The Weight Loss Effects of an LCHF Diet. A Statistical Meta-Analysis of Meta-Analyses

Content

Abstract

Background

Purpose of the Study

Study Design and Methods

Research Question

Literature Review

Macronutrient- and Micronutrient-Related Effects

Method Search

Data Analysis

Findings and Discussion

Effect Size Calculations

Limitations

Discussion of Findings with Respect to Theory and Empirical Research

Conclusion

Recommendations

References

Appendix A

Appendix B

Abstract

Given the existing problem of obesity, the purpose of this statistical meta-analysis was to measure the weight-loss effectiveness of LCHF diets. Using Cohen’s d as the measure, it was found that the mean effect size for LCHF diets was higher (M = 0.5333, SD = 0.29209) than the mean effect size for non-LCHF diets at an Alpha of 0.10, p = 0.058. Additionally, it was found that the mean effect size for LCHF diets (M = 0.5333, SD = 0.29209) in comparison to control (non-diet) groups (M = 0.0358, SD = 0.03470), p < 0.001 was also higher. These findings provide empirical support for the claim that LCHF diets are effective, both in isolation and in comparison to other dietary interventions.

Keywords: LCHF, low carbohydrate high fat, obesity, weight loss.

The Weight Loss Effects of an LCHF Diet: A Statistical Meta-Analysis of Meta-Analyses

Obesity has been described as a premier public health problem in developed countries, America in particular (Frederick, Snellman, & Putnam, 2014; Karp et al., 2014; Kay & McLaughlin, 2014; Zilanawala et al., 2015; Zukiewicz-Sobczak et al., 2014). Obesity has been demonstrated to play a causal role in the development of many ailments, including cardiovascular disease and diabetes. Public health authorities, policy-makers, scientists, non-profit and for-profit businesses around the world have made attempts to reduce the obesity problem by recommending modifications of diet to the members of their communities’. In the United States, awareness of the need to consciously manipulate diet to protect against obesity dates from the 19th century, when the increasing availability of calories and the decreasing need for physical labor indicated the beginning of a major shift in the body mass index (BMI) of Americans (Kaplan, Hill, Lancaster, & Hurtado, 2000). However, for well over a century, there has been continuous and contentious scholarly dispute over the effectiveness of specific diets (Zuk, 2013).

A low-carbohydrate, high-fat (LCHF) diet has been practiced in many settings. There is some evidence that an LCHF diet was the cornerstone of aboriginal diets in many regions (including Canada, Alaska, and Scandinavia) in which fat from animals was a more readily available source of calories (Fumagalli et al., 2015; Thorburn, Macia, & Mackay, 2014). In contemporary settings different iterations of the LCHF diet have been adopted by individuals, either on the basis of scientific recommendations or as a selective return to alleged ancestral eating patterns (Zuk, 2013).

Background

Despite the existence of empirical studies on the effectiveness of an LCHF diet, specifically the effectiveness of LCHF diets (in terms of weight loss) this is limited in regards to retrospective analyses. The absence of retrospective reviews of the effectiveness of LCHF diets could mean that additional evidence could be realized about the effectiveness of such diets, especially in comparison to other diets. The opportunity for performing this meta-analysis is to gain a better understanding of the effects of an LCHF diet on weight loss through the synthetization of prior LCHF research data.

Purpose of the Study

The purpose of this study is to measure the weight-loss effectiveness of LCHF diets by conducting a meta-analysis of meta-analyses, to test the strength of prior evidence gathered. This purpose will be achieved in two ways. First, the effect size of LCHF diets on weight loss will be calculated in isolation. Second, the effect size of LCHF diets on weight loss will be compared to the effect sizes of other kinds of diets on weight loss. The findings will thus offer two complementary measures, relative and absolute, of the weight-loss effectiveness of LCHF diets.

Study Design and Methods

Scholars have long called attention to the superior reliability of meta-analyses. The theory of statistical meta-analysis is based on the central claim that the pooled average of multiple effect sizes is superior to a single study’s reported effect size (Cochrane, 1972). One specific advantage of meta-analytical pooling is that effect sizes can be given different weights based on the quality of the studies from which they are derived; thus, for example, more weight can be ascribed to results based on the level of evidence they display (Campbell, 1988; Cochrane, 1972). LCHF diets and weight loss are particularly appropriate for statistical meta-analysis because of the existence of a large number of existing studies on this topic. Therefore, instead of attempting to generate new primary research on the relationship between LCHF diets and weight loss, it is more expedient to pool existing research to derive a measure of effect size. Thus a retrospective review and analysis of meta-analyses to identify effectiveness of LCHF diets was performed.

To proceed from information to knowledge, it is necessary, first of all, to define what the information is. In terms of the effects of LCHF on weight loss, there are many different results in the empirical literature and therefore a necessity for scholars to make sense of these results in a manner that can inform practice and scholarship. By sorting through existing information to reach a single, more reliable conclusion, statistical meta-analysis plays an important role in providing a transition from information to knowledge.

Research Question

A research question for a statistical meta-analysis ought to be based on a population, intervention, comparator, and outcomes (PICO) approach (Dickson & Cherry, 2014). In existing studies of diets and weight loss, there is substantial variation in PICO components. In terms of population, some studies sample children or adolescents (Pulgaron, E. R., & Delamater, A. M. (2014); whereas, others sample adults (Soltani, S., Shirani, F., Chitsazi, M. J., & Salehi‐Abargouei, A. (2016). In terms of the intervention, there is variability in terms of (a) how the diets effects are measured and (b) how the diets are defined. In some studies (Peters, H. P., Bouwens, E. C., Schuring, E. A., Haddeman, E., Velikov, K. P., & Melnikov, S. M. (2014), only the effect of one kind of diet is measured; in other studies (Tonstad, S., Malik, N., & Haddad, E. (2014), the effects of several different kinds of diets are compared. In addition, researchers have defined diets in different ways, for example, in terms of calories below maintenance or the macronutrient composition percentages of LCHF diets. Weight loss is a common outcome measurement, but so is change in body fat percentage and change in body mass index (BMI).

The research question for this study is a synthesis of questions asked by previous studies. Specifically, this study asks: “what is the effect of an LCHF diet on weight loss among non-hospitalized American adults as measured after a minimum of 6 months?”

Literature Review

Physiology does not provide a complete theoretical account of the relationship between diets and weight loss, for the reason that different dietary approaches affect calorie consumption in other ways. While physiology provides the underlying theoretical foundation—often abbreviated CICO, or Calories In, Calories Out—quantifying the relationship between calorie consumption and weight loss, physiology alone does not explain why, for instance, some diets have a higher rate of compliance than that of others (Chambers, McCrickerd, & Yeomans, 2015; Faith, Heo, Kral, & Sherry, 2013; Halford & Harrold, 2012; Poulsen et al., 2014; Westerterp-Plantenga, Lemmens, & Westerterp, 2012). More specific biological theories must therefore be brought forward to investigate LCHF diets, or, indeed, any diet (Cooper, Stevenson, & Paton, 2016; Gibbons et al., 2013; Gluck, Yahav, Hashim, & Geliebter, 2014; Hill, Rolls, Roe, De Souza, & Williams, 2013; Jaremka et al., 2015; Sarker, Franks, & Caffrey, 2013).

Macronutrient- and Micronutrient-Related Effects

Switching to an LCHF diet has other effects that can also have an influence on the trajectory of weight loss. One possible influence involves the role of the macronutrient of protein. If an LCHF diet is adopted as part of a purposive weight loss plan, then it is highly likely that some of the missing carbohydrate calories will be replaced by protein (Westerterp-Plantenga et al., 2012). As a macronutrient, protein has a more satiating effect than either carbohydrates or fat. In addition, protein has a higher metabolic cost, that is, a higher percentage of calories expended in the attempt to digest and process protein. Some estimates suggest that only 80-85% of protein calories are stored, with the remaining 10-15% of protein calories expended in processing (Mitchell et al., 2015; Schneider, 2013). By contrast, sugar and simple carbohydrates are processed remarkably efficiently and quickly. Thus, isocaloric protein-only and carbohydrate-only meals would have different impacts on body composition, as perhaps 80-85 calories of the protein meal, but 95%-97% calories of the carbohydrate meal, would be absorbed by the body.

Thus, one of the benefits of an LCHF diet is the transference of formerly carbohydrate calories into protein calories. Assuming that an LCHF diet is also a higher-protein diet, each gram of protein that replaces each gram of carbohydrate would result in roughly 0.5 fewer net calories absorbed by the body (Mitchell et al., 2015; Schneider, 2013). If a dieter on an LCHF template were to replace 100 grams of carbohydrate with 100 grams of protein, but keep his or her diet isocaloric, he or she would thereby store 50 fewer calories a day—a trivial amount, but one that, over a year, represents 5.2 pounds of bodyweight.

The inefficient nature of protein digestion is a boon to the LCHF dieter who replaces carbohydrates with protein. However, given protein’s satiety effects (Westerterp-Plantenga et al., 2012), the inevitable increase of protein following a reduction in carbohydrate is likely to reduce calorie consumption even further. Thus, one intriguing mechanism for the weight loss effectiveness of an LCHF could be such a diet’s subsequent association with increased protein consumption.

Reducing carbohydrates could also benefit weight loss through the mechanism of nutritional sufficiency. People with high-carbohydrate diets, particularly as consumed in the United States, include substantial amount of simple carbohydrates in their diets. These simple carbohydrates are nutritionally empty. Therefore, no matter how many such grams of carbohydrate are consumed, the body continues to experience a depletion of certain micronutrients and vitamins, which in turn triggers hunger signals. This phenomenon has been described as the intersection of obesity and malnutrition (Horvath, de Castro, Kops, Malinoski, & Friedman, 2014; Panagopoulou, Fotoulaki, Nikolaou, & Nousia‐Arvanitakis, 2014; Prentice, 2006; Tanumihardjo et al., 2007; Wells, 2012, 2013; Zapatero et al., 2013). As obese individuals might have low quantities of certain micronutrients and vitamins, their brains might continue to send out hunger signals in the hope a body will consume these micronutrients and vitamins (Horvath et al., 2014; Panagopoulou et al., 2014; Prentice, 2006; Tanumihardjo et al., 2007; Wells, 2012, 2013; Zapatero et al., 2013).

Among people who adopt an LCHF diet, one natural outcome is the reduction of calories from simple carbohydrates, which, in theory, would prime the body to seek out and consume micronutrients and vitamins from other sources, such as complex carbohydrates. Because complex carbohydrates are lower in calories than isovoluminous amounts of simple carbohydrates, it is likely that they will be featured more prominently in LCHF diets. Once individuals who substitute complex carbohydrates for simple carbohydrates, they are likely to consume more necessary micronutrients and vitamins, which, in turn, will improve satiety. Moreover, the satiety effect of the switch to a greater volume of complex carbohydrates is complemented by the satiety of a higher-protein diet. Thus, there are likely to be positive and complementary effects of micronutrient and macronutrient profile changes brought on by an LCHF diet.

One fascinating aspect of nutritional science is the stark divergence between the laws of physiology that underlie CICO and the immense variability in how humans respond to diets. While all diets will yield weight loss if they follow the CICO principle (Crino et al., 2015; Hopkins et al., 2016; Phillips, 2014; Thomas et al., 2014; Thomas et al., 2013; Weijenberg et al., 2013; Wells, 2013), not all diets are as easy to maintain. The review of findings from the literature indicated that there are three main mechanisms by which LCHF diets might be more readily associated with fat loss. First, LCHF diets might promote satiety through the added consumption of animal fats and fiber from complex carbohydrates (Berti et al., 2015; Fallaize et al., 2013; Gibbons et al., 2016). Second, LCHF diets might promote a switch to a fat-burning metabolism associated with reduced water and glycogen weight as well as with the depletion of stored body fat (Berti et al., 2015; Fallaize et al., 2013; Gibbons et al., 2016). Third, LCHF diets might result in an increased consumption of protein and also micronutrients from complex carbohydrates, both of which could help to decrease hunger signals and thereby promote weight loss or weight management (Horvath et al., 2014; Panagopoulou et al., 2014; Prentice, 2006; Tanumihardjo et al., 2007; Wells, 2012, 2013; Zapatero et al., 2013). Cumulatively, these reasons suggest that LCHF diets have multiple, complementary ways to promote a long-term energy imbalance as a result of which stored body fat can be metabolized.

Method Search

A search is a preliminary search of the literature on a meta-analytical topic (Dickson & Cherry, 2014; Dundar & Fleeman, 2014; Torgerson, 2004). Scoping searches are followed by a more detailed description of how identified studies were culled for purposes of meta-analysis (Dickson & Cherry, 2014; Dundar & Fleeman, 2014; Torgerson, 2004). In this section, both the search and the more detailed study selection processes have been discussed.

The search was conducted on MEDLINEEMBASE, and the Cochrane Library Database of Clinical Trials. The following provides a couple examples of the search strings used: “weight loss” AND “low-carbohydrate, high fat” and “fat loss” AND “low-carbohydrate, high fat”. The complete list of search strings that were used can be found in Appendix A.

These results were used to identify the initial studies loaded into the PRISMA flow Diagram. The PRISMA flow diagram presented in Figure 1 outlines the entire process of study selection for the statistical meta-analysis, beginning from the records identified through the scoping search and continuing through to the selection of the 12 studies included in the meta-analysis. It should be noted that the inclusion of 10-15 high-quality and relevant sources is considered acceptable in statistical meta-analysis (Dickson & Cherry, 2014; Dundar & Fleeman, 2014; Torgerson, 2004).

In the PRISMA diagram located in Appendix B, it should be noted that an effort was made to identify relevant studies that were not meta-analyses. This effort was undertaken because the number of recent meta-analyses of LCHF effects on weight loss appeared to be somewhat low. Ultimately, only three non-meta-analyses were included in the meta-analysis. The inclusion of these non-meta-analytical studies raised the quality of the meta-analysis, as, without them, there would have been only nine studies included.

Data Analysis

Perhaps the common element of statistical meta-analyses is the testing of effect size (McGraw & Wong, 1992; Rosenthal, 1994; Thalheimer & Cook, 2002). However, there are many ways to calculate effect size. Even the effect sizes reported in a study are likely to represent one of several alternate ways in which effect sizes can be calculated (Thalheimer & Cook, 2002). One common measure of effect size is Cohen’s d. Cohen’s d has the advantage of being widely reported, calculable from many common descriptive statistics, including some combination of F statistics, the number of subjects, means, standard errors and standard deviations.

Table 1 captures the basic information (Cohen, 2013) needed to calculate Cohen’s d.

Abbildung in dieser Leseprobe nicht enthalten

The existence of two groups in Cohen’s d means that such an approach cannot, without complex modifications, be used to compare effect sizes from more than two forms of dieting. The two-group structure of Cohen’s d means that it is well-suited to comparing weight loss in a within-subjects model in which the two groups can be treated as matched pairs, with the ‘before’ state representing someone before embarking on a LCHF diet and the after state representing the same individual after a LCHF diet. Cohen’s d can also be used to compare effect sizes across LCHF and non-LCHF groups. Table 2 offers an idea of how Cohen’s d can be derived when making comparisons within an all-LCHF intervention; Table 3 demonstrates how Cohen’s d can be calculated for LCHF and non-LCHF groups.

Table 2

Abbildung in dieser Leseprobe nicht enthalten

To calculate Cohen’s d, the data from Table 2 would be used to calculate a mean, standard deviation, and sample size for each column. This information could then be placed into an automated Cohen’s d calculator (IDP, 2016) to calculate the effect size. The resulting effect size would be a control-intervention-based estimate of the effectiveness of an LCHF diet in the absence of other diets types to compare. This structure can be easily modified to accommodate studies in which LCHF and non-LCHF diet effects were provided. Such a structure has been provided in Table 3 below.

Table 3

Abbildung in dieser Leseprobe nicht enthalten

Using the approach in Table 3, Group 1 would consist of all individuals with the LCHF diet as an intervention; whereas, Group 2 would consist of all individuals with a diet other than the LCHF diet as an intervention. In such an approach, the means, standard deviations, and sample sizes would be calculated separately for Groups 1 and 2, and the results would be used to calculate Cohen’s d. Note that this approach can be modified to calculate the Cohen’s d of a study in which some individuals are assigned to an LCHF diet group and other individuals comprise a control group.

Cohen’s d was chosen as the measurement of effect size for this statistical meta-analysis. Cohen’s d was chosen because it is both common and simple to calculate, on the basis of data structures such as those presented in Tables 1, 2, and 3 above. However, it should be noted that Cohen’s d has some important limitations. In particular, Cohen’s d is not well-suited to calculate effect sizes for studies in which there are two groups (such as an LCHF and non-LCHF group) and two experimental states (such as a control and an intervention group). The effect size measurement of Morris’s d was created specifically for this situation). The absence of Morris’s d and other more advanced forms of effect size calculation can be considered one of the limitations of the present study. However, given that many of the studies discussed and used as the basis of analysis, do not use advanced designs, the absence of Morris d calculations might not necessarily be a limitation of this study.

Having settled on Cohen’s d as a measure of effect size, it is also necessary to discuss possible approaches to pooling. The simplest of all approaches is simply to average the Cohen’s d sizes and to use the result as an estimate (Cohen, 2013). In more complex approaches (Cohen, 2013), effect sizes from different studies are accorded different weights based on factors such as the quality of the study, which could include level of evidence, participant size, and other factors. In this study, simple weighted averaging was used.

It should also be noted that all but three of the 12 studies included in this meta-analysis were meta-analyses. All meta-analyses report effect sizes. However, some meta-analyses do not report the Cohen’s d effect size. For this reason, it was necessary to calculate Cohen’s d based on the descriptive (and, in some cases, inferential) statistics reported in the meta-analyses and, where applicable, in the primary research articles.

[...]