点击显示 收起
1 From the Departments of Preventive Medicine (AHW, C-CT, DOS, and MCP) and Obstetrics and Gynecology (FZS) and the General Clinical Research Center (Los Angeles County Medical Center) (CM), University of Southern California, Keck School of Medicine, Los Angeles, CA; Food Science and Human Nutrition, Iowa State University, Ames, IA (SH and PM); and the Department of Obstetrics and Gynecology, Chulalongkorn University, Bangkok, Thailand (SC)
2 Supported by the Susan G Komen Breast Cancer Foundation (BASIC99-00328), the Whittier Foundation, a USC Cancer Center Core Support grant (2P30 CA14089-26), and a NIEHS grant (P30 ES07048). 3 Reprints not available. Address correspondence to AH Wu, University of Southern California/Norris Comprehensive Cancer Center, 1441 Eastlake Avenue, MC 9175, Los Angeles, CA 90089–9175. E-mail: annawu{at}usc.edu.
ABSTRACT
Background: Low intake of dietary fat and high intake of soy foods have been suggested to partly explain the lower breast cancer rates in Asia, perhaps because of lower endogenous estrogens.
Objective: The objective was to assess the hormonal and nonhormonal effects of diets resembling an Asian diet in terms of total fat and soy food contents.
Design: Fifty-seven postmenopausal women participated in a randomized, controlled, dietary intervention study. The subjects consumed a very-low-fat diet (VLFD; 11% of energy as fat), a Step I diet (25% of energy as fat) supplemented with soy food (SFD; 50 mg isoflavones/d), or a control Step I diet (CD; 27% of energy as fat) with no soy food. All diets were prepared at the General Clinical Research Center of the University of Southern California. Serum hormones and other markers were measured at baseline and every 2 wk during the 8 wk of intervention.
Results: There were no significant differences in total estradiol and sex hormone binding globulin at the completion of the intervention between women in the SFD and VLFD groups and those in the CD group. Serum insulin decreased significantly in the SFD group, and leptin decreased significantly in the SFD and VLFD groups; however, these changes did not differ significantly from the changes in the CD group.
Conclusions: This study does not provide evidence that ingestion of soy food or a VLFD significantly reduces estrogen concentrations in postmenopausal women. However, short-term changes in diet may have significant and beneficial effects on blood insulin and leptin concentrations.
Key Words: Controlled randomized trial • low-fat diet • soy food • blood biomarkers
INTRODUCTION
The role of diet in the etiology of breast cancer remains controversial in observational epidemiologic studies. Pooled analyses of cohort studies conducted in Western countries, which used food-frequency questionnaires, showed no relation between fat intake and breast cancer risk (1). However, in the Norfolk component of the European Prospective Investigation of Cancer study, which used a 7-d food diary, risk of breast cancer increased significantly with increasing intakes of total and saturated fat (2). These differences in cohort findings emphasize the importance of measurement error in dietary assessment (3). The role of soy in the etiology of breast cancer is also inconclusive. Breast cancer risk was unrelated to soy intake in studies conducted in Western populations, where intake was low (median daily intake <1 mg soy isoflavones) (4, 5). Although breast cancer risk was significantly inversely associated with soy intake in several studies conducted in Asian populations with substantial soy intakes, and this finding remained after adjustment for dietary and nondietary risk factors (6, 7), the possibility of residual confounding cannot be ruled out for certain. Because of findings of stimulatory effects of dietary genistein (a main isoflavone in soybeans) in human breast (8) and breast cancer cell growth in MCF-7 cells and in an ovariectomized athymic mice model (9), there are also concerns regarding the safety of soy intake, particularly in postmenopausal women.
There is compelling evidence that estrogen concentrations are a critical determinant of breast cancer risk (10). Dietary fat reduction (11) and soy supplementation intervention studies (12) have been conducted with the rationale that a reduction in endogenous estrogen concentrations in short-term settings will lend support to a role of dietary fat or soy on breast cancer risk. However, the quality of these intervention studies, particularly the dietary fat reduction studies, has been questioned (13). To further examine the influence of dietary fat reduction and soy food supplementation on circulating hormone concentrations in postmenopausal women, we conducted a randomized, controlled, dietary intervention study. We investigated the separate effects on hormonal endpoints of a soy food–supplemented Step I diet (30% of energy as fat, 50% of energy as carbohydrate, 20% of energy as protein and designed to provide 50 mg isoflavones/d; SFD) and a very-low-fat, high-carbohydrate diet (designed to provide 12% of energy as fat, 68% of energy as carbohydrate, 20% of energy as protein, and no soy food; VLFD) compared with a control Step I diet (30% of energy as fat, 50% of energy as carbohydrate, 20% of energy as protein, and no soy food; CD) in free-living postmenopausal women.
SUBJECTS AND METHODS
Subjects
Postmenopausal women were recruited between July 2000 and August 2002 through flyers and newsletters that were distributed on campus and at the University of Southern California (USC) Health Fair, public service announcements on local radio stations, and a one-time advertisement on a local television station. To be eligible for inclusion, subjects had to be postmenopausal (1 y since the last menstrual period), 50 y of age, and noncurrent users of menopausal hormone therapy (ie, stopped use 6 mo before entering study). Women were excluded if they were consuming a special diet (eg, low-fat, high-fiber) or had a history of cancer (other than nonmelanoma skin cancer), diabetes mellitus, or other chronic disease.
A total of 274 women completed an initial telephone screening; 130 were eligible and 70 completed the baseline assessment and consented to be randomly assigned to 1 of 3 dietary arms (see baseline assessment below). Six women withdrew within the first 2 wk of entering the study because they did not wish to adhere to our diet protocol. We excluded another 7 women (1 in the CD group, 5 in the SFD group, and 1 in the VLFD group) because their baseline circulating estrogen concentrations suggested that they were not postmenopausal. The final analysis included 57 women (20 in the CD group, 17 in the SFD group, and 20 in the VLFD group) who represented the racial-ethnic diversity of the study area (17 white, 22 Hispanics, 11 African Americans, and 7 Asians). The study protocol was approved by the USC Institutional Review Board. Written informed consent was obtained from all study participants.
Diets
All the diets in this study were prepared in the Bionutrition Department's Research Kitchen at the General Clinical Research Center (GCRC) at the Los Angeles County USC Medical Center. Eight-day cycle menus for each of the 3 diets were developed by the nutritionist at the GCRC (see below).
During the 8 wk of study, participants received all foods to be consumed from the research kitchen at the GCRC. Each daily menu included foods for breakfast, lunch, dinner, and a morning and an evening snack. The food supply for each week was packed and stored in insulated coolers that the participants took home. Written instructions regarding food safety and reheating methods were provided. Participants were given a daily log to check off all of the foods on the menu that they consumed and to record any extra foods eaten that were not on the menu. They were also asked to return to the kitchen all uneaten food or leftovers in their original containers, which were weighed and deducted from their daily nutrient intake.
The control diet followed the American Dietetic Association guidelines for a healthy balanced diet, including 50% of energy from carbohydrates, 30% from fat, and 20% from protein (14). The VLFD, which provided 12% of energy from fat, was developed by using mostly legumes as a source of protein along with fish, lean chicken, and low-fat dairy products. The SFD, which provided 50 mg isoflavones/d (15 g soy protein), was developed by using some modified recipes provided by Mori-Nu tofu (Morinaga Nutritional Foods, Torrance, CA) and recipes developed in the GCRC Research Kitchen so that the tofu was disguised in soups, smoothies, sauces, and other dishes. The nutrient contents of the diets were evaluated by using nutritional software (Pronutra version 2.0.1; Viocare, Princeton, NJ, formerly of Princeton Multimedia Technologies). Subjects were assigned to treatment groups at the time of the preentry visit and were blinded to the diet to which they were randomly assigned, and this was largely successful (see below). The isoflavone content of the SFD was monitored throughout the study (see below).
Interested subjects who met the eligibility criteria were mailed a 3-d food record with instructions to record their food intake on 2 typical days of the week and on 1 weekend day. Because the objective of the study was for participants to maintain their body weight during the intervention period, each participant's caloric requirements were calculated by using the Harris Benedict equation [655 + 4.3 x weight (lb) + 4.3 x height (in) – 4.7 x age (y)]. The value obtained with this equation was multiplied by an activity factor of 1.3 for ambulatory sedentary participants (15). Participants were weighed every 2 wk. If they had lost or gained >2 kg from their initial baseline weight, they were placed on a higher or lower caloric diet as needed. During the 8 wk of intervention, daily nutrient intake was computed based on the daily logs of foods eaten, ie, food provided by the metabolic kitchen, minus foods that were returned, and plus any extra foods consumed. All food records were analyzed by using Pronutra version 2.0.1.
Baseline assessment and data and sample collection
At the initial screening interview, we administered a baseline questionnaire that asked about menstrual, reproductive, and menopausal factors. Body weight, blood pressure measurements, and blood samples were obtained at baseline and after 2, 4, 6, and 8 wk of the intervention. Blood specimens were collected between 0600 and 1100 after the subjects had fasted for 12 h. Serum and plasma were separated by centrifugation (2500 x g, 15 min, 4 °C). On the day of the blood draw, subjects were asked to collect an overnight urine specimen into plastic bottles that contained 1 g ascorbic acid. Urine specimens were separated into 100-mL aliquots and stored at –20 °C.
Daily food records were used as a measure of compliance. In addition, urinary isoflavone concentrations were determined at baseline, at least once during the intervention, and at the completion of the study. Urinary isoflavone concentrations were used as a measure of compliance with the SFD. Although an objective measure of compliance for the CD or the VLFD is not available, we used reductions in HDL cholesterol and increases in triacylglycerol concentrations to confirm a decrease in fat intake and an increase in carbohydrate intake, as was done in other dietary fat reduction trials (16).
Blood hormone, lipid, insulin-like growth factor I, and insulin-like growth factor binding protein 3 and other analyses
Estradiol, estrone, testosterone, and androstenedione were quantified in serum by previously described radioimmunoassay (RIA) methods (17, 18). The intra- and interassay CVs were between 7% and 16% for estradiol, estrone, and androgens. Free testosterone and free estradiol were calculated on the basis of measured total testosterone and total estradiol concentrations, respectively; sex hormone binding globulin (SHBG) concentrations; and an assumed constant for albumin (19). This method has been found to have high validity (r = 0.97) compared with direct measurements (20). Other analytes were measured in serum by highly specific direct immunoassays. SHBG was measured by chemiluminescent immunometric assay on the Immulite analyzer (Diagnostic Products Corporation, Inglewood, CA). The SHBG intraassay and interassay CVs were 7% and 10%, respectively. A two-site chemiluminescent immunoassay was used to measure insulin-like growth factor I (IGF-I) on the Nichols Advantage Specialty System (Quest Diagnostics Nichols Institute, San Juan Capistrano, CA). The IGF-1 intraassay and interassay CVs were 2% and 10%, respectively. A competitive RIA that used kits obtained from Quest Diagnostics Nichols Institute was used to measure insulin-like growth factor binding protein 3 (IGFBP-3); the intraassay and interassay CVs were 7% and 12%, respectively. Insulin concentrations were measured by chemiluminescent immunometric assay on the Immulite analyzer (Diagnostic Products Corporation). The intraassay and interassay CVs were 6% and 8%, respectively. Leptin was quantified by RIA with kits from Linco Research Inc (St Charles, MO); the intraassay and interassay CVs were 8% and 6%, respectively.
Serum concentrations of lipids and lipoproteins were determined on the Vitros analyzer by using the following methods: total cholesterol was quantified colorimetrically after hydrolysis of cholesterol esters and subsequent oxidation of the free cholesterol. HDL cholesterol was measured colorimetrically after precipitation of LDL cholesterol and VLDL cholesterol by using dextran sulfate and magnesium chloride. Triacylglycerols were also quantified colorimetrically, and LDL cholesterol was calculated as follows:
RESULTS
Women in the 3 diet groups did not differ significantly in age, body weight, or daily intake of calories, macronutrients, or fiber at baseline (Table 1). During the 8 wk of study, there was a small but significant reduction in body weight in all 3 dietary arms: –1.4 kg in the CD group (P = 0.001), –2.0 kg in the SFD group (P < 0.0001), and –2.1 kg in the VLFD group (P < 0.0001). This reduction did not differ significantly between the 3 diet groups. During the study period, intake of fat, carbohydrates, and protein as a percentage of energy and intake of fiber differed significantly between the 3 diet groups (Table 1).
View this table:
TABLE 1. Characteristics of the study population at baseline and on average during the 8 wk of intervention1
Baseline urinary isoflavone concentrations were low (1.2–4.0 µmol/d) and not different significantly between subjects in the 3 diet arms. During the intervention, total urinary isoflavone increased significantly to 32.4 (95% CI: 18.4, 56.9) µmol/d among subjects in the SFD group but remained low among women in the VLFD (1.3 µmol/d) and the CD (1.0 µmol/d) groups. In the SFD group, increases were observed for each specific isoflavone; concentrations of genistein, daidzein, and glycitein increased 14-, 9-, and 6-fold, respectively, during the intervention compared with baseline values (data not shown).
Baseline blood concentrations of cholesterol (total, HDL, and LDL cholesterol) and triacylglycerols were not significantly different in the 3 diet groups (Table 2). At the end of 8 wk of intervention, total cholesterol decreased significantly in all 3 groups, HDL cholesterol decreased significantly only in the VLFD and SFD groups, and LDL cholesterol decreased significantly only in the VLFD group. Triacylglycerol concentrations decreased in the SFD and CD groups but increased in the VLFD group; none of these changes were statistically significant and they did not differ significantly between the 3 diet groups (Table 2). These changes (baseline compared with week 8) and differences between groups were not statistically significant after adjustment for changes in body weight (data not shown).
View this table:
TABLE 2. Geometric mean concentrations of blood lipids at baseline and at the completion of the intervention
There were no significant differences in any of the sex-steroid hormones at baseline between the 3 diet groups (Table 3). In association with the intervention, changes in serum estradiol, serum estrone, and the ratio of estradiol to estrone were not statistically significant, and SHBG concentrations increased nonsignificantly in all 3 diet groups. Changes in estrogen and SHBG concentrations also did not differ significantly between the 3 groups. Androgen concentrations increased nonsignificantly in the SFD group and decreased nonsignificantly in the VLFD and CD groups. Although total and free testosterone concentrations decreased significantly in the CD group, these changes in androgen concentrations did not differ significantly between the 3 groups. None of these results were significantly affected by additional adjustment for changes in body weight (data not shown).
View this table:
TABLE 3. Geometric mean serum hormone concentrations at baseline and at the completion of the intervention
Baseline concentrations of IGF-1 (P = 0.048) and IGFBP-3 (P = 0.047) were significantly higher in the SFD compared with the CD and the VLFD groups (Table 4). In association with the intervention, IGF-1 concentrations increased in the 3 dietary arms and was statistically significant in the VLFD group. Concentrations of IGFBP-3 did not change significantly in any of the 3 dietary arms. Baseline insulin and leptin concentrations did not differ significantly among the 3 groups. Insulin concentrations decreased in all 3 groups, but significantly only in the SFD (–32.7%) and the CD (–20.8%). Leptin concentrations decreased significantly in all 3 groups, from 32.7% in the VLFD group to 47.2% in the CD group. Changes in IGF-I, IGFBP-3, insulin, and leptin concentrations did not differ significantly between the 3 groups. These patterns of changes were not significantly altered by further adjustment for changes in body weight (data not shown).
View this table:
TABLE 4. Geometric mean concentrations of serum growth factors, insulin, and leptin at baseline and at the completion of the intervention1
The statistical significance or nonsignificance obtained by using the random-effects model, which used all time points, was not unlike those that used the baseline and week 8 data (see the baseline and percentage change data for P2 and P3 in Tables 2 and 3 and P3 and P4 in Table 4). We observed no significant treatment x time interaction effects.
DISCUSSION
The results of this 8-wk intervention study indicate that a VLFD and an SFD produced very small reductions in serum estrogen concentrations in postmenopausal women; these changes did not differ significantly from those observed in women after a CD. These results were unexpected because the fat intake in the VLFD was very low (11.3% of total energy; Table 1), akin to fat intakes in Japan during the 1950s (23). The amount of soy isoflavones added (50 mg isoflavones or 15 g soy protein in the form of tofu) in the SFD was not dissimilar to soy intakes in Japan (7). The fat intake in the CD and SFD groups did not differ significantly from the fat intake in a Step I diet, which has <30% of energy as fat and <10% of energy as saturated fat (24).
Our findings on the hormonal effects of soy food are not unlike the results of a 3-mo soy intervention study in postmenopausal women, which found minimal changes in concentrations of gonadotropin and SHBG (25). In a meta-analysis of 6 soy-intervention studies in postmenopausal women (26-31), we calculated that circulating estrogen concentrations decreased 9.7% in the soy group compared with 5.8% in the control group (4 of these studies had a control group). Although a high soy intake was suggested to decrease urinary genotoxic estrogen metabolites in a 3-mo intervention study in postmenopausal women (32), no association was reported in a 6-mo soy-intervention study (30). We did not measure urinary concentrations of estrogen metabolites in this study. Duncan et al (33) proposed that the ability to produce equol after ingesting soy may be an important determinant of hormonal response to soy intake. However, urinary equol concentrations were found in only 3 of the 17 women in the SFD arm, which limited our ability to conduct meaningful analysis separately for equol producers and nonproducers in this study. Whereas soy does not appear to have any major effects on blood estrogen concentrations, at least in short-term settings, these observations do not necessarily contradict epidemiologic observations of an inverse association between soy food and breast cancer risk because longer-term eating habits of soy are likely to have been captured in these studies (6, 7).
The slight reduction in serum estrogen concentrations in the VLFD group of this study differs from our meta-analysis of 13 studies on dietary fat reduction in which serum estradiol concentrations decreased in premenopausal (–7.4%) (based on 9 studies) and postmenopausal (–23.0%) women (based on 4 studies) (11). In the 4 previous intervention studies in postmenopausal women, the range of fat intake was 10% (34) to 24% of energy (35), and weight loss ranged from 0 (36) to 3.5 kg (34) in 1 mo. The intervention period in our study was short but not unlike that of previous studies; 3-5 mo in one study (37) and 3 wk to 2 mo in the other 3 studies (34-36). An important difference between the present study and the previous studies we reviewed (11) was that dietary intake in the intervention and control groups was tightly controlled during the 8 wk of this study. Recent results from 2 longer-term dietary fat reduction studies in premenopausal women (38, 39) suggest that the hormonal effects of a low-fat diet may be weak or nonexistent.
Serum IGF-I concentrations increased significantly in the VLFD group, but this change did not differ significantly from that in the CD group. Circulating IGF-I and IGFBP-3 concentrations have not been consistently associated with intake of dietary fat, carbohydrates, and soy in cross-sectional studies conducted in Western (40-42) and eastern populations (17, 43). Insulin concentrations decreased in all 3 diet groups, significantly in the SFD and the CD groups, but these reductions did not differ significantly between the diet groups. Significant reductions in insulin concentration in association with soy supplementation were reported in women with type 2 diabetes (44) but not in soy-supplementation studies of nondiabetic women (27, 30). Whereas leptin concentrations decreased significantly in all 3 diet groups, these changes did not differ significantly between the 3 diet groups. Reduction in leptin concentrations remained after adjustment for changes in body weight during the intervention. Most studies published suggest that fasting and refeeding may change blood leptin concentrations, but less is known about the effects of specific nutrients on leptin concentrations (45). No effect of soy on leptin concentrations was reported in a previous soy-intervention study (46). It is of interest that in a large cross-sectional study in the United States, leptin concentrations increased significantly in a stepwise manner with increasing body mass index (47). Mean leptin concentrations were 26.0 ng/mL in women with a body mass index (kg/m2) between 30 and <35 and were lower (19.8 mg/mL) in women with a body mass index between 27.5 and <30. Thus, one interpretation of the current results is that even moderate changes in diet (ie, Step 1 control diet) may have beneficial effects, ie, in terms of changes in leptin and insulin concentrations.
Did noncompliance play any role in these findings? We examined adherence based on urinary isoflavone concentrations in the SFD group and serum lipid changes in all 3 arms. Compliance in the SFD group appeared good because urinary isoflavone concentrations, an accepted specific biomarker of soy intake, increased significantly during the 8 wk of intervention (from 4.0 to 32.4 µmol/d), whereas no significant changes were observed between subjects in the other dietary arms. Lipid changes in the VLFD and CD groups were not significantly different from published results. Specifically, women in the VLFD group showed significant decreases (–10.8% to –17.3%) in total cholesterol, LDL cholesterol, and HDL cholesterol and a nonsignificant increase (5.1%) in triacylglycerol concentrations. Lipid changes of these magnitudes were reported in other low-fat (10–18% of energy as fat), short-term (3–6 wk) intervention studies (48, 49), although larger reductions (20%) in total cholesterol and LDL-cholesterol concentrations have been reported in some studies with more substantial weight loss (34). Decreases in HDL cholesterol and increases in triacylglycerol in association with reductions in dietary fat and increases in carbohydrate intake are now well-established and have been used as biomarkers of such dietary changes (16). Reductions in total cholesterol, LDL-cholesterol, HDL-cholesterol, and triacylglycerol concentrations (–6.7% to –8.9%) in the CD group were not significantly different from changes reported with other Step I diets (24). In the SFD group, total cholesterol and HDL cholesterol decreased significantly, but there was no reduction in LDL cholesterol. Note that, at baseline, the study population essentially had total and LDL-cholesterol concentrations within the high end of the normal range. Thus, reductions in these lipid concentrations in the soy-intervention group would not be necessarily expected. Moreover, the amount of soy we added (15 g soy protein) was lower than the 25 g approved by the US Food and Drug Administration for its cholesterol-lowering action and substantially lower than the amount of soy used in most other clinical trials (50). In another trial that used 20 g soy protein/d, significantly decreased non-HDL-cholesterol concentrations were observed after 6 wk but not after 3 wk of the study, but subjects in this study were men who were moderately hypercholesterolemic at baseline (51).
Finally, several strengths and limitations of this study should be mentioned. An important strength is that all the meals were prepared in the metabolic kitchen of the USC GCRC, and all foods and beverages were provided to participants in containers appropriate for microwaving (if applicable). Subjects made weekly visits to the GCRC to pick up the meals for the following week and to return uneaten foods, and they donated blood specimens every 2 wk for biomarker measurements. Participants were blinded to the diet regimen to which they were randomly assigned. When we asked subjects at the completion of the study to speculate about the diet regimen to which they were assigned, 25% in the CD group, 36% in the SFD group, and 70% in the VLFD group correctly identified their diet. Thus, only the VLFD group identified their diet more accurately than at random.
There were also several limitations in this study. The baseline measurement of blood hormone, lipid, IGF, insulin, and leptin concentrations was based on a single sample collection. Despite our close monitoring of subjects during the 8 wk of dietary intervention, subjects in all 3 dietary arms experienced a small but statistically significant weight loss (1.4-2.1 kg). We have to assume that our assessment of dietary intake at baseline (based on 3-d food records) was an underestimate of the actual intake because all 3 groups showed significant weight loss despite lower reported caloric intakes at baseline than during the intervention period. During the intervention, fat intake in the CD and SFD groups was 26.2% and 24.7% of energy, respectively—lower than the amount (30% of energy as fat) we had planned. On the basis of the daily diet records and uneaten foods that were returned, many of the participants unexpectedly skipped items such as salad dressing, mayonnaise, and margarine, which were included for their addition to foods. Nevertheless, fat intake in the VLFD group (11.3% of energy as fat) was still 15% lower than that in the CD group (26.2% of energy as fat), which suggests that we should have had reasonable power to detect significant changes in lipid between these 2 diet groups. This would not have influenced our findings in relation to estrogen and SHBG because the magnitude of changes did not differ significantly between the 3 dietary arms. The intervention was only 8 wk long and, thus, the longer-term effects of these dietary changes are not known. Finally, although we may have captured the very-low-fat and soy food contents in our VLFD and SFD groups, our intervention diets may not have reflected the traditional Asian diets in terms of sources of fat, protein, and carbohydrates.
In summary, the present study does not provide evidence that short-term ingestion of soy food or a VLFD, resembling intakes in a traditional Asian diet, significantly reduces estrogen concentrations in postmenopausal women. In all 3 diet groups, insulin and leptin concentrations decreased. Results on the effects of a low-fat intervention diet in the Women's Health Initiative will be extremely informative in elucidating the longer-term effects of dietary fat reduction in postmenopausal women.
ACKNOWLEDGMENTS
We are extremely grateful to all the study participants and the data collection team at the General Clinical Research Center, including Yolanda Stewart, Carla Flores, and others, and in the Department of Preventive Medicine, including Rachel Waasdorp, June Yashiki, and Mei-Ying Lai.
AHW contributed to the conception and design. AHW and MCP obtaining funding. CM and AHW collected data. FZS, SH, PM, and SC measured the outcomes of interest. C-CT managed the data. C-CT, DOS, MCP, and AHW conducted the statistical analysis. AHW, FZS, and MCP interpreted the study results. AHW prepared the results. FZS, CM, C-CT, SH, PM, SC, DOS, and MCP reviewed the manuscript. The authors had no conflicts of interest.
REFERENCES