Two Patterns of Adipokine and Other Biomarker Dynamics in a Long-Term Weight Loss Intervention

RESEARCH DESIGN AND METHODS

Statistical analyses

For intention-to-treat analyses, we included all 322 participants and used the most recent values for weight and blood pressure. We characterized the study population across quartiles of BMI and stratified by sex for leptin and adiponectin because their levels significantly differed between sexes. The pattern analysis, however, was similar in men and women. We tested for statistically significant differences between different time points within dietary intervention groups by t test and for given time points between dietary groups by ANOVA. We used SPSS software, version 18 (SPSS Inc., Chicago, IL), and Stata software, version 9 (StataCorp LP, College Station, TX), for the statistical analysis.

Methods of the nonbiased approaches: trend analysis and clustering.

In the trend analysis, we define the trends correspondence measure (TCM) to represent the correspondence between the trend in the first (0–6 months) and second (7–24 months) parts of the trial for each biomarker. This analysis represents the fraction of the change in the second part of the intervention that continues the trend observed in the first part. Formally, it is defined as follows: Let represent the change in variable i in individual from time 0 to time t, in percentages of its value at time 0:

where represents the value of that variable in that individual at time t. Let be the vector connecting the points (0, 0) and (6, ). Let be the vector connecting the points (6, ) and (6, ). TCM is defined as the cosine of the angle between and the line containing (Supplementary Fig. 1). Hierarchical clustering variables were used. The h-clust function in R was used with default parameters. A distance matrix was derived from the TCM values of every two variables using the following:

RESULTS

The mean age of the study population was 52 ± 7 years with mean BMI of 30.9 ± 3.6 kg/m². Most of the participants (86%) were men. Higher baseline BMI was significantly associated with increased proportion of women and higher baseline fasting insulin, hsCRP, leptin, and chemerin serum concentrations and lower levels of HDL-C (Table 1). A significant sex difference was found for leptin and adiponectin. The three diet interventions resulted in two segments: a rapid weight loss phase during the first 6 months followed by partial or complete reversal at the 7–24 months interval (weight maintenance/regain phase) (Fig. 1A).

We identified two distinct major patterns in the dynamics of biomarker serum concentrations measured at baseline and at 6 and 24 months of intervention (Fig. 1B). The first, pattern A, included biomarkers whose dynamics closely corresponded to changes in body weight, exhibiting a rapid change between 0 to 6 months, during the weight loss phase, followed by partial or complete reversal at the 7–24 months interval, during the weight maintenance/regain phase. The following biomarkers exhibited this pattern: fasting plasma insulin, TGs, leptin, chemerin, MCP-1, and RBP4 serum concentrations (Fig. 1B). All biomarkers showed a significant difference (P < 0.05 for all) between baseline and 6 months and between 6 and 24 months, with chemerin and MCP-1 exhibiting a full return at 24 months to the baseline levels and leptin, RBP4, vaspin, progranulin, and fetuin-A remaining significantly different between baseline and 24 months (P < 0.05).

In sharp contrast, a second pattern of dynamics, pattern B, could be observed for HDL-C, hsCRP, HMW adiponectin, fetuin-A, progranulin, and vaspin during the 24 months of dietary intervention (Fig. 1B). Each of these parameters displayed a continuing, cumulative change in the same direction throughout the 24 months of dietary intervention, irrespective of weight stabilization or partial weight regain. Consistently, all of these biomarkers were significantly (P < 0.05 for all) improved after 24 months as compared with baseline, and except for HMW adiponectin, all exhibited a significant change between 6 and 24 months (P < 0.05), reflecting additional improvement during the weight stabilization/regain phase. It is interesting that we found significant intercorrelations at baseline between several parameters of pattern A, whereas for parameters of pattern B, the only significant correlations were found between adiponectin and CRP, adiponectin and HDL-C, and CRP and HDL-C (Supplementary Table 1).

We next used a nonbiased hierarchical clustering approach to analyze the dynamics of the various parameters tested at baseline and at 6 and 24 months and to further understand the extent of similarity of trends within the clusters (i.e., to quantify the distance from the weight change pattern). A hierarchical clustering dendrogram depicts the results of this analysis (Supplementary Fig. 2), revealing that leptin is the parameter whose dynamic most closely corresponded during the 24-month period of dietary intervention to that of body weight, followed by insulin, chemerin, and MCP-1 (Supplementary Fig. 2). Of interest, several parameters showed a distinct pattern that differed from the pattern of weight change, including hsCRP, HDL-C, RBP4, and the adipokines vaspin and adiponectin, whose dynamic patterns closely corresponded to each other, suggesting a cumulative improvement in the same direction (pattern B). However, our statistical approach does not allow defining whether the distance of a specific biomarker pattern from the weight change pattern is statistically significant.

The DIRECT included low-fat, Mediterranean, and low-carbohydrate dietary intervention arms in which unique dietary compositions with respect to carbohydrate, total fat, cholesterol, and monounsaturated-to-saturated fat intake were documented. We next assessed whether the dynamics in the aforementioned parameters depended on the dietary strategy (Fig. 2). Although there were (for some biomarkers) significant differences in the extent of the effects of the dietary strategies on the changes of specific biomarkers (3), the three diets similarly exhibited the distinct two patterns of dynamics. Both patterns were also similar across sex and type 2 diabetic groups (data not shown).

CONCLUSIONS

In this study, we analyzed the circulating levels of traditional biomarkers and common and more recently discovered adipokines and their dynamics during 2 years of dietary intervention, using both a nonbiased mathematical modeling approach and qualitative analysis. We found that the adipokines leptin, chemerin, and MCP-1 correspond to body weight dynamics, similar to plasma insulin and serum TG levels. In contrast, HMW adiponectin, fetuin-A, progranulin, and vaspin exhibited cumulative change despite partial weight regain, similar to the changes in hsCRP and HDL-C.

Pattern B dynamics may reflect either a delayed effect of the initial weight loss or a continuous beneficial response to switching to healthier dietary patterns. The two distinct biomarker patterns suggest different underlying biological mechanisms through which these biomarkers cluster together during different phases of weight loss and maintenance. Even the classical adipokines leptin (pattern A) and adiponectin (pattern B) seem to reflect different biological processes associated with weight loss and weight regain: Changes in circulating leptin are a function of changes in fat mass, whereas adiponectin serum concentrations may reflect changes in adipose tissue function beyond the effects on fat mass. We therefore hypothesize that biomarkers cluster in pattern A because they directly reflect body (fat) mass dynamic. This hypothesis is further supported by several intercorrelations among parameters of pattern A. In contrast, parameters in pattern B may reflect alterations in other or in addition to fat mass–related biological mechanisms, including improved chronic subclinical inflammation (as reflected by intercorrelations with CRP), adipose tissue function (as reflected by intercorrelations with adiponectin), or lipid metabolism (as reflected by intercorrelations with HDL-C) upon diet interventions.

Regardless of the yet-to-be-discovered mechanisms linking healthful diet and beneficial health effects independent of body weight dynamics, we demonstrate that 1) sustained moderate weight loss is sufficient to improve insulin and TG levels, and 2) adhering to long-term dietary intervention, even when experiencing body weight plateau or partial regain, results in clinically relevant cumulatively improved HDL-C, adiponectin, and hsCRP levels.

Our study has several limitations, including the fact that only a few female participants were studied. The statistical approach does not allow defining whether the distance of a specific biomarker pattern from the weight change pattern is statistically significant or not. We relied on self-reported dietary intake, but we validated the dietary assessment in two different dietary-assessment tools (FFQs and food diaries) and one subjective questionnaire, asking the participants about the extent of their adherence to the assigned diet, and used electronic questionnaires to minimize missing data. The unique nature of the workplace in this study, which permitted a closely monitored dietary intervention for 2 years, makes it difficult to generalize the results to other free-living conditions. However, we believe that similar strategies to maintain adherence could be applied elsewhere and should be used. Finally, in these analyses, the end points constituted changes in biomarkers, which are only surrogate markers for cardiometabolic disease risk. Yet some of these biomarkers (hsCRP, adiponectin, and HDL-C) are well-documented biomarkers of disease risk, whereas data of the clinical relevance of other markers is still accumulating. The strengths of the study include the one-phase design in which all participants started simultaneously, the long duration of the study, the relatively large study-group size, the high rate of adherence, and the comprehensive list of traditional biomarkers and novel adipokines.

Our data suggest that reduction in circulating TGs and insulin could be achieved only if weight loss, even if moderate, is maintained. TGs, as a sensitive biomarker of lifestyle, were recently reviewed in a new scientific statement of the American Heart Association (17), which concluded that lifestyle interventions (diet and exercise) are the key factors for treating hypertriglyceridemia. Weight loss is an established treatment option to improve insulin sensitivity in patients with type 2 diabetes (18). It is noteworthy that fasting plasma insulin concentration has been shown to be the best predictor of higher baseline carotid vessel wall volume in the DIRECT study population (19).

Our finding that increases in circulating HDL-C and HMW adiponectin and decreases in hsCRP levels continued in a cumulative manner throughout the entire study period may be clinically important. Although it is clear that adiponectin and CRP are strong markers for the development of chronic disease, the best strategy to favorably alter the inflammatory response is uncertain (20). We previously reported that increased HDL-C and apoA1 are associated with a significant regression in carotid vessel wall volume achieved within 2 years of the DIRECT (19).

As suggested by the hierarchical clustering analysis, leptin has the closest dynamic to weight changes, followed by insulin. Adipokines may play a role as signals from adipose tissue to the brain, directly contributing to weight regain, or at least as predictors of body weight dynamics. Although metabolic, behavioral, neuroendocrine, and autonomic responses all contribute to the maintenance of body energy stores at a defined state, part of the opposition to sustained weight loss can be attributed to leptin (21). Circulating leptin levels and leptin mRNA expression highly correlate with adipose tissue mass and, thus, can be used as surrogate parameters for changes in the amount of adipose tissue (22). We recently reported that plasma leptin reduction, combined with the degree of initial weight loss and with genetic variations in the leptin gene, constitutes a significant predictor of subsequent long-term weight regain (6). Additional adipokines signal the functional status of adipose tissue to other tissues and frequently reflect weight changes.

In our study, chemerin, MCP-1, and RBP4 serum concentrations corresponded to body weight dynamics. Our data of chemerin dynamic are in contrast to a recent study demonstrating that a decrease in circulating chemerin concentrations occurs despite weight maintenance after bariatric surgery (23). Chemerin, a proinflammatory protein, is highly expressed in the liver and adipose tissue and is associated with impaired insulin sensitivity, altered glucose, and lipid metabolism, with a dual role in inflammation and metabolism (7). MCP-1, primarily secreted by monocytes and macrophages, is a potent proinflammatory chemokine. It is synthesized by adipose tissue, elevated in obesity, and has a role in monocyte recruitment and, apparently, in the development of type 2 diabetes (8). RBP4, secreted by adipocytes and hepatocytes, correlates with the degree of obesity, insulin resistance, visceral fat accumulation, and fasting glucose (11–13). In our study, chemerin, MCP-1, and RBP4 serum concentration patterns closely followed the body weight pattern, suggesting that these adipokines reflect rather than cause changes in body weight.

In contrast, the adipokines fetuin-A, progranulin, and vaspin exhibited cumulative improvement despite partial weight regain throughout the study. Fetuin-A, a glycoprotein synthesized by hepatocytes, is associated with insulin resistance and fat accumulation in the liver in humans (10). Progranulin levels are associated with macrophage infiltration into omental adipose tissue and with markers of inflammation, hyperglycemia, and predominantly visceral fat accumulation (9). Vaspin, suggested as a visceral adipose tissue–derived factor, is associated with obesity and impaired insulin sensitivity (14). Our data on the dynamics of fetuin-A, progranulin, and vaspin during the long-term diet intervention suggest that these adipokines are directly linked to reduced food intake and/or diet intervention effects, which are independent of body weight dynamics. Because insulin sensitivity and carotid vessel wall volume (19) are improved during the entire DIRECT, fetuin-A, progranulin, and vaspin serum concentrations may provide a mechanistic link between healthful diet and improvements in these obesity-related parameters.

The two patterns of dynamics described herein were similarly observed among the three intervention arms that were proven to be distinct in dietary composition. Although our study may be underpowered to detect minor differences between the three arms, it is also plausible that factors common to the three diets are major drivers of the change in the various parameters. One obvious common factor is the degree of caloric and total food weight deficit achieved (5), which was similar between the three dietary groups despite the fact that the low-carbohydrate group was not instructed to restrict caloric intake. Yet additional dietary changes were common to the three arms, including decrease in trans fatty acids and sweetened beverages and increase in vegetables and fiber. Although we were unable to link a specific dietary component to a particular parameter, this is an important area for future research, because specific dietary factors could then be implemented in larger public health measures to combat obesity-associated morbidity, regardless of whether they are part of a more comprehensive dietary change.

In conclusion, we demonstrate for the first time that two major classes of dynamics of biomarkers can be described during long-term diet interventions. One pattern closely reflects weight change, and the other is suggestive of cumulative beneficial effects, alternatively as a delayed response to the initial weight loss or perhaps to continued healthful dieting. Thus, weight reduction is not the sole indicator of the beneficial effects of healthful dieting, which may be discernable by measuring a specific group of biomarkers despite partial weight maintenance/regain.

Supplementary Material