Improving the scientific rigour of nutritional recommendations for adults with type 2 diabetes: A comprehensive review of the American Diabetes Association guideline‐recommended eating patterns

Abstract Aims The global rate of type 2 diabetes (T2D) continues to rise. Guidelines that influence the worldwide treatment of this disease are central to changing this trajectory. We sought in this review to evaluate the appropriateness of sources cited in the American Diabetes Association's (ADA) guidelines on eating patterns for T2D management, identify additional relevant sources, and evaluate the evidence. Materials and Methods We reviewed the evidence behind the ADA's recommendations on eating patterns in the 2018 and 2019 ADA Standards of Care and the 2014 ADA Nutrition Therapy Recommendations for Adults with Diabetes. Additionally, we conducted a comprehensive search to identify any additional studies not included in the cited evidence. To determine appropriateness of inclusion in the guidelines, the following criteria were applied: 1) it was a clinical trial or systematic review/meta‐analysis of clinical trials; 2) it involved persons with T2D; 3) one of the study arms followed one of the eating patterns currently recommended; 4) its reported outcomes included glycaemic control; 5) outcomes were reported separately for persons with T2D. Results We found a wide variation in the evidence for each eating pattern. Issues that have hampered the guideline process include: lack of a rigorous literature review, resulting in the omission of pertinent studies; an overreliance on prospective cohort studies; inconsistent standards for evidence; inclusion of studies not on persons with T2D; and bias. Conclusions The ADA Guidelines recommended eating patterns fall short of rigorous standards of scientific review according to state‐of‐the‐art systematic review and guideline creation practices.

). Despite the availability of standards to improve the development of clinical practice guidelines, there is still wide concern among the scientific community that even the most wellrespected guidelines lack sufficient rigour. [2][3][4][5][6] Over half of adults in the United States now have type 2 diabetes (T2D) or prediabetes, 7 and worse, this multifactorial epidemic is now worldwide and shows no signs of slowing, with rates of both T2D and T2D-related health complications rising. 8 When advising people with T2D on food choices, many healthcare providers rely on nutrition guidelines provided by the American Diabetes Association (ADA), and these guidelines influence standard recommendations made around the globe. 9-11 Given these alarming trends, it is of paramount importance to review the treatment guidelines to ensure they are based on rigorous, accepted scientific methods.
The ADA's approach to the evidence in developing its guidelines has been to employ a grading system to rate the strength of evidence.
An "A" rating is given to well-conducted randomized controlled trials (RCTs) that are adequately powered, as well as to meta-analyses that incorporate quality ratings. "B" ratings are given to well-conducted cohort studies, "C" ratings are for poorly controlled trials or uncontrolled studies, and a score of "E" is for expert consensus or clinical experience. This approach does not follow any of the widely accepted standards or "guidelines for guidelines" such as Agree II, GRADE, or those from the National Academy of Sciences, Engineering and Medicine (Table S1).
Several concerns prompted our review of the evidence cited by the ADA in support of its recommendations for eating patterns in the management of T2D: (a) a strong reliance by the ADA on sources that they rate as B, C and E 12,13  We conducted a review of the sources cited for currently recommended eating patterns in the ADA's Standards of Medical Care in Diabetes (Table 1) (2018 and 2019 standards), 12,14 and the ADA's Nutrition Therapy Recommendations for Adults with Diabetes (2014 recommendations), 13 which helped inform the 2018 standards. In addition, a comprehensive search was conducted to identify any studies that would have been appropriate to include in a rigorous review.
The review considers the strength of the evidence but does not assign a grade to each study.

| MATERIALS AND METHODS
The present review includes studies newly cited in the 2019 standards (Tables S2-S5), as well as sources cited in the 2014 recommendations and the 2018 standards (Table S7). ketogenic. We also found other articles by reviewing references cited in relevant studies. A flow diagram of the search can be found in Figure S1.
Two co-authors independently conducted the searches and evaluated all studies for appropriateness. In cases of disagreement, the two co-authors and a third co-author discussed the findings and reached agreement. All studies deemed appropriate for inclusion are presented in Tables S2-S5.
Third, we evaluated the evidence from all of the assembled studies, those cited by the ADA (Table S7) as well as those we had identified (Tables S2-S5). We did not assign a grade to each study but rather, on a prima facie basis, assessed whether or not the cited study provided evidence of benefit.  which either reported a low incidence of T2D in those following the DASH diet or recommended the diet for blood pressure control.

| Additional evidence
We identified a post hoc analysis of the Exercise and Nutritional Inter-  during the study period in participants without T2D or prediabetes in the DASH arm, more than with the control and DASH diet + exercise. 23 The RCT by Paula et al 24 compared the DASH diet + exercise to a diet based on ADA guidelines that did not include exercise. The significance of change from baseline and in a comparison of interventions was mixed; DASH + exercise resulted in a greater reduction in blood pressure but no difference in glycaemic control when compared to usual care; however, the effect of the DASH diet without exercise was unknown. 24

| Summary of evidence
To our knowledge, clinical research on the DASH diet that provides outcomes for people with T2D consists of two RCTs, of 4 and 8 weeks' duration, and a post hoc analysis. 15,23,24 Only one of the two trials showed glycaemic improvement that can be attributed to the DASH diet alone. According to our evaluation, the other cited sources provide limited to no support for the DASH diet for people with T2D in improving glycaemic control for the reasons already cited: these studies were not clinical trials or systematic reviews, or did not pro-  26 for which data were reanalysed with essentially the same results in 2018, 27 reported a significant reduction of major cardiovascular events in both versions of the Mediterranean diet studied, compared with the control. Two systematic reviews 29,30 found limited evidence that the Mediterranean diet is effective for glycaemic control, but more robust support for CVD risk reduction. Also cited was a commentary favouring the Mediterranean diet that was based on a non-systematic selection of articles. 31

| Additional evidence
We identified 12 other studies on the Mediterranean diet worthy of consideration: four RCTs, two RCT follow-up studies, and six systematic reviews with meta-analysis (Table S3). [32][33][34][35][36][37][38][39][40][41][42][43] One RCT found that this diet significantly improved HbA1c and body mass index in postmenopausal women with T2D, but the diet was not superior to usual care for improving blood pressure and lipids. 32 A 2-year RCT 36 comparing low-fat, low-carbohydrate and Mediterranean diets in obese people with T2D, with data available for 36 persons with T2D, found that the Mediterranean diet improved FBG, but not HbA1c levels, compared to a low-fat and low-carbohydrate diet. Two studies 33,34 followed up Esposito 2009, 25 which was included in the ADA-cited evidence ( Table 1). Both studies found longer times to medication requirement in the Mediterranean diet arm versus the low-fat diet arm, as well as increased partial remission and improved FBG and CVD risk markers. One of two smaller 12-week RCTs found a statistically significant HbA1c reduction favouring a Mediterranean diet over a typical diet; the other did not find a difference between the Mediterranean diet and a low-fat diet. 35,37 Neither of these trials resulted in between-group statistical significance for CVD risk factor markers including body mass index, blood pressure and lipids, but one found improvement in inflammation markers and flow-mediated dilation in the Mediterranean diet arm only. 35 Four systematic reviews with meta-analysis [38][39][40][41] and two with network meta-analysis 42,43 concluded that the Mediterranean diet is superior to other eating patterns for glycaemic control, weight loss, lipid profile, and reduced need for diabetes medication.

| Summary of evidence
The ADA-cited sources combined with additional ones identified through our search resulted in a total of seven RCTs, two follow-up RCT studies, and seven systematic reviews (including five with metaanalysis) that are appropriate for consideration in developing nutrition guidelines for T2D. Among the included trials are several large-scale studies, one with 3614 participants 26,27 and one with more than 200 participants. 25,33,34 Longer-term studies include one lasting 12 months, 28 one lasting 24 months, 36 and two lasting longer than 4 years. [25][26][27]33,34 As recommended by the ADA guidelines, we found that the Mediterranean eating pattern has demonstrated effectiveness in improving glycaemic control 25,28,[32][33][34][38][39][40][41][42][43] as well as CVD risk factors and even in reducing CVD events. 22,23,26,27,29,30,33,34,[38][39][40][41][42][43] This diet appears to be appropriately considered helpful for T2D management; its inclusion in the recommended eating patterns is warranted. However, questions remain about which components of the Mediterranean diet contribute to its effectiveness on all of these outcomes. Some studies suggest that it is the diet's more moderate carbohydrate content (<50% total energy intake) that accounts for reductions in weight and CVD risk, 44 while others suggest that the high monounsaturated fat content in the diet plays an important role in improving insulin sensitivity, glycaemic control, and inflammation. 45,46 Research in these areas will strengthen future nutritional recommendations and provide more in-depth guidance on how the Mediterranean diet can be used for T2D management.

| Additional evidence
We identified nine studies 40,56-62 not included in the ADA review, three of which were published after the 2018 standards (Table S4).
Three RCTs found reductions in HbA1c from baseline, [56][57][58] and two found the test diet superior compared to the control diet. 56,57 In these studies, the plant-based diets were compared to an energy-restricted diet, the recommended Korean Diabetes Association diet, and the participants' usual diet. However, in all three studies, a slight increase in triglycerides was observed in the intervention arms, with one study reporting a statistically significant change. 57 This study also reported significant decreases in weight, as well as in total, LDL and HDL cholesterol levels in the intervention arm. 57 A follow-up study 59  RCTs, five found a significant between-group advantage for the lowcarbohydrate arm for glycaemic control. 28,69,71,83,84 Of the eight that did not show a between-group glycaemic advantage, all but one found a reduction from baseline, and three had greater reductions in medication use. 73,74,82 Of the seven trials with a duration of ≥1 year, three showed sustained clinically significant improvements in HbA1c at 1 year, 28,69,82 and two showed sustained meaningful benefit at 2 years. 36,78 Another 1-year study found the low-carbohydrate diet resulted in decreased glucose variability, which has been found to be an independent CVD risk factor, making it an important overall consideration. 85 An isocaloric trial found the low-carbohydrate arm had a significant decrease in insulin and visceral fat accumulation compared to a high-carbohydrate arm. 70 Of the 10 studies that reported on lipids, five found significant improvements in triglycerides with a low-carbohydrate diet 74 (Table S5).  All 27 studies reported outcomes data for people with T2D and thus were appropriate for consideration in the development of nutritional recommendations for T2D management. Of the 10 RCTs, all of which reported on glycaemic control, nine found that a low-carbohydrate diet resulted in a significant change from baseline to end of study 86,87,89,90,[92][93][94]96 ; six also found a superior betweengroup reduction favouring the low-carbohydrate diet. 86,87,90,92,94,96 While some studies found that the control diet also improved glycaemic control significantly from baseline, none found the control diet superior to the low-carbohydrate diet. All 12 single-arm and nonrandomized trials found that a low-carbohydrate diet significantly improved glycaemic control from baseline to end of study; the two studies that made between-group comparisons found the lowcarbohydrate diet superior to the control diet. 99 98 Another longer trial also found sustained improvement in glycaemic control at 44 weeks. 103 All of these studies assessed HbA1c as the primary glycaemic marker, 86 and/or elimination of glycaemic control medications in the lowcarbohydrate arm. Five of six studies that conducted between-group comparisons of medication use found the low-carbohydrate diet to be superior, 86,88,89,91,99 and one study 92 found that both diets reduced usage significantly from baseline with no between-group difference.

| Additional evidence
No study found the control diet to be superior although there was some reduction in medication use from baseline in two of the studies in the control group. 88,90 Overall a favourable result was seen with regard to triglycerides and HDL cholesterol. No study found the control diet to be superior or that a low-carbohydrate diet significantly worsened triglycerides or HDL cholesterol. The additional evidence is mixed regarding the lowcarbohydrate diet's effects on LDL cholesterol. Eight studies found no significant change within group from baseline, [87][88][89]91,93,95,96,106 whereas five other studies found that the diet resulted in significant improvement 101,105,107 or showed superiority to a control diet. 97,98 In another study, the diet improved LDL cholesterol significantly in women but not in men. 102 Two studies found that the diet resulted in significant worsening from baseline. 99 HbA1c reduction compared to other eating patterns, but that a Mediterranean diet was superior for reduction of FBG. 43
Three of the four 2-year studies reporting on diabetes medication use found significant reductions with a low-carbohydrate diet compared to a control diet 88,91,97 ; this includes the one study that did not sustain HbA1c reduction at 2 years. 88 Evidence from 30 trials and 10 follow-up studies shows that a low-carbohydrate diet is an effective dietary approach for addressing dyslipidaemia. More than half of the studies that reported triglyceride levels found a significant improvement from baseline with a lowcarbohydrate diet; eight also showed superiority over a control diet. 28,71,84,91,96,99,101,103 Similarly, the evidence consistently showed significant improvements in HDL cholesterol with a low-carbohydrate diet, with 10 studies finding a significant increase over control diet. 28,70,71,73,88,89,91,99,101,103 It is also worth again noting that two 99

| SUMMARY
Treatment guidelines must be based on rigorous scientific standards that are consistently applied in order to ensure that guidelines are both reliable and credible. In reviewing the evidence cited in support of the ADA recommendations on eating patterns for T2D management, we found multiple reasons for concern. Although the ADA does provide a rubric for grading studies to include in its evidence review, not apparent in the 2018 or 2019 standards or the 2014 recommendations is a description of the process used to guide final selection decisions. Perhaps that is the source of the issues we find concerning; for example, studies were cited as evidence that by the ADA's own rubric were not A-rated sources or that were not conducted in people with T2D, were not clinical trials, or were not based on a systematic review of the evidence.
Our literature searches added considerably to the body of credible evidence worthy of consideration for a thorough review of the ADA recommendations on eating patterns. We found two additional studies to include on the DASH diet, 12 studies on the Mediterranean diet, nine on plant-based diets, and 27 on low-carbohydrate diets.
Almost all of these additional studies were published prior to the documents reviewed in the present paper.
We would like to note several things in the ADA documents that could be interpreted as evidence of bias, one of which is the inclusion of opinion pieces or reviews favouring the DASH, plant-based and Mediterranean eating patterns that were not based on a systematic approach to the literature. 22 vegan studies "did not consistently improve glycaemic control or CVD risk factors except when energy intake was restricted, and weight was lost." In contrast, the 2014 recommendations and both the 2018 and 2019 standards raise concerns about lack of sustainability with a lowcarbohydrate diet over the long term. While adherence is a common behaviour change problem, it is not unique to low-carbohydrate diets, and the long-term data on this approach are supportive.
Our review is based only on studies in which glycaemic control in people with T2D is an endpoint, because of its central importance to T2D management. The aim has been to produce a review and presentation (Tables S1 and S2) of a more complete body of evidence that is objective, fair and easily accessible to most readers and may prove useful in the creation of future iterations of the ADA guidelines.
Another section of the ADA guidelines on HbA1c target guidance was recently reviewed and assessed by the American College of Physicians when they issued new HbA1c target guidance. Using the Agree II instrument for evaluation, the American College gave a score of 3.7 out of 7 for the ADA guidelines, the second-lowest of six guidelines scored. Additionally, the ADA guidelines scored significantly lower than all others in "rigor of development." Table S6 provides our assessment of the ADA guidelines using the National Academies of Sciences, Engineering, and Medicine's Clinical Practice Guidelines We Can Trust evaluation method, along with recommended steps for improving the overall process. Additionally, another review evaluated the evidence for CVD prevention in the 2016 edition of the Standards of Care. 4 The prior two and current reviews of separate sections of the ADA guidelines all raise the same underlying concern regarding the rigour of the guideline development process. Given this, we believe our review is a critically important document that reinforces the need for a process change.

| CONCLUSION
In order to change the current global trajectory of T2D, it is imperative that health organizations be willing to invest resources in creating objective guidelines based on rigorous and unbiased scientific review.
Guidance from the ADA is valuable on many fronts; however, the present review of the current standards and recommendations, specifically on recommended eating patterns, finds significant shortcomings regarding scientific review methodologies, which are likely to translate to suboptimal clinical care decisions for people with T2D.

ACKNOWLEDGMENTS
We thank Dr James McCarter and Nina Teicholz for their edits, which greatly improved the manuscript.