Race differences in the effect of obesity on inflammation among Black and White older adults who quit smoking: a longitudinal analysis of the Health and Retirement study

Abstract

Importance

Race differences in inflammation have been attributed to higher body mass index (BMI); however, few studies have comprehensively examined whether changes in obesity influence changes in C-reactive protein (CRP) by race, particularly following smoking cessation.

Objective

As CRP declines after quitting smoking, the current study examined whether longitudinal changes in multiple measures of obesity were differentially associated with CRP change among non-Hispanic Black (NHB) and non-Hispanic White (NHW) older adults who formerly smoked.

Design

This is a longitudinal analysis of the Health and Retirement Study, nationally representative study of US adults. Participants were matched on self-reported demographic, socioeconomic, and behavioral factors using propensity score analysis to control for the effects of social disadvantage. C-reactive protein, waist circumference, body weight, and BMI were assessed in 2006, 2010, and 2014. Linear mixed models examined race differences in the effects of time-varying obesity on CRP change (n = 899 observations).

Results

C-reactive protein significantly declined over time (β = −0.23, p < .0001). However, no differences in CRP change were found between NHBs (n = 146) and NHWs (n = 245). Only weight was positively associated with CRP change over time (β = 0.002, p < .001).

Conclusions and Relevance

In our matched sample, there were no race differences in CRP decline among older adults, and only weight was associated with a lessened decline over time. Our findings suggest that weight alone may be an important predictor of CRP in older adults rather than BMI, which has been studied extensively in previous research.

Introduction

Racial disparities in obesity have persisted for decades in the United States, with African Americans having a higher prevalence of obesity (49.5%) and overweight (76.7%) than their non-Hispanic White peers (34.3% and 66.7%, respectively).^1,2 Obesity, the second leading cause of preventable death and disability in the United States, is linked to chronic disease independently³ and through a pathway of chronic inflammation.⁴ Given excess body weight directly contributes to the promotion of proinflammatory cytokines,⁵ adults with obesity are more likely to have elevated levels of C-reactive protein (CRP), an acute-phase protein and common marker of both systemic inflammation,⁶ as well as repeated measurements of clinically-elevated CRP (e.g., >10 mg/L), than their counterparts without obesity.⁷ Racial disparities have further been observed in CRP, with African Americans in particular, having higher levels of CRP compared with their White peers.^8–10 Some studies have established that the racial differences observed in CRP were mostly due to body mass index (BMI),^8,9 although body weight did not fully explain the association.⁸ One study using a nationally representative sample found the association between BMI and CRP was similar across race, but the association between larger waist circumference and higher CRP varied by race/ethnicity in women particularly.¹¹

Smoking history may also contribute to observed racial disparities in CRP. Chronic inflammation is not only a key pathway connecting obesity with chronic disease but also a pathway between tobacco smoke exposure and chronic conditions.¹² CRP has been examined largely in humans in tobacco research as a biomarker of chronic inflammation^12,13 with smoking inducing elevations of CRP.¹² Therefore, smoking cessation is associated with a decrease in CRP,^13–15 with a larger decrease the longer an individual has been smoke-free.¹³ However, studies have shown that following long-term smoking cessation (e.g., 20 years), CRP levels do not return to levels similar to those who have never smoked.^14–16 In addition, about 80% of adults who quit using tobacco gain weight, and African Americans have higher rates of this post-cessation weight gain over time.^17,18 Previous longitudinal studies showed that African Americans gained an average of 4-10 pounds (lbs.; 1.8-4.5 kg) after quitting for 1 year,¹⁹ and there were no differences in short-term weight gain (<6 months) between African American and White smokers.²⁰ Similarly, a prospective study of treatment-seeking adults with a 5-year follow up did not find differences between races in post-cessation weight gain.²¹ However, cohort studies with a longer follow-up period of 10-15 years have shown that African Americans are twice as likely to gain weight and gain excessively (i.e., >10 lbs.; 4.5 kg) after quitting smoking,^17,18 indicating that disparities in post-cessation weight gain may become apparent several years after smoking cessation. This disproportionate weight gain over time may contribute to inflammation-related chronic disease risk. Therefore, we sought to examine racial disparities in obesity and CRP in a sample of older adults who formerly smoked.

A majority of research on race-related differences in CRP has been conducted cross-sectionally and in nonsmoking samples.¹¹ Thus, there is a need to understand racial disparities in long-term inflammation change and how obesity contributes to these disparities. Our recent research longitudinally examined race differences in smoking history and CRP and found that smoking status, BMI, age, sex, and education were associated with CRP change in a sample of Black and White older adults. Yet, after stratifying the sample by race, smoking status and socio-demographic variables remained significant in White adults, but BMI was the only significant predictor of CRP for Black adults.²² While BMI is one of the most widespread measures of obesity due to its safety, simplicity, and overall convenience, existing evidence points to differences in cardiometabolic disease risk at differing BMI cutoffs across racial and ethnic groups;^23,24 therefore, it is necessary to use multiple measures of obesity, beyond BMI alone, to examine race differences in CRP.

Prospective studies that include longitudinal measurements of obesity are needed for direct comparisons of inflammation in African American and White older adults who formerly smoked. Further, understanding the unique effects of body weight and body composition is necessary to understand potential race differences in inflammation, a mechanism linking obesity to chronic disease. In our previous analysis, BMI was a unique predictor of CRP among Black adults, independent of socio-demographic variables; however, other measures of obesity were not examined. Thus, in the present analyses, we sought to identify unique effects of obesity that may explain these disparities so that we can better inform interventions and screening guidelines to reduce disparities.

Methods

Study methods and results are reported following the Strengthening the Reporting of Observational Studies in Epidemiology Statement for cross-sectional studies (Supplementary File).

Study design

The current study utilizes publicly available de-identified data from the Health and Retirement Study (HRS), a longitudinal, nationally representative study of US adults born between 1924-1941.²⁵ Original HRS participant recruitment occurred in 1992 with biannual waves of core data collection. A total of 16 924 participants provided biomarker data at 4-year intervals between 2006 and 2016, with half of the sample providing data every 2 years. The study sample was limited to 3551 adult participants who: (1) self-identified as non-Hispanic, White/Caucasian (NHW) or non-Hispanic, Black/African American (NHB), (2) have self-reported being quit for at least 1 year, and (3) denied ever having or developing cancer in 2006 (baseline).

Measures

Inflammation

The HRS obtained CRP data [micrograms per milliliter (µg/mL)] from high-sensitivity assays of dried blood spots from finger prick samples,²⁶ with greater CRP values indicating greater inflammatory response. CRP data from 3 assessment waves (2006, 2010, and 2014) were included in study analysis.

Smoking status

At each wave, participants self-reported whether they were currently smoking cigarettes (yes vs. no). Former smoking was defined as those who (1) reported history of smoking prior to 2006, (2) have quit smoking for at least 1-year, and (3) denied any smoking at each assessment point between 2006 and 2014.

Obesity

Physical measurements were obtained from participants at each HRS assessment. Weight was measured in pounds using a digital scale, and height and waist circumference were measured to the nearest quarter inch by trained research staff. Body mass index was calculated from height and weight data in the HRS [(weight/(height)² × 703)]. A categorical (yes vs. no) obesity variable was defined as having a BMI of 30 or greater.

Control variables

Socio-demographics (i.e., race, age, sex, education, household income), smoking behaviors (i.e., cigarettes per day, years since quitting, and age started smoking), physical activity (i.e., self-report of whether participants engaged in moderate and vigorous activity at least 1 time per month) and physical health history (i.e., diagnosis of heart condition or diabetes) were assessed or verified by participants at each wave. Participants were categorized as having diabetes if they self-reported that a doctor had ever told them they had diabetes. Participants were categorized as having a heart condition if they self-reported that a doctor had ever told them they had a heart attack, coronary heart disease, angina, congestive heart failure, or other heart problems. Socio-demographic variables and smoking behaviors were controlled for through propensity score matching (discussed below) because they are closely associated with race and may obscure the true “effect” of race if included as covariates in the same analytical model. Chronic disease covariates were not included in the propensity score analyses because they are potential mediators in the pathway linking obesity and CRP, rather than confounders of the race-CRP association.

Statistical plan

We used propensity score matching to isolate any effects of race on obesity and inflammation change and to control for potential confounding in our analyses.²⁷ Socioeconomic status (SES; e.g., education, income) is documented as a primary reason for racial disparities in chronic diseases,^28–30 and this SES-chronic disease link may be due to access to health care, the built environment, exposure to carcinogens, and unhealthy behaviors.³¹ Further, adults with lower SES have higher rates of smoking and obesity than those with higher SES. Research also shows that African Americans begin smoking at a later age³² and smoke fewer cigarettes per day compared to their White counterparts.³³ Although it is common practice to control for socio-demographics by including them as covariates in a regression model, regression models with variables on race and other socio-demographic control variables must be carefully interpreted because these variables are so closely associated with each other;^34,35 including race and control variables in the same analytical model might obscure the true “effect” of race given that the control variables capture some of the relationship between race and the outcome variable. Because of the role of SES influencing behavioral risk factors, and the disproportionate rate of African Americans having a lower SES status in the United States, we calculated propensity scores to match Black and White adults on SES and other social and behavioral factors. This matching allowed us to use a more parsimonious model, which may increase statistical power and precision, by balancing the sample on socio-demographic and behavioral factors that are associated with race rather than including each variable as a covariate in the final model (see Supplementary Table for before and after matching comparisons of SES and behavioral variables by race). Previous research has used this approach in health disparities research, and studies have calculated propensity scores for race to match cases on characteristics such as demographic factors, geographic location, and healthcare access, to determine the degree to which racial/ethnic disparities in certain outcomes are explained.^36–39 In the current study, SAS PROC PSMATCH⁴⁰ was used to estimate propensity scores for race with a greedy algorithm, then NHW and NHB subjects were approximately matched 1:2 by specifying a caliper distance of 0.5 on the logit of the propensity scores, creating a final dataset of N = 391 (n = 245 NHW and n = 146 NHB). To calculate propensity scores, we used age (continuous), sex (0 = male, 1 = female), education level (ordinal), household income (ie, total annual household income in US dollars received during the previous calendar year; continuous), number of cigarettes smoked per day (continuous), years smoked (continuous), and age started smoking (continuous).

Next, using the matched dataset, linear mixed effects longitudinal modeling, using SAS PROC MIXED and restricted maximum likelihood estimation, was used to examine race differences and obesity effects on changes in inflammation among adults who formerly smoked. This method allowed for the simultaneous examination of intra-individual and inter-individual change within a single statistical test, and we were able to examine individual change in CRP levels while accounting for race or between-subjects, effects on CRP. Random subject intercept and time effects were included to account for the nesting of repeated observations within subjects; the base year of 2006 was recoded as 0 with the square root of differences between subsequent years used to create a linear relationship. In cases of incomplete data across time, all available data and weight measurements were used. To determine race differences in the effect of obesity on CRP change, 3 separate models were analyzed for BMI (continuous), waist circumference (continuous), and body weight (continuous). Each obesity variable was centered and entered as a time-varying covariate to account for changes in obesity over time. While potential confounding variables (e.g., age, sex, education, and income) were controlled through propensity score matching, we also included covariates that were strongly associated with CRP in the literature in our linear mixed effects longitudinal models. Physical activity (0 = no, 1 = yes) was initially examined as a covariate in the model, as well as self-reported diagnoses of diabetes and heart condition at any assessment point between 2006 and 2014. We initially examined these chronic disease covariates (i.e., diabetes and heart condition) as time-varying controls in each model because diabetes remained significantly different between groups after matching, and heart disease risk increases with age. Similarly, though controlled for through propensity score matching, smoking severity and history (i.e., cigarettes per day and years since quitting; continuous) have strong associations with the outcome, CRP, and were, therefore, examined as covariates in the model to further improve precision of the effect estimate of race (although these coefficients were not interpreted). Next, a race × obesity measure interaction term was included to test whether race predicted the association between obesity change, and a race × obesity measure × time interaction term was included to test whether it predicted CRP change over time, holding constant the effects of the other covariates.

For all of our final models, we did not include physical activity because it was not associated with CRP at any time point and did not improve model fit. We removed race × obesity measure × time interaction as it was not significant (see explanation in Results section) and led to a poorer model fit. To improve model fit, we kept the chronic disease covariates and smoking severity and history in the final models. The following model was specified and interpreted in the final analyses:

Level 1: CRP _it = π_0i+π_1i (Time_it) + π_2i (Obesity measure_it ) + π_3i(Diabetes_it ) + π_4i (Heart condition_it ) + e_it

Level 2: π_0i = β₀₀ + β₀₁(Race_i )+ β₀₂(Race_i x Obesity measure_i ) + β₀₃(Cigarettes per day_i )+ β₀₄(Time since quit_i ) + r_0i

π_1i = β₁₀+ β₁₁(Race_i )+ β₁₂(Time since quit_i )+r_1i

π_2i = β₂₀

π_3i = β₃₀

π_4i = β₄₀

Combined: CRP_it = β₀₀ + β₀₁ (Race_i ) + β₀₂ (Race_i x Obesity measure_i )  + β₀₃ (Cigarettes per day_i ) + β₀₄ (Time since quit_i ) + β₁₀ (Time_it )   + β₁₁ (Race_i x Time_it ) + β₁₂ (Time since quit_i x Time_it )   + β₂₀ (Obesity measure_it ) + β₃₀ (Diabetes_it )   + β₄₀ (Heart condition_it ) + r_0i + r_1i (Time_it ) + e_it

Definition of terms: β_0i = average baseline CRP (intercept); β_1i = average rate of CRP change (slope); β_2i = effect of obesity measures on CRP over time; β_3i = effect of diabetes on CRP over time; β_4i = effect of heart condition on CRP over time; r_0i = error component of intercept; r_1i = error component of slope; e_it = overall random error.

Results

The final matched dataset included 391 adults who had at least 1 measurement of CRP over the study period (899 total observations). Participant characteristics can be found in Table 1. In this matched sample, participants were balanced on all socio-demographic and behavioral variables, with the exception of diabetes diagnosis. Particularly, a significantly higher proportion of NHB adults had diabetes (47%) compared to NHW adults (22%; p < .001). Baseline race differences were observed in CRP, waist circumference, and weight (ps < 0.05).

Table 1.

Sample characteristics.

Characteristic	White (n = 245)	Black (n = 146)	t or χ  ²	p
Sex (n, % female)	148 (60.4%)	86 (58.9%)	0.029	0.77
Education (n, % high school or less)	165 (67.3%)	114 (71.5%)	5.16	0.64
Grade school	40 (16.3)	30 (20.6)
High school	125 (51.0)	74 (50.7)
College degree	55 (22.5)	31 (21.2)
Graduate school	25 (10.2)	11 (7.5)
Age (M, SD in years)	70.6 (7.5)	70.4 (8.0)	0.25	0.79
Household income (M, SD in USD)	13,480.1 (23,048.6)	10,931.2 (19,479.9)	1.17	0.24
Diagnosis of diabetes (n, %)	54 (22.0%)	55 (37.7%)	11.11	0.0009*
Diagnosis of heart condition (n, %)	82 (33.5%)	42 (28.8%)	0.93	0.33
Obesity measures
Body mass index (BMI; M, SD)	28.4 (5.9)	29.5 (5.9)	−1.78	0.11
Body weight (M, SD in pounds)	177.6 (43.3)	188.9 (43.7)	−2.49	0.01*
Waist circumference (M, SD in inches)	39.9 (5.9)	41.2 (5.6)	−2.11	0.035*
Outcome
Baseline C-reactive protein (µg/mL; M, SD)	4.95 (8.2)	9.01 (19.7)	−2.84	0.019*
History of cigarette use
Age started smoking (M, SD in years)	21.4 (14.5)	20.3 (12.8)	0.76	0.43
Cigarettes smoked per day (M, SD)	10.4 (9.2)	8.6 (8.4)	1.58	0.11
Time since quitting smoking (M, SD in years)	20.6 (13.2)	20.7 (13.9)	−0.06	0.95

p-values denote significant group differences using t-tests and chi-square analyses between White vs. Black groups. t-tests were used for continuous variables (age, household income, BMI, body weight, waist circumference, CRP, age started smoking, cigarettes smoked per day, and time since quitting) and χ² tests were used for categorical variables (sex, education, diagnosis of diabetes, and diagnosis of heart condition)

Abbreviations: USD = US dollar, BMI = body mass index.

^* p<.05.

Controlling for smoking behavior/history (i.e., cigarettes per day and time since quitting) and chronic disease covariates, 3 separate models were fitted to analyze race differences in the effect of obesity measures on CRP: model 1 = BMI, model 2 = waist circumference, and model 3 = weight (see full results in Table 2). In all models, there was a significant decrease in CRP over time (ps < 0.0001). The race × time effect was not significant, indicating that there were no race differences in the rate of CRP change. In model 1, BMI positively predicted baseline CRP (P<.001), but was not associated with change in CRP over time. Similarly, in model 2, waist circumference was positively associated with initial CRP (P < .001) but not with rate of change. In model 3, weight predicted both baseline CRP (P < .05) and change in CRP over time (P < .01). Additionally, the race × obesity variable was not associated with baseline CRP in any models (ps > 0.05). Of note, although there was no significant interaction of race and obesity measure in any of our 3 models, we still examined the race × obesity measure × time, which was not significant in any models (ps > 0.05), indicating that the effect of obesity on change in CRP over time was not moderated by race.

Table 2.

Effects of obesity measures on CRP.

Variable	BMI		Waist circumference		Weight
	Estimate (SE)	p	Estimate (SE)	p	Estimate (SE)	p
Intercept	1.22 (0.15)	<.0001*	1.23(0.16)	<.0001*	1.24 (0.15)	<.0001*
Time (years)	−0.25 (0.06)	<.0001*	−0.29 (0.06)	<.0001*	−0.25 (0.06)	<.0001*
Race (NHB)	0.17 (0.15)	0.25	0.14 (0.16)	0.35	0.21(0.15)	0.16
Race (NHB) × time	−0.014 (0.06)	0.83	0.01 (0.07)	0.83	−0.03 (0.06)	0.67
Obesity measure (baseline)	0.05 (0.01)	0.0002*	0.04 (0.01)	0.001*	0.004 (0.002)	0.013*
Obesity measure × time	0.010 (0.005)	0.06	0.008 (0.006)	0.18	0.002 (0.001)	0.001*
Obesity measure × race (NHB)	−0.014 (0.016)	0.41	0.005 (0.006)	0.16	−0.002 (0.003)	0.49
Heart condition	0.045 (0.12)	0.70	0.07 (0.12)	0.55	0.034 (0.12)	0.77
Diabetes	−0.045 (0.13)	0.72	−0.20 (0.13)	0.13	−0.04(0.13)	0.75
Cigarettes per day	0.002 (0.006)	0.79	0.006(0.006)	0.33	0.003 (0.006)	0.59
Time since quitting	−0.014 (0.005)	0.005*	−0.014(0.005)	0.009*	−0.014 (0.005)	0.005*
Time since quitting  ×  time	0.003 (0.002)	0.26	0.003(0.002)	0.22	0.002 (0.002)	0.30

“Obesity measure” represents variable of interest in each model identified at the top of table (ie, BMI, waist circumference in inches, or weight in pounds).

Abbreviations: NHB = non-Hispanic Black, BMI = body mass index.

^* p<.05 (bolded).

Exploratory analysis

Given the differences in weight distribution between men and women, we explored the interactive effect of race and sex in our models. First, we tested a race × sex interaction term in our 3 final models. Next, we stratified our sample by sex, then tested our 3 models separately by sex. Our results showed that there was no significant interaction of race and sex in the overall models for BMI (p = .76), waist circumference (p = .88), or weight (p = .80), and sex was not significantly associated with CRP (ps > 0.05). Stratified analyses did not show any unique effects of race and obesity (particularly waist circumference; p = .59) on CRP, and results were similar in women to the overall models.

Discussion

Prior literature examining race differences in CRP has primarily focused on cross-sectional, nonsmoking populations. There remains a need to better understand potential racial disparities in the changes in CRP over time and how different measurements of obesity may contribute to these disparities, particularly among individuals with a history of smoking who may have heightened disease risk. We have previously examined smoking status and longitudinal changes in CRP in an older adult, unmatched sample between 2006 and 2014, and found that CRP declined over time, although the decline halted for NHB between 2010 to 2014; current smoking, age, sex, BMI, and education all predicted CRP levels. Interestingly, however, the model held for NHW adults when the sample was stratified by race and showed a significant decline in CRP over time, but did not for NHB adults—only BMI was significant in predicting CRP in this subsample.²² In the current sample restricted to those who have quit smoking and matched on socioeconomic variables and previous tobacco use history, we sought to further explore race and CRP change and delineate the impact of different types of obesity on CRP, taking into account the long-term impact of smoking cessation. We found no differences in changes in CRP between NHW and NHB older adults holding obesity and other covariates constant. While our results showed that body weight was significantly associated with both baseline CRP and CRP change over time among NHW and NHB older adults who formerly smoked, other measures of obesity—specifically BMI and waist circumference—did not predict CRP change over time. Notably, race also did not alter the effect of any obesity measure on CRP in this study.

Numerous population level studies have pointed to disparities in CRP levels across racial groups, with previous findings consistently showing higher levels of CRP among NHBs compared to NHWs.^41–44 Previous research examining the HRS have found that elevated levels of CRP among Black respondents remained, even after adjusting for multiple socio-demographic factors (e.g., age, sex, residential setting) and smoking history,⁴² with one study finding that racial differences in longitudinal CRP was mediated by educational disadvantage and baseline smoking status.⁴⁴ Altogether, this research has suggested that these racial differences in CRP are due in part to socio-environmental factors within the built environment, in addition to factors such as obesity. Presently, our study examined multilevel longitudinal models and capitalized on the statistical technique of propensity score matching on socioeconomic and behavioral factors. As NHB have held a disproportionate burden of social disadvantage and unhealthy behaviors, this method allowed for the distribution of baseline socio-demographic and smoking behavior characteristics to be similar between NHB and NHW participants so that the race “effect” on CRP can be isolated. While results of a t-test (see Table 1) showed race differences in CRP at baseline, there were no differences in initial CRP controlling for obesity, chronic disease covariates, and smoking history. There was also no interactive effect of race and obesity measures on CRP among adults that were matched on key socioeconomic and behavioral variables. Although we are interpreting with caution, our results may support conclusions from previous literature suggesting socioeconomic factors are important drivers of racial differences in CRP above and beyond obesity.

We observed race differences in specific obesity measures within our sample at baseline, with NHB adults having higher body weight and larger waist circumference than NHW adults, while BMI did not differ significantly by race. Within the broader literature, however, there has been evidence suggesting ethno-racial differences in body fat distribution and composition among NHW and NHB adults. For example, BMI has been shown to be higher among Black than White adults, with a general trend pointing to higher BMI among Black women in particular.⁴⁵ Other researchers have consistently pointed to higher levels of visceral adiposity among White adults compared with their Black peers when using more refined measures of body fat distributions.^46–49 However, we did not see any differences in obesity (BMI, waist circumference, and weight) and CRP in women compared to men in our matched sample.

Moreover, BMI and waist circumference have been used extensively for decades as a measurement of obesity to determine body fat and central adiposity, and both have been associated with SES variables (e.g., education, income, and neighborhood poverty).^50–52 Interestingly, our study showed that body weight was significantly associated with change in CRP among NHW and NHB older adults who formerly smoked, but other common measurements of obesity (i.e., BMI, waist circumference) were not associated with changes in CRP. The association between an individual’s body mass and CRP may be differentially affected by factors related to post-cessation weight gain that were unmeasured in the current study. For example, a review of reviews noted that post-cessation weight gain can differentially trigger chronic conditions (e.g., diabetes) depending on various factors, such as race/ethnicity, SES, and eating behaviors.⁵³ Although we attempted to control for such potential factors, future studies are still needed to clarify these associations–particularly among individuals who formerly smoked. Furthermore, menthol cigarette use may also contribute to race differences in inflammation among former smokers, as menthol cigarette use has been disproportionately higher among Black adults.⁵⁴ Some research also suggests that menthol cigarette use may be associated with greater post-cessation weight gain⁵⁵ and obesity,⁵⁶ potentially compounding obesity-related inflammatory risk. Although menthol cigarette use was not directly measured in the HRS, its disproportionate use historically among African Americans and its biological and behavioral effects warrant further consideration as an unmeasured contributor to long-term inflammation among older adults who formerly smoked.

Strengths and limitations

The current study assessed multiple measures of obesity to prospectively examine race differences in obesity and CRP among older adults with a history of smoking. This was the first study to our knowledge to isolate the effects of race and obesity on inflammation using propensity score matching, a method that has been increasingly used to account for and control for the strong association between race and social disadvantage. We also examined a sample of older adults who formerly smoked, a group at high risk of developing chronic disease, to allow us to explore any lasting effects of smoking cessation and possible contributors to racial disparities over time. However, results should be interpreted in light of the study’s limitations. Given our analysis of older adults and using a matched dataset, our results may not be generalizeable to younger or middle-aged adults, nor to a broader range of socioeconomic statuses. Although we assessed multiple measures of obesity to explore differences in general vs. central obesity, BMI, a measure developed in predominately White male populations, does not distinguish between lean body mass and fat mass,⁵⁷ and waist circumference does not distinguish between visceral and subcutaneous abdominal adipose tissue, despite having been shown to be a much better predictor of cardiometabolic disease, including increased Type 2 Diabetes risk and cardiovascular disease events,⁵⁸ beyond BMI alone. Body weight alone gives no specific insight into actual body composition, including fat distribution. With a greater understanding of these markers of obesity among individuals who formerly smoked, we will be able to further elucidate mechanisms and potential preventive strategies. In addition, self-reported measures of physical activity were not significantly associated with CRP at baseline, so they were excluded from the final analyses. Therefore, we did not have robust measures of physical activity, or other measures of energy expenditure such as diet, to control for in our analyses. Although we accounted for diagnoses of diabetes and cardiovascular disease, future work is needed to delineate how changes in metabolic variables affect inflammation among older adults who formerly smoked, especially given the observed race differences in diabetes diagnosis despite propensity score matching.

Conclusion

The current study attempted to further explore the association between race, obesity, and inflammation by taking a closer look at the effects of specific types of obesity on longitudinal CRP change. Our findings suggest that weight alone may be an important predictor of CRP in older adults rather than BMI, which has been studied extensively in previous research. Given race did not modify the associations examined in the current analyses when NHB and NHW adults were matched on socio-demographic and behavioral factors, our results point to the importance of continuing to take into account factors related to social disadvantage when conducting future research that examines racial differences in obesity and CRP. Future studies are still needed to clarify these relationships among individuals who formerly smoked, as the scientific knowledge regarding the full long- term effects of smoking cessation in relation to CRP and disease risk is still evolving.

Author contributions

Marcia M. Tan (Conceptualization [lead], Data curation [lead], Funding acquisition [lead], Investigation [lead], Methodology [lead], Project administration [lead], Resources [lead], Writing—original draft [equal], Writing—review & editing [equal]), Tiwaloluwa A. Ajibewa (Writing—original draft [equal], Writing—review & editing [equal]), John Cursio (Data curation [lead], Formal analysis [lead], Writing—original draft [supporting]), Donald Hedeker (Formal analysis [supporting], Investigation [supporting], Writing—original draft [supporting]), Andres Mauricio Garcia Sierra (Writing—review & editing [supporting]), Dedeepya Konuthula (Writing—review & editing [supporting]), and Nancy C. Jao (Investigation [supporting], Visualization [lead], Writing—review & editing [equal])

Supplementary material

Supplementary material is available at Annals of Behavioral Medicine online.

Funding

Funding for this study and manuscript preparation was supported by the National Cancer Institute under grant number R21CA249357 to MMT. NCJ was supported by National Heart, Lung, and Blood Institute (K01HL164670).

Conflicts of interest

The authors declare that they have no conflict of interest. The content is solely the responsibility of the authors and does not necessarily represent the official views of National Institutes of Health (NIH). NIH had no role in the study design, collection, analysis or interpretation of the data, writing the manuscript, or the decision to submit the paper for publication.

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Transparency statement

Study registration

This study was not formally registered.

Analytic plan pre-registration

The analysis plan was not formally pre-registered.

Data Availability

The data analyzed during the current study are available from The University of Michigan Health and Retirement Study (HRS) and publicly available. Restrictions apply to the availability of some of these data, which were used under license for the current study, and so are not publicly available. Data are, however, available from the authors upon reasonable request and with permission of The University of Michigan HRS.

Analytic code availability

Analytic code used to conduct the analyses presented in this study are not available in a public archive. They may be available by emailing the corresponding author.

Materials availability

Materials used to conduct the study are not publicly available.