Dietary Weight Loss and Exercise Effects on Serum Biomarkers of Angiogenesis in Overweight Postmenopausal Women: A Randomized Controlled Trial

Overweight and obesity is associated with increased risk of a variety of cancers (1), and regular physical activity is associated with reduced cancer risk (2). A largely unexplored mechanism that could link obesity with cancer risk is angiogenesis, a process where new blood vessels form from preexisting vessels allowing tissues to expand, regulated by maintaining a balance between proangiogenic and antiangiogenic factors (3). Angiogenesis plays an important role in obesity; adipose tissue is highly plastic, and requires vascularization to expand (4). Here, we investigated the proangiogenic VEGF and plasminogen activator inhibitor type-1 (PAI-1), and the antiangiogenic pigment epithelium-derived factor (PEDF), as they have been extensively studied in overweight/obesity, and have also been implicated in tumorigenesis in both animal models and in epidemiologic studies, extending our previous investigation of physical activity on these biomarkers (5).

Angiogenesis blockade is an active area of clinical and translational research (6, 7). Most of the developed drugs target VEGF and its related pathways, including bevacizumab, an mAb that inhibits VEGF-A, which has been approved for use by the FDA for treatment of a number of cancers (8). However, given the potential adverse effects of these compounds (9), they have not been proposed for the cancer prevention setting. Low-risk and low-cost methods for reducing angiogenesis, therefore, could have important public health benefits.

VEGF is a key regulator of angiogenesis and vascular permeability (10). VEGF is elevated in the obese state (11, 12), although mice overexpressing VEGF were protected against diet-included obesity and insulin resistance (13). VEGF can also promote the growth, survival, migration, and invasion of cancer cells (14). PAI-1 is a serine protease inhibitor (serpin) and is elevated in overweight and type II diabetic patients (15, 16). It promotes angiogenesis (17, 18), and is a prognostic marker for poor outcome in a number of cancers including breast (19). PEDF, an adipokine and serpin, has antiangiogenic properties and is active against a wide range of angiogenic stimuli, including VEGF (20). It has broad antitumor activity, and reduced levels of PEDF are associated with a worse prognosis in a variety of different cancers (21). However, it is also associated with the presence of the metabolic syndrome, and contributes to the development of insulin resistance in obesity (22, 23).

Excessive accumulation of adipose tissue creates a protumorigenic environment, characterized by inflammation, macrophage invasion, and increased angiogenesis (24). In this environment, tumor cells can circumvent inhibitory signals and harness these dysregulated processes to proliferate, and form new blood vessels, resulting in inappropriate growth and tissue invasion (24, 25). Further expansion of established dormant avascular tumors requires initiation of angiogenesis, or the “angiogenic switch,” allowing the tumor to transition to exponential growth (26).

Previously, we found that women who lost at least 1.85% of baseline fat mass (corresponding to median levels) in a 12-month exercise intervention experienced significant reductions in some biomarkers of angiogenesis (5), compared with sedentary controls. To extend this research, we investigated the independent and combined effects of dietary weight loss and exercise on circulating levels of VEGF, PAI-1, and PEDF, in the context of a completed 12-month randomized controlled trial (RCT), the Nutrition & Exercise for Women (NEW) trial. We randomly assigned 439 postmenopausal overweight/obese women to a reduced calorie dietary weight loss program, an aerobic exercise program, a combined dietary weight loss plus exercise program, or to a control arm. We hypothesized that women randomized to dietary weight loss, with or without exercise, would have significant decreases in VEGF and PAI-1, and increases in PEDF, compared with controls.

This study is ancillary to the NEW (www.clinicaltrials.gov NCT00470119) study, a 12-month RCT that tested the effects of caloric restriction and/or exercise on circulating sex steroid hormones in healthy overweight postmenopausal women. The study was carried out in the Fred Hutchinson Cancer Research Center (FHCRC; Seattle, WA), and performed with the approval of the FHCRC Institutional Review Board, in accordance with an assurance filed with and approved by the U.S. Department of Health and Human Services. Written informed consent was obtained from each participant.

The trial is described in detail elsewhere (27). Briefly, 439 postmenopausal, healthy overweight [body mass index (BMI) > 25 kg/m²], sedentary women, ages 50–75 years, not taking hormonal therapy, were recruited through media and mass mailings and were enrolled in the study between 2005 and 2008; 12-month follow-up for all participants was completed in 2009. Eligible participants were randomly assigned to a (i) reduced-calorie dietary modification intervention (N = 118); (ii) moderate-to-vigorous intensity aerobic exercise intervention (N = 117); (iii) combined diet and exercise intervention (N = 117); or (iv) control (no intervention; N = 87). Exclusion criteria included: >100 minutes/week of moderate physical activity; diagnosed serious medical condition(s); postmenopausal hormone use; consumption of >2 alcoholic drinks/day; current smoking; participation in another structured weight loss program; contraindication to participation (e.g., abnormal exercise tolerance test). Permuted block randomization was used to achieve a proportionally smaller control group, stratified according to BMI (≥ or <30 kg/m²) and race/ethnicity. The random assignment was generated by a computerized program, written by the study statistician, and run by the study manager to assign eligible participants to study arms. Investigators and laboratory staff were blinded to randomization arm.

The dietary intervention was a modification of the Diabetes Prevention Program and LookAHEAD lifestyle behavior-change programs with goals of 1,200 to 2,000 kcal/day, <30% daily calories from fat, 10% weight loss by 6 months, and weight maintenance thereafter. Participants had at least two individual meetings with a dietician followed by weekly group meetings for 6 months; thereafter, they attended monthly with biweekly phone/email contact. Intervention adherence was defined by percent of in-person nutrition session attendance.

The exercise intervention goal was 45 minutes of moderate-to-vigorous [≥4 metabolic equivalents (MET)] intensity exercise at a target heart rate of 70%–85% observed maximum, 5 days/week by week 7. Participants attended three facility-based supervised sessions/week and exercised 2 days/week at home. They recorded exercise mode, duration, peak heart rate, and perceived exertion at each session. Activities of ≥4 METs (28) counted toward the prescribed target.

Controls were asked not to change their diet or exercise habits.

Fasting (12 hours) venous blood samples were collected during clinic visits at baseline (prerandomization) and 12 months, when the study was completed. Participants refrained from alcohol (48 hours), vigorous exercise or NSAID use (24 hours) prior to fasting venous blood collection (50 mL) at baseline and 12 months. Blood was processed within 1 hour, and stored at −70°C.

VEGF, PEDF (serum), and PAI-1 (plasma) were assayed at the Clinical and Epidemiologic Research Laboratory, at the Department of Laboratory Medicine, Boston Children's Hospital, Boston, MA, using ELISA from R&D Systems. Duplicate pooled blood samples were included for quality assurance (QA) purposes and to assess interassay and intraassay coefficient of variation (CV). Baseline and 12-month samples from each individual were included in the same batch, and participants' samples were randomly placed across batches. Laboratory personnel were blinded with regard to subject and QA sample identity. Interassay and intraassay CVs for each assay were: VEGF 7.3% and 5.9%; PAI-1 8.2% and 7.9%; and PEDF 9.3% and 6.0%.

All study measures were obtained at baseline and 12 months by trained personnel blinded to participants' randomization status. Height and weight were measured and BMI (kg/m²) calculated. Body composition (fat mass and percent body fat) was measured by dual-energy X-ray absorptiometry (DXA) whole-body scanner (GE Lunar). Cardiorespiratory fitness (VO₂ max) was assessed using a maximal graded treadmill test according to a modified branching protocol (29).

Questionnaires collected information on demographics, medical history, dietary intake, supplement use, and physical activity patterns. Other covariates including adipokines, sex steroid hormones, biomarkers of inflammation, lipids, blood counts, and telomere length were measured as described previously (30–35).

Partial Pearson correlation coefficients were calculated between baseline biomarker measures, with Bonferroni correction for multiple testing (0.05/99 = significant at P < 0.0005).

Descriptive data are presented as geometric means [95% confidence intervals (CI)]. Mean changes in analytes from baseline to 12 months, stratified by arm, were computed. The intervention effects on these variables were examined on the basis of the assigned treatment at randomization, regardless of adherence or study retention (i.e., intent-to-treat). Mean 12-month changes in the intervention arm were compared with controls using the generalized estimating equations (GEE) modification of linear regression to account for intraindividual correlation over time. Intervention effects are presented as both absolute and relative change. Bonferroni correction adjusted for multiple comparisons (two-sided α = 0.05/3 = 0.02 for three comparisons) for the primary analysis. We used the method of last observation carried forward to deal with missing data at 12 months.

Changes in body composition and VO₂ max levels were calculated and used to stratify observed changes in analytes between arms at 12 months. Weight loss was categorized as no change/gained any weight (referent); lost <5% of baseline weight; and lost >5% of baseline weight. Participants with missing 12-month data were categorized as no change/gained weight. Changes in VO_{2 max} were calculated and categorized as tertiles (increased <3.5%; increased >3.5%–14.3%; increased >14.3%); participants with missing VO₂ max data at 12 months were classified as increased <3.5%. Percent fat loss was categorized as: no change/gaining any fat; decreasing <2.6%; decreasing 2.6%–6.4%; and decreasing >6.4%, corresponding to tertiles of change. Participants missing 12-month percent body fat, were categorized as no change/gaining fat. Fat loss, weight change, and VO_{2 max} levels in the control group were added as separate categories.

In secondary analyses, we examined changes in analytes for each intervention arm in subgroups of participants compared with controls by the above categories of weight loss, change in VO₂ max, and change in percent body fat.

All models were adjusted for age, baseline BMI (<30 kg/m², >30 kg/m²) and race/ethnicity. All statistical tests were two sided. Statistical analyses were performed using SAS software (version 8.2, SAS Institute Inc.).

At 12 months, 399 of 439 participants completed physical exams and provided a blood sample, 397 underwent a DXA scan, and 371 completed a VO₂ max test; 39 did not complete the study (Fig. 1). One participant randomized to diet + exercise was excluded from analysis due to missing baseline blood measures. At baseline, participants were on average 57.9 years, with an average BMI of 30.9 kg/m², and were predominantly non-Hispanic Whites (Table 1).

Data on intervention adherence, weight loss, and body composition changes in this trial have been reported previously (27). The mean weight change was −2.4% (P = 0.03; exercise arm), −8.5% (P < 0.001; diet), and −10.8% (P < 0.001; diet + exercise) vs. −0.8% among controls. Women in all intervention groups significantly reduced percent body fat (all P < 0.001) compared with controls.

Percent of daily calories from fat decreased in both the diet and diet + exercise arms (−6.7% and −8.0%, respectively). In both diet groups, women attended an average of 27 diet counseling sessions (86%). Women randomized to exercise participated in moderate-to-vigorous activity for a mean (SD) of 163.3 (70.6) minutes/week, whereas women randomized to diet + exercise participated for 171.5 (62.9) minutes/week. Both groups significantly increased average pedometer steps/day and VO₂ max compared with baseline.

After Bonferroni correction, both PAI-1 and PEDF correlated strongly and statistically significantly (hereon “significantly”; all P < 0.0005) with the majority of markers associated with overweight (anthropometrics), triglyceride levels, and insulin dysregulation, and negatively with adiponectin, ghrelin, and HDL cholesterol (Table 2). PEDF correlated positively and significantly with C-reactive protein (CRP) and Serum amyloid protein A (SAA). Both had significant and negative associations with sex hormone binding globulin; with red blood cell counts and hematocrit levels (PAI-1 only); and with white blood cell counts (PEDF only); all P < 0.001. VEGF did not correlate significantly with any baseline variables, with the exception of blood platelets (P < 0.0001).

Participants randomized to the diet + exercise arms had statistically significantly greater reductions in PAI-1 at 12 months compared with controls (−19.3% vs. +3.48%, respectively, P < 0.0001; Table 3). Participants randomized to the diet (−9.20%) and diet + exercise (−9.90%) arms had significantly greater reductions in PEDF compared with those randomized to the control arm (+0.18%), all P < 0.0001. Finally, participants randomized to the diet (−8.25%, P = 0.0005) and diet + exercise (−9.98%, <0.0001) arms had significantly greater reductions in VEGF compared with those randomized to the control arm (−1.21%). There were no statistically significant differences in any of the analytes in participants randomized to the exercise arm, compared with controls.

There was a statistically significant linear trend of reductions in PAI-1, PEDF, and VEGF with increasing weight loss among participants randomized to both the diet and diet + exercise arm: PAI-1, P_trend = 0.0003; P_trend < 0.0001, respectively; PEDF both P_trend < 0.0001; and VEGF both P_trend < 0.0001 (Table 4). In addition, participants who were randomized to the exercise arm had a statistically significant linear trend of reductions in PEDF with increasing levels of weight loss (P_trend = 0.002), but not in PAI-1 or VEGF.

Participants randomized to the diet + exercise arm and who increased their VO₂ max levels by any level (Table 5) had significantly greater decreases in PAI-1 compared with controls (P_trend = 0.0006). Similar trends were seen for PEDF and VEGF (both P_trend < 0.0001). There were no statistically significant changes in any analyte in participants who were randomized to the exercise arm only and who increased their VO₂ max compared with controls.

Finally, similar to weight loss, increasing quantities of percent body-fat loss were associated with a statistically significant linear trend in reductions in PAI-1 in the diet and diet + exercise arms (both P_trend < 0.0001); in PEDF [P_trend < 0.0001 (diet); P_trend = 0.009 (exercise); P_trend < 0.0001 (diet + exercise)]; and in VEGF in the diet and diet + exercise arms (both P_trend < 0.0001; Supplementary Table S1).

This study compared the effects of dietary weight loss, exercise, or their combination on circulating levels of regulators of angiogenesis, in a large sample of healthy, overweight/obese postmenopausal women. While correlation does not imply causality, the strong and statistically significant associations between PAI-1, PEDF, and the majority of the anthropometric markers and circulating adipokines, suggest that they are interlinked with other factors associated with overweight and obesity. Of interest, despite PEDFs' categorization as an antiangiogenic factor, all of the correlations were in the same direction as for PAI-1, confirming similar results from a study in 125 men (36). In contrast, VEGF levels did not correlate significantly with any variable, with the exception of platelet levels and white blood cell counts. Platelets are themselves regulators of angiogenesis; they have a proliferative effect on cancer cells both in vitro and in vivo (37) and can guide formation of early metastatic niches (38).

Women randomized to the diet and diet + exercise arms had statistically significant reductions in PEDF (−9.20% and −9.90%, respectively) and VEGF (−8.35% and −9.98%, respectively); and statistically significant reductions in PAI-1 in women randomized to the diet + exercise arm only (−19.3%). Weight loss appeared to account for the majority of these changes, with increasing quantities of weight loss and reductions in percent body fat both associated with statistically significant linear decreases in all analytes. This effect was seen only in women randomized to the diet and diet + exercise arms for PAI-1 and VEGF, but in all three intervention arms for PEDF. However, changes in VO₂ max in the exercise arm alone were not associated with statistically significant reductions in any analyte. We confirmed that despite being an antiangiogenic factor, weight loss is significantly associated with reductions in PEDF.

While some studies have also reported effects of exercise and weight loss on these angiogenic factors (39, 40), the studies have been small (41, 42), cross-sectional (43), or limited to men (44). In one study, 79 participants were randomized to a 12-week trial comparing an exercise intervention, a hypocaloric diet, and a combined arm, found that VEGF-A was nonsignificantly reduced by 10% to 22% in the weight loss groups. However, the study lacked a control arm (39). In our previous exercise trial in overweight/obese postmenopausal women, a 12-month exercise intervention produced a significantly greater reduction in PEDF levels (−3.7%), compared with control condition (+3.0%; P = 0.009), and that above median loss of body fat was associated with greater reduction in PEDF (5). The reasons for this difference are unclear—baseline levels of analytes in the current study were similar to those in our earlier investigation (444.9 pg/mL, VEGF; 11.8 μg/mL, PEDF; and 6.1 ng/mL, PAI-1; ref. 5). Dietary weight loss was not tested in that trial. A small study of 33 obese men and women also reported that weight loss was statistically significantly associated with reductions in circulating PEDF levels (36). A variety of studies have characterized the antitumorigenic effects of PEDF (21), so the reductions in PEDF in response to weight loss appear to be counterintuitive. However, its actions appear to be tissue-specific, having, for example, anti-inflammatory effects in the retina, and proinflammatory effects in adipose tissue and macrophages (21).

In our previous studies investigating the role of weight loss and exercise on circulating biomarkers, the majority of the biomarkers that were reduced (or in some cases elevated) in response to weight loss were also reduced or elevated, albeit to a lesser degree, by exercise alone (30–32, 34, 35). Our findings here indicate that exercise alone, either by intervention arm, or when stratified by changes in VO₂ max, had no effect on PEDF, PAI-1, or VEGF. With the exception of PEDF, weight loss in the exercise arm was not associated with alterations in levels of these analytes. Exercise increases both skeletal muscle mass and circulation; both processes require upregulation of angiogenesis. It is unclear why the exercise intervention or increases in VO_{2 max} in the exercise arm alone, did not affect circulating levels of VEGF, PAI-1, or PEDF. However, while there are few data on this topic in the literature, it appears that expression of angiogenic factors may be localized to muscle tissue, and may contribute little to circulating levels. A small RCT of 79 obese men and women randomized to a diet, exercise, or control arm found no statistically significant differences in circulating VEGF levels in the exercise arm at 6 months (39). Finally, the degree of weight loss experienced by participants in our study in the exercise arm (−2.4%) compared with −8.5% in the diet arm, and −10.8% in the diet + exercise arm (27), may have been insufficient to influence circulating levels of these analytes. While it is impossible to ascertain whether reductions in circulating levels of VEGF and other angiogenic factors could impact tumor-level angiogenesis, systemic reductions in these markers might conceivably be associated with a less favorable milieu for tumor growth and proliferation. It has been hypothesized that, as a consequence of metabolic syndrome, upregulation of PAI-1 expression predisposes breast cancer to more aggressive stages, partially by affecting angiogenesis (45, 46). Proangiogenic factors such as VEGF secreted by adipose stem cells have been implicated in tumor growth by promoting vascularization (47, 48). Indeed, a review suggested that a chemopreventive approach targeting both angiogenesis and inflammation in healthy individuals (termed angioprevention) may prevent tumor cell growth and progression by blocking vascularization of indolent tumors (6). Suggested interventions include metabolic regulators such as metformin, anti-inflammatory agents, and a variety of phytochemicals and their derivatives (6). However, many antiangiogenic drugs, including metformin (49), have significant side effects. Weight loss may represent a safe and effective method of improving the angiogenic profile in both patients with cancer and in healthy individuals.

Study strengths include the randomized trial design, the inclusion of dietary weight loss, exercise, and combined weight loss and exercise interventions, the excellent adherence to interventions, and high quality and valid biomarker assays. Limitations include the relatively homogenous population, which may limit generalizability, assays of only a select number of angiogenesis markers, and measurement of angiogenesis only in blood rather than in target tissue. Linear analyses may potentially incorporate observational study weaknesses, such as confounding. However, we used the GEE modification of linear regression analyses, which has been demonstrated to overcome potential observational study weaknesses such as confounding, effect modification, and correlation within individuals over time.

In conclusion, we report that weight loss reduced circulating VEGF, PEDF, and PAI-1, and suggest that weight loss in overweight or obese postmenopausal women may reduce risk for cancer in part through altering angiogenesis. Further investigations into the unexpected reductions of PEDF levels with weight loss are warranted.

No potential conflicts of interest were disclosed.

C. Duggan and A. McTiernan had full access to all data and take responsibility for its integrity and accuracy of the data analysis.

Conception and design: C. Duggan, A. McTiernan

Development of methodology: A. McTiernan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Duggan, A. McTiernan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Duggan, J. de Dieu Tapsoba, C.-Y. Wang

Writing, review, and/or revision of the manuscript: C. Duggan, J. de Dieu Tapsoba, C.-Y. Wang, A. McTiernan

Study supervision: A. McTiernan

This work was supported by grants from the NCI at the NIH: R01 CA105204-01A1 and U54-CA116847, and from the Breast Cancer Research Foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.