The prevalence of obesity in the UK is increasing, with 41 % of adult men and 33 % of women being overweight and an additional 25 % and 20 % respectively being obese(1). As the health, economic and social costs of obesity have become increasingly apparent, the number and variety of weight-loss programmes has multiplied. Numerous dieting programmes are currently available that focus on different strategies for reducing energy intake, by restricting specific macronutrients (fat or carbohydrate), controlling portion size or replacing meals with substitutes.
Obesity is one of the major modifiable risk factors for CVD. Weight loss per se results in favourable changes in blood lipid profiles, decreasing circulating LDL cholesterol (LDL-C) and TAG levels and increasing HDL cholesterol (HDL-C) if weight loss is maintained(Reference Klein, Burke, Bray, Blair, Allison, Pi-Sunyer, Hong and Eckel2). However, restriction of specific macronutrients can produce differential effects on lipid profiles. High-carbohydrate, low-fat diets, while decreasing LDL-C, have a tendency to increase plasma TAG concentrations and decrease HDL-C(Reference Roche3), whereas low-carbohydrate diets, though effective in lowering plasma TAG concentrations, can be inherently high in saturated fat and cholesterol and show variable effects on LDL-C(Reference Samaha, Iqbal, Seshadri, Chicano, Daily, McGrory, Williams, Williams, Gracely and Stern4–Reference Westman, Yancy, Edman, Tomlin and Perkins8).
Obesity is frequently associated with raised plasma TAG and a predominance of small, dense LDL particles. This type of LDL, or LDL phenotype, is also found in CHD patients and has been associated with up to a threefold increase in risk of CHD(Reference Austin, King, Vranizan and Krauss9, Reference Griffin, Freeman, Tait, Thomson, Caslake, Packhard and Shepherd10). There is evidence to suggest that the cholesterol-lowering effect of both low-fat (high-carbohydrate) and high-fat (low-carbohydrate) diets is much greater in people with a small, dense LDL phenotype(Reference Krauss and Dreon11). While few studies have been carried out on the effect of weight loss on different LDL phenotypes, low-carbohydrate, low-energy diets have also been shown to increase LDL particle size; but these studies have been of short duration and with small subject numbers(Reference Dreon, Fernstrom, Miller and Krauss12–Reference Volek, Sharman, Gomez, DiPasquale, Roti, Pumerantz and Kraemer14). Furthermore, with the notable exception of one recent study(Reference Dansinger, Gleason, Griffith, Selker and Schaefer15), there have been few direct comparisons of the efficacy of commercially available weight-loss programmes in modifying blood lipids and lipoproteins. The present multi-centred study, the largest in the UK to date, was designed to examine the efficacy of four popular weight-loss programmes, in a randomised controlled design, on a range of outcome variables related to weight loss, body composition and biochemical variables that have been reported elsewhere(Reference Boley, Herriot, DeLooy, Fox, Bonham, MacDonald, Millward, Morgan and Truby16, Reference Truby, Baic, DeLooy, Fox, Macdonald, Morgan, Taylor and Millward17). The current paper reports the effects of these diets on plasma lipid and lipoproteins as markers of CVD risk.
Methods
Weight-loss programmes
The trial involved five study groups: four weight-loss regimes that were chosen as being representative of the major different approaches to weight management in the UK and a control group. The diets were: (i) The Slim-Fast Plan (a low-fat meal replacement approach); (ii) Weight Watchers Pure Points programme (an energy-controlled low-fat healthy eating diet); (iii) Dr Atkins’ New Diet Revolution (a low-carbohydrate diet); and (iv) Rosemary Conley’s ‘Eat yourself Slim’ Diet and Fitness Plan (an energy-controlled low-fat healthy eating diet + weekly group exercise class). The control was a delayed treatment group which involved no dietary intervention until after 6 months.
Study setting, subjects and design
Three hundred overweight and obese men and women were recruited via a BBC advertising campaign at five university centres across the UK, i.e. Surrey (Guildford), Nottingham, Ulster (Coleraine), Bristol and Edinburgh (Queen Margaret University College). Each centre aimed to recruit a cohort of sixty participants, to allow twelve in each diet group plus a further twelve in a control group. Subjects were chosen from people aged 18 to 65 years who lived within 30 miles of a test centre and had a self-reported BMI between 27 and 40 kg/m2. Those who fulfilled these preliminary inclusion criteria were deemed eligible for the study after confirmation from their general practitioner, based on the following exclusion criteria: prior history of CHD, known type 1 or 2 diabetes, liver or respiratory failure, gout, lipid-lowering or anti-hypertensive medication, history of obesity with known cause (i.e. Cushing’s syndrome, hypothyroidism), previous gastric or weight-loss surgery, taking any weight-losing drugs (including Orlistat or Sibutramine), clinical depression, eating disorders, drug or alcohol abuse, any malabsorptive state (including lactose intolerance), treatment for a malignancy, pregnancy or breast-feeding. Participants gave their informed written consent to take part in the study. Ethical approval was obtained from the South East Multi-centre Research Ethics Committee.
Participants were stratified by gender (only 30 % of participants were male) and randomly allocated to any of the five groups (four diets and control). Each participant undertook the diet to which they had been assigned for 6 months. For the group-based programmes (Weight Watchers and Rosemary Conley), participants arranged to attend the most geographically convenient class and the costs of joining and attending one class per week for 6 months were reimbursed on presentation of receipts. Both parent companies of Weight Watchers (www.weightwatchers.co.uk) and Rosemary Conley (www.rosemary-conley.co.uk) signed a contract committing to the provision of standard care. For Slim-Fast, the cost of up to two meal replacements per day was reimbursed on presentation of receipts, and a copy of the Slim-Fast Support Pack was provided. The Atkins group was given a copy of Dr Atkins’ New Diet Revolution (Reference Atkins18). Control group subjects were asked to maintain their current diet and exercise pattern and were offered any of the diets for 6 months at the end of study, free of charge. All participants were able to claim reimbursement of travel costs. Diet group participants were instructed to follow the specific guidelines for each dietary programme, and every effort was made to avoid investigator management of, or interference in, participants’ food intakes.
Participants attended the test centres on a 4-weekly basis, where they were weighed in light clothing and their blood pressure and waist circumference was measured. A fasting venous blood sample was taken from all participants at baseline, and after 8 and 24 weeks for the measurement of insulin, glucose and plasma lipids. Additional blood samples were taken monthly from the Atkins group to monitor renal function (urea, electrolytes and cystatin C).
Dietary analyses
Participants were asked to complete a 7 d diet diary at baseline, 8 and 24 weeks, using a previously validated diary based on household measures. Diaries were checked for completeness and energy and macronutrient composition were calculated using nutrient analysis software (Windiets Research Version; The Robert Gordon University, Aberdeen, UK).
Biochemical analyses
Plasma glucose, TAG, HDL-C and LDL-C were measured by standard automated spectrophotometric methods (reagent kits from Randox, County Antrim, UK). The inter-assay CV were <5 % for these assays. Insulin was measured by an immunochemiluminometric assay (Molecular Light Technologies, Cardiff, UK). The inter-assay CV was <10 %. Insulin sensitivity was calculated by the homeostasis model assessment (HOMA)(Reference Matthews, Hosker, Rudenski, Naylor, Treacher and Turner19). LDL particle size was determined by iodixanol density gradient centrifugation as previously described(Reference Davies, Graham and Griffin20, Reference Griffin, Caslake, Yip, Tait, Packard and Shepherd21). Cystatin C was measured by an automated immunoturbidimetric method(Reference Newman, Thakkar, Edwards, Wilkie, White, Grubb and Price22). Blood samples from all visits (baseline, 2 and 6 months) were analysed for each individual within a single batch for all biochemical analyses.
Statistical analyses
Analyses are reported for all completing participants, defined as those with data available at each of three time points: baseline, week 8 and week 24. The final drop-out rate was 28 %. ANOVA showed no effect of test centre on either anthropometric or demographic measures, total weight loss in completing participants, or total drop-out rate. Data for all participants were therefore analysed together. Changes in metabolic variables over time were examined by repeated measures ANOVA (with ‘time’ as the within-subject factor and ‘diet group’ as the between-subject factor). Comparisons between diets for metabolic variables at 6 months were examined by analysis of covariance, with diet group and gender as between-subject factors and baseline measurements as covariates. Post hoc comparisons for differences between groups were located using the Tukey Honest Significant Difference test. Associations between weight loss, plasma lipids and insulin were tested using the Pearson’s correlation. Data are reported as means and their standard errors unless otherwise stated and P values of <0·05 were considered statistically significant. Data were analysed using the Statistica statistical software package (Statsoft, Tulsa, OK, USA).
Results
Baseline characteristics
Demographic and anthropometric measures of the groups are shown in Table 1. There were no significant differences in baseline characteristics between the groups.
M, males; F, females.
Dietary analyses
A summary of macronutrient and energy intakes at baseline, 2 and 6 months, for those subjects returning complete 7 d diaries, is shown in Table 2. The dietary data have also been partly reported elsewhere(Reference Dansinger, Gleason, Griffith, Selker and Schaefer15). The diet at baseline comprised on average 37 % fat, 42 % carbohydrate, 16 % protein and 5 % alcohol (percentage of energy), and showed no differences between groups. Predicted changes in macronutrient composition occurred over the course of the study. Thus, there was a marked reduction (66–80 %) in carbohydrate intake on the Atkins diet and reductions of between 30 and 50 % in fat intake on the other three diets. On average, mean daily energy intake fell by 30 % on the four diets compared with baseline values, but decreased much less (10 %) in the control group after 2 months of dieting.
Changes in body weight
Absolute weights at baseline, 2 and 6 months are shown in Table 3. There was significant weight loss between baseline and both 2 and 6 months in all four dieting groups (P < 0·001). However, no significant differences in weight loss between diets were observed at 6 months. There was no significant change in weight in the control group. A full examination of weight outcomes has been reported previously(Reference Truby, Baic, DeLooy, Fox, Macdonald, Morgan, Taylor and Millward17). Markers of renal function were maintained within normal limits for all Atkins participants.
LDL-C, LDL cholesterol; HDL-C, HDL cholesterol.
Mean values were significantly different from baseline: *P < 0·05, **P < 0·01.
Relationships between reported dietary intake and weight loss
Within those participants for whom food intake was successfully measured, there was no significant difference between diet groups in the fall in energy intakes (the difference between energy intake at baseline and at 2 and 6 months) or weight loss at 2 or 6 months. An analysis of the relationship between individual weight loss and the fall in energy intakes showed, for the cohort as a whole, that at 2 months weight loss was significantly related to the fall in energy intake (r = 0·40, P < 0·001, n 162) but at 6 months the weight loss was not significantly related to the fall in energy intake (r = 0·23, P = 0·115, n 47), although food intake was successfully measured only in forty-seven subjects at 6 months. Within individual dietary groups, weight loss at 2 months was related to the fall in energy intakes only for Rosemary Conley (r = 0·50, P < 0·005) and Slim-Fast (r = 0·51, P < 0·002) groups, while at 6 months there was a significant relationship only for the Slim-Fast group (r = 0·63, P = 0·049).
Changes in plasma lipids and lipoproteins within and between dietary groups
Mean values of plasma lipids and lipoproteins are shown in Table 3. Plasma LDL-C concentrations decreased significantly in the Weight Watchers, Slim-Fast and Rosemary Conley groups after 6 months compared with basal values, but were unchanged for the Atkins group. Plasma TAG concentrations decreased significantly in the Atkins, Weight Watchers and Rosemary Conley groups, but were unchanged in the Slim-Fast group. HDL-C was unchanged at 6 months compared with basal values in the Atkins groups, but decreased significantly in all other groups, including the control group.
Changes in plasma lipid and lipoprotein concentrations between 0 and 6 months are shown in Fig. 1. There was a significant effect of diet on plasma LDL-C (P = 0·0005), TAG (P < 0·021) and HDL-C (P = 0·018) after adjustment for baseline values, but no effect of gender. There was no significant effect of diet on the ratio of total cholesterol to HDL-C.
Changes in LDL particle size at 6 months between groups; association with plasma LDL cholesterol and TAG
LDL peak density decreased within all dietary groups, indicating an increase in LDL particle size (Fig. 2). There was a significant effect of diet on LDL density (P = 0·003) across groups after adjusting for baseline values but no effect of gender. The decrease in LDL peak density (increase in LDL particle size) was significantly greater than the control only in the Atkins group. While the effects of the four diets on LDL-C were unrelated to LDL particle size, the reductions in plasma TAG after 6 months were respectively three- and ninefold greater in subjects with a predominance of small, dense LDL at baseline on the Atkins (P = 0·003) and Weight Watchers (P < 0·001) diets (Fig. 3).
Changes in plasma insulin and glucose within and between dietary groups
Mean plasma insulin was decreased after 6 months in all dieting groups, compared with a small rise in insulin in the control group. There were no differences between the dieting groups but evidence of a diet–gender interaction (P = 0·045). Changes in glucose levels showed no effect for either diet or gender, although there was a trend (P = 0·053) towards a dietary effect, with the largest mean glucose fall found for the Weight Watchers and Rosemary Conley diets. Similarly, insulin sensitivity (%S), calculated by HOMA, showed no effect for either diet or gender, although there was a trend (P = 0·095) towards a dietary effect, with an increase in insulin sensitivity in the dieting groups.
Associations between changes in weight or waist circumference and changes in plasma insulin, glucose and lipids
Associations between changes in weight or waist circumference (0–6 months) and changes in plasma insulin, glucose and lipids are shown in Table 4. Weight loss was significantly associated with changes in insulin, glucose and in the lipid profile at 6 months. Changes in waist circumference showed similar associations to those for weight loss with the exception of glucose. Correlations were generally stronger for men than for women.
LDL-C, LDL cholesterol; HDL-C, HDL cholesterol.
Values are Pearson correlation coefficients (r).
Discussion
The present study was a multi-centred, randomised controlled trial of commercially available weight-loss strategies as they are likely to be followed by the public. Thus every effort was made to avoid investigator management of, or interference in, participants’ food intakes. As discussed elsewhere(Reference Truby, Baic, DeLooy, Fox, Macdonald, Morgan, Taylor and Millward17), there were no differences in diet, centre or gender in terms of the proportions of participants who completed or withdrew. The analysis of completers showed that, within each group, there was variable compliance resulting in a large range of weight loss for each of the four regimes, but all four diets tested were equally effective at promoting weight loss over 6 months.
Participants achieved approximately 70 % of their final weight reduction after 2 months, and this was reflected in significant improvements in most metabolic variables by this time. Overall, weight loss was positively correlated with improvements in blood lipids with respect to plasma TAG, insulin, glucose, LDL-C values and LDL particle size. These findings are in agreement with many other studies(Reference Dansinger, Gleason, Griffith, Selker and Schaefer15, Reference Poobalan, Aucott, Smith, Avenell, Jung, Broom and Grant23, Reference Dattilo and Kris-Etherton24) and emphasise the beneficial impact of negative energy balance and weight loss on metabolic factors related to CVD risk in overweight and obese subjects. The present study shows that this can be achieved through engagement with popular, commercially available weight-loss programmes in the UK. The similar efficacy of the four diets in terms of weight loss also afforded an opportunity to identify any effects of the macronutrient composition of the diets on CVD risk markers that were independent of weight loss per se. The carbohydrate-restricted Atkins diet was markedly different in macronutrient composition compared with the other three diets (Weight Watchers, Rosemary Conley and Slim-Fast). When macronutrient intake was expressed as a percentage of total energy intake, the Atkins diet reduced carbohydrate intake substantially after 2 months and maintained a shift in the contribution of energy from carbohydrate to total fat and protein over 6 months. However, because the fall in energy intake resulted primarily from a reduction in carbohydrate-rich foods that were not replaced by protein or fat, in absolute terms, the intake of protein, total and saturated fat was largely unchanged from baseline values(Reference Boley, Herriot, DeLooy, Fox, Bonham, MacDonald, Millward, Morgan and Truby16, Reference Truby, Baic, DeLooy, Fox, Macdonald, Morgan, Taylor and Millward17). The Weight Watchers, Rosemary Conley and Slim-Fast diets showed similar shifts in macronutrient composition, with an initial decrease in the percentage energy from fat and an increase in the percentage energy from carbohydrate after 2 months, with some participants returning to baseline values after 6 months. The energy derived from protein showed a slight rise over time and remained higher than baseline values by the end of the study. These changes in dietary intake would indicate that the subjects were adhering to the diets to which they were assigned.
The increasing popularity of low-carbohydrate diets has prompted much debate over their potential to produce adverse effects on blood lipids and lipoproteins(Reference Acheson25–Reference Volek, Sharman and Forsythe27). The evidence regarding LDL-C is equivocal. Early work demonstrated significant increases in LDL-C when obese subjects consumed a diet ‘devoid of carbohydrate containing foods’ for 4 weeks followed by a progressive intake of 5–8 g carbohydrate for the next 4 weeks(Reference Larosa, Fry, Muesing and Rosing6). Other short-term studies (1–3 months) have shown no change in LDL-C levels(Reference Sharman, Gomez, Kraemer and Volek13, Reference Volek, Sharman, Gomez, DiPasquale, Roti, Pumerantz and Kraemer14). The majority of longer-term studies have also shown no change in LDL-C after 6 months(Reference Samaha, Iqbal, Seshadri, Chicano, Daily, McGrory, Williams, Williams, Gracely and Stern4, Reference Foster, Wyatt, Hill, McGuckin, Brill, Mohammed, Szapary, Rader, Edman and Klein5, Reference Brehm, Seeley, Daniels and D’Alessio7), with one study reporting a significant decrease in LDL-C(Reference Westman, Yancy, Edman, Tomlin and Perkins8). However, HDL-C levels have generally been shown to be maintained(Reference Samaha, Iqbal, Seshadri, Chicano, Daily, McGrory, Williams, Williams, Gracely and Stern4, Reference Larosa, Fry, Muesing and Rosing6, Reference Sharman, Gomez, Kraemer and Volek13–Reference Dansinger, Gleason, Griffith, Selker and Schaefer15) or even improved(Reference Foster, Wyatt, Hill, McGuckin, Brill, Mohammed, Szapary, Rader, Edman and Klein5, Reference Brehm, Seeley, Daniels and D’Alessio7, Reference Westman, Yancy, Edman, Tomlin and Perkins8) on a low-carbohydrate diet. This finding is in contrast to the effects of low-fat, high-carbohydrate diets, which have been shown to consistently lower LDL-C but also decrease HDL-C(Reference Poppitt, Keough, Prentice, Williams, Sonnemans, Valk, Robinson and Wareham28), even when accompanied by significant weight loss(Reference Pelkman, Fishell, Maddox, Pearson, Mauger and Kris-Etherton29). There is general agreement that low-carbohydrate diets are effective in lowering plasma TAG levels, especially if accompanied by weight loss(Reference Zulet, Berkenpas and Martinez30). Such diets are more effective at lowering TAG during weight loss than are low-fat, high-carbohydrate diets(Reference Samaha, Iqbal, Seshadri, Chicano, Daily, McGrory, Williams, Williams, Gracely and Stern4, Reference Foster, Wyatt, Hill, McGuckin, Brill, Mohammed, Szapary, Rader, Edman and Klein5, Reference Sharman, Gomez, Kraemer and Volek13). While the latter diets have been shown to either increase or decrease plasma TAG in the absence or presence of active weight loss respectively(Reference Yu-Poth, Zhao, Etherton, Naglak, Jonnalagadda and Kris-Etherton31), variations in study design, length of intervention (often 8 weeks or less) and dietary composition have confounded interpretation of this effect. Moreover, few studies have included a parallel control group or intervened in free-living subjects with commercially available diets. The dietary effects on blood lipids in the present study are in broad agreement with previous literature. Significant differences were clearly apparent between the low-carbohydrate Atkins group and two of the reduced-fat diets, Rosemary Conley and Weight Watchers. The latter had more pronounced effects in lowering LDL-C, while plasma TAG was reduced more effectively in the Atkins group. HDL-C was reduced in all groups with the notable exception of Atkins, but was unaffected by any of the diets in comparison to the control. Similarly, LDL-C was unaffected by the Atkins and Slim-Fast diets, whereas plasma TAG was unaffected by the Slim-Fast and Rosemary Conley diets, relative to the control. There were non-significant changes in some lipid parameters in the control group over 6 months that may have influenced between-group comparisons. Although there was no significant weight loss over time in the control group, other aspects of lifestyle such as exercise and activity were uncontrolled and subjects may have altered their behaviour in this respect. This might be expected given that all subjects in the control group had initially intended to lose weight as a prerequisite in volunteering for the study. The limited efficacy of these commercial weight-loss programmes in improving blood lipids may also be due, in part, to the very variable weight loss observed in all four dieting groups under free-living conditions, with some individuals being extremely successful and others less so(Reference Truby, Baic, DeLooy, Fox, Macdonald, Morgan, Taylor and Millward17). Moreover, the weekly exercise class in the Rosemary Conley programme conferred no additional benefit in terms of weight loss or changes in blood lipids over a diet-alone programme of similar macronutrient composition, such as Weight Watchers.
High-fat, low-carbohydrate and low-fat, high-carbohydrate diets have been shown to produce differential effects on plasma LDL-C and TAG that depend on the initial type of LDL or LDL phenotype. High-fat, low-carbohydrate diets have been shown to increase LDL particle size and produce greater reductions in LDL-C and TAG in comparison to low-fat, high-carbohydrate diets in patients with a predominance of small, dense LDL. Likewise, the LDL-C-lowering capacity of a low-fat, high-carbohydrate diet is enhanced, and the potentially adverse TAG-raising effects of this diet limited, in individuals with a preponderance of this abnormal type of LDL(Reference Krauss and Dreon11, Reference Dreon, Fernstrom, Williams and Krauss32). In the present study, there was no evidence of an association between dietary effects on LDL-C and the predominant type of LDL. This may be explained by the surprisingly low expression of small, dense LDL in this overweight population at the outset of the study, and thus inadequate statistical power to identify such a relationship. In contrast, significantly greater reductions in plasma TAG were associated with a predominance of small, dense LDL particles in both the Atkins and Weight Watchers groups at baseline. This finding can be attributed to the close metabolic interrelationship between raised plasma TAG and the production of a more atherogenic species of small, dense LDL(Reference Austin, King, Vranizan and Krauss9, Reference Griffin, Freeman, Tait, Thomson, Caslake, Packhard and Shepherd10).
Central weight gain is an established underlying cause of lipid abnormalities associated with increased CVD risk that can be corrected effectively through weight loss. Our study has demonstrated that commercial weight-loss programmes can help people with uncomplicated obesity and that relatively modest weight loss is associated with improvement in CVD risk factors. It was, however, relatively short-term (6-months) in terms of sustainable weight loss. The importance of strategies to maintain weight loss, in addition to successful dieting strategies, cannot be overemphasised. In this context, a 12-month follow-up of these participants, reported elsewhere(Reference Truby, Baic, DeLooy, Fox, Macdonald, Morgan, Taylor and Millward17), pointed to an advantage of programmes based on group support in maintaining weight loss. While the present study showed all diets to be equal in their ability to promote weight loss, there was some disparity in their effects on blood lipids and lipoproteins. Despite concern over the possible adverse effects of low-carbohydrate diets, no detrimental effects were observed in subjects who were actively losing weight but also consuming a significantly greater proportion of their total energy as fat. Our findings also suggest that the cholesterol- and TAG-lowering efficacy of these diets could be significantly improved by the recognition and targeting of subjects with a small, dense LDL phenotype.
Acknowledgements
We wish to thank Dr Ian G. Davies for the blood analysis; Rebecca Hiscutt, Anne Herriot and Manana Stanley for the diet analysis; and Professor M.B.E. Livingstone for her assistance with the study. The authors wish to acknowledge the British Broadcasting Corporation for sponsorship of this study, and all the participants who persevered with their dietary interventions. None of the authors has any conflicts of interest.