Plasma Levels of Triglycerides and IL-6 Are Associated With Weight Regain and Fat Mass Expansion.

CONTEXT: Long-term weight loss (WL) maintenance is the biggest challenge for overweight and obesity because of the almost unavoidable phenomenon of partial or even total weight regain (WR) after WL.
OBJECTIVE: In the present study we investigated the relations of (the changes of) adipocyte size and other risk biomarkers with WR during the follow-up of the Yoyo dietary intervention.
METHODS: In this randomized controlled study, 48 overweight/obese participants underwent a very-low-calorie diet to lose weight, followed by a weight-stable period of 4 weeks and a follow-up period of 9 months. Anthropometric measurements, adipocyte volume of abdominal subcutaneous adipose tissue, and plasma metabolic parameters (free fatty acids [FFAs], triglycerides [TGs], total cholesterol, glucose, insulin, homeostasis model assessment of insulin resistance [HOMA-IR], interleukin 6 [IL-6], angiotensin-converting enzyme [ACE] activity, retinol binding protein 4 [RBP4]) at the beginning and the end of follow-up were analyzed.
RESULTS: Our results show that changes of TGs, IL-6, HOMA-IR, and ACE are significantly positively correlated with WR. Multiple linear regression analysis shows that only TG and IL-6 changes remained significantly correlated with WR and increased body fat mass. Moreover, the change in HOMA-IR was tightly correlated with the change in TGs. Surprisingly, change in adipocyte volume during follow-up was not correlated with WR nor with other factors, but positive correlations between adipocyte volume and HOMA-IR were found at the beginning and end of the follow-up.
CONCLUSION: These results suggest that TGs and IL-6 are independently linked to WR via separate mechanisms, and that HOMA-IR and adipocyte volume may indirectly link to WR through the change of plasma TGs.

Prevention of weight regain (WR) after successful weight loss (WL) is one of the biggest challenges for individuals with overweight or obesity because of the almost unavoidable phenomenon of WR after WL. Studies have indicated that WL maintenance can improve comorbidities such as insulin resistance (IR), dyslipidemia, hypertension, and sleep apnea (1, 2), while WR can easily curtail these benefits and even induce detrimental conditions once again (3). Unfortunately, the knowledge of biomarkers of WR is still limited, which makes it difficult to understand the mechanism of WR, preventing the introduction of mitigating measures.

Adipose tissue (AT), especially the inherent adipocytes in this tissue, has recently received attention as a prominent contributor to the risk of WR. Mariman et al (4, 5) have proposed that increased stress between the extracellular matrix and shrunken adipocytes during diet-induced WL could be a driving factor for WR, and that the most feasible remedy for the adipocyte to get rid of this mechanical stress is to return to its original state by restoring fat. In line with this, studies have shown that the expression of certain stress genes and extracellular remodeling genes is linked to risk for WR (6, 7). In addition, during WL the reduction in size of adipocytes seems to alter their metabolic and inflammatory characteristics that facilitate the storage and clearance of ingested energy. As an integral part of the metabolic and inflammatory activity of the AT, this could influence WR after WL (5, 8). Furthermore, a number of plasma factors have been identified as predictors for WR (9-12). These include retinol binding protein 4 (RBP4) (9, 13), angiotensin-converting enzyme (ACE) activity (13, 14), and free fatty acids (FFAs) (13, 15). Wong and colleagues (16) reported that during calorie restriction greater improvements in insulin sensitivity (decreased fasting insulin and homeostasis model assessment for insulin resistance [HOMA-IR]) were inversely and independently correlated with WR.

The Yoyo study is a dietary WL intervention with a 9-month follow-up executed to investigate factors predicting WR. Previous analyses from this prospective study show that changes of biomarkers such as RBP4, FFAs, and ACE during diet-induced WL were independently correlated with WR, and thus could be predictors of WR in people with overweight and obesity (13). However, the relations of their changes during the follow-up period with weight have not been analyzed before. In the present study we selected these and other parameters to study their possible relation to WR during the follow-up period. It includes parameters linked to lipid storage (body fat mass, adipocyte volume, triglycerides [TGs], FFAs, and total cholesterol), glucose handling (glucose, insulin, HOMA-IR), previously identified WR predictors (FFAs, ACE, RBP4), and interleukin 6 (IL-6) as a proinflammatory marker.

Materials and Methods

Participants and Study Design

The present study is the follow-up to the previously described Yoyo study, from which the changes of the parameters during the follow-up were investigated for their relations with WR (Fig. 1). Forty-eight individuals with overweight or obesity were included in the present study. In short, 61 individuals with overweight or obesity (body mass index 27-36) were recruited by advertisement via local media. Exclusion criteria were described previously (17). Participants were randomly assigned to either a very-low-calorie diet (VLCD, rapid WL) or a low-calorie diet (LCD, slow WL) group. Participants in the VLCD group underwent a 5-week diet of 500 kcal/day by consuming 3 meal replacements/day (Modifast; Nutrition et Santé Benelux). Participants in the LCD group underwent a 12-week diet of 1250 kcal/day, which was designed by a dietitian. The aim of both regimens was an average WL of around 10% in both groups. At the end of the WL period (T2), on average 9.2 ± 2.5% of body weight was lost with no difference between the diet groups. Following WL, all participants underwent a 4-week weight-maintenance diet based on their individual energy requirements. This weight-stable period (T2-T3) was designed to investigate the effect of WL without the interfering effect of a pronounced negative energy balance. The dietitian provided dietary advice to both groups to assist in WL during the WL period and to assist in remaining weight stable throughout the weight-stable period. After the weight-stable period, each participant’s body weight was monitored monthly for 9 months (follow-up, T3-T4). During this follow-up period, participants did not receive advice on monitoring or limiting food intake to mimic nonrestricted, free-living conditions.

Figure 1.

Schematic overview of study design. Variables were obtained and measured at baseline (T1), at the end of weight loss (T2), end of weight stable (T3) and end of follow-up (T4). The dietary intervention period (T1-T3) was composed of the weight loss and weight-stable periods.

Anthropometric measurements from T1 to T4 were performed as previously described (17). Body fat mass was determined by using the Bod Pod device (Cosmed) for air-displacement plethysmography to measure body volume, and calculate body density and body composition according to the 2-compartment model, as previously reported (17). Samples including subcutaneous AT biopsies and blood samples were taken at the start of the study (T1) and at the end of each period (T2, T3, T4). Throughout the entire study, 4 participants during dietary intervention and 4 during follow-up withdrew because of personal circumstances or cancer, 2 participants started prescribed medication that could influence body weight during follow-up, and in 3 participants we could not collect enough biopsy material. These participants were excluded. Accordingly, 48 participants aged 52.27 ± 8.30 years (mean ± SD) were included in the present study for further analysis. The characteristics of the remaining 48 participants from T1 to T4 are shown in Table 1.

Table 1.

Characteristics of participants at baseline, end of weight loss, weight-stable, and follow-up periods

Items	Baseline (T1)	End of WL (T2)	End of WS (T3)	End of follow-up (T4)
Diet, LCD/VLCD	25/23
Sex, female/male	23/25
Anthropometric parameters
Weight (kg)	93.58 ± 10.04	84.76 ± 9.42a	84.48 ± 9.67b	88.82 ± 10.49a
BMI	31.05 ± 2.21	28.12 ± 2.25a	28.02 ± 2.29b	29.46 ± 2.42a
Waist circumference, cm	102.63 ± 9.86	94.64 ± 8.90a	94.92 ± 9.21b	98.38 ± 9.21a
Hip circumference, cm	110.64 ± 6.02	105.23 ± 6.15a	104.38 ± 5.88a,b	105.88 ± 7.93a
Waist-Hip ratio	0.93 ± 0.09	0.90 ± 0.09a	0.91 ± 0.08b	0.93 ± 0.09a
Body fat, %	39.13 ± 8.70	33.93 ± 10.21a	32.91 ± 10.18a,b	35.66 ± 9.53a
Fat free mass, kg	58.07 ± 13.57	56.81 ± 12.77a	57.53 ± 13.24a,b	58.12 ± 13.65a
Plasma biomarkers
TGs, µmol/L	1239.35 ± 514.40	822.92 ± 304.46a	1047.87 ± 409.38a,b	995.75 ± 398.94
FFA, µmol/L	476.48 ± 180.39	645.53 ± 234.37a	455.85 ± 229.32a	546.16 ± 196.21a
Free glycerol, µmol/L	91.44 ± 32.21	112.34 ± 37.42a	92.65 ± 36.96a	97.32 ± 35.58
Total cholesterol, mmol/L	6.01 ± 0.88	5.23 ± 0.89a	5.75 ± 0.85a,b	5.9 ± 0.86
Glucose, mmol/L	5.29 ± 0.45	5.04 ± 0.50a	5.12 ± 0.53b	4.98 ± 0.47a
Insulin, µU/mL	15.8 ± 5.94	11.25 ± 4.02a	12.18 ± 4.05b	10.82 ± 4.88
HOMA-IR	3.8 ± 1.62	2.64 ± 1.20a	2.8 ± 1.09b	2.45 ± 1.28
Adipokines
IL-6	0.58 ± 0.21	0.71 ± 0.33a	0.67 ± 0.29	0.59 ± 0.23a
ACE	46.56 ± 18.98	41.95 ± 18.80a	44.77 ± 17.91a	47.17 ± 21.50a
RBP4	27.83 ± 8.05	24.11 ± 8.59a	26.30 ± 7.54a,b	27.55 ± 7.67
Adipocyte size
Adipocyte diameter, µm	77.79 ± 5.86	69.77 ± 6.64a	74.78 ± 6.29a,b	77.12 ± 6.06a
Adipocyte volume, ×105 µm3	2.47 ± 0.57	1.78 ± 0.53a	2.19 ± 0.54a,b	2.40 ± 0.58a

All variables are described as mean ± SD.

Abbreviations: ACE, angiotensin-converting enzyme; BMI, body mass index; FFA, free fatty acids; HOMA-IR, homeostasis model assessment of insulin resistance; IL-6, interleukin 6; LCD, low-calorie diet; RBP4, retinol binding protein 4; TGs, triglycerides; VLCD, very-low-calorie diet; WL, weigh loss; WS, weight stable.

Significant change (P < .05) between this time point and the previous time point with paired t test.

Significant change (P < .05) during diet intervention (T1-T3) period.

This study was carried out according to the Declaration of Helsinki guidelines and was registered at clinicaltrials.gov (registration No. NCT01559415). All procedures involving human participants were approved by the Central Committee on Human Research and by the Medical Ethical Committee of Maastricht University, the Netherlands. Written informed consent from all participants was obtained before participation in the study.

Adipose Tissue Biopsy

Abdominal subcutaneous AT needle biopsies (approximately 1 g) were collected 6 to 8 cm laterally from the umbilicus under local anesthesia (2% lidocaine) after an overnight fast at baseline, end of WL, end of weight stable, and end of follow-up. Biopsies were immediately rinsed with sterile saline and visible blood vessels were removed with sterile tweezers. Then a small sample of AT was fixed overnight in 4% buffered paraformaldehyde and embedded in paraffin for analysis of adipocyte size.

Adipocyte Size Measurement

Histological sections (8 µm) were cut from paraffin-embedded tissue, mounted on microscope glass slides, and dried overnight in an incubator at 37 °C. The sections were stained with hematoxylin and eosin. Digital images were captured with the use of a Leica DFC320 digital camera at 20× magnification (Leica DM3000 microscope). Computerized morphometric analysis (Leica QWin V3) of individual adipocytes was performed in a blinded fashion as previously described (18). Adipocyte volume (V) was calculated from the mean of diameter (d) and SD of measurements of adipocyte diameter using the Goldrick equation: V = (19, 20). Approximately 400 adipocytes per sample were measured.

Biochemical Assays

Biochemical analysis of blood samples was performed as described previously (13). In short, blood samples were collected into EDTA and serum tubes and were centrifuged for 15 minutes at 1000g at 4 °C and 21 °C, respectively. Aliquots were immediately frozen in liquid nitrogen and subsequently stored at –80 °C until analysis. Plasma glucose, FFAs, TGs, and total cholesterol were analyzed with standard enzymatic methods (ABX Pentra 400 autoanalyzer, Horiba ABX). Plasma insulin concentrations were analyzed with commercially available radioimmunoassay (RIA) kits (Human insulin–specific RIA, Millipore Corp, Billerica, Millipore catalog No. EZHI-14K, RRID: AB_2800327). Plasma concentrations of IL-6 (Meso Scale Discovery, catalog No. K151B3S, RRID: AB_2909464), RBP4 (R&D Systems, catalog No. DRB400, RRID: AB_2813876) were determined by enzyme-linked immunosorbent assay. Serum ACE activity was measured with a standard enzymatic assay (Bühlmann, catalog No. KK-ACK).

Calculations

To analyze WR during the follow up, we calculated the Δ changes of body weight as follows: WR (kg) = body weight at T4 – body weight at T3. Similarly, during follow-up the changes in absolute fat mass were named “fat mass expansion” and calculated as follows: fat mass expansion (kg) = absolute fat mass at T4 – absolute fat mass at T3.

In parallel, the changes in variables during follow-up were described as Δ values and were calculated as follows: Δ = absolute value at T4 – absolute value at T3.

Insulin Resistance Assessment

The HOMA-IR serves as a surrogate measurement of IR (21-24). The HOMA-IR was calculated as follows: HOMA-IR = [fasting insulin (µU/mL)] × [fasting glucose (mmol/L)] ÷ 22.5 (24, 25), and was used as an index of central IR status in further analyses.

Statistical Analysis

All variables were checked for normal distribution by Shapiro-Wilk test and histogram, P greater than .05 was considered as the threshold for normal distribution. For variables with P less than .05, the Shapiro-Wilk test, histogram, and Q-Q plot were checked and outliers, if any, were removed before further analyses. Variables are presented as mean ± SD. To determine possible changes over the intervention and follow-up, a 2-tailed dependent t test was carried out. In statistical analyses, P less than .05 was considered to be significant unless otherwise stated. Statistical analyses were conducted using SPSS version 22.0 for Windows 10 (SPSS Inc).

For relationships between parameters or with WR percentage, Spearman rho was calculated. Scatterplots were extra-checked between significant variables before further analyses to double check whether a linear relationship was applicable. Subsequently, multiple linear regress analyses were performed to investigate the associations between the significant factors and WR.

Results

Participant Characteristics

Clinical characteristics of all the participants (n = 48) at baseline (T1), end of WL (T2), end of weight-stable phase (T3) and follow-up (T4) are displayed in Table 1. WL from T1 to T2 was similar in the LCD group (n = 25, –8.37 ± 2.67 kg) and the VLCD group (n = 23, –9.32 ± 1.86 kg) (P = .16). The sex ratio (female:male) of LCD and VLCD was 12:13 and 11:12, respectively. Adjusting for sex did not influence the WL results between these 2 groups nor within each group (P > .05). In addition, body weight change was comparable in the 2 groups during the weight-stable phase (T2-T3, P = .91) and follow-up (T3-T4, P = .81). Thus, the rate of dietary intervention-induced WL did not affect the subsequent WR during the 9 months in a nonrestricted. free-living environment. Participants from both diet groups were pooled in the subsequent analyses. The individual evolution of the parameters studied in the trial along all time points is shown in Supplementary Fig. S1 (26).

Anthropometric parameters, body weight, body mass index, waist and hip circumferences, waist-hip ratio, body fat (%), and fat free mass all were lowered during the dietary intervention (T1-T3) while they increased significantly during follow-up (T3-T4) (see Table 1).

Metabolic parameters plasma TGs and total cholesterol both significantly decreased during WL and then significantly increased during the weight-stable period, while FFAs and glycerol increased during WL and decreased during the weight-stable phase. During follow-up, only FFAs increased significantly, whereas the levels of TGs, glycerol, and total cholesterol remained unchanged. Fasting glucose had a persistent significant decrease during the dietary intervention and follow-up; both fasting insulin level and the insulin resistance assessed by HOMA-IR decreased during the dietary intervention and stayed unchanged during follow-up (see Table 1).

Adipocyte Volume

As we expected that adipocyte volume would be related to fat mass and as such to body weight and WR, we paid extra attention to adipocyte size. The volume of adipocytes during the whole study was recorded and the changes can be seen in Fig. 2. It was observed that the average volume of adipocytes had shrunken 28% during WL (T1-T2, P .001), rebounded 23% during the weight-stable phase (T2-T3, P < .001), and increased 10% during follow-up (T3-T4, P = .011). The average volume at the end of the dietary intervention (T3) was significantly lower (P < .001) compared to the average baseline volume, whereas at the end of follow-up (T4), adipocyte volume had already rebounded to the baseline level (P = .486). But no significant correlations were observed between the changes of adipocyte volume and the changes of body fat mass (r = –0.116, P = .439) nor between the changes of the adipocyte volume and body weight (r = 0.001, P = .994) during follow-up (T3-T4).

Figure 2.

A, Adipocyte volume at each time point. B, Scatterplots of adipocyte volume at the start of the present study (T1), end of weight loss (T2), end of weight stable (T3) and end of follow-up (T4) are shown with the mean ± SD. To determine possible changes of adipocyte volume over the intervention and follow-up periods, 2-tailed dependent t tests were carried out. *P less than .05, **P less than .01, ***P less than .001. HOMA-IR, homeostasis model assessment of insulin resistance.

Correlations Between Parameter Changes During Follow-up and Weight Regain

On average, during follow-up (T3-T4) the participants gained 4.34 ± 3.45 kg of body weight. To determine which of the aforementioned parameters were related to the WR, correlation analysis was performed between the changes of these parameters during the follow-up and the change of body weight. This showed that changes of glucose, insulin, HOMA-IR, TGs, ACE, and IL-6 during follow-up were significantly positively correlated with the changes in body weight (Fig. 3), whereas no significant correlations were observed with adipocyte volume, FFAs, total cholesterol, and RBP4 (Supplementary Table S1) (26). Next, we tested for correlations between the parameters. This showed that the change in TGs was significantly positively correlated with the change of HOMA-IR (r = 0.521, P = .001) during follow-up. No other correlations between different parameters were detected (see Supplementary Table S1) (26).

Figure 3.

Several factors during follow-up significantly correlated with weight regain. Scatter plots of the changes of parameters from T3 to T4 with body weight changes: A, glucose; B, insulin; C, homeostasis model assessment of insulin resistance (HOMA-IR); D, triglycerides (TG); E, interleukin 6 (IL6); and (F) angiotensin-converting enzyme (ACE), respectively.

Subsequently, to further investigate if and how WR was linked to the correlated parameters, multiple linear regression was performed. As HOMA-IR is calculated from fasting glucose and insulin, we imputed HOMA-IR instead of glucose and insulin. The changes of HOMA-IR, ACE, TGs, and IL-6 were set as the independent factors; weight change during follow-up was set as dependent factor. Table 2 shows the associations of WR with the significantly correlated parameters. In the unadjusted model (model 1), the changes in HOMA-IR, IL-6, and TGs were positively associated with weight regain, whereas ACE was not. When adjusted for age, sex (model 2), and additionally adjusted for diet (model 3), IL-6 and TGs were still positively associated with WR, whereas the change in HOMA-IR during follow-up reached borderline significance in model 2 (P = .05) and model 3 (P = .049). Meanwhile, in the fully adjusted model (model 3), 1 SD increase in HOMA-IR during the follow-up period was associated with an average 0.29 kg WR, 1 SD increase in IL-6 was associated with about 0.28 kg WR, and 1 SD increase in TGs was associated with 0.38 kg WR. Thus, the positive associations of TG (β = 0.376, P = .01) was a little bit stronger than both HOMA-IR (β = 0.292, P = .049) and IL-6 (β = 0.287, P = .037) with WR.

Table 2.

Linear regression analyses between changes of factors with weight regain (T3-T4)

Variables	Model 1			Model 2			Model 3
	β	95% CI	P	β	95% CI	P	β	95% CI	P
								lower
ACE	0.222	–0.017 to 0.275	.082	0.215	–0.025 to 0.275	.099	0.224	–0.022 to 0.282	.091
HOMA-IR	0.291	0.049 to 1.731	.039	0.282	–0.012 to 1.741	.053	0.292	0.004 to 1.781	.049
IL-6	0.302	0.616 to 6.519	.019	0.285	0.219 to 6.525	.037	0.287	0.211 to 6.577	.037
TGs	0.388	0.001 to 0.006	.007	0.383	0.001 to 0.006	.013	0.376	0.001 to 0.006	.015

Model 1: unadjusted. Model 2: adjusted for age and sex. Model 3: adjusted for age, sex, and diet.

Abbreviations: ACE, angiotensin-converting enzyme; HOMA-IR, homeostasis model assessment of insulin resistance; IL-6, interleukin 6; TGs, triglycerides.

Although fat-free mass (such as fluids, muscle, and bone) is involved in the development of WR as well, the most prominent part is the expansion of fat mass. To see whether the factors were also associated with the changes of fat mass, multiple-regression analysis was further performed by setting the changes in HOMA-IR, ACE, TGs, and IL-6 as independent factors, and the change of absolute fat mass as a dependent factor (Table 3). Significant positive correlations between the increase in IL-6 and TGs with fat mass were observed in the unadjusted model (model 1), the minimally adjusted model (model 2, adjusted for age and sex), and the fully adjusted model (model 3, adjusted for age, sex, and diet). In contrast, a significant association of HOMA-IR with increased body fat mass was not observed (P > .05).

Table 3.

Linear regression analyses between change of factors and change in fat mass (T3-T4)

Variables	Model 1			Model 2			Model 3
	β	95% CI	P	β	95% CI	P	β	95% CI	P
								lower
ACE	0.212	–0.025 to 0.266	.101	0.214	–0.028 to 0.271	.109	0.222	–0.026 to 0.278	.101
HOMA-IR	0.252	–0.085 to 1.59	.077	0.265	–0.082 to 1.672	.074	0.274	–0.070 to 1.711	.070
IL-6	0.382	1.470 to 7.359	.004	0.398	1.450 to 7.758	.006	0.40	1.434 to 7.812	.006
TGs	0.355	0.001 to 0.006	.015	0.328	0.000 to 0.006	.034	0.322	0.000 to 0.006	.039

Model 1: unadjusted. Model 2: adjusted for age and sex. Model 3: adjusted for age, sex, and diet.

Abbreviations: ACE, angiotensin-converting enzyme; HOMA-IR, homeostasis model assessment of insulin resistance; IL-6, interleukin 6; TGs, triglycerides.

As mentioned earlier, both the changes in TGs and IL-6 remained positively associated with WR and fat mass expansion; HOMA-IR merely kept significant with WR but not increased body fat mass. This indicates that TGs and IL-6 may contribute to the development of WR via 2 distinct mechanisms based on the expansion of fat mass.

Adipocyte Volume Is Associated With Homeostasis Model Assessment for Insulin Resistance

Although no correlations were observed between the changes in adipocyte volume and WR nor between changes in adipocyte volume and other parameters, we were still curious if there could be indirect links between adipocyte volume and WR through metabolic parameters. Therefore, we further explored the relations between the absolute values of adipocyte volume with the parameters for fatty acid and glucose handling at the beginning of follow-up (T3; Supplementary Table S2) and end of follow-up (T4; Supplementary Table S3) (26). Insulin and HOMA-IR significantly correlated with adipocyte size both at T3 (r = 0.398, P = .007; r = 0.343, P = .028, respectively) and T4 (r = 0.359, P = .018; r = 0.379, P = .017, respectively).

The plots between adipocyte volume and HOMA-IR at T3 and T4 are shown in Fig. 4. Adipocyte volume positively correlates with HOMA-IR before and after follow-up (P ≤ .05, Supplementary Table S4) (26) according to Spearman rho. To better understand the relations between volume and HOMA-IR, linear regression was performed with HOMA-IR as an independent factor. A significant association of adipocyte size with HOMA-IR was detected at T4, but not at T3 (Table 4). The positive association at T4 remained basically unchanged in the minimally adjusted model (model 2) and in the fully adjusted model (model 3, β = 0.346, P = .041). This result indicates that 1 SD higher of adipocyte volume at the end of follow-up was associated with 0.346 units higher of HOMA-IR.

Figure 4.

Correlations between adipocyte volume and homeostasis model assessment of insulin resistance (HOMA-IR) at the end of diet intervention (T3) and at the end of follow-up (T4).

Table 4.

Linear regression analyses between adipocyte size and homeostasis model assessment of insulin resistance before and after the follow-up period

Variables	Model 1		Model 2		Model 3
	β	P	β	P	β	P
Volume at T1	0.0515	.749	0.0382	.814	0.0589	.711
Volume at T3	0.2713	.086	0.2413	.152	0.2427	.139
Volume at T4	0.3477	.030	0.3276	.048	0.3467	.041

Model 1: unadjusted. Model 2: adjusted for age and sex. Model 3: adjusted for age, sex, and diet.

Discussion

The present study focused on determining the relation between WR and various parameters during the 9-month follow-up of the Yoyo project. These parameters were selected based on their involvement in the adipocyte model for WR, or how their WL-induced changes predicted WR. We report for the first time that changes in TGs and IL-6 (during the free-living state right after dietary intervention) were independently positively associated with WR in individuals with overweight or obesity, whereas HOMA-IR and ACE were not. Robust correlations between TG and HOMA-IR were observed throughout the study (P < .01). Although no significant correlations between changes of adipocyte volume and WR or HOMA-IR were observed, adipocyte volume correlated with HOMA-IR at T3 (postintervention) and T4 (end of follow-up). These results together suggest that TGs and IL-6 act on separate underlying mechanisms of WR, and that adipocyte volume and HOMA-IR are more likely linked with WR through a bridge function of TGs (Fig. 5). These results provide an important indication about WR development, and hopefully will animate scientists to look into the underlying mechanisms of WR.

Figure 5.

Relations between various factors and WR. HOMA-IR, TG, IL6, and ACE during follow-up are significantly correlated with WR, of which TG and IL6 are independently associated with WR and increased fat mass. Adipocyte volume is not significantly correlated with WR but associated with HOMA-IR. HOMA-IR positively correlates with TG. This indicates that WR is influenced by TG and IL6 via separate mechanisms involving fat mass, and that adipocyte volume and HOMA-IR have an influence on WR via plasma TG. ACE may influence WR via fat-free mass. ACE, angiotensin-converting enzyme. HOMA-IR, homeostasis model assessment for insulin resistance; IL6, interleukin 6; TG, triglycerides; WR, weight regain.

Adipocyte Size and Weight Regain

Adipocyte size is a surrogate value for the adipocyte TG content in line with the morphological changes of adipocytes (27). Surprisingly, significant correlations between changes in adipocyte volume and WR or increased body fat mass were not detected in the present study.The volume was calculated from the measured diameter (d) of the cell and the SD of the measured diameter by using the Goldrick formula of V = (19). When diameter was used instead of volume, still no significant correlations could be found. Nevertheless, an independent positive association of adipocyte volume with HOMA-IR was observed at T1, T3, and T4, and HOMA-IR significantly correlated with WR. In the present study, subcutaneous AT was analyzed, which may differ from other types of AT. Other studies have shown that both subcutaneous and visceral adipocyte size are positively correlated with whole-body IR (28-30). Even in early studies of human AT, increased fat cell size was shown to correlate with impaired whole-body metabolic regulation (31) and systemic IR (32). However, in these studies the relation was not investigated during WR. Interestingly, Bahceci et al (33) reported that enlarged adipocytes are not only less insulin responsive, but their size is also positively correlated with inflammatory factors such as nuclear factor-κB, tumor necrosis factor γ, and IL-6. In the present study, IL-6 was associated with WR, but we were unable to find a relation between IL-6 and changes or absolute values at different time points of adipocyte volume. It could be argued that adiponectin could influence the relation between (changes in) insulin sensitivity and adipocyte volume, but no significant associations were observed with plasma adiponectin levels (not shown). Potential explanations as to why no significant associations of changes in adipocyte size with the changes of any of the parameters were observed may be the following: First, the AT biopsies were obtained only from the subcutaneous AT depot, which ignored possible disproportionate changes in visceral AT as a result of WL and WR. Clinically, individuals tend to rapidly regain more visceral fat than subcutaneous fat during overnutrition, and subcutaneous fat tends to play a metabolically protective role whereas visceral fat does not. The depot-dependent difference in metabolic functions (34, 35), specifically in glucose homeostasis and inflammation, could explain our observations. Second, because HOMA-IR represents the status of only hepatic IR instead of whole-body IR status, it may explain why no positive correlation between the changes in HOMA-IR and adipocyte volume was detected during the follow-up period. Although we could not demonstrate a direct relation between adipocyte volume and WR, our observations suggest that adipocyte size may have an indirect link with WR via relations with HOMA-IR.

Triglycerides and Weight Regain

In this study, the change in plasma TGs was found to correlate positively with the development of WR and of fat mass. This has also been observed in other studies (36, 37). Changes in TGs also correlated significantly with HOMA-IR, but in a regression model with WR as the dependent factor and including TGs and HOMA-IR as independent factors, TGs were associated with WR and HOMA-IR was not. HOMA-IR is a parameter that can be used for whole-body or peripheral IR because it is strongly associated with euglycemic clamp results (r = 0.73, P < .001) (38), but in principle it is intended to estimate the IR of the liver (25, 39). After WL, whole-body insulin sensitivity usually improves as also indicated by the reduction in HOMA-IR after WL in this study. In individuals with a more pronounced IR after WL, higher production of TGs may occur from FFAs released from AT and from dietary fatty acid intake, resulting in increased plasma TG levels (40, 41). At the same time, increased levels of plasma glucose and insulin will create an optimal condition for adipocyte and ectopic TG-derived fatty acid uptake and storage, that is, insulin for promoting glucose uptake and glucose for providing intracellular glycerol-3-phosphate to allow esterification of fatty acids, all in line with increased WR. A relation between insulin sensitivity and WR has been reported before (42, 43). Here we propose that IR is indirectly involved in regaining weight by supporting TG levels in plasma and uptake by peripheral tissues.

Interleukin 6 and Weight Regain

IL-6 is known as one of the major proinflammatory cytokines involved in various biological processes. AT is a major source of circulating IL-6 (44), and a positive correlation between circulating IL-6 with adiposity has repeatedly been found (45). There seems to be a positive relation between fasting IL-6 concentration and fat mass, body fat percentage, and waist circumference (46, 47). In line with this, it has been shown that increased systemic IL-6 levels are associated with elevated fat mass, not only in rodent models, but also in obese humans (48, 49). Recently, human studies have identified IL-6 as a factor associated with or predicting WR (50, 51). Our study shows that the change in IL-6 during follow-up is positively correlated with WR and in particular with change in fat mass. Intriguingly, the association of IL-6 with change in fat mass (β = 0.40, P = .006) was even stronger than with TGs (β = 0.32, P = .039) in the unadjusted model, minimally adjusted model, and fully adjusted model. Obesity is often accompanied by low-grade inflammation (52, 53). Data from our group based on the same cohort have suggested that risk for WR is related to the ability of people to reduce the inflammatory activity of the AT during the weight maintenance period (T2-T3) (54). Persistent inflammation after WL will result in a higher WR. In this respect, higher IL-6 plasma levels may be part of the mechanism as it has been shown that IL-6 can act as a Th2 cytokine in obesity by stimulating macrophage M2 polarization and local AT macrophage proliferation (55).

Angiotensin-converting Enzyme and Weight Regain

ACE is known as a component of the renin-angiotensin-aldosterone system. It regulates blood pressure by converting inactive angiotensin I to active angiotensin II, a potent vasoconstrictor, and it can break down the vasodilator bradykinin (56, 57). ACE serves in many other processes of the body including regulation of body fluid content and electrolyte concentrations (58). A number of studies indicate that in rodents ACE modifies adipocyte growth and differentiation (59, 60). In human adipocytes angiotensin II increases lipogenesis (61), but an inhibitory effect on adipogenesis has also been reported (62). ACE could also influence tissue lipid content by modifying tissue blood flow. Genetic studies indicate a role for ACE in the risk of developing obesity (39, 40). Moreover, an increase in the plasma concentration of ACE during WL has been proposed as predictor of WR both in males and females (10, 63). On the other hand, the Yoyo study reported that WL-induced change in ACE correlated negatively with risk of WR (16). Overall, there seem to be sufficient indications to suppose involvement of ACE in the risk of (re)gaining weight. In the present study, the concentration of ACE increases significantly during follow-up. At the same time, in this study ACE was found to correlate positively with WR, although not as an independent factor. However, our data are not sufficient to reveal the underlying mechanism. It is also possible that the correlation between ACE and WR is a false-positive finding.

Limitations

The present results are based on a limited number of individuals and the results may depend on the design of the intervention. More extensive studies using similar and other parameters (eg, physical activity, anxiety, and sleeping may affect outcomes) should be undertaken to validate the present findings. In addition, our results are merely associations and do not provide evidence for a causal relationship (64). In this respect, the model of Fig. 5 remains hypothetical and causal studies are needed to consolidate it. Nevertheless, it can be used to point out what types of studies are most relevant and what parameters have to be measured.

Conclusions

The present study has demonstrated that changes in HOMA-IR, TGs, IL-6, and ACE during follow-up are significantly positively correlated with WR, of which TGs and IL-6 remain positively associated with WR and fat mass increase in a multiple regression analysis. We found no evidence for an association between changes in adipocyte volume and WR, but adipocyte volume was associated with HOMA-IR. Moreover, changes in HOMA-IR positively correlated with changes in TGs. These results suggest that TGs and IL-6 may influence WR via separate mechanisms, and that HOMA-IR and adipocyte volume are indirectly linked to WR, probably through TGs. Changes in ACE were no longer associated with WR in the multiple regression analysis. These findings are compiled in the model shown in Fig. 5, which remains hypothetical until causal studies confirm the present findings.