Efficacy of an intermittent energy restriction diet in a primary care setting

Patients with obesity (BMI > 30 kg/m²) aged 18–65 years who had been referred to RIO by their general practitioner were recruited into the study and description of the referral criteria can be found elsewhere [7]. The study obtained a favourable opinion from the South Yorkshire NHS Ethics Committee (ref: 14-YH-0018) and was conducted in accordance with the guidelines laid down in the Declaration of Helsinki. All participants provided written informed consent. The trial was prospectively registered in ISRCTN (ISRCTN31465600).

The present study was a parallel-armed cohort study comparing IER to CER. The weight loss intervention was 6 months, with follow-up at 1 year (i.e. 6 months following cessation of the intervention) for those participants still engaged with the practice. Participants were not randomised, but instead self-selected their dietary intervention group to promote patient autonomy and facilitate long-term successful lifestyle change. Dietary advice was provided by specialist obesity nurses.

Measurements were taken before participants started their diet and then serially over the course of the study by either the specialist nurse or healthcare assistant. These measurements included weight, total body fat, fat-free mass (FFM), waist circumference, systolic and diastolic blood pressure and an overnight fasted blood sample. Body composition was assessed by bioimpedance (TANITA MC-180MA; Tanita Corp, Tokyo, Japan). Waist circumference was measured at the midpoint between the lower margin of the lowest palpable rib and top of the iliac crest at the end of normal expiration. Blood pressure was measured using an automated sphygmanometer following a 5-min rest (7670-16767; Welch Allyn, Skaneateles Falls, NY, USA).

Biochemical analyses (with the exception of insulin) were conducted at two accredited hospital laboratories (Rotherham District General Laboratory; Royal Surrey County Hospital Pathology Partnership, UK) by personnel blinded to group assignment. Fasting insulin was measured in batch upon study completion via radioimmunoassay using a commercially available kit (Merck Millipore, MA, USA; inter/intra assay CVs < 10%) by a study investigator (RA) who was blinded to group assignment at the time of the analyses.

Participants completed validated 7-day diet diaries at 3 time points: baseline (prior to commencing diet), then months 3 and 6 whilst on the diet. Diaries included pictorial guides to aid portion size estimations. All dietary analyses were carried out in diet plan 7 (Forestfield Software, Horsham, UK) using the McCance and Widdowson’s composition of foods integrated dataset. For the purpose of this study, a compliant ‘fast’ day was defined as one where energy intake was ≤ 3347 kJ, which corresponds to the very low energy threshold defined by the NICE Obesity Guidelines [9].

Data were statistically analysed using SPSS v23 (IBM, Chicago, USA). Data were first checked for normality using the Shapiro–Wilk test, with non-normally distributed data normalised via log transformation if required to permit parametric testing. The primary analysis was a completer-only analysis, owing to the high attrition rates exhibited by both groups. The alternative approach is intention to treat (ITT) analyses which include every subject who is randomized, ignoring noncompliance, protocol deviations and withdrawal which thus maintains prognostic balance generated from the original random treatment allocation. However, in the present study complete outcome data were not available for all randomized participants, i.e. study measurements were not taken for participants who dropped out. Hence, assumptions must be made, e.g. using last observation carried forward (LOCF) or baseline observation carried forward (BOCF) which affect the validity of study outcomes, particularly in the context of the 66% attrition rate. ITT analyses for anthropometric data (for which intermediate outcomes were available) were, therefore, conducted as secondary analyses (presented in the supplementary materials).

Of the 197 participants (IER = 99, CER = 98) who started the study, 66 (IER = 27, CER = 39) completed the 6-month weight loss phase. Baseline characteristics of these study completers are presented in Table 1. The groups were matched for BMI, adiposity, gender and were primarily Caucasian. CER participants tended to be older (56 vs 50 years, p = 0.055). 131 participants withdrew from the study, with reasons for dropout detailed in the study consort diagram (Fig. 1). More dropouts occurred in the IER group (n = 72), of which a number of dropouts were attributed to diet-related adverse effects including faintness on fast days (n = 1), constipation (n = 1), self-reported hypoglycaemia (n = 1), and n = 18 reported that they could not tolerate the diet. Non-completers tended to have a slightly higher BMI (41.5 ± 0.7 vs. 39.6 ± 0.7 kg/m²; p = 0.070, unpaired t test); no other significant differences were noted. Of the completers of the weight loss phase, follow-up data at 1 year from study commencement were available for 47 (IER = 17, CER = 30) patients.

As depicted in Fig. 2, among completers mean weight loss was significantly greater in the IER group (5.4 ± 1.1%) versus the CER group (2.8 ± 0.6%) at 6 months (diet*time p = 0.001), with a mean difference of − 1.8 kg [95% CI − 3.2 to − 0.4 kg]. There was a significantly greater reduction in fat mass in the IER group (mean difference − 2.7 kg [95% CI − 3.5 to − 1.8 kg]; diet*time p < 0.001), as well as in waist circumference (mean difference − 1.6 cm [95% CI − 2.9 to − 0.3 cm]; diet*time p = 0.005). FFM was comparably reduced in both groups (diet*time p = 0.120).

Statistical outcomes for LOCF and BOCF ITT statistical analyses on anthropometric data are presented in Supplementary Table 1. Both ITT statistical approaches found that reductions in FM tended to be greater (diet*time p ≤ 0.081) in the IER group versus the CER group. Changes in body weight (IER − 2.0% ± 0.4% and CER − 1.6 ± 0.3%; p ≥ 0.264), waist circumference (p ≥ 0.253) and FFM (p ≥ 0.460) were statistically comparable between the dietary groups.

Changes in cardiometabolic risk factors are presented in Table 3. With regard to glycaemic control indices, there were no significant between-group differences in the changes in fasting glucose (diet*time p = 0.143) or insulin (diet*time p = 0.873). Neither intervention resulted in significant changes in total cholesterol (TOTC) or LDL cholesterol (diet*time p ≥ 0.349), whereas HDL cholesterol (p = 0.007 paired t test) and triacylglycerol (p = 0.008 paired t test) increased in the IER group only although neither achieved between-group significance (diet*time both p ≥ 0.127). Next, for systolic blood pressure, the relative reduction was greater in the IER group (diet*time p < 0.001) with a mean difference of − 8.6 mmHg [95% CI − 12.7 to − 4.4 mHg], whereas changes in diastolic blood pressure were comparable (diet*time p = 0.701). Lastly, high sensitivity CRP (an inflammatory marker) did not change significantly in either group (diet*time p = 0.873).

1 year after the start of the intervention, there were 17 (17%) and 30 (31%) patients still engaged with RIO who had completed the IER and CER interventions, respectively.

At one-year follow-up (Fig. 2), anthropometric measures (weight, waist circumference) had returned back to baseline levels among IER participants (p ≥ 0.781 baseline vs 1 year, paired t test). This contrasted with CER participants, who tended to maintain the reductions in these anthropometric measures (p ≤ 0.054 baseline vs 1 year, paired t test).

Some clinical data, including blood pressure, were available at the 1 year follow-up. Relative to baseline, for both groups, blood pressure parameters returned close to baseline values (p > 0.05, paired t test), whilst systolic blood pressure was greater at 1 year in the CER group versus baseline (p = 0.015 paired t test).

This study aimed to compare the effects of IER versus CER on anthropometry and cardiometabolic risk factors within a primary care based Tier 3 NHS weight management service. Importantly, the study also followed up those patients who were still actively engaged with RIO six months after the end of the intervention period to explore whether either diet was able to provide any long-lasting benefits. The study found greater degrees of weight loss with IER, but also greater attrition and weight regain following a 6-month follow-up period. It should be noted that the high attrition rates and subsequent underpowering of the study represent significant limitations of the study. Nonetheless, findings from the present study provide useful insights into the implementation of IER within a “real-life” NHS setting and highlight several areas worthy of further investigation. Key findings shall be discussed in turn.

Both CER and IER exhibited a significant weight loss of 3–5%, respectively, surpassing weight loss thresholds deemed to be clinically significant by NICE [9]. Indeed, both groups exhibited an improvement in at least one cardiometabolic health parameter. The IER group experienced a greater reduction in weight and associated adiposity after the 6-month intervention and exhibited superior improvements in some cardiometabolic health parameters including reductions in systolic blood pressure and in fasting triacylglycerol.

Initial weight trends are consistent with self-reported dietary intake data, which showed that IER facilitated a greater overall reduction in energy intake. Unfortunately, dietary intakes were not assessed at 1 year follow-up, but presumably IER participants resumed normal eating patterns which might explain this weight regain. Although both groups were provided with healthy eating advice and supported, more research is required to assess how to optimise IER diets and dietary advice to promote improved tolerance and weight maintenance within such setting.

These data are not in agreement with the popular speculation that IER might prove easier to follow as individuals need not energy restrict every day. Our data also contrast with a recent study of alternate day fasting which showed no differences in weight regain following 6 months of follow-up [6]. A key difference is that the current study involved patients with complex obesity who were referred to a specialist obesity service within the NHS. Moreover, it is also worth noting that most patients attending RIO are those patients who have previous struggled to lose weight or failed on more traditional Tier 2 weight loss interventions.

The blood lipid data are in keeping with research demonstrating that IER can positively modulate triacylglycerol metabolism [5]. With regard to systolic blood pressure, the majority of prior research shows comparable changes with CER, and so it is interesting that the present study and another utilising near identical dietary protocols [5] have shown IER to exert a greater reducing effect. However, much like weight trends, these improvements in systolic blood pressure proved to be transient, depreciating over the course of the follow-up period.

A particular strength of the study is that it was conducted in clinical practice and thus fills a vital knowledge gap in the IER research field. In addition, the 1 year follow-up period for those patients still engaged with the weight loss service provides some insight into the longer-term sustainability of IER.

The study does have some limitations which are important to note. Firstly, the high (66%) attrition rates within this cohort with complex obesity limit the external validity of the study. Moreover, by limiting the available sample size for the study, the statistical power of the study was also reduced which increased the risk of type 2 error (false negative). It should be noted, however, that attrition rates within the CER group, which reflects engagement with standard care, are on a par with published attrition rates for RIO and other Tier 3 specialist obesity services [10] and these findings ultimately represent “real world” data. Attrition rates, combined with a lack of follow-up data for participants who did not complete the dietary interventions, also limited the ITT analyses (included as supplementary materials) due to the assumptions required when accounting for missing observations. Thus, a completer-only analysis was used as the primary statistical method.

Next, personnel conducting the anthropometric assessments (through interaction with participants) were unblinded to diet assignment. Additionally, a lack of details on confounding factors such as medication, smoking and physical activity mean results cannot directly prove the association between the dietary interventions and study outcomes were causal. Furthermore, with a higher proportion of females in a predominantly Caucasian population, applicability of these results to the wider population is limited. Lastly, is the use of bioimpedance which sensitive to hydration status and hormonal status. Moreover, ensuring standard conditions are met before anthropometric measurements, such as prior physical activity, dietary intake and using the same observer are essential for accuracy and reproducibility, however were not always practical given the nature and location of the study.

Overall, results from the present study comparing the long-term effects of IER with CER within an NHS weight management setting show that IER may be more effective at reducing body weight and improving some cardiometabolic risk factors in the short term. However, it was also associated with higher rates of attrition and weight regain following cessation of the 6-month weight loss phase when compared with standard care in this cohort of patients with complex obesity. Considerations for future research include a multiple-centre trial to assess the effectiveness of IER in the wider population, including its efficacy across various ages and ethnic groups. But first, future research is required to establish factors affecting short- and long-term acceptability of IER, and how best to optimise dietary advice to facilitate long-term compliance and weight maintenance.