquantitative food-frequency questionnaire (DietQ; Tinuviel Software, Warrington, United Kingdom) (5). They were also weighed, their height was measured, and a venous blood sample was taken. After venesection, serum samples 1 From the Division of Epidemiology and Public Health, University of Nottingham, Nottingham, United Kingdom (AWF, CG, SAL, TMM, JRB), and the Department of Clinical Chemistry, Nottingham City Hospital, Nottingham, United Kingdom (SJ). 2 Supported by the Wellcome Trust (TMM), Asthma UK, the British Lung Foundation, and the University of Nottingham. 3 Address reprint requests to AW Fogarty, Division of Epidemiology and Public Health, University of Nottingham, Clinical Sciences Building, City Hospital, Nottingham NG5 1PB, United Kingdom. andrew.fogarty@ nottingham.ac.uk. Received June 20, Accepted for publication August 24, Am J Clin Nutr 2008;87:30 5. Printed in USA American Society for Nutrition

2 WEIGHT CHANGE AND INFLAMMATION 31 were separated by centrifugation (room temperature, 3000 rpm, 5 min), typically within 15 min, and stored at 80 C. These samples were subsequently defrosted, and highly sensitive CRP was measured with the use of an automated, immunoturbidimetric assay on an Olympus AU5400 analyzer (Miami, FL). All analyses were performed by the same professional in The normal range in apparently healthy persons in the United States is 1.5 mg/dl (9). Statistical analysis For the statistical analysis, smoking status was defined in 4 categories: current smoker (those who had smoked within 1 mo of the study in 1991 or 2000), long-term ex-smoker (those who had not smoked for 1 mo before the appointment in 1991), shorter-term ex-smoker (those who were current smokers in 1991 but had not smoked for 1 mo in 2000), and never smokers. The number of pack-years of smoking exposure was estimated for participants from their reported age at starting smoking and the usual amount smoked during this period. The cross-sectional data for serum CRP were not normally distributed and so are presented as the median values with interquartile ranges, whereas the cross-sectional data for weight were normally distributed and are presented as means with SDs. We calculated the change in weight and change in CRP between 1991 and Because these values were both normally distributed, parametric statistical tests were used for the analysis. Initially, we analyzed the association of the change in weight with change in serum CRP with the use of multiple linear regression with adjustment for a priori confounding factors of age, sex, smoking status, baseline body mass index (BMI; in kg/m 2 ), and total of cigarette pack-years smoked. Potential confounding factors, including baseline weight, baseline CRP, vitamin C intake, vitamin E intake, and polyunsaturated fatty acid intake, were added to the model and were considered as potential confounding factors if their inclusion altered the size of the effect by 10%. Because none of these variables met this criterion, only a priori confounders were considered in the final model. We examined the linearity of the relation between change in weight and change in serum CRP by categorizing weight in deciles and assessed this association both graphically and by using the likelihood ratio test. Because the relation is linear, results are presented both as change in milligram per liter of serum CRP per kilogram increase in weight and graphically in deciles. We tested for interactions between sex, age, smoking status, and baseline BMI and baseline CRP on the relation between change in weight and change in CRP, and considered these significant at the value of P The analyses were performed with STATA version 9 (Stata Corporation, College Station, TX). By using an estimate of within-subject SD of 7 mg/l for the change in CRP over 9 y derived from our own data, the available maximum sample size of 1222 participants provides 90% power to detect a linear trend between change in weight and change in CRP with a difference in change in CRP of 2 mg/l between the lowest and highest quintiles of change in weight (NQUERY, 4.0; Statistical Solutions, Saugus, MA). RESULTS Of the 2633 persons who participated in the original survey in 1991, 2614 participants provided complete baseline data, including weight. Of those, 1334 (51%) provided follow-up data in Four percent of those who had participated in 1991 had died, 7% had moved from the area, 23% declined or were unable to participate in the study, and we were unable to establish contact with the other 15%. Ninety-three percent, or 2425, of those who participated in 1991 and 97%, or 1301, of those who participated in 2000 provided a blood sample, and 1222 participants provided paired data for both weight and serum CRP in 1991 and 2000, which were available for subsequent analyses. These data were generally representative of the total study population in both years (Table 1). Compared with all participants who responded in 2000, there was a slightly lower response rate of those in the lowest quintile for weight in 1991, with 47% of this subgroup providing data in 2000 (Figure 1). A similar differential in the response rate was seen in those who were in the top quintile for CRP in 1991, of whom 47% provided samples in 2000 (Figure 1). The geometric mean for serum CRP increased significantly (P 0.001) from 1.22 mg/l (95% CI: 0.03, mg/l) in 1991 to 1.76 mg/l (95% CI: 0.09, 62.0 mg/l) in Change in serum CRP was normally distributed, and the mean increase in serum CRP was 0.85 mg/dl (95% CI: 91.1, 44.0 mg/dl). Mean weight also increased significantly from kg to kg between 1991 and 2000 (mean increase: 2.9 kg; 95% CI: 2.6, 3.2 kg). The increase in weight was generally higher in women, smokers, and younger participants (Table 2). After adjusting for the a priori confounding factors of sex, age, baseline BMI in 1991, smoking status in both 1991 and 2000, and pack-years of cigarettes consumed, there was a significant linear association between change in weight and change in serum CRP between 1991 and 2000, such that a 1-kg increase in weight was associated with an increment of 0.09 mg/l (95% CI: 0.02, 0.16 mg/l) in serum CRP during the period of the study. This represents an increase in serum CRP in the highest decile of weight gain that was 2.14 mg/l (95% CI: 0.28, 4.01 mg/l) larger than the increase in serum CRP in the lowest decile of weight gain (Figure 2). Eliminating outlying values by restricting the analysis to those with a CRP of 50 mg/l had a small effect when outliers from 1991 were excluded, such that a 1-kg increase in weight was associated with an additional increment of 0.12 mg/l (95% CI: 0.07, 0.17 mg/l) in serum CRP, but there was no substantial change after excluding outliers from 2000 (0.10 mg/l; 95% CI: 0.03, 0.16 mg/l). No interactions were observed with sex, age, smoking status, or baseline weight or serum CRP in 1991 (Table 3). DISCUSSION This study shows that in a randomly selected population of adults an increase in body weight during a 9-y interval was positively associated with a change in CRP, a marker of systemic inflammation. The relation of CRP to weight gain was linear, with each 1-kg gain in weight being associated with an increase in CRP of 0.09 mg/l (05% CI: 0.02, 0.16 mg/l) after adjusting for potential confounding factors. With a risk stratification for cardiovascular disease derived from healthy American men and women (9), this would equate to an estimated relative increase in risk of a cardiovascular event of 30% for those in the lowest quintile of baseline serum CRP who gain 4 kg, assuming a causal relation between inflammation and cardiovascular disease and ignoring any additional effects of weight gain that are independent of inflammation.

3 32 FOGARTY ET AL TABLE 1 Baseline characteristics of study population in Cross-sectional population, 1991 Provided blood Cross-sectional population, 2000 Provided blood Study population who provided both weight and CRP measures in Total (n 2614) for CRP (n 2425) Total (n 1346) for CRP (n 1301) 1991 and 2000 (n 1222) Men [n (%)] 1308 (50) 1220 (50) 667 (50) 647 (50) 613 (50) Age (y) Height (m) Weight (kg) BMI (kg/m 2 ) Men Women BMI (kg/m 2 ) [n (%)] 125 (5) 111 (5) 38 (3) 33 (3) 43 (4) [n (%)] 1190 (46) 1117 (46) 462 (34) 451 (35) 562 (46) [n (%)] 992 (38) 919 (38) 585 (43) 564 (43) 485 (40) 30 [n (%)] 307 (12) 278 (11) 261 (19) 253 (19) 132 (11) Smoking status [n (%)] Never 1292 (49) 1189 (49) 680 (51) 654 (50) 642 (53) 3 Ex 728 (28) 679 (28) 510 (38) 494 (38) 355 (29) 3 Current 594 (23) 557 (23) 156 (12) 153 (12) 225 (18) 3 CRP (mg/l) 1.14 ( ) ( ) 1 In 1991 participants who did not give a blood sample were significantly younger (mean age: 41.2 y) than those who did give a blood sample (mean age: 44.7 y) (P 0.001, linear regression). There were no other significant differences between participants who gave blood and those who did not give blood in age, height, weight, BMI, or smoking status in 1991 or CRP, C-reactive protein. 2 x SD (all such values). 3 Values from Median; interquartile range in parentheses (all such values). Methodologic considerations Our findings are based on a randomly selected population from the electoral register of a Local Authority Area in Nottingham and are therefore likely to be representative of the general population. Storage of the serum samples at 80 C and testing the assays by the same professional in the same time period permits confidence in the accuracy of the highly sensitive CRP assay (10). Although participation in the current study was potentially biased by survival, nonmigration, and motivation to participate, our data suggest that the participants in 2000 were broadly similar to the original population in terms of diet, smoking history, initial lung function, and history of respiratory disease (8). Participants who donated a blood sample in both surveys were generally similar to participants who did not, with the exception that participants who did not give a sample in 1991 were slightly younger than participants who did, and we do consider this unlikely to influence our findings, although it may slightly modify the generalizability of our findings. Our response rate of 51% of the original study population raises concerns about response bias, particularly because persons having higher BMI values (11) or serum CRP concentrations (12) have increased rates of mortality. In our study population we were able to account for 85% Participants (n) % 50% 55% 52% 50% 50% 53% 54% 53% 47% Weight (in kg) ranked as quintiles in CRP (in mg/dl) ranked as quintiles in 1991 FIGURE 1. Participants who participated in both 1991 and 2000 in quintiles of baseline weight and serum C-reactive protein (CRP) in 1991.

4 WEIGHT CHANGE AND INFLAMMATION 33 TABLE 2 Summary of change in weight in total study population from 1991 to Participants Change in weight n kg Total population Sex Men Women Smoking status in 1991 Never Ex Current Age in 1991 (y) Weight in quintiles in 1991 (kg) 1 ( ) ( ) ( ) ( ) ( ) Significant comparisons between sex (P 0.02, t test), smoking status (P 0.19, ANOVA), age (P 0.001, test for trend), and weight (P 0.03, test for trend). 2 x SD (all such values). of the original cohort, of whom 4% had died. A small differential was observed in response rates for baseline weight with 47% of participants in the lowest quintile for weight in 1991 responding in 2000 and also for baseline CRP with 47% of participants in the highest quintile for CRP in 1991 participating again in the second survey in However, for our association between change in weight and change in CRP to be a consequence of differential response rates leading to bias, participants who did not respond in the follow-up study would have to have had a strong inverse association between weight gain and CRP. This is known to occur, particularly in the context of clinical disease such as infection and advanced malignancy. However, we would anticipate that, although advanced clinical disease will have affected some of our study population, it would have been necessary for this to have affected the majority of the nonresponders to create an artifactual association between change in weight and change in CRP which we consider unlikely in our population who were aged y in Because we do not have validated data available on the incidence of inflammatory diseases, the concurrent use of medications, or lifestyle exposures such as exercise (13) in our dataset, we are unable to exclude the possibility that these may constitute unadjusted confounding factors for the association between change in weight and change in systemic inflammation. Finally, our use of a cohort study design means that we are unable to eliminate the possibility that our results are a consequence of reverse causality, ie, increased systemic inflammation results in weight gain, although the short-term weightloss studies suggest otherwise (4). Previous studies of weight change and serum CRP The size of effect found in our study of an increase in weight of 1 kg resulting in an increment of CRP of 0.09 mg/l is similar to that seen in the recent meta-analysis by Selvin et al (4), which examined 28 lifestyle interventions that aimed at reducing weight. The investigators concluded that weight loss was associated with a decline in CRP concentrations across all types of interventions, with an overall mean change in CRP of 0.13 mg/l per 1-kg loss of weight. Although this size of effect for change in CRP per kilogram decrease in weight is within our CIs of 0.02 to 0.16 mg/l, and thus consistent with our data, the data from the weight-loss studies have necessarily looked at the effect of a decrease in weight on CRP, whereas our data from a community population over 9 y showed that most persons gained weight with an average change in weight of 2.9 kg (95% CI: 2.6, 3.2 kg). Furthermore, Figure 2 shows that even participants who lost weight over the 9yofourstudy did not, on average, experience a decrease in serum CRP. This discrepancy between our data and the weight-loss intervention studies may be a consequence of a variety of factors. The most notable factor is the period of follow-up, because we resurveyed our original population after 9 y, whereas the longest follow-up available in the weight-loss intervention studies was 24 mo, with most of the studies being less than a year in duration (4). Alternatively, other environmental or temporal factors in addition to an increase in weight may also result in the higher CRP values observed over the9yofourstudy. However, the consistency of the size of effect Change in CRP (mg/l) FIGURE 2. Association of weight change in deciles with change in C-reactive protein (CRP) from 1991 to Data are shown as size of effect relative to bottom decile of change in weight with 95% CIs. Multiple linear regression model was adjusted for sex, age (continuous), BMI in 1991, smoking status (categorical), and smoking pack-years in 1991.

5 34 FOGARTY ET AL TABLE 3 Effect of change in weight on C-reactive protein (CRP) between 1991 and Participants Change in serum CRP P for interaction 2 n mg/l (0.02, 0.16) 3 Stratified by sex 0.22 Men (0.02, 0.24) Women ( 0.03, 0.15) Stratified by age in 1991 (y) ( 0.03, 0.28) ( 0.04, 0.13) ( 0.03, 0.25) ( 0.29, 0.13) (0.06, 0.31) Stratified by smoking status 0.51 Never ( 0.07, 0.16) Ex, stopped before (0.05, 0.23) Ex, stopped between 1991 and ( 0.09, 0.29) Current (0.08, 0.30) Stratified by BMI in 1991 (kg/m 2 ) ( 0.30, 0.35) (0.02, 0.20) (0.05, 0.24) ( 0.48, 0.14) 1 Multiple linear regression model was adjusted for sex, age (continuous), smoking status (categorical), smoking pack-years in 1991, and BMI in Test for effect modification of the relevant exposure on the association between change in weight and change in CRP. 3 Incremental change; 95% CI in parentheses (all such values). between our data and the weight-loss intervention studies suggest that adipose tissue is an important factor in modulating systemic inflammation, possibly against the background of as yet unidentified environmental exposures, although we cannot discount the possibility that a change in lean mass may also influence systemic inflammation. Our data complement previous studies showing that baseline markers of systemic inflammation are elevated in those who both gain (14 16) and lose weight (16), suggesting that the association between systemic inflammation and body weight is complex. However, consideration should be given to maintaining body weight in adults after they have attained an appropriate weight based on existing recommendations for BMI (17). Our observation that participants who increase their weight have an associated linear increase in systemic inflammation is consistent with the observation that women who have even a modest weight gain during adulthood have a greater risk of death independent of physical activity. In the cohort from the Nurses Heath Study, an increase in weight was associated with an excess of death from cardiovascular disease and death from cancer, compared with those who did not increase their weight (11). Similarly, in a population of Swedish men whose BMI increased by 15%, an excess of pancreas and renal cell cancers were seen (18). Chronic low-grade inflammation provides a unifying explanation for both of these observations having been implicated in the pathogenesis of both cancer (19, 20) and cardiovascular disease (9, 21). Although the mechanisms that drive the link between adiposity and chronic inflammation are not fully delineated, the production by fat cells of both proinflammatory cytokines, such as interleukin-6 and tumor necrosis factor-, and adipocytokines, such as adiponectin and lectin, may be important for many aspects of inflammation and immunity (22, 23). Conclusion In conclusion, we have shown for the first time in a prospective population-based study that an increase in weight in adults is associated with an increase in systemic inflammation during a 9-y interval, after correction for likely confounding factors. We thank all persons who took part in the study and Professor Ian Mac- Donald for comments on the manuscript. The author s responsibilities were as follows AWF (guarantor): developed the hypothesis, performed statistical analyses, and wrote the final version of the manuscript; JRB, SAL, and TMM collected the data and reviewed and critiqued the manuscript; SJ did the biochemical analyses and reviewed and critiqued the manuscript; CG performed the initial statistical analyses and reviewed and critiqued the manuscript. None of the authors had a personal or financial conflict of interest. REFERENCES 1. Visser M, Bouter L, McQuillan G, Wener M, Harris T. Elevated C-reactive protein levels in overweight and obese adults. JAMA 1999; 282: Mokdad A, Serula M, Dietz W, Bowman B, Marks J, Koplan J. The continuing epidemic of obesity in the United States. JAMA 2000;284: Haslam D, Sattar N, Lean M. Obesity: time to wake up. Br Med J 2006;333: Selvin E, Paynter N, Erlinger T. The effect of weight loss on C-reactive protein. A systematic review. Arch Intern Med 2007;167: Britton J, Pavord I, Richards K, et al. Dietary magnesium, lung function, wheezing, and airway hyper-reactivity in a random adult population sample. Lancet 1994;344: Britton J, Pavord I, Richards K, et al. Dietary sodium intake and the risk of airway reactivity in a random adult population. Thorax 1994;49: Britton J, Pavord I, Richards K, et al. Dietary antioxidant vitamin intake and lung function in the general population. Am J Respir Crit Care Med 1995;151:

6 WEIGHT CHANGE AND INFLAMMATION McKeever T, Scrivener S, Broadfield E, Jones Z, Britton J, Lewis S. Prospective study of diet and decline in lung function in a general population. Am J Respir Crit Care Med 2002;165: Ridker P. High-Sensitivity C-Reactive Protein. Potential adjunct for global risk assessment in the primary prevention of cardiovascular disease. Circulation 2001;103: Aziz N, Fahey J, Detels R, Butch A. Analytical performance of a highly sensitive C-reactive protein-based immunoassay and the effects of laboratory variables on levels of protein in blood. Clin Diagn Lab Immunol 2003;10: Hu F, Willett W, Li T, Stampfer M, Colditz G, Manson J. Adiposity as compared with physical activity in predicting mortality amongst woman. N Engl J Med 2004;351: Man P, Connett J, Anthionisen N, Wise R, Tashkin D, Sin D. C-reactive protein and mortality in mild to moderate chronic obstructive pulmonary disease. Thorax 2006;61: Mora S, Lee I-L, Buring J, Ridker P. Association of physical activity and body mass index with novel and traditional cardiovascular biomarkers in women. JAMA 2006;295: Engstrom G, Hedblad B, Stavenow L, Lind P, Janzon L, Lindegarde F. Inflammation-sensitive plasma proteins are associated with future weight gain. Diabetes 2003;52: Barinas-Mitchell E, Cushman M, Meilahn E, Tracy R, Kuller L. Serum levels of C-reactive protein are associated with obesity, weight gain, and hormone replacement therapy in healthy postmenopausal women. Am J Epidemiol 2001;153: Barzilay J, Forsberg C, Heckbert S, Cushman M, Newman A. The association of markers of inflammation with weight change in older adults: the Cardiovascular Health Study. Int J Obes (Lond) 2006;30: Willett W. Guidelines for a healthy weight. N Engl J Med 1999;341: Samanic C, Chow W-H, Gridley G, Jarvholm B, Fraumeni J. Relation of body mass index to cancer risk in 362,552 Swedish men. Cancer Causes Control 2006;17: Akyon Y. Effect of antioxidants on the immune response of Helicobacter pylori. Clin Microbiol Infect 2002;8: Calle E, Kaaks R. Overweight, obesity and cancer: epidemiological evidence and proposed mechanisms. Nat Rev Cancer 2004;4: Ridker P, Buring J, Rifai N, Cook N. Development and validation of improved algorithms for the assessment of global cardiovascular risk in women. JAMA 2007;297: Tolg H, Moschen A. Adipocytokines: mediators linking adipose tissue, inflammation and immunity. Nature 2006;6: Van Gaal L, Mertens I. Mechanisms linking obesity with cardiovascular disease. Nature 2006;444: