Pharmacogenetic interactions of medications administered for weight loss in adults: a systematic review and meta-analysis

Abstract

Aim

To analyze roles of single nucleotide variants (SNVs) on weight loss with US FDA-approved medications.

Materials & methods

We searched the literature up until November 2022. Preferred Reporting Items for Systematic reviews and Meta-Analyses guidelines were followed.

Results

14 studies were included in qualitative analysis and seven in meta-analysis. SNVs in CNR1, GLP-1R, MC4R, TCF7L2, CTRB1/2, ADIPOQ, SORCS1 and ANKK1 were evaluated relative to weight loss with glucagon-like peptide-1 agonists (13 studies) or naltrexone–bupropion (one study). CNR1 gene (rs1049353), GLP-1R gene (rs6923761, rs10305420), TCF7L2 gene (rs7903146) were associated with weight loss in at least one study involving glucagon-like peptide-1 agonist(s). The meta-analysis did not identify any consistent effect of SNVs.

Conclusion

Pharmacogenetic interactions for exenatide, liraglutide, naltrexone–bupropion and weight loss were identified, but the directionality was inconsistent.

The prevalence of obesity has been steadily increasing [1,2] and reached 42% between 2017 and 2020 in the USA [1]. Overweight and obesity are often comorbid with other noncommunicable conditions including Type 2 diabetes, hypertension or coronary artery disease and are associated with an increased risk of all-cause and cardiovascular mortality [3,4]. Consequently, obesity is now regarded as a public health concern that is accompanied by a high social and economic burden [5]. In 2016, the estimated direct and indirect costs of obesity and overweight people in the USA was US$1.72 trillion and they accounted for 47.1% of chronic diseases-related expenditures [6]. The management of obesity includes identification of possible etiologies, determination of obesity phenotype [7], screening for comorbid conditions, lifestyle modifications, pharmacological interventions as well as bariatric procedures, including surgery [8].

Pharmacological agents are efficacious [9,10] and are increasingly used for the management of obesity and its complications. The drug classes that have been approved by the US FDA for the long-term management of obesity in the general population are glucagon-like peptide-1 receptor agonists (GLP-1 RAs) (semaglutide and liraglutide), naltrexone–bupropion, phentermine–topiramate and orlistat [11]. A network meta-analysis demonstrated that the most efficacious GLP-1 RA agents are subcutaneous weekly semaglutide or daily liraglutide [10]. While GLP-1 RAs were initially approved for the treatment of Type 2 diabetes, their effects on weight loss are thought to be secondary to both increased satiety mediated centrally and delayed gastric emptying [12]. Naltrexone–bupropion stimulates pro-opiomelanocortin neurons and reduces food intake at the level of the midbrain [13]. The combination of the sympathomimetic phentermine and anticonvulsant topiramate suppresses appetite and enhances satiety [14]. Orlistat inhibits pancreatic lipase, leading to fat malabsorption [15]. In addition, setmelanotide, a melanocortin-4 receptor agonist, has been approved for the treatment of obesity in patients aged 6 years or older who have specific gene mutations in pro-opiomelanocortin, proprotein convertase subtilisin/kexin type 1 or the leptin receptor as well as Bardet–Biedl syndrome [16,17].

Despite significant advances in the management of obesity, utilization of antiobesity medications remains suboptimal with barriers including a variable response to the medications, the high cost and unsatisfactory insurance coverage [18,19].

There has been increased interest in identifying predictors of response to weight loss medication, namely GLP-1 RA. For example, a recent study showed that significant slowing of gastric emptying of solids and weight loss > 1 kg at 5 weeks were significant predictors of weight loss > 4 kg at 16 weeks in patients with obesity who were treated with liraglutide [20]. Hence, given that a significant proportion of patients fail to achieve therapeutic targets, it is conceivable that precision medicine could potentially maximize the benefit of specific interventions administered to patients with a certain genetic profile [21].

This study addresses the hypothesis that genetic variation significantly impacts the weight loss in response to pharmacological treatment of obesity. Pharmacogenomics studies the influence of single nucleotide variants (SNVs) on the effectiveness of medications. The genes of interest are most often related the medication's pharmacokinetics or pharmacodynamics [22]. Thus, our objective was to analyze the published literature on the potential role of SNV on the pharmacological effects of agents used in the treatment of obesity, notably GLP-1 RA, naltrexone–bupropion, phentermine–topiramate and orlistat.

Materials & methods

We conducted a systematic review and meta-analysis in adherence with the Preferred Reporting Items for Systematic reviews and Meta-Analyses 2020 checklist [23].

Search methodology

A comprehensive search of several databases from each database's inception to 15 November 2022, with no language limitation, was conducted by a medical librarian (LJ Prokop). The databases included Ovid MEDLINE(R) and Epub Ahead of Print, In-Process and Other Non-Indexed Citations, and Daily; Ovid EMBASE, Ovid Cochrane Central Register of Controlled Trials, Ovid Cochrane Database of Systematic Reviews and Scopus. The search strategy was designed and conducted by an experienced librarian with input from the study's investigators. We used controlled vocabulary supplemented with keywords to search for studies that evaluated the impact of SNVs on weight change in adult patients receiving medications approved for the treatment of obesity. The actual strategy listing all search terms used and how they are combined is available in the Supplementary File.

Inclusion criteria

We included observational or interventional studies that evaluated the effect of SNVs on the weight loss efficacy of FDA-approved weight loss medications, namely naltrexone–bupropion, phentermine–topiramate, orlistat and several GLP-1 RAs (albiglutide, dulaglutide, exenatide, semaglutide, liraglutide, lixisenatide). We only included medications that are used to treat obesity that is not secondary to specific rare genetic conditions; thus, our study excluded setmelanotide which is approved for chronic weight management in adult and pediatric patients 6 years of age and older with monogenic or syndromic obesity due to Bardet–Biedl syndrome. We considered studies conducted in adults (> 18 years) with or without obesity. Studies were required to have documented comparison of post-treatment or change in weight, BMI or body fat percent in individuals with and without SNVs. As we aimed to explore gene variation by pharmacotherapy interactions in different patient populations, and we did not limit the review by duration of treatment, comorbidities or concomitant weight or glucose reduction treatments. We excluded case reports and case series.

After the removal of duplicate references, two investigators (K Vosoughi, J BouSaba) independently reviewed each title, abstract and full-text article based on prespecified eligibility criteria. A citation was advanced to full-text review if one of the investigators included it. In cases of disagreements between the two primary reviewers, adjudication was conducted by a third reviewer (M Camilleri) who was blinded to the comments from other reviewers.

Data extraction

Two investigators (K Vosoughi, J BouSaba) independently extracted data using a piloted data extraction form. Information on study setting, baseline characteristics (number of patients, BMI, comorbidities, concomitant treatment) and study outcome (change in weight or BMI) were included and discrepancies were resolved by discussion in pairs or referral to the wider reviewer team of four investigators. Authors were contacted for missing data or when clarification was required (e.g., when central tendency and/or dispersion statistics were not reported).

Meta-analysis

All SNVs studied in at least three studies were included in the quantitative meta-analysis. Although all included studies compared weight loss between those with genetic variants and reference alleles, the literature evaluation demonstrated that outcomes were presented using varying scales. Specifically, studies presented the following outcomes: absolute weight loss, percentage weight change, likelihood of having a > 5% weight change. In view of the heterogeneity of outcomes, the effects of different treatments were standardized using previously described methods, as follows [24]. For studies utilizing a continuous outcome measure (e.g., absolute weight loss), standardized mean difference (SMD) in weight loss between those with allelic variant compared with the homozygous reference alleles were computed as Hedges' g by calculating the difference in means between the groups and dividing by the pooled standard deviations. A Hedges' g, as opposed to a Cohen's d, was calculated to correct for small sample sizes seen in some studies. For binary outcomes of clinically significant weight loss estimated as an odds ratio, we conducted a linear transformation of the natural log of the odds ratio to a SMD. Pooled effects were grouped by variant and estimated using inverse variance weighting with a random effects model. All calculations of SMD were conducted in Excel (Microsoft^®) and pooled estimates for meta-analyses were conducted in RevMan 5 (Review Manager [RevMan], Version 5.4, Cochrane Collaboration, 2020) [25].

Results

Among 515 citations retrieved by the electronic and manual searches, 14 studies fulfilled inclusion criteria (Figure 1). These included two placebo-controlled randomized clinical trials (RCTs) [26,27], one post hoc analysis of prior RCTs [28] and 11 observational studies [29–39].

Figure 1. . Flow chart summarizing study retrieval and identification.

The included studies evaluated pharmacogenetic interactions in patients with obesity (four studies) [26–29,35], patients with Type 2 diabetes who were overweight or obese (eight studies) [29–31,33,34,36–39], and patients with polycystic ovary syndrome (PCOS) and obesity (one study) [32]. The interaction of genetic variants with the weight loss effect of GLP-1 RAs and naltrexone–bupropion was reported in 13 studies and one study, respectively. No study of the other FDA-approved medications fulfilled the inclusion criteria. Description of the included studies and genetic interactions with weight loss medications are summarized in Table 1 & Figure 2 and detailed in the next section.

Table 1. . Studies evaluating the association of single nucleotide variants with response to weight-lowering medications.

Gene	Genetic variant(s)	Treatment and duration, allowed CT	Study population and distribution by genotype or allele	Main outcome	Ref.
Glucagon-like peptide-1 receptor agonist
GLP-1R	rs6923761 G > A	Liraglutide 1.8 mg q.d. for 14 weeks, NA	90 patients with Type 2 diabetes and overweight (BMI > 25 kg/m2) GG = 51, GA-AA = 39	Carriers of A allele achieved significantly greater weight loss in response to treatment (-4.52 ± 4.6 kg in the GA or AA group compared with -2.78 ± 2.8 kg in the GG group; p < 0.05)	[31]
GLP-1R	rs6923761 G > A	Liraglutide 1.2 mg q.d. for 12 weeks, CT not allowed	57 premenopausal females with polycystic ovary syndrome and obesity (BMI ≥30 kg/m2) GG = 27, GA-AA = 30	Carriers of A allele showed marginally greater weight loss compared with patients with GG genotype (OR of 5% weight change from baseline: 3.06; 95% CI: 0.96–9.74; p = 0.058)	[32]
GLP-1R	rs6923761 G > A	Liraglutide 3 mg q.d. for 16 weeks, CT not allowed	59 patients with obesity (BMI ≥30 kg/m2) GG = 31, GA-AA = 28	No significant association between genetic variants and change in weight (-5.6 ± 3.2 kg in the GG group vs - 6.1 ± 4.6 kg in the AA-AG genotype group; p > 0.05)‡	[27]
GLP-1R	rs6923761 G > A	Exenatide 5 μg b.i.d. for 4 weeks, CT not allowed	Ten patients with obesity class I and II (BMI 30–40 kg/m2) and accelerated gastric emptying (T1/2 <79 min or gastric emptying 1 h > 35%) GG = 4, AA-AG = 6	No significant association between genetic variants and change in weight (-1.9 ± 0.84 kg change in GG vs -0.9 ± 1.25 kg in AA-AG; p > 0.05)‡	[28]
GLP-1R	rs10305420 C > T	Liraglutide 1.2 mg q.d. for 12 weeks, CT not allowed	57 premenopausal females with polycystic ovary syndrome and obesity (BMI ≥30 kg/m2) CC = 25, CT-TT = 31	Carriers of T allele showed higher proportion of ≥5% weight change (OR: 0.27; 95% CI: 0.09–0.85; p = 0.025)	[32]
GLP-1R	rs10305420 C > T	Exenatide 20 μg q.d. or liraglutide 1.2 mg q.d. for 12 weeks, CT not allowed	156 patients with Type 2 diabetes (exenatide and liraglutide combined) CC = 106, CT-TT = 50	No significant effect on change in BMI (-1.6 ± 1.7 kg/m2 in CC group vs -1.7 ± 2.1 kg/m2 in the CT-TT; p = 0.271)	[29]
GLP-1R	rs10305420 C > T	Exenatide 5 μg b.i.d.† for 26 weeks, CT not allowed	285 patients with poorly controlled Type 2 diabetes (HbA1c 8.5–12%) and obesity (BMI > 30 kg/m2) T allele = 13.5% (single nucleotide variant in Hardy–Weinberg equilibrium in the sample)	Carriers of T allele achieved significantly greater weight loss in response to treatment (-1.27 kg; 95% CI: [-2.2, -0.3]; p = 0.009)	[34]
GLP-1R	rs3765467 G > A	Exenatide 20 μg q.d. or liraglutide 1.2 mg q.d. for 12 weeks, CT not allowed	156 patients with Type 2 diabetes (exenatide and liraglutide combined) GG = 114, GA-AA = 42	No significant effect on change in BMI (-1.7 ± 2.2 kg/m2 in GG group compared with -1.4 ± 1.5 in the GA-AA group; p = 0.187)	[29]
GLP-1R	rs3765467 G > T	Exenatide 5 μg b.i.d.† for 26 weeks, CT not allowed	285 patients with poorly controlled Type 2 diabetes (HbA1c 8.5%–12%) and obesity (BMI > 30 kg/m2) T allele = 22.8% (single nucleotide variant in Hardy–Weinberg equilibrium in the sample)	No significant association between genetic variant and change in weight (OR of weight loss among T allele carriers: 0.2; p = 0.58)	[34]
TCF7L2	rs7903146 C > T	Liraglutide 3 mg q.d. for 16 weeks, CT not allowed	59 patients with obesity (BMI ≥30 kg/m2) treated with liraglutide CC = 39, CT-TT = 20	No significant association between genetic variants and change in weight (-5.91 ± 4.06 kg in the CC group vs -5.81 ± 3.94 kg in the CT-TT group)‡	[27]
TCF7L2	rs7903146 C > T	Exenatide 5 μg b.i.d. for 4 weeks, CT not allowed	Ten patients with obesity class I and II (BMI 30–40 kg/m2) and accelerated gastric emptying (T1–2 <79 min or GE 1 h > 35%) CC = 3, TT-CT = 7	Carriers of T minor allele achieved a statistically significant greater weight loss (-1.65 ± 1.03 kg in CT-TT group vs -0.47 ± 0.66 kg in CC group; p < 0.05)‡	[28]
TCF7L2	rs7903146 C > T	Exenatide 10 μg b.i.d. for 8 weeks, insulin or DPP-4 inhibitors not allowed	56 patients with Type 2 diabetes and BMI 25–35 kg/m2 CC = 26, CT-TT = 30	No significant association between genetic variants and change in weight (-2.16 ± 2.2 kg in CC group vs -2.49 ± 6.8 kg in CT-TT group; p > 0.5)	[33]
CTRB1-2	rs7202877 T > G	Liraglutide 1.2 mg or 1.8 mg q.d. for 26 weeks, CT not allowed	116 patients with Type 2 diabetes TT = 97, TG-GG = 19	No significant association between genetic variants and change in weight (OR of losing >3% of baseline weight for TT vs TG/GG: 1.12; 95% CI: 0.4–3.2; p = 0.84)	[36]
CNR1	rs1049353 G > A	Liraglutide 1.8 mg q.d. for 14 weeks, all patients received metformin ± sulfonylurea	86 patients with obesity (BMI > 30 kg/m2) and poorly controlled Type 2 diabetes (HbA1c > 7%) GG = 51, GA-AA = 35	No significant association between genetic variants and percentage of weight change from baseline (3.2% ± 10.5% weight change in the GA-AA vs 3.4% ± 3.7% in the GG genotype group; p > 0.05)	[31]
CNR1	rs1049353 G > A	Liraglutide 0.6 mg q.d. then 1.2 mg q.d. for 16 weeks, CT not allowed	230 patients with Type 2 diabetes, BMI 23–32 kg/m2. GG = 188, GA-AA = 42	Statistically significant difference in change in BMI between patients with and without minor A allele (3.1 kg/m2 for GG vs 3.4 kg/m2 for GA-AA; p = 0.026)	[37]
ADIPOQ	rs2241766 T > G	Liraglutide 0.6 mg q.d. then 1.2 mg q.d. mg q.d. for 14 weeks, NA	95 patients with Type 2 diabetes and BMI > 24 kg/m2 TT = 46, TG-GG = 49	No significant effect on change in BMI (4.2 kg/m2 in TT and 4 kg/m2 in TG-GG groups; p > 0.05)	[38]
SORCS1	rs1416406 A > G	Exenatide 10 mcg b.i.d. for 4 weeks then 10 mcg b.i.d. for 48 weeks, NA	101 patients with newly diagnosed Type 2 diabetes and BMI 20–35 kg/m2 AA = 39, GG-GA = 62 (GG = 7, GA = 55)	No significant effect on change in BMI (-1.4 ± -1.1 kg/m2 in the GG group, -1.4 ± - 1.5 kg/m2 in the GA group and -1.1 ± -1.2 kg/m2 in the AA group)	[39]
MC4R	rs13447324 105 C > A rs13447325 110 A > T rs13447333 542 G > A rs121913567 656 C > T rs13447329 335 C > T	Liraglutide 3 mg q.d. for 16 weeks, CT not allowed	42 patients with obesity (BMI > 28 kg/m2) Pathogenic MC4R mutations = 14, matched control = 28	No significant difference in weight change between pathogenic MC4R and control group (-6.8 ± 6.73 kg in MC4R mutation group vs -6.1 ± 6.35 kg in control group; p = 0.9)	[26]
Naltrexone–bupropion
ANKK1	rs1800497 G > A	Naltrexone–bupropion 32 mg–360 mg q.d. for 12–16 weeks (mean follow-up time 14.7 weeks), CT not allowed	33 patients with BMI 27–45 kg/m2 GG = 18, AG-AA = 15	No significant association between genetic variants and change in weight (percent of weight loss: 5.9% ± 3.2 for AA-AG vs 4.2% ± 4.2 for GG; p = 0.20)	[35]

Data reported are mean ± standard deviation, unless stated otherwise.

^†The dose was increased to 10 μg b.i.d. in those with an HbA1c reduction of less than 0.5% after 2 months.

^‡The amount of weight change in each group or difference in weight change was not available but data were obtained.

b.i.d.: Twice a day; CT: Concomitant treatment; HbA1c: Hemoglobin A1c; q.d.: Once a day; MC4R: Melanocortin-4 receptor; NA: Not available; OR: Odds ratio.

Figure 2. . Flow diagram highlighting the interaction between various SNVs and weight-lowering medications.

Each small square represents a different study in the literature, and the outcome of interest is weight loss or change in BMI, expressed as significant effect of the variant or reference allele (shown as ‘in favor’ in the figure). Some studies evaluated the effects of different SNVs and are included multiple times.

*Evaluated MC4R mutations: rs13447324 105 C > A, rs13447325 110 A > T, rs13447333 542 G > A, rs121913567 656 C > T, rs13447329 335 C > T.

SNV: Single nucleotide variant.

GLP-1 receptor agonists

Gene polymorphisms involved in the pharmacogenetic studies of liraglutide and exenatide are detailed in the following sections.

SNVs in the GLP-1R gene

rs6923761

Four studies, with a total of 216 patients, evaluated the interaction of GLP-1 RAs and rs6923761 SNV (G > A), with inconsistent results which are summarized here.

A prospective single-arm study of 90 patients with Type 2 diabetes showed greater weight loss in patients with minor allele A after treatment with liraglutide [31]. Another prospective study of females with PCOS and obesity treated with liraglutide also found an association between A allele carrier status and greater weight loss, with borderline statistical significance (p = 0.058) [32]. However, in a subanalysis of 59 participants with obesity treated with liraglutide in a prior RCT, the minor A allele was not associated with a significant change in weight [27].

In contrast, in a post hoc analysis of a small RCT of ten patients treated with exenatide, patients with GG genotype showed a greater weight loss after treatment compared with patients with AA-AG genotypes (A allele carriers) [28].

rs10305420

Results on response to treatment with GLP-1 RAs were also inconsistent with rs10305420 SNV (C > T). A study of 57 females with PCOS and obesity showed that patients with minor T allele carriers achieved a higher proportion of ≥5% weight loss in response to liraglutide [32]. In a more extensive study of 285 patients with poorly controlled diabetes treated with exenatide, individuals with CC genotype achieved a mean of 1.27 kg greater weight loss compared with carriers of T allele (CT-TT genotypes) [34]. However, in a study of 156 patients with diabetes treated with either liraglutide or exenatide, minor allele T was not associated with change in the BMI after treatment [29].

rs3765467

In two studies that evaluated the effect of exenatide (n = 285) [34] and exenatide or liraglutide (n = 156) [29] in patients with Type 2 diabetes, there was no differential weight loss between patients with different SNV for rs3765467 (G > A).

SNVs in the TCF7L2 gene

The effect of rs7903146 (C > T), a TCF7L2-related SNV, was tested in three studies. In a post hoc analysis of prior RCTs, patients on exenatide who carried the CT-TT genotype (T allele carriers) had greater weight loss compared with patients with the CC genotype [28]. However, two other studies of liraglutide [27] and exenatide [33] did not show any association between TCF7L2 gene rs7903146 SNV and differential weight loss based on different alleles.

SNVs in other genes

A study of 116 patients with Type 2 diabetes showed that CTRB1-2 gene rs7202877 (T > G) SNV was associated with a 12% higher rate of ≥3% weight loss compared with baseline in response to liraglutide, without reaching statistical significance [36].

Two observational studies investigated the interaction between CNR1 gene rs1049353 (G > A) and weight loss in patients on liraglutide. In one study with 230 patients, patients with A minor allele (GA-AA) had significantly lower BMI at the end of the study compared with patients with GG SNV [37]. Another study with 86 patients did not identify any differential weight loss between patients with or without the minor A allele [30].

In a study with 95 patients on liraglutide, there was not a significant change in BMI between patients with and without the minor G allele for ADIPOQ rs2241766 T > G [38].

In another study with 101 patients on exenatide, there were no significant differences in the change in BMI between patients with and without the minor G allele for SORCS1 rs1416406 [39].

In a placebo-controlled trial with 42 patients on liraglutide, there was no significant weight change between patients with and without pathogenic mutations in the MC4R gene (rs13447324 105 C > A, rs13447325 110 A > T, rs13447333 542 G > A, rs121913567 656 C > T, rs13447329 335 C > T) [26].

Naltrexone–bupropion

In a retrospective analysis of 33 individuals who were treated with naltrexone–bupropion, ANKK1 gene SNV rs1800497 G > A gene was not associated with a significantly greater weight loss in response to treatment [35].

Meta-analysis

Three SNVs met criteria for inclusion in the meta-analysis, specifically at least three studies documented in the literature: GLP-1R rs6923761 and rs10305420 and TCF7L2 rs7903146. The meta-analysis (Figure 3) did not identify any significant differential weight loss between allele carriers pertaining to these three SNVs. Pooled estimates were as follows: GLP-1R gene rs6923761 (SMD: 0.09 [-0.62, 0.80]; p = 0.81; I² = 72%), GLP-1R gene rs10305420 (SMD: -0.22 [-0.74, -0.29]; p = 0.39; I² = 57%) or TCF7L2 gene rs7903146 (SMD: 0.52 [-0.33, 1.36]; p = 0.23; I² = 57%). Analyses suggested a moderate to high level of heterogeneity between the studies.

Figure 3. . Forest plot evaluating the association of glucagon-like peptide-1 agonists with single nucleotide variants in the GLP-1R gene (rs692376, rs10305420) and TCF7L2 gene (rs7903146).

df: Degrees of freedom.

Discussion

This systematic review and meta-analysis identified pharmacogenetic interactions for some of the weight loss medications, but the directionality of the associations was not consistent across studies, as documented in the meta-analysis.

While we searched the literature for all studies evaluating the effect of SNV on the weight loss efficacy of FDA-approved weight loss medications, namely GLP-1 RA, naltrexone–bupropion, phentermine–topiramate and orlistat, only studies investigating liraglutide, exenatide and naltrexone–bupropion fit prespecified inclusion criteria and were included.

Table 2 describes the expected biological effects of the different genes and SNVs reported in the literature as well as the summary of the current analysis. Among the SNVs tested, the one associated with the greatest SMD was the variant allele of TCF7L2 gene rs7903146, compared with the reference allele. However, the results in the three relevant studies were inconsistent.

Table 2. . Biological effects of the different single nucleotide variants studies in this systematic review and meta-analysis and summary of current analysis.

Gene	Single nucleotide variant; minor allele frequency	Expected/proven biological effect	Prior reports and summary of current analysis	Ref.
GLP-1 agonists
GLP1R	rs6923761 (G > A); 30%	GLP1 receptor; decreased response to infused GLP-1	Pooled estimates of four studies did not identify allelic effect	[27,28,31,32,41–43]
GLP1R	rs10305420 (C > T); 26%	GLP1 receptor; unclear effect	Pooled estimates of three studies did not identify allelic effect	[29,32,34,41,44,45]
GLP1R	rs3765467 (G > A, T, C); A: 18%	GLP1 receptor; decreased insulin secretion in response to GLP-1	Two studies did not identify allelic effect	[29,34,41,46]
TCF7L2	rs7903146 (C > T); 29%	Transcription factor that mediates GLP-1 secretion; development of Type 2 diabetes	Two studies did not identify a difference One study in favor of variant allele Pooled estimates of three studies did not identify allelic effect	[27,28,33,40–43,47–51]
CNR1	rs1049353 (G > A); 18%	Regulation of food intake both centrally and peripherally	Previous systematic review and meta-analysis: variant allele associated with lower BMI One study in favor of variant allele One study did not identify allelic effect	[30,37,41,45,52]
MC4R	rs13447324 105 C > A, rs13447325 110 A > T, rs13447333 542 G > A, rs121913567 656 C > T, rs13447329 335 C > T	Hypothalamic receptor that plays a central role in regulating metabolic expenditures; loss of function mutation: early onset obesity; release of GLP-1 from enteroendocrine L cells	One study did not identify allelic effect	[26,53–55]
CTRB1-2	rs7202877 (T > G); 10%	Proteolytic activation into chymotrypsin; allelic variant protective against the development of Type 2 diabetes; increased GLP-1-mediated insulin secretion	One study did not identify allelic effect	[36,41,56,57]
ADIPOQ	rs2241766 (T > G); 11%	Adiponectin: antidiabetic, antiatherogenic and antiinflammatory properties; increased risk of dyslipidemia	In children: increased body weight One study did not identify allelic effect	[38,41,58–60]
SORCS1	rs1416406 (A > G); 39%	Central control of metabolism; increased glucose-stimulated insulin secretion in patients with polycystic ovary syndrome	One study did not identify allelic effect	[39,41,61–63]
Naltrexone–bupropion
ANKK1	rs1800497 (G > A); 27%	Taq1A polymorphism adjacent to the DRD2 gene; associated with addictive, impulsive behavior, overeating; minor A allele associated with diminished striatal density of DRD2.	Increased risk of obesity One study did not identify allelic effect	[35,41,64,65]

SNV: Single nucleotide variant.

Strengths & limitations

The strengths of this systematic review and meta-analysis include having a medical librarian conduct a comprehensive search of the literature and two investigators screen studies and extract data independently with adjudication using a blinded third author when needed. Importantly, we used SMD to provide outcome estimates in our meta-analysis. This approach assumes the difference in standard deviations between studies is due to differences in outcome measures rather than differences between the study designs and samples and allowed us to pool results from studies that used different measures to report the same construct. However, one should be cautious about interpreting measures presented as SMD, as a key assumption in these transformations is that the various outcomes are measuring the same clinical construct. In the case of our review, all outcomes were related to weight loss, so this assumption is reasonably met. Limitations include using inverse variance weighting in the meta-analysis for outcomes transformed to SMD, which gives more weight to studies in which the patient population is very homogenous due to strict inclusion criteria, which results in smaller variances in the outcome [29]. Another limitation of the meta-analysis is grouping GLP-1 RA together, which limits the ability to determine the pharmacogenomic effects of individual medications within this class.

Appraisal of role of pharmacogenetics in use of GLP-1 agents based on published data

Based on our review, there is insufficient evidence to suggest genetic testing prior to prescribing GLP-1 RAs. More pharmacogenetic studies involving patients prescribed GLP-1 RAs are needed to more precisely estimate the functional impact of genetic variations and identify patients likely to receive benefit. Continuing to look at targeted or candidate genes that have physiological justification with regards to GLP-1 RA mechanism of action should be prioritized to maximize power in the evaluation of drug effects on weight loss in phase IIb or phase III clinical trials. Furthermore, it is worth noting that the studies included in our review evaluated SNVs, which represent only one type of genetic variation (i.e., missense changes). Continued advancement in genetic sequencing and bioinformatics may be better suited to evaluate multiple types of functional variants, including copy number and canonical splice-site variants, as opposed to SNVs in exons.

Although genome-wide association study (GWAS) approaches have been widely used in other studies exploring association with disease states, we believe that such an approach would not be feasible for our specific question. Indeed, a GWAS approach would require very large sample sizes to achieve appropriate power, detailed documentation of the prescribed medication as well as other concomitant medications that may impact pharmacokinetics of the medication of interest, in addition to follow-up regarding the outcome of weight loss. The latter two limitations are not generally available in most large biobanks used for GWAS studies, but they could conceivably be included in pivotal phase IIb or phase III clinical trials.

Another approach that could be considered is to use polygenic scores (PGS), which have emerged as promising tools for complex trait risk prediction. PGSs have been applied in studies of pharmacogenomics. Nevertheless, a recent systematic review identified significant pitfalls in the studies using PGSs in pharmacogenomics, including insufficient population diversity compromising generalizability; substantial variability in the methods used to develop PGSs, with between 3 and 6.6 million variants included in the PGSs; and significant inconsistencies in the reporting of PGSs analyses and results, particularly in terms of risk model development and application, coupled with a lack of data transparency and availability [40].

Conclusion

In summary, we identified 14 studies reporting weight loss in patients on weight-lowering medications with or without allelic variants of interest. 13 out of the 14 studies investigated GLP-1 RA and one study naltrexone–bupropion. A meta-analysis of three SNVs did not identify any significant results in favor of the variant or the reference alleles. It is too early to recommend pharmacogenetic testing in clinical practice. Further studies are needed to deepen our understanding of the pharmacogenetic interactions of weight-lowering medications, including evaluation of candidate genes, advances in genetic sequencing and bioinformatics and potentially use of PGS, in order to guide the practice of individualized medicine in the field of obesity.

Summary points.

• There is growing interest in identifying predictors of response to weight-loss medications.

• Pharmacogenomics studies the influence of single nucleotide variants (SNVs) on the effectiveness of medications.

• Systematic search of the literature for pharmacogenomic interactions of glucagon-like peptide (GLP-1) agonists, naltrexone–bupropion, phentermine–topiramate and orlistat identified 363 citations.

• Fourteen studies were included in the qualitative analysis and seven studies in the meta-analysis, based on prespecified eligibility criteria.

• The studies evaluated the association of the CNR1, GLP-1R, MC4R, TCF7L2, CTRB1/2, ADIPOQ, SORCS1 and ANKK1 SNVs with the weight-lowering effects of GLP-1 agonists (13 studies) or naltrexone–bupropion (one study).

• Alleles of CNR1 gene (rs1049353), GLP-1R gene (rs6923761, rs10305420) and TCF7L2 gene (rs7903146) were associated with differential weight loss in at least one study involving GLP-1 agonist(s).

• Meta-analysis did not identify any consistent effect of these SNVs on weight loss with the same medications.

• Further studies are required before the recommendation of pharmacogenetic testing in clinical practice management of drug-induced weight loss.

• Further studies are needed to deepen our understanding of the pharmacogenetic interactions of weight-lowering medications, including evaluation of candidate genes, advances in genetic sequencing and bioinformatics and potentially use of polygenic scores.