Metabolic Impact of Intermittent Fasting in Patients With Type 2 Diabetes Mellitus: A Systematic Review and Meta-analysis of Interventional Studies

Abstract

Context

Intermittent fasting (IF) has been proposed as a weight-loss strategy with additional cardiometabolic benefits in individuals with obesity. Despite its growing popularity, the effect of IF in patients with type 2 diabetes (T2DM) remains unclear.

Objective

We conducted a systematic review and meta-analysis to evaluate the metabolic impact of IF compared to standard diet in patients with T2DM.

Methods

Embase, PubMed, and clinicaltrials.gov between 1950 and August 12, 2020 were searched for randomized, diet-controlled studies evaluating any IF intervention in adults with T2DM. We examined the impact of IF on weight loss and glucose-lowering by calculating pooled estimates of the absolute differences in body weight and glycated hemoglobin A_1c (HbA_1c) compared to a control group using a random-effects model.

Results

Seven studies (n = 338 participants; mean body mass index [BMI] 35.65, mean baseline HbA_1c 8.8%) met our inclusion criteria. IF induced a greater decrease in body weight by –1.89 kg (95% CI, –2.91 to –0.86 kg) compared to a regular diet, with no significant between-study heterogeneity (I² 21.0%, P = .28). The additional weight loss induced by IF was greater in studies with a heavier population (BMI > 36) (–3.43 kg [95% CI, –5.72 to –1.15 kg]) and in studies of shorter duration (≤ 4 months) (–3.73 kg [95% CI, –7.11 to –0.36 kg]). IF was not associated with further reduction in HbA_1c compared to a standard diet (HbA_1c –0.11% [95% CI, –0.38% to 0.17%]).

Conclusion

Current evidence suggests that IF is associated with greater weight loss in patients with T2DM compared with a standard diet, with a similar impact on glycemic control.

intermittent fasting, type 2 diabetes, glycated hemoglobin, body weight

Obesity plays a role in 61% to 79% of cases of type 2 diabetes mellitus (T2DM) (1) and confers additional morbidity as increased body mass index (BMI) has been linked to poorer cardiovascular risk profile and increased mortality in this patient population (2-5). Although lifestyle interventions represent an important pillar of diabetes management, maintaining weight loss and obtaining sustained glycemic control with nonpharmacological approaches is a difficult goal to achieve (6).

Intermittent fasting (IF), in which energy consumption is repeatedly and intentionally interrupted or markedly reduced for a period of time, has been a focus of recent research and proposed as a weight-loss strategy with additional cardiometabolic benefits in individuals with overweight and obesity. These additional benefits include reduction in total cholesterol and blood pressure, and improved insulin sensitivity (7-12). Although IF interventions have not been standardized, common regimens include time-restricted feeding (TRF), in which feeding is allowed for only a window of 4 to 8 hours per day with 16 to 20 hours of fasting (13-16), and intermittent or short-term energy restriction through very-low calorie diets (VLCDs), in which caloric consumption remains between 300 and 600 kcal/day (17).

Despite the growing popularity of IF in the lay media, limited research has been conducted in patients with T2DM (18). Previous reports in individuals with T2DM have suggested that IF interventions can induce similar weight loss and reduction in glycated hemoglobin A_1c (HbA_1c) as standard dietary recommendations (19-23). However, the small sample sizes preclude definitive conclusions based on these individual studies, indicating the need for a robust and systematic evaluation of the effect of IF in T2DM. Thus, the purpose of this systematic review and meta-analysis is to evaluate the metabolic impact of IF interventions in patients with T2DM.

Material and Methods

This systematic review and meta-analysis is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement and was registered with the International Prospective Register of Systematic Reviews (http://www.crd.york.ac.uk/prospero/; CRD42020159009) (24).

Data sources and searches

We selected relevant studies published between 1950 and August 12, 2020. We searched PubMed, Embase, and clinicaltrials.gov using the following combined text and Medical Subject Heading (MeSH) terms: “type 2 diabetes,” “intermittent fasting,” “time restricted feeding,” “very low calorie diet.” The complete PubMed search was as follows: ((((intermittent fasting [Text Word] OR time restricted feeding [Text Word] OR very low calorie diet [Text Word]) OR))) AND (“diabetes mellitus, type 2” [MeSH Terms] OR type 2 diabetes mellitus [Text Word]). All potentially eligible studies were considered for review, regardless of the primary outcome or language. We also conducted a manual search using references of key articles published in English.

Study selection

Studies were eligible for inclusion if they: (1) were interventional studies, which could be either randomized parallel-arm trials or crossover trials conducted in adults with T2DM, (2) compared any IF intervention to a standard diet consisting of either a healthy pattern dietary recommendation with caloric deficit or normal caloric intake (control group), and (3) reported changes in body weight or HbA_1c. Exclusion criteria were as follows: studies that did not report a control group, retrospective studies, or observational studies. If a study was reported in more than one publication, we included the data reporting the primary outcome.

Intervention investigated and outcome measurements

We evaluated any IF intervention, which includes (i) 24-hour complete fasting, (ii) intermittent restricted energy intake (25% total caloric intake), and (iii) TRF (feeding allowed for only a window of 4 to 8 hours daily with 16 to 20 hours of fasting). The IF intervention could be applied on alternate days, twice weekly, or during a continuous period of time and were compared to a standard dietary recommendation consisting of regular eating hours. The primary outcomes were mean differences in the changes in (i) body weight and (ii) HbA_1c, between baseline and the end of the intervention. The following secondary outcomes were assessed: changes in fasting glucose, total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, triglycerides, systolic and diastolic blood pressure, and waist circumference.

Data extraction and quality assessment

Two independent investigators (E.B. and C.K.K.) reviewed study titles and abstracts. Studies that satisfied the inclusion criteria were retrieved for full-text assessment. Studies selected for review by both investigators had an agreement value (k) of 98.3%; disagreements were resolved by a third investigator (J.M.).

The following data were extracted from each study: age, percentage of male participants, total number of participants, duration of intervention, baseline BMI, baseline HbA_1c, mean changes in body weight (mean [SD]), mean changes in HbA_1c (mean [SD]), and mean changes in other metabolic parameters (secondary outcomes listed earlier). The risk of bias was evaluated according to the PRISMA recommendations (24).

Data synthesis and analysis

We examined the impact of IF on 2 primary outcomes: (1) weight reduction, as assessed by mean change in body weight (kg), and (2) glucose-lowering effect, as assessed by mean change in HbA_1c (%). We calculated pooled estimates of the absolute differences between arithmetic means before and after interventions as compared to control group using a random-effects model (DerSimonian-Laird method). The I² value was used to evaluate the magnitude of heterogeneity between studies, with values greater than 50% indicating moderate-to-high heterogeneity (25). Recognizing that the duration of intervention and baseline characteristics of the study population could affect the impact of IF on the outcomes, we performed stratified analyses by the mean values for each of the following variables: (i) study duration (≤ 4 months and > 4 months), (ii) baseline BMI (≤ 36 and > 36 kg/m²), and (iii) baseline HbA_1c (≤ 9.0% and > 9.0%). To further evaluate the impact of IF on metabolic profile of patients with T2DM, sensitivity analyses were performed: (i) using random-effects models and (ii) after excluding Kahleova et al (26), because this study was the only randomized crossover trial included in our analysis with the remainder consisting of randomized parallel-arm trials. In addition, Kahleova et al (26) was the only study that evaluated a TRF protocol.

The possibility of publication bias was evaluated using a funnel plot of effect size against the SE for each trial. Funnel plot asymmetry was evaluated by Begg and Egger tests, with significant publication bias defined as a P value less than .1 (27). The trim-and-fill computation was used to estimate the effect of publication bias (28).

All analyses were performed using Stata 14.0 (Stata Corp.

Results

Study characteristics

We identified 210 studies, of which 194 were excluded on the basis of title and abstract. Sixteen studies were retrieved for detailed assessment, 9 of which were excluded (4 for lack of a control group and 5 that were considered duplicates because there were additional publications from included trials). Seven studies (21-23, 26, 29-31) (with data from 338 participants) met our inclusion criteria; 6 were randomized parallel-arm trials (21-23, 29-31) and 1 was a randomized crossover trial (26). Included studies were published between 1991 and 2018, with median study duration of 24 weeks (range, 19-260 weeks). Patients had a mean age of 56.3 years (range, 51.2-65 years), 24.2% to 54% were male, baseline BMI of 35.65 (range, 32.4-37.9), and mean baseline HbA_1c of 8.8% (range, 7.2%-10.4%), representing a population largely obese with poor glycemic control (Table 1). Two studies reported duration of diabetes (mean duration of diabetes of 8.0 years). Five studies (21, 23, 26, 29, 30) reported data on background diabetic treatment, with the majority of participants taking oral antidiabetic medications or on diet only (75%-100%) (21, 23, 29, 30), while 20% to 25% were also taking insulin (21, 23, 29, 30).

Table 1.

Characteristics of included studies

Authors	Publication year	Sample size	Study design	Study intervention	IF duration, d	Total study duration, wks	Mean age, y	Male, %	Baseline mean BMI	Baseline HbA1c, %
Wing et al (29)	1991	36	Randomized parallel- arm	Intervention arm: VLCD (400 kcal/day) wks 5-12 (run-in period 0-5), remainder of wks behavior therapy + l000-1500 kcal/d Control arm: behavior therapy + 1000-1500 kcal/d for 20 wks	56	20a	51.2	24.2	37.7	10.4
Wing et al (23)	1994	93	Randomized parallel- arm	Intervention arm: VLCD (400-500 kcal/d) wks 0-12 and 12-24, LCD (1000-1200 kcal/d) remainder of wks Control arm: LCD (1000-1200 kcal/d) throughout	168	50	51.8	35.5	37.9	10.4
Williams et al (22)	1997	54	Randomized parallel- arm	Intervention arm 1: VLCD (400-600 kcal) 5 consecutive days during wk 2, then VLCD 1×/wk and regular diet (1500- 1800 kcal) 6×/wk for remaining 15 wks Intervention arm 2: VLCD (400-600 kcal) 5 consecutive days during wks 2, 7, 12, and 17, and regular diet (1500-1800 kcal) remaining wks Control arm: regular diet (1500-1800 kcal) 7×/wk for study duration	20	20	52.2	38.9	35.9	8.1
Paisey et al (30)	2002	45	Randomized parallel- arm	Intervention arm: VLCD (450 kcal/d for women, 650 kcal/ day for men) for 6 wks, followed by slow reintroduction of standard eating Control arm: intensive conventional diet and exercise (diet/exercise sessions 2×/wk for 6 mos)	42	260	53.9	36.7	36.8	Not reported
Kahleova et al (26)	2014	54	Randomized crossover	Intervention arm: 12 wks of 2 meals/d (breakfast 6-10 am; lunch 12-4 pm) Control arm: 12 wks of 6 meals/d (breakfast/lunch/dinner + 3 snacks)	84	24	59.4	54.0	32.6	7.2
Li et al (31)	2017	46	Randomized parallel- arm	Intervention arm: 2 days prefasting (1200 cal) followed by 7 d VLCD (300 cal/d), followed by slow reintroduction of Mediterranean diet Control arm: Mediterranean diet, normal calorie intake	7	16	65.0	NA	32.4	7.7
Carter et al (21)	2018	137	Randomized parallel- arm	Intervention arm: intermittent restricted diet (500-600 kcal/d) 2×/wk, regular diet remaining 5×/wk Control arm: continuous restricted diet (1200-1500 kcal/day) 7×/wk	104	52	61.0	43.8	36	7.3

Abbreviations: BMI, body mass index; cal, calories; HbA_1c, glycated hemoglobin A_1c; IF, intermittent fasting; kcal, kilocalories; LCD, low-calorie diet; NA, not available; VLCD, very low-calorie diet.

^aOutcomes also available after 1 year.

The IF intervention adopted in these studies varied, with one study evaluating TRF (26), 2 studies evaluating intermittent energy restriction (21, 22), and 4 studies assessing short-term energy restriction through VLCD (23, 29-31). Table 1 gives a detailed overview of the IF interventions followed in each study. The included studies did not evaluate macronutrient content. Regarding physical activity, 3 studies (22, 23, 30) recommended participants increase physical activity level without further assessing adherence to this recommendation. Kahleova et al (26) evaluated physical activity using a pedometer, but did not find any differences between the intervention arms.

A risk of bias assessment is shown in Table 2. All 7 studies reported adequate randomization and none were stopped early. Two trials reported intention-to-treat analyses (21, 26), while the remaining 5 reported results per protocol (22, 23, 29-31). The dropout rates varied from 8.3% to 44% with no differences between the intervention and control arms (see Table 2).

Table 2.

Risk of bias

Author	Intention to treat analysis	Intermittent fasting efficacy evaluated	Stopped early	Dropout rate, %			Outcome assessment accurate
				Overall	Intervention arm	Control arm
Wing et al (29)	No	Yes	No	8.3	0	15.8	Yes
Wing et al (23)	No	Yes	No	15.0	15.6	14.6	Yes
Williams et al (22)	No	Yes	No	16.7	16.7	16.7	Yes
Paisey et al (30)	No	Yes	No	16.7	13.3	20.0	Yes
Kahleova et al (26)	Yes	Yes	No	6.7	5.8	7.5	Yes
Li et al (31)	No	Yes	No	30.4	30.4	30.4	Yes
Carter et al (21)	Yes	Yes	No	29.2	27.1	31.3	Yes

Impact of intermittent fasting on weight loss

Six studies assessed the change in body weight between IF and standard diet (21-23, 26, 29, 31). Pooling the data from these studies showed a significant decrease in body weight by 1.89 kg (95% CI, –2.91 to –0.86 kg) in the IF arm compared to control, with no significant between-study heterogeneity (I² = 21.0%, P = .28) (Fig. 1A). In this analysis, there was evidence of publication bias on the Egger test (P = .034); however, the trim-and-fill test showed that effect estimates were not significantly affected as no trimming was necessary to adjust the overall estimate.

Figure 1.

Meta-analysis of the mean difference in A, body weight (kg), and B, glycated hemoglobin A_1c (HbA_1c; %) between intermittent fasting interventions and standard diet. Weight (%) represents the weighted average of the effect sizes.

Sensitivity analyses were performed stratifying by baseline BMI (≤ 36 vs > 36) and study duration (≤ 4 vs > 4 months) (Fig. 2). Compared to standard diet, the additional weight loss induced by IF was greater in studies with a heavier population (BMI > 36) (–3.43 kg [95% CI, –5.72 to –1.15 kg]) than in those with a BMI less than or equal to 36 (–1.42 kg [95% CI, –1.90 to –0.95 kg]) (Fig. 2A). In addition, IF interventions led to more pronounced weight loss compared to standard diet both in short-term (–3.73 kg [95% CI, –7.11 to –0.36 kg]) and long-term study durations (–1.44 kg [95% CI, –1.91 to –0.97 kg]) (Fig. 2B). In these analyses, there was no significant between-study heterogeneity.

Figure 2.

Meta-analysis of the mean difference in body weight (kg) between intermittent fasting interventions and standard diet in analyses stratified by baseline A, body mass index (BMI), and B, study duration. Weight (%) represents the weighted average of the effect sizes.

We repeated the analysis excluding Kahleova et al (26) given that this study was the only crossover trial evaluating a TRF intervention whereas the other studies consisted of parallel design evaluating very restricted energy intake. In this analysis, there was a decrease in body weight by 2.6 kg (95% CI, –4.18 to –1.02 kg) in the IF arm compared to the control arm.

Impact of intermittent fasting on glycemic control

The pooled analysis of the 6 trials (21-23, 26, 29, 31) that assessed the change in HbA_1c showed a nonsignificant decrease of –0.11% (95% CI, –0.38% to 0.17%) in the IF arm compared to control, with no significant between-study heterogeneity (Fig. 1B). There was no publication bias in this analysis as assessed by the Egger test (P = .30).

Recognizing that IF interventions may differentially affect those with a higher baseline HbA_1c, and that changes in HbA_1c are more evident over a period of several months, we repeated this analysis stratifying by baseline HbA_1c (Figure 3A) and study duration (Fig. 3B). Similarly to the results observed in the pooled analysis, there were no significant differences in HBA_1c reduction conferred by IF interventions compared to standard diet detected among subgroups, and no significant between-study heterogeneity (Fig. 3).

Figure 3.

Meta-analysis of the mean difference in glycated hemoglobin A_1c (HbA_1c; %) between intermittent fasting interventions and standard diet in analyses stratified by baseline A, HbA_1c, and B, study duration. Weight (%) represents the weighted average of the effect sizes.

Furthermore, we repeated the analysis excluding Kahleova et al (26) and demonstrated no significant reduction in HbA_1c comparing IF with control arm (–0.26% [95% CI, –0.73% to 0.22%]).

Impact of intermittent fasting on additional metabolic parameters

To further evaluate the metabolic impact of IF interventions compared to a standard diet, we pooled the data of studies that assessed changes in fasting glucose, lipid profile, blood pressure, and waist circumference. IF was not associated with additional positive effects on any of these parameters compared to a standard diet (Table 3).

Table 3.

Meta-analysis comparing metabolic parameters (mean difference) between intermittent fasting interventions and standard diet in patients with type 2 diabetes

Clinical characteristic	No of studies	Absolute mean difference	95% CI	I  2
Fasting glucose, mmol/L	4	–0.75	–2.24 to 0.75	85.7
Total cholesterol, mmol/L	6	0.070	–0.13 to 0.26	36.5
LDL cholesterol, mmol/L	5	0.040	–0.04 to 0.12	0
HDL cholesterol, mmol/L	6	–0.05	–0.17 to 0.08	88.1
Triglycerides, mmol/L	6	0.090	–0.03 to 0.21	0
Systolic blood pressure, mm Hg	3	–0.68	–14.85 to 13.49	89.1
Diastolic blood pressure, mm Hg	3	–1.80	–15.02 to 11.41	93.2
Waist circumference, cm	2	–2.44	–7.94 to 3.06	0

Abbreviations: HDL, high-density lipoprotein; LDL, low-density lipoprotein.

Discussion

Our results suggest that, compared to a standard diet, IF induces additional weight loss (~ 1.9 kg) while having similar effect on HbA_1c, lipid profile, and blood pressure in adults with T2DM representing a population largely obese with poor glycemic control. The positive impact of IF on weight reduction was more pronounced in heavier patient populations and in studies of shorter duration.

Although weight loss plays a central role in the treatment of T2DM associated with obesity (32), poor adherence to physical activity recommendations and sustained healthy diet has limited the effectiveness of nonpharmacological strategies in adults with T2DM (33). Our results demonstrated that IF may represent a therapeutic alternative, as this intervention significantly decreased body weight when compared to a standard diet, an effect observed irrespective of baseline BMI or study duration. Reinforcing the results of our analyses, in the longest trial evaluating IF in T2DM (12 months’ duration), Carter et al (21) observed a –6.8-kg weight loss in the IF arm compared to –5.0 kg in the control group, with HbA_1c reduction of –0.5% vs –0.2%, respectively. Despite the steeper impact on weight loss achieved, Carter and colleagues (21) demonstrated a similar reduction in fat-free mass in the IF arm compared to the control group. Notably, the additional reduction in body weight induced by IF was modest in the context of the baseline BMI of our study population (35.6); however, this additional weight loss may facilitate the overall goal of 5% reduction in total body weight in obese individuals with T2DM (32, 34-41).

Despite the greater weight loss induced by IF interventions, our results demonstrate a neutral impact in HbA_1c reduction. A previous meta-analysis evaluating the impact of IF in the general population (n = 545 participants) (42) demonstrated a modest reduction in fasting glucose of –4.16 mg/dL in addition to a significant reduction in BMI. It should be noted, however, that despite the lack of statistical superiority of IF compared to a standard diet on glycemic control in individuals with T2DM, there may still be important clinical implications. Williams et al (22) found no difference in HbA_1c values between the IF and standard diet arms after a 20-week intervention; however, more participants in the IF group achieved the target HbA_1c than in the control group (22). Another aspect to consider is the concomitant use of antidiabetic medications. Wing and colleagues found that, despite similar reduction in HbA_1c among study arms, there was a significant difference in the number of participants taking glucose-lowering medications 1 year post intervention (45% of participants in the IF arm and 69% of participants in the control arm) (23), suggesting that IF may represent another lifestyle option to optimize diabetes care, reducing the requirement of antidiabetic medications.

Other potential positive effects of IF beyond weight loss and glycemic control have been described in the literature, possibly due to its potential to shift the preferential use of glucose from glycogenolysis to fatty acids and fatty acid–derived ketones. Specifically, during the period of prolonged fasting when glycogen stores in hepatocytes are depleted, an accelerated lipolysis produces fatty acids and glycerol in a metabolic pathway independent of pancreatic insulin secretion (43). By inducing this change, IF could promote positive metabolic changes independent of weight loss. This concept was demonstrated in a crossover trial evaluating the effect of IF in prediabetic men while maintaining weight. In that study, IF improved insulin sensitivity, β-cell responsiveness, blood pressure, and oxidative stress levels, independently of weight loss (44).

A limitation of our meta-analyses is the heterogeneity among the IF protocols. Currently, dietary regimens consisting of very restrictive caloric intake such as 400 to 500 kcal daily as well as complete fasting have been described in the literature as IF protocols (18, 45). There is a lack of standardization of IF interventions as fasting interventions have differed in the number of calories allowed per day, the timing and duration of the fasting window, the number of fasting days per week, and the length of the intervention. However, this is not expected to affect our results as both fasting and severely restricted diets induce a ketogenetic state characterized by an increase in free fatty acids and ketone bodies acetoacetate and β-hydroxybutyrate, which is the likely mechanism through which the benefits of this intervention are obtained (45). Another limitation is that the long-term adherence to IF is uncertain as the majority of included trials lasted between 1 and 24 weeks. In addition, detailed information on how caloric deficit was achieved in these studies is lacking. Finally, in our analyses, it was not possible to evaluate the safety of IF in T2DM, which is particularly relevant for patients taking insulin. In this context, data from a previous report demonstrated that the rates of hypoglycemia in individuals with T2DM undergoing both 2 consecutive and nonconsecutive days of IF are acceptable and decreased with proper education and titration of antidiabetic medications (20).

In conclusion, given the historical challenges of nonpharmacological approaches in T2DM, our results support IF as a noninferior alternative strategy for weight loss. Our analyses demonstrate the therapeutic potential of IF as a weight-reduction strategy in T2DM and highlight the need for further research in this field. Specifically, there is a need for trials comparing the different IF protocols with longer duration of follow-up and with deeper phenotyping of the clinical and metabolic effects of this intervention in patients with T2DM.

Abbreviations

Abbreviations
 
• BMI

  body mass index

 
• HbA_1c

  glycated hemoglobin A_1c

 
• IF

  intermittent fasting

 
• LCD

  low-calorie diet

 
• PRISMA

  Preferred Reporting Items for Systematic Reviews and Meta-Analyses

 
• T2DM

  type 2 diabetes mellitus

 
• TRF

  time-restricted feeding

 
• VLCD

  very low-calorie diet

Acknowledgments

C.K.K. holds the Canadian Diabetes Association Clinician-Scientist award.

Financial Support: This work was supported by intramural funds.

Author Contributions: C.K.K. conceived the systematic review and analysis plan. E.B., C.K.K., and J.M. selected studies for inclusion and abstracted data. E.B. performed the statistical analyses, interpreted the data, and wrote the first draft. E.B., C.K.K., and J.M. critically revised the manuscript for important intellectual content and approved the final draft.

Additional Information

Disclosures: Dr Kramer reports grants from Boehringer Ingelheim outside the submitted work. The other authors have nothing to disclose.

Data Availability

Data sets generated during and/or analyzed during the present study are not publicly available but are available from the corresponding author on reasonable request.