Comparison of the Effectiveness of Low Carbohydrate Versus Low Fat Diets, in Type 2 Diabetes: Systematic Review and Meta-Analysis of Randomized Controlled Trials

The clinical benefit of low carbohydrate (LC) diets compared with low fat (LF) diets for people with type 2 diabetes (T2D) remains uncertain. We conducted a meta-analysis of randomized controlled trials (RCTs) to compare their efficacy and safety in people with T2D. RCTs comparing both diets in participants with T2D were identified from MEDLINE, Embase, Cochrane Library, and manual search of bibliographies. Mean differences and relative risks with 95% CIs were pooled for measures of glycaemia, cardiometabolic parameters, and adverse events using the following time points: short-term (3 months), intermediate term (6 and 12 months) and long-term (24 months). Twenty-two RCTs comprising 1391 mostly obese participants with T2D were included. At 3 months, a LC vs. LF diet significantly reduced HbA1c levels, mean difference (95% CI) of −0.41% (−0.62, −0.20). LC diet significantly reduced body weight, BMI, fasting insulin and triglycerides and increased total cholesterol and HDL-C levels at the short-to-intermediate term, with a decrease in the requirement for antiglycaemic medications at intermediate-to-long term. There were no significant differences in other parameters and adverse events. Except for reducing HbA1c levels and adiposity parameters at short-to-intermediate terms, a LC diet appears to be equally effective as a LF diet in terms of control of cardiometabolic markers and the risk of adverse events in obese patients with T2D.


Introduction
Type 2 diabetes (T2D) is a chronic and life-changing metabolic disorder that occurs when the pancreas does not produce enough insulin, or the body cannot effectively use the insulin it produces to regulate blood sugar. It is characterised by hyperglycaemia and associated with an unhealthy lifestyle [1][2][3]. Glycated hemoglobin also known as HbA1c is a preferred diagnostic test for T2D because it reflects an individual's average blood glucose levels over the previous 3 months [4]. Reduction in levels of HbA1c is associated with reduction in T2D complications such as damage to the heart, nerves, blood vessels, eyes and kidneys, and death [4,5]. T2D is a major global public health concern due to the rising prevalence and its impact on the health of affected individuals, their families and the substantial costs associated with its management. According to the World Health Organisation, [3] the number of people with diabetes has quadrupled since 1980, with T2D as the vast majority (over 95%) of cases. In addition, the prevalence of T2D is now rising rapidly in both adults and children, and in low-and middle-income countries than in high-income countries. It is now the ninth leading cause of death, with over 6 million people dying from the disease in 2021 [6].

Eligibility Criteria
Studies were eligible if they were randomized controlled, open or blinded trials that: (i) enrolled adults classified as prediabetes and those living with T2D regardless of medication use, glucose and glycated haemoglobin (HbA1c) levels, and comorbidities; (ii) compared a LC diet (<26% of total energy or <130 g of carbohydrate a day) with a LF diet (>26% of total energy or >130 g of carbohydrates a day); (iii) reported at least 12 weeks duration of the trial; (iv) and reported on any of the outcomes below.

Data Extraction and Risk of Bias Assessment
Two authors (TA and SKK) independently extracted data, with inconsistencies resolved by discussion. A predesigned data extraction form was used to extract all the relevant information on publication date, geographical location, study design characteristics (e.g., randomization, allocation concealment, blinding, duration), population (e.g., baseline age, percentage of males, baseline BMI and HbA1c), intervention and comparator, and outcomes. Outcome data were extracted for the specific time points reported by the trials. For multiple publications of studies using data from the same trial, non-overlapping data based on the most comprehensive results were extracted. We used the Cochrane Collaboration's risk of bias tool to assess the risk of bias of included RCTs [23]. This tool evaluates seven possible sources of bias: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting and other bias. For each individual domain, studies were classified into low, unclear and high risk of bias. We also used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) tool to assess the quality of the body of evidence, based on study limitations, inconsistency of effect, imprecision, indirectness and publication bias [24].

Statistical Analysis
Summary measures of effect were presented as relative risks (RRs) (95% CIs) for binary outcomes and mean differences (95% CIs) for continuous outcomes and. Relative risks and 95% CIs were estimated from the extracted raw counts for the interventions and comparators. For studies that reported data such as medians (ranges and 95% CIs) and means (SDs and standard errors), these were converted to means and standard deviations using methods described by Hozo and colleagues [25]. For each outcome, effect estimates (RRs and mean differences) were estimated for the time points of 3 months (±1 month), 6 months (+2 months), 12 months (±3 months), and 24 months (±6 months), based on the distribution of the time points reported by the eligible studies and to maintain some consistency with that of a previous review [20]. The time points were categorised as short-term (3 months), intermediate term (6 and 12 months) and long-term (24 months). Measures of effect were pooled using random effects models to minimize the effect of heterogeneity [26]. Where appropriate, fixed effects models were used in parallel analyses. We planned to investigate sources of heterogeneity using subgroup analysis and random effects meta-regression [27] as well as assess for small study effects using formal tests such as Begg's funnel plots [28] and Egger's regression symmetry test [29]. However, these could not be performed because of the limited number of studies (<10) for each outcome assessed. All analyses were conducted using Stata version MP 16 (Stata Corp, College Station, TX, USA). For outcomes that could not be pooled, a narrative synthesis was used to summarise the results.

Study Identification and Selection
Our initial search of relevant databases and manual scanning of reference lists identified 19,029 potentially relevant citations. After screening based on titles and abstracts, 38 articles remained for full text evaluation. Following detailed assessments, 13 articles were excluded because (i) intervention/comparator was not relevant (n = 8); (ii) they were not randomized studies (n = 2); (iii) population was not relevant (n = 2); and (iv) was based on a conference presentation (n = 1). The remaining 25 articles  which were based on 22 unique RCTs, met our inclusion criteria and were included in the meta-analysis ( Figure 1).

Measures of Glycaemia
Comparing LC with LF diets, there were no significant differences in fasting glucose levels at 3, 6, 12, and 24 months: mean differences (95% CIs) of −2.05 (

Measures of Glycaemia
Comparing LC with LF diets, there were no significant differences in fasting glucose levels at 3, 6, 12, and 24 months: mean differences (95% CIs) of

Body Composition
At 3 and 6 months, a LC vs. LF diet significantly reduced body weight, mean differences (95% CIs) of  Figure 6). Results from single reports showed no significant differences in fat free mass at 3, 6, and 12 months comparing a LC with a LF diet ( Figure S2).

Blood Pressure
When LC and LF diets were compared, except for a reduction in SBP at 3 months: mean difference (95% CI) of −4.53 mmHg (−8.57, −0.48) (Figure 7), there were no significant differences in SBP and DBP at all other time periods (Figure 7; Figure S3).

Blood Pressure
When LC and LF diets were compared, except for a reduction in SBP at 3 months: mean difference (95% CI) of −4.53 mmHg (−8.57, −0.48) (Figure 7), there were no significant differences in SBP and DBP at all other time periods (Figure 7; Figure S3).

Lipids
A comparison of LC and LF diets showed there were significant increases in total

Measures of Inflammation
There were no significant differences in CRP levels between the LC and LF diets at 3, 6, 12, and 24 months ( Figure S9). Results from single reports showed that a LC diet reduced interleukin-6 (IL-6) levels at 6 months, with no significant difference at 3 months ( Figure S10).

Other Cardiovascular Risk Markers
The LC diet significantly reduced fasting insulin levels compared with the LF diet at 3 months: mean difference (95% CI) of −2.83 µIU/mL (−4.73, −0.93), but with no significant differences at 6, 12, and 24 months ( Figure S11). Results from single reports showed no significant differences in HOMA2-IR at all time points comparing a LC with a LF diet ( Figure S12). The LC diet significantly reduced HOMA-IR compared with the LF diet at 3 months: mean difference (95% CI) of −0.71 (−1.05, −0.37) ( Figure S13). A single report showed no significant difference in HOMA2-%B at all time points comparing a LC with a LF diet ( Figure S14).

Liver Function
At 3 months, a LC vs. LF diet significantly reduced ALT levels, mean difference (95% CI) of −8.86 U/L (−17.09, −0.62). Results from a single report showed that at 6 months, a LC vs. LF diet reduced ALT levels ( Figure S15). Comparing a LC with a LF diet, there were no significant differences in AST and GGT levels at 3 and 6 months (Figures S16 and S17).

Renal Function
Based on a single report, a LC vs. LF diet significantly increased creatinine levels at 6 months; with no differences at 3 and 24 months ( Figure S18). Results from single reports showed no significant differences in estimated GFR at 3, 6, and 24 months comparing a LC with a LF diet ( Figure S19). Results from single reports showed no significant differences in microalbumin at 3, 6, and 12 months comparing a LC with a LF diet ( Figure S20). Results from single reports showed a significant increase in urea at 3 months, with no significant differences at 6 and 12 months comparing a LC with a LF diet ( Figure S21). Results from single reports showed no significant differences in urinary albumin at 6 and 24 months comparing a LC with a LF diet ( Figure S22). Results from single reports showed a significant increase in uric acid at 3 months with no difference at 24 months comparing a LC with a LF diet ( Figure S23).

Medication Changes
The LC diet achieved a significant reduction in antiglycemic MES compared with the LF diet at 12 and 24 months, with no significant difference at 6 months ( Figure S24).

Adverse Events
With respect to the two diets, there was no significant difference in risk of adverse events (e.g., musculoskeletal ailments with exercise training, hypoglycaemia and gastrointestinal complaints): RR (95% CI) of 1.27 (0.74-2.18; p = 0.38) ( Figure S25).

GRADE Summary of Findings
The GRADE working group recommends up to 7 patient-important outcomes to be listed in the "summary of findings" tables in systematic reviews [24]. In addition to the primary outcomes of fasting glucose and HbA1c levels, we selected body weight, BMI, SBP, and total cholesterol based on their frequency of reporting. We also included the outcome of adverse events since it is recommended that the 7 selected outcomes should include a safety outcome. GRADE ratings for the outcomes are reported in Table S4. GRADE quality of the evidence ranged from moderate to very low.

Key Findings
Given the persisting uncertainty regarding the net clinical benefits of LC diets compared with LF diets for people with T2D remains uncertain, we conducted an aggregate meta-analysis to compare the efficacy and safety of LC with LF diets in people with T2D. For the primary endpoints, LC compared with LF diet reduced HbA1c levels only at the 3-month time point, with no differences in fasting glucose levels at all time points. For body composition measures, LC diet reduced body weight, BMI, and waist circumference mostly at short and intermediate terms. LC diet reduced SBP at 3 months, with no significant differences in SBP and DBP at other time points. Findings for lipid parameters were inconsistent: at short and intermediate terms, there were reductions in levels of triglycerides and total cholesterol/HDL-C ratio; increases in total cholesterol and HDL-C; and no significant differences in levels of LDL-C. There were no significant differences for CRP, AST, GGT, and the risk of adverse events comparing LC with LF diet. LC diet significantly reduced fasting insulin and ALT at short-to-intermediate terms and antiglycemic MES at intermediate-tolong terms. Results from single reports showed LC diet reduced IL-6 levels; increased levels of creatinine and urea; with no differences for fat free mass, HOMA2-IR, HOMA2-%B, estimated GFR, microalbumin, and urinary albumin between the two interventions. The GRADE quality of evidence for the 7 relevant outcomes ranged from moderate to very low.

Comparison with Previous Studies
Similar to the findings of the review by Goldenberg et al. [20] we found moderate evidence for a beneficial effect of a LC diet on weight and BMI at short to intermediate term (3-6 months). However, according to SACN [18], this beneficial effect only occurred at 3 months. Although the quality of evidence for HbA1c in our study was 'very low' compared to 'adequate' in the SACN review [18] and 'high' as reported by Goldenberg et al. [20] all three reviews showed a beneficial effect of a LC diet over the control diet in the short and intermediate term for this glycaemic marker. The short to intermediate term reduction in fasting insulin observed from a LC diet is consistent with the findings of Goldenberg et al. [20] and SACN [18]. In contrast to these two reviews, we observed beneficial effects of a LC diet over LF diet on triglyceride and HDL-C levels, and total cholesterol/HDL-C ratio. The increase in levels of total cholesterol when a LC diet was compared with a LF diet was inconsistent with the findings of Goldenberg et al. [20] who found no effect. Similar to the findings of SACN [18] and Goldenberg et al. [20] we found no significant difference in adverse events between the two diets, but a LC diet resulted in greater reduction in MES compared to a LF diet in our review, whereas our review specifically compared a LC diet with a low LF diet (>26% of total energy or >130 g of carbohydrates a day), the review by Goldenberg et al. [20] compared low and very low carbohydrate diets with a wide range of diets including dietary programs higher in carbohydrates (≥26%), palaeolithic diet as well as no treatment; however, most of the trials included in their review used low fat diets as their control parameters. Furthermore, our review considered only patients with T2D and was based on RCTs, which are the gold standard for evaluating the effectiveness of interventions; whereas, other reviews have included observational studies, people with type 1 diabetes, and a variety of comparator diets [55][56][57][58]. In addition, we evaluated a comprehensive list of cardiometabolic outcomes including blood pressure and markers of renal and liver function which were not evaluated by previous reviews. We also showed that a LC diet was associated with significant reduction in waist circumference at 12 months; an outcome not considered by the reviews of SACN [18] and Goldenberg et al. [20].

Explanations for Findings
Dietary protein is associated with greater satiety and therefore reduction in calorie intake [59]. The satiety effect would explain the significantly higher reduction in weight and BMI from the LC compared to LF diet in the short-to-intermediate term. In addition, our findings suggest that weight loss irrespective of carbohydrate and fat restriction resulted in reduction in BMI, serum lipids, and measures of inflammation and glycaemia. Therefore, it was not possible to distinguish between the impact of carbohydrate compared to fat restriction on these outcomes. The improvement in the clinical outcomes may be attributed to calorie-restriction and associated weight loss rather and macronutrient restriction. Thus, a calorie-restricted balanced diet could produce similar favourable clinical outcomes. This is supported by the results from the ongoing DiRECT study (Diabetes Remission Clinical Trial), where a daily calorie intake of 825-853 kcal/day resulted in remission to a nondiabetic state and off antidiabetic drugs for over half the study participants [60]. The short-to-intermediate term beneficial effect of the LC compared to the LF could be due to the difficulty with adherence to the LC, and therefore unsustainable weight loss in the long term. This is also supported by evidence from the DiRECT programme, where sustained remissions at 24 months, was associated with sustained weight loss [7,61]. The DiRECT study involves total meal replacement, stepped food introduction and structured support for long term weight maintenance. Thus, a calorie-restricted balanced diet accompanied by regular behavioural support could be superior to a LC at improving clinical outcomes including remission for people living with T2D.

Implications of Findings
Based on our analysis, a short-to-intermediate term LC diet rather than a LF diet could be recommended for overweight and obese adults with uncontrolled T2M to achieve glycaemic control and weight loss. The LC diet also decreased the requirement for antiglycaemic medications at intermediate-to long-term. Though no adverse findings were observed at long-term, the LC diet may not be beneficial in the long term given no significant evidence of differences between the two diets at 24 months with respect to all outcomes. These findings are in contrast to previous studies, which have reported that LC diets have adverse effects on lipids, blood pressure and renal function [33,34,36,38,45,46,62]. LC diets which are also higher in dietary protein loads cause accumulation of ketones, resulting in abnormal metabolic functioning. However, it should be acknowledged that the literature on the metabolic effects of LC diet comprises heterogenous studies with small sample sizes. By pooling relevant literature on the topic, our findings suggest that LC diets may only be suitable for short term control of glycaemia and weight loss. Given that a LC diet is characterised by the consumption of large amounts of saturated fat and small amounts of fruits, vegetables and fiber, there is a potential for LC diets to adversely impact on lipid profiles, which are major risk factors for coronary heart disease [63]. Indeed, our results showed that LC significantly increased levels of total cholesterol at short-to-intermediate term. Given that patient-centred care is a key aspect of T2D management, [64] patients who choose a LC diet could be supported to manage their diabetes effectively. Patients on this diet should be advised to base their carbohydrates on foods rich in fibre, variety of fruits and vegetables, given their beneficial effect on glycaemic control and cardiometabolic risk factors [65]. In line with general healthy eating advice, limited intake of salt, trans and saturated fats and regular hydration should also be recommended [66].

Strengths and Limitations
Based on evidence from 20 unique RCTs, our review represents an up-to-date comprehensive systematic review and meta-analysis evaluating the efficacy and safety of LC compared with LF diets. Other strengths of the current review included (i) the evaluation of a comprehensive panel of outcomes, which were reported according to time points; (ii) the utilisation of several meta-analytic approaches including ensuring consistency to enhance pooling of most of the data; and (iii) detailed assessment of the risk of bias of included trials and quality of the evidence using the Cochrane risk of bias and GRADE tools, respectively (iv) the evaluation of the effects on blood pressure, renal and liver function, and other markers which were not included in the most recent comprehensive reviews. The limitations were mostly inherent to the studies and included: (i) the inconsistencies in outcome definitions, time points and assessments; (ii) the results of some outcomes were based on single reports; (iii) all trials had a high risk of bias in the domains of blinding of participants and personnel (iv) limited information provided on types of carbohydrates consumed (v) inability to generalise our findings to other populations such as black ethnicities, and (vi) inability to conduct subgroup analysis by relevant characteristics such as age, sex, geographical location, ethnicity, and BMI as prespecified in the protocol, due to the limited studies available for pooling for each outcome and lack of specific data such as ethnic-specific data analyses.

Conclusions
Except for reducing HbA1c levels and body composition measures at short-to-intermediate term and decreasing the requirement for antiglycaemic medications at intermediate-tolong term, a LC diet appears to be equally effective as a LF diet in terms of control of cardiometabolic markers and the risk of adverse events in obese patients with T2D. The current evidence suggest that LC diets may not be beneficial over the long-term.

Key Points
Question: Is a low carbohydrate diet more effective for control of cardiometabolic markers and the risk of adverse events in obese patients with T2D compared to a low fat diet?
Findings: Except for reducing HbA1c levels and adiposity parameters at short to intermediate terms, a LC diet appears to be equally effective as a LF diet in terms of control of cardiometabolic markers and the risk of adverse events in obese patients with T2D.
Meaning: A short to intermediate term LC diet could be recommended for overweight and obese adults with uncontrolled T2M to achieve glycaemic control and weight loss.