Variability in and predictors of glycaemic responses after 24 weeks of treatment with exenatide twice daily and exenatide once weekly.

The range of glycated haemoglobin (HbA1c) responses and characteristics associated with above-average response to exenatide twice daily and once weekly were examined. Data were pooled from 8 exenatide-twice-daily and 5 exenatide-once-weekly studies. A baseline HbA1c-corrected measure of change in HbA1c after 24 weeks identified high, average and low responses. Multiple linear regression and multivariate generalized estimating equation models identified factors associated with high response. Among 2355 participants (exenatide twice daily, n = 1414; exenatide once weekly, n = 941), baseline HbA1c correlated with change in HbA1c (P < .0001). Across baseline HbA1c levels, the 25th to 75th percentile of HbA1c change ranged from -0.3% to -3.2% with exenatide twice daily and from -0.5% to -3.6% with exenatide once weekly. Asian ethnicity and older age were significantly associated with high response to exenatide twice daily; no factors were significantly associated with response to exenatide once weekly. These data provide clinically useful information for estimating the likelihood that, depending on baseline HbA1c, an individual can achieve HbA1c goals. The association between Asian ethnicity, age and high response to exenatide twice daily may relate to the specific effects of exenatide twice daily on postprandial glucose.

INTRODUCTION

Personalization of treatment is a central tenet of care for type 2 diabetes (T2D),1 yet little information is available about which patients are more likely to respond to a given drug. Furthermore, responses to drugs are typically described as mean changes in glycated haemoglobin (HbA1c) without details on the range of changes for individuals, which would assist clinicians in determining the likelihood a specific drug will lower HbA1c enough to achieve target.

Exenatide is a glucagon‐like peptide‐1 (GLP‐1) receptor agonist, available in twice‐daily and once‐weekly formulations, that improves glycaemic control.2, 3 Median peak plasma concentrations are observed in 2.1 hours with exenatide twice daily, whereas exenatide once weekly peaks in the first few hours, at week 2 and weeks 6 to 7, as a result of gradual release of exenatide from poly (d,l‐lactide‐coglycolide) microspheres.4 In head‐to‐head studies, exenatide once weekly led to greater improvements in HbA1c and fasting plasma glucose levels, whereas exenatide twice daily led to greater improvement in 2‐hour postprandial glucose (PPG).5, 6, 7

Most studies of T2D show significant between‐participant variability in response to glucose‐lowering drugs.5, 6 Baseline HbA1c is an important contributor to observed variability; patients with higher baseline HbA1c exhibit greater overall change in HbA1c,8 however, significant variability exists beyond this relationship, suggesting that other characteristics also contribute to response.9, 10 Identifying such characteristics would be of great clinical utility.

The aim of the present analysis was to (1) describe the distribution of HbA1c responses to exenatide twice daily and exenatide once weekly after 24 weeks of therapy and (2) identify baseline characteristics associated with above‐average glycaemic response to exenatide twice daily or once weekly after adjustment for baseline HbA1c.

METHODS

Individual participant data were pooled from separate randomized controlled exenatide studies of 24 weeks to 48 months (Table S1 in File S1). The 24‐week completer population of each study and glycaemic response at week 24 was used in the analyses. Included participants received exenatide twice daily or exenatide once weekly for at least 168 days and had baseline self‐monitored blood glucose (SMBG) measurements.

The spread of HbA1c reductions was assessed via plots of distribution of HbA1c change. To identify factors other than baseline HbA1c that were related to change in HbA1c, the analysis was corrected for baseline HbA1c. The linear relationship between baseline HbA1c and change in HbA1c was determined from all participant data for each therapy. Participants were then divided into tertiles based on residuals to the regression line: for a given baseline HbA1c, tertile 1 included participants with the greatest HbA1c reductions (‘high responders’); tertile 2 included average reductions (‘average responders’); and tertile 3 included the smallest reductions (‘low responders’).

Variables assessed were based on availability of the information in routine practice and potential relevance (Table S2 in File S1). Correlations between each variable and change in HbA1c after 24 weeks of treatment were examined using univariate regression. Variables with r 007E; 0.1 and P < .05 were considered to be potentially associated with treatment. Because of collinearity, mean daily blood glucose was excluded from exenatide twice daily assessments in favour of pre‐ and postmeal blood glucose (believed to be more relevant to exenatide twice daily's time course of action). For longer‐acting exenatide once weekly, mealtime glucose was excluded and mean daily blood glucose included.

Baseline characteristics identified from univariate regression analyses were then included in stepwise multivariate linear regression models. Variables with a significance level P < .15 were subsequently modelled via multivariate generalized estimating equation (GEE) models to estimate the odds of being in tertile 1. Significantly increased odds were defined as P < .05.

RESULTS

Data were pooled from 8 studies of exenatide twice daily and from 5 studies of exenatide once weekly (Table S1 in File S1) with 1640 and 1092 participants in the exenatide arms, respectively. Of those, 1414 and 941 participants, respectively, with all baseline data were included in the analysis.

There was a larger mean HbA1c reduction for exenatide once weekly than exenatide twice daily (Figure S1A in File S1), potentially attributable in part to higher mean baseline HbA1c in the exenatide‐once‐weekly vs exenatide‐twice‐daily groups (8.50% vs 7.98%). For both formulations, baseline HbA1c was significantly correlated with change in HbA1c (P < .0001), with extensive variability around the mean response (Figure S1B and S1C in File S1). With increasing baseline HbA1c, the 25th percentile of HbA1c change ranged from −0.3% to −1.4% with exenatide twice daily and −0.5% to −1.5% with exenatide once weekly (Figure 1). The 75th percentile of HbA1c reduction ranged from −1.1% to −3.2% with exenatide twice daily and −1.3% to −3.6% with exenatide once weekly.

Figure 1

Box and whiskers plot of the change in HbA1c for a given baseline HbA1c. Midline represents the 50th percentile, the box represents the 25th and 75th percentiles, and whiskers represent the 10th and 90th percentiles

Baseline characteristics (Tables S3 and S4 in File S1) were generally similar across responder tertiles. For exenatide twice daily, participants in tertile 1 were older and more often Asian than participants in tertile 3 (P < .05 for both). For exenatide once weekly, there were more Asian participants in tertile 3 than in tertile 1 (P < .05). For both formulations, baseline HbA1c was lowest in tertile 2 (P < .05). There were small differences among SMBG measurements.

Stepwise multivariate linear regression analysis and GEE analysis

For exenatide twice daily, greater baseline pre‐breakfast blood glucose, baseline PPG excursion for breakfast, age and Asian ethnicity were all (P ≤ .0001) associated with greater HbA1c reduction after 24 weeks. For exenatide once weekly, greater baseline mean daily blood glucose and baseline fasting blood glucose were both (P < .05) associated with a greater HbA1c reduction. Using factors identified by multivariate linear regression models, GEE models were run to determine the odds of being in tertile 1 (Table 1). For exenatide twice daily, the odds of being in tertile 1 were 2.1 times greater for Asian than non‐Asian participants (P < .0001), and increased by 12.4% for every 5‐year increase in age (P < .0001). The odds of being in tertile 1 increased slightly (+2.2%; P = .06) for every 0.56‐mmol/L increase in baseline PPG excursion for breakfast and decreased slightly (−2.4%; P = .07) for every 0.56‐mmol/L increase in baseline fasting blood glucose.

Table 1

Baseline characteristics associated with being in the high‐responder tertile (tertile 1) to exenatide twice daily and exenatide once weekly

Baseline characteristic	Estimated odds ratio	95% confidence intervals for estimated odds ratio		P
Exenatide twice daily
Fasting blood glucose (per 0.56‐mmol/L increase)	0.98	0.95	1.00	.07
PPG excursion for breakfast (per 0.56‐mmol/L increase)	1.02	1.00	1.05	.06
Ethnicity (Asian vs non‐Asian)	2.11	1.51	2.95	<.0001
Age (per 5‐years older)	1.12	1.07	1.18	<.0001
Exenatide once weekly
Fasting blood glucose (per 0.56‐mmol/L increase)	1.00	0.96	1.04	.96
Mean daily blood glucose (per 0.56‐mmol/L increase)	0.98	0.95	1.02	.37

For glucose measurements, to convert mmol/L to mg/dL multiply by 18.

For exenatide once weekly, the GEE analysis found no baseline measures that were significantly associated with being in tertile 1 (Table 1).

DISCUSSION

The aim of the present retrospective analysis was to demonstrate the range and distribution of HbA1c responses to exenatide and identify characteristics associated with optimum glycaemic response. First, this analysis demonstrated the wide range of HbA1c‐lowering responses with both exenatide formulations, indicating that the mean reduction is of only modest value in assessing likely responses. Second, GEE analysis found that Asian ethnicity (predominantly Japanese in these studies) and increasing age were the only predictors of HbA1c reduction for exenatide twice daily. No independent predictors of response to exenatide once weekly were identified.

There are numerous reports of mean HbA1c reductions associated with glucose‐lowering drugs; however, without an appreciation of the spread of reductions, and how that distribution relates to baseline HbA1c, it is difficult to provide individualized advice to patients about likely treatment outcomes. To the best of our knowledge, this is the first study to report such data, and the information in Figure 1 can be applied directly to patient‐care decisions about the likelihood that an individual patient will reach goal when one of these drugs is commenced.

Previous studies in the same database suggested Asian participants with T2D respond to exenatide twice daily better than white participants.11 This analysis extends these findings by showing this effect is independent of baseline HbA1c. It has been proposed that loss of PPG control, either through physiological or dietary means, is the primary glucose defect experienced by Asians with T2D.12 Exenatide twice daily, which has a stronger effect on PPG relative to longer‐acting treatments,5 may therefore be more efficacious for Asian patients; however, our finding that PPG is, at best, a weak predictor of response needs to be considered. The glycaemic profile of a single day may be too limited to describe longer‐term glycaemic patterns. It should also be noted that the lack of relationship of baseline fasting blood glucose and PPG with HbA1c reduction does not imply that exenatide does not improve these variables but rather that, once baseline HbA1c is known, information about fasting blood glucose and PPG is not useful in predicting HbA1c reduction.

This study also showed that age had an effect on participant responsiveness to exenatide twice daily but not exenatide once weekly. The trend of increased effectiveness in older participants has also been observed with dipeptidyl peptidase‐4 inhibitors.13 Interestingly, the present analysis did not find that longer duration of diabetes, concomitant medication use, body weight, or body mass index were important predictors of treatment response; thus, although both exenatide formulations are generally effective treatments, our findings suggest that exenatide twice daily may be especially effective for Asian and older patients.

Other multivariate analyses have examined participant characteristics associated with treatment success (as opposed to above‐average responsiveness) for exenatide twice daily. These analyses have found that greater weight loss and higher baseline estimated β‐cell function or C‐peptide levels are associated with treatment success,14, 15, 16 greater disease duration reduces likelihood of treatment success, and sex, race and ethnicity affect treatment response as well.14, 17 However, these analyses failed to account for the substantial impact of baseline HbA1c.

The present analysis has a number of limitations. It was a retrospective pooled analysis that assumed all participants were compliant with therapeutic recommendations during the studies. Potential study heterogeneity was not formally assessed nor controlled for in the statistical model because of potential confounding with other baseline variables. Regarding the significant findings relating to Asian ethnicity, most Asian participants in the analysis of exenatide twice daily were in a single study,18 but the mean HbA1c reduction in the placebo arm of this study was intermediate between the HbA1c reductions in the other 2 placebo arms of studies on exenatide twice daily.19, 20 Regarding the finding in relation to age, within‐study age variation was much greater than between‐study age variation; thus, from both an analytical and a study design perspective, we believe that it is unlikely that study heterogeneity explains our findings. The analysis population was limited by characteristics of the trial populations; results may differ in broader patient populations. SMBG measurements included in the present analysis relied on participant‐reported measurements without standardized meals and were performed on a single day, so may have limited capacity to reflect daily glucose profiles. In addition, we selected participants who completed therapy; our results may not apply to participants with less exposure to exenatide or to those excluded from clinical trials. Factors not considered as part of the analysis, such as β‐cell function and insulin sensitivity, or factors not measured, such as plasma drug concentration, diet composition and GLP‐1 receptor characteristics, may influence response to exenatide treatment. Multiplicity was not controlled for, which may increase the risk of false‐positive results. Cross‐validation from an independent data set and a controlled study would be needed to confirm the findings.

In conclusion, the present study quantified the wide range of HbA1c responses to treatment with exenatide twice daily and exenatide once weekly, to inform treatment choices in clinical practice. While baseline HbA1c affects response to exenatide, Asian ethnicity and higher participant age were also associated with larger HbA1c reductions for exenatide twice daily. No predictors were associated with an improved response to exenatide once weekly. This analysis may assist clinicians to individualize treatment by identifying patients with characteristics that may result in improved response to exenatide twice daily.

File S1.

Click here for additional data file.