Dulaglutide (DU) 1.5 mg was associated with improved composite renal outcomes that included new-onset macroalbuminuria in people with type 2 diabetes with previous cardiovascular disease or cardiovascular risk factors in the REWIND (Researching cardiovascular Events with a Weekly INcretin in Diabetes) trial. This exploratory post hoc analysis evaluated kidney function–related outcomes, excluding the new-onset macroalbuminuria component, among the REWIND participants.
Intent-to-treat analyses were performed on REWIND participants (n = 4,949 DU, n = 4,952 placebo). Time to occurrence of a composite kidney function–related outcome (≥40% sustained decline in estimated glomerular filtration rate [eGFR], per the Chronic Kidney Disease Epidemiology Collaboration 2009 equation, end-stage renal disease, or renal-related death), and mean annual eGFR slope were examined. Analyses were conducted overall and within subgroups defined by baseline urinary albumin-to-creatinine ratio (UACR <30 or ≥30 mg/g) and baseline eGFR (<60 or ≥60 mL/min/1.73 m2).
The post hoc composite kidney function–related outcome occurred less frequently among participants assigned to DU than placebo (hazard ratio [HR] 0.75, 95% CI 0.62–0.92, P = 0.004), with no evidence of a differential DU treatment effect by UACR or eGFR subgroup. A ≥40% sustained eGFR decline occurred less frequently among participants assigned to DU than placebo (HR 0.72, 95% CI 0.58–0.88, P = 0.002). The mean annual decline in eGFR slope was significantly smaller for participants assigned to DU than placebo (−1.37 vs. −1.56 mL/min/1.73 m2/year, P < 0.001); results were similar for all subgroups.
The estimated 25% reduced hazard of a kidney function–related outcome among participants assigned to DU highlights its potential for delaying or slowing the development of diabetic kidney disease in people with type 2 diabetes.
Introduction
Findings from cardiovascular (CV) outcomes trials (CVOTs) of several glucagon-like peptide 1 receptor agonists (GLP-1 RAs) in people with type 2 diabetes, including exenatide (EXSCEL [Exenatide Study of Cardiovascular Event Lowering]) (1), liraglutide (LEADER [Liraglutide Effect and Action in Diabetes: Evaluation of Cardiovascular Outcome Results]) (2), semaglutide (SUSTAIN-6 [Trial to Evaluate Cardiovascular and Other Long-term Outcomes with Semaglutide in Subjects with Type 2 Diabetes]) (3), and efpeglenatide (AMPLITUDE-O [Effect of Efpeglenatide on Cardiovascular Outcomes]) (4), suggest that GLP-1 RAs can reduce the risk of macroalbuminuria onset or slow decline in the estimated glomerular filtration rate (eGFR). Over a median follow-up period of 5.4 years in the REWIND (Researching cardiovascular Events with a Weekly INcretin in Diabetes) CVOT of adults with type 2 diabetes and previous CV disease or risk factors, a prespecified secondary kidney disease–related outcome, defined as the first occurrence of a sustained eGFR decline of ≥30%, incident macroalbuminuria, or chronic renal replacement therapy, occurred less frequently among participants assigned to treatment with the GLP-1 RA dulaglutide (DU) than among those assigned to placebo (hazard ratio [HR] 0.85, 95% CI 0.77–0.93, P = 0.0004); eGFR for this prespecified secondary outcome was calculated per the Modification of Diet in Renal Disease (MDRD) equation (5). Furthermore, in an exploratory analysis of the REWIND study, treatment with DU versus placebo was associated with a lower risk of an outcome of ≥40% sustained eGFR decline, again with eGFR as calculated per the MDRD equation (6). A recent meta-analysis synthesizing findings from these and other GLP-1 RA CVOTs showed that participants assigned to GLP-1 RAs experienced a lower risk of a broad composite kidney-related outcome that included development of new-onset macroalbuminuria, with results mainly driven by a reduction in urinary albumin excretion (7).
Evidence also supports a relationship between DU and outcomes other than CV events. The AWARD-7 (A Study Comparing Dulaglutide With Insulin Glargine on Glycemic Control in Participants With Type 2 Diabetes [T2D] and Moderate or Severe Chronic Kidney Disease [CKD]) trial, a 1-year, international, multicenter, phase III, randomized, open-label clinical trial, compared treatment with once-weekly DU to that with daily insulin glargine in participants with type 2 diabetes and moderate-to-severe CKD. In that trial, DU was associated with a reduced 1-year decline in eGFR and reduced urinary albumin-to-creatinine ratio (UACR) compared with insulin glargine (8). Moreover, exploratory analysis of the AWARD-7 trial demonstrated that DU versus insulin glargine was associated with a reduced risk of a composite kidney-related outcome that excluded changes in the UACR: a sustained eGFR decline of ≥40%, incident end-stage renal disease (ESRD), or renal-related death (9).
This exploratory post hoc analysis of the REWIND study evaluated the effect of once-weekly DU 1.5 mg versus placebo on a composite kidney function–related outcome. Composite kidney function–related outcomes are conventionally used in renal outcomes trials and are usually defined as a composite end point of eGFR-related outcomes, ESRD, and renal-related death. For this post hoc analysis, the composite kidney function–related outcome was defined as the first incidence of a sustained eGFR decline (≥40%), ESRD, or renal-related death. These components are commonly used in kidney outcomes studies (10,11) compared with definitions that include new onset of macroalbuminuria as a component. In addition, eGFR for this current analysis was calculated using the 2009 Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation, a more precise and accurate value compared with the MDRD-based value (12).
Research Design and Methods
Study Design and Participants
Details of the REWIND study design have been published previously (5,13). In brief, this multicenter (371 sites in 24 countries), randomized, double-blind, placebo-controlled CVOT enrolled men and women aged ≥50 years with established or newly detected type 2 diabetes and with a previous CV event or CV risk factors. For inclusion in the REWIND study, participants were required to have type 2 diabetes, glycated hemoglobin (HbA1c) of ≤9.5% (80 mmol/mol), a BMI of at least 23 kg/m2, and to be on a stable antihyperglycemic regimen (antihyperglycemic drug naive, basal insulin alone, or up to two oral glucose-lowering drugs, with or without basal insulin therapy) for at least the past 3 months (13). Key exclusion criteria were a coronary or cerebrovascular event within the previous 2 months, renal dialysis or eGFR <15 mL/min/1.73 m2, severe hypoglycemia within the past year, previous pancreatitis, bariatric surgery or known abnormal gastric emptying, prior participation in any study investigating DU, and cancer within the past 5 years (13). The REWIND trial protocol (ClinicalTrials.gov identifier: NCT01394952, registered on 15 July 2011) was approved by research ethics boards at all sites, and all participants provided written informed consent.
Randomization and Masking
Procedures
After randomization, participants were seen at 2 weeks, 3 months, 6 months, and then every 6 months for detailed assessments. UACR and eGFR were measured in local laboratories at baseline and then every 12 months. Management of renal-protective medications, blood pressure, and CV risk was left to the discretion of the investigator throughout the trial, as per current and/or local guidelines.
End Points
The main end point for this post hoc analysis was time to event for the composite renal outcome, defined as a sustained eGFR decline ≥40% from baseline (i.e., at two or more consecutive visits), ESRD, or renal-related death. ESRD was defined based on investigator-reported events that included chronic dialysis, kidney transplant, adverse events reported as ESRD, or sustained eGFR <15 mL/min/1.73 m2. Chronic dialysis was defined as reported by investigators or an adverse event with Medical Dictionary for Regulatory Activities (MedDRA; version 21)–preferred term of “dialysis, dialysis device insertion, hemodialysis, hemofiltration, peritoneal dialysis, continuous hemodiafiltration, or artificial kidney device user.” Kidney transplant was defined as reported by investigators or with an adverse event with the MedDRA-preferred term of “kidney and liver transplant, kidney and pancreas transplant, kidney replacement therapy, or kidney transplant.” Additional ESRD events were identified using the MedDRA-preferred term of “diabetic end-stage renal disease, renal failure, end-stage renal disease.” Sensitivity analyses further evaluated alternative composite outcomes per the components mentioned above, substituting CV-related death, renal- or CV-related death, or all-cause death in place of the renal-related death component; all-cause death was included to rule out the possibility of competing risk. Additionally, annual eGFR slope (i.e., change in eGFR over time) was calculated to compare the extent of eGFR decline observed in the two arms. The CKD-EPI creatinine equation was used to calculate eGFR for all analyses.
These outcomes were evaluated in the overall population as well as in eGFR and UACR subgroups, first as defined by baseline albuminuria status (normoalbuminuria [UACR <30 mg/g] and albuminuria [UACR ≥30 mg/g]) and second by baseline eGFR categories (normal [eGFR ≥60 mL/min/1.73 m2] and reduced eGFR [eGFR <60 mL/min/1.73 m2]). These subgroup analyses were done to investigate whether the renal outcomes or decline in kidney function in participants treated with DU 1.5 mg or placebo differed by participants’ baseline renal function.
Statistical Analyses
Sample size calculations for the REWIND trial have been reported elsewhere (5,13). In this analysis, we report findings for renal-specific and other composite outcomes, including CV-related death, CV- or renal-related death, or all-cause death, the individual components of the composite outcomes, and subgroup analysis of these composite outcomes, and compare change in eGFR from baseline to end of follow-up in participants assigned to DU versus placebo.
Analyses were done according to the intention-to-treat approach that included all randomly assigned participants. Baseline demographic data were summarized using means and SDs for continuous variables and percentages for categorical variables. Treatment group comparisons were assessed for the time to first event of the composite renal end point and for each of its components (≥40% sustained eGFR decline, ESRD, or renal-related death) using a Cox proportional hazards regression model, and associated Kaplan-Meier curves were generated. Cox proportional hazards models included fixed effects for treatment group, baseline eGFR category, baseline HbA1c, baseline eGFR, and baseline systolic blood pressure. The potential differential effect of DU versus placebo across subgroups (baseline UACR <30 vs. ≥30 mg/g, baseline eGFR <60 vs. ≥60 mL/min/1.73 m2) was explored by adding terms for subgroup and the subgroup-by-treatment interaction. The same approach was applied for the sensitivity analyses. The CKD-EPI creatinine equation was used to calculate eGFR for all analyses. Baseline eGFR data outside the range of 1 to 150 mL/min/1.73 m2 were excluded, as such measurements are likely incorrect. Additional sensitivity analyses for time-to-event outcomes were done using the MDRD equation to calculate eGFR to confirm consistency with the main analysis using the CKD-EPI equation.
The annual eGFR slope, or the mean rate of change of eGFR per year, during the trial period when participants received study drug, was based on a mixed-effects model for repeated measures, adjusted for baseline systolic blood pressure, with a log transformation, where individual slopes were estimated via a random intercept/random coefficient. Analyses were repeated within subgroups as defined by baseline UACR (<30 vs. ≥30 mg/g) and baseline eGFR (<60 vs. ≥60 mL/min/1.73 m2).
Tests of treatment group differences were performed at a two-sided α = 0.05, and 95% CIs are reported. In all subgroup analyses, a subgroup-by-treatment interaction P value <0.05 was considered statistically significant. Analyses presented were exploratory and hence not controlled for type I error. Analyses were performed using SAS 9.4 software (SAS Institute, Cary, NC).
Data and Resource Availability
Data are available and will be disclosed only upon request and approval of the proposed use of the data by a review committee.
Results
Baseline Characteristics
Between 18 August 2011 and 14 August 2013, a total of 9,901 REWIND participants were randomly assigned to a once-weekly subcutaneous injection of DU 1.5 mg (n = 4,949) or placebo (n = 4,952). At baseline, 3,114 participants (31.5%) had CV disease, 3,467 (35.0%) had albuminuria (i.e., UACR ≥30 mg/g [3.4 mg/mmol]), 791 (8.0%) had macroalbuminuria (i.e., UACR >300 mg/g [33.9 mg/mmol]), and the mean ± SD eGFR was 77.4 ± 19.5 mL/min/1.73 m2. Median participant follow-up was 5.4 years (interquartile range 5.1–5.9 years) for a total of 51,830 person-years. Baseline demographics and clinical characteristics were largely balanced between the two arms (Supplementary Table 1).
Composite Renal Outcomes in the Overall Population
In the overall population, the composite renal outcome of ≥40% sustained eGFR decline, ESRD, or renal-related death occurred in 181 participants (3.7%) in the DU arm and in 238 participants (4.8%) in the placebo arm (HR 0.75, 95% CI 0.62–0.92, P = 0.004) (Figs. 1 and 2). In a sensitivity analysis when eGFR was calculated using the MDRD equation, the composite outcome was also less frequent among participants in the DU arm than in the placebo arm, with an estimated 28% risk reduction (Supplementary Fig. 1).
Effect of DU 1.5 mg vs. placebo on the composite kidney function–related outcome (≥40% sustained eGFR decline, ESRD, or renal-related death), additional composite outcomes, and composite end point components. eGFR was calculated using the 2009 CKD-EPI equation (eGFR units: mL/min/1.73 m2). The HRs, 95% CIs, and P values are based on Cox proportional hazards models with fixed effects for treatment group, baseline eGFR category, baseline HbA1c, baseline eGFR, and baseline systolic blood pressure.
Effect of DU 1.5 mg vs. placebo on the composite kidney function–related outcome (≥40% sustained eGFR decline, ESRD, or renal-related death), additional composite outcomes, and composite end point components. eGFR was calculated using the 2009 CKD-EPI equation (eGFR units: mL/min/1.73 m2). The HRs, 95% CIs, and P values are based on Cox proportional hazards models with fixed effects for treatment group, baseline eGFR category, baseline HbA1c, baseline eGFR, and baseline systolic blood pressure.
Effects of DU 1.5 mg treatment vs. placebo on the composite kidney function–related outcome of ≥40% sustained eGFR decline, ESRD, or renal-related death in all REWIND participants. Note: eGFR was calculated using the 2009 CKD-EPI equation (eGFR units: mL/min/1.73 m2). The HR, 95% CI, and P value are based on a Cox proportional hazards model with fixed effects for treatment group, baseline eGFR category, baseline HbA1c, baseline eGFR, and baseline systolic blood pressure.
Effects of DU 1.5 mg treatment vs. placebo on the composite kidney function–related outcome of ≥40% sustained eGFR decline, ESRD, or renal-related death in all REWIND participants. Note: eGFR was calculated using the 2009 CKD-EPI equation (eGFR units: mL/min/1.73 m2). The HR, 95% CI, and P value are based on a Cox proportional hazards model with fixed effects for treatment group, baseline eGFR category, baseline HbA1c, baseline eGFR, and baseline systolic blood pressure.
Sensitivity analyses that replaced the renal-related death component with CV-related death, renal- and CV-related death, or all-cause death yielded results consistent with the main outcome findings (HR 0.84, 95% CI 0.74–0.95 for the composite end point with CV-related death; HR 0.84, 95% CI 0.75–0.95 for the composite end point with renal- or CV-related death; and HR 0.83, 95% CI 0.75–0.92 for the composite end point with all-cause death) (Fig. 1).
Composite Renal Outcomes in Subgroups Defined by Albuminuria and eGFR
For the composite renal end point including renal-related deaths, there was no evidence of a differential treatment effect according to baseline UACR (<30 vs. ≥30 mg/g, interaction P = 0.077) or baseline eGFR (<60 vs. ≥60 mL/min/1.73 m2, interaction P = 0.306) (Fig. 3 and Supplementary Fig. 2). Sensitivity analyses on different composite end points or using the MDRD equation also did not show any differential treatment effect by baseline UACR or eGFR.
Effect of DU 1.5 mg on the composite kidney function–related outcome of ≥40% sustained eGFR decline, ESRD, or renal-related death overall and by baseline albuminuria and eGFR status. Note: Only participants with baseline UACR and baseline eGFR values were included in the subgroup analyses, respectively. eGFR was calculated using the 2009 CKD-EPI equation (eGFR units: mL/min/1.73 m2). The HRs, 95% CIs, and P values overall and within subgroups are based on Cox proportional hazards models with fixed effects for treatment group, baseline eGFR category, baseline HbA1c, baseline eGFR, and baseline systolic blood pressure. Interaction P values were derived from overall models with additional terms for subgroup and the subgroup-by-treatment interaction.
Effect of DU 1.5 mg on the composite kidney function–related outcome of ≥40% sustained eGFR decline, ESRD, or renal-related death overall and by baseline albuminuria and eGFR status. Note: Only participants with baseline UACR and baseline eGFR values were included in the subgroup analyses, respectively. eGFR was calculated using the 2009 CKD-EPI equation (eGFR units: mL/min/1.73 m2). The HRs, 95% CIs, and P values overall and within subgroups are based on Cox proportional hazards models with fixed effects for treatment group, baseline eGFR category, baseline HbA1c, baseline eGFR, and baseline systolic blood pressure. Interaction P values were derived from overall models with additional terms for subgroup and the subgroup-by-treatment interaction.
Components of Composite Renal Outcome
A significantly lower proportion of participants assigned to DU than to placebo had ≥40% sustained eGFR decline (3.2% vs. 4.3%; HR 0.72, 95% CI 0.58–0.88, P = 0.002) (Fig. 1). There were no significant between-group differences in rates for other components of the composite outcomes (ESRD: HR 0.78, 95% CI 0.49–1.25; renal-related death: HR 1.63, 95% CI 0.61–4.31; CV death: HR 0.91, 95% CI 0.78–1.06; all-cause death: HR 0.90, 95% CI 0.80–1.01) (Fig. 1).
Mean Annual eGFR Slope
In the overall population, the adjusted annual mean decline in eGFR was significantly slower in participants assigned to DU than to placebo (−1.37 vs. −1.56 mL/min/1.73 m2/year, P < 0.001) (Fig. 4 and Supplementary Table 2).
Effect of DU 1.5 mg vs. placebo on eGFR over the study period for the overall population. Note: Estimates are based on a mixed-model repeated-measures model, adjusted for baseline eGFR, baseline systolic blood pressure, and age, with eGFR values (mL/min/1.73 m2) calculated per the 2009 CKD-EPI equation.
Effect of DU 1.5 mg vs. placebo on eGFR over the study period for the overall population. Note: Estimates are based on a mixed-model repeated-measures model, adjusted for baseline eGFR, baseline systolic blood pressure, and age, with eGFR values (mL/min/1.73 m2) calculated per the 2009 CKD-EPI equation.
When the mean annual eGFR slope was analyzed by albuminuria subpopulations, the DU arm had significantly slower annual decline than the placebo arm in both those with baseline UACR <30 mg/g (−1.22 vs. −1.33, P = 0.005) and those with baseline UACR ≥30 mg/g (−1.73 vs. −2.05, P < 0.001) (Supplementary Table 2). Similarly, when the mean annual eGFR slope was analyzed by eGFR subpopulations, the DU arm showed significantly slower annual decline than the placebo arm in the baseline normal eGFR group and the reduced eGFR group (Supplementary Table 2).
Conclusions
In this post hoc analysis of the REWIND study, participants receiving once-weekly DU 1.5 mg had an estimated 25% reduced risk of occurrence of the composite renal outcome (≥40% sustained eGFR decline, ESRD, or renal-related death) compared with those receiving placebo over a median follow-up of 5.4 years. As expected, participants with albuminuria or reduced eGFR at baseline demonstrated a greater incidence of this composite renal outcome regardless of treatment assignment (Fig. 3). The risk of occurrence of the composite outcome in the DU group compared with the placebo group was consistently reduced regardless of baseline albuminuria or baseline eGFR. Furthermore, in sensitivity analyses replacing the renal-related death component with CV-related death, renal- or CV-related death, or all-cause death, results were overall consistent with the main findings (Fig. 1). The kidney function–related outcomes used in these analyses are consistent with outcomes that were previously used in other recent kidney outcomes studies (10,11).
The mean annual eGFR slope analysis showed significantly slower eGFR decline in the DU group over a median follow-up of 5.4 years. These results were seen not only in the overall population but also in participants at a higher risk of progressive kidney disease as defined by higher UACR and lower eGFR values at baseline. These data further support prior observations from both the REWIND and the AWARD-7 trials of a potential delay in the development or progression of diabetic kidney disease in adults with type 2 diabetes and CV risks and/or moderate-to-severe CKD when treated with DU 1.5 mg (6,8). The possibility of the contribution of differences in baseline systolic blood pressure in the observed mean annual eGFR decline was circumvented by adjusting the eGFR slope analysis with baseline systolic blood pressure, ensuring the results observed were not related to baseline differences in blood pressure.
While changes in eGFR among participants assigned to DU in this same study population have been previously reported, it is important to note a key difference in the evaluation of eGFR change over time. The previous REWIND paper did not evaluate eGFR slope. Rather, change in eGFR was evaluated using the overall change in least squares means of eGFR from baseline through the median 5.4 years of follow-up (6); that outcome cannot be directly compared with this current analysis examining an outcome of annual change in eGFR slope. The slope analysis presented here reflects the eGFR decline trajectory over time. The adjusted annual mean decline in eGFR was significantly slower in participants assigned to DU than to placebo, which is consistent with the fewer kidney events observed for the composite kidney outcome. Indeed, recent literature supports the utility of the eGFR slope as a valuable surrogate for progression of CKD as it is strongly predictive of hard clinical kidney outcomes (14,15).
Strengths of this analysis include the evaluation of a stringent composite renal outcome, one that does not include new onset of macroalbuminuria, over a median follow-up of 5.4 years in a large, representative, and broad population with type 2 diabetes with a wide range of baseline renal function and CV risk factors, and the inclusion of a sizeable percentage of women. We used a narrow composite kidney function–related outcome that has been conventionally used in kidney outcomes studies. Notably, the main finding was consistent among analyses using either the MDRD or CKD-EPI equations to calculate eGFR. Very limited data exist on kidney outcomes in adults with type 2 diabetes who have an eGFR >60 mL/min/1.73 m2, making these results of great interest to both clinicians and patients.
Limitations for this analysis include its post hoc nature, lack of adjudication for these kidney outcomes, the limited numbers of participants with CKD, the use of local laboratories to measure albuminuria and serum creatinine level for estimation of eGFR, and the relatively infrequent laboratory draws over the study period. While a new CKD-EPI equation that does not include a race variable was developed in 2021 (16), these analyses used the older 2009 equation. Including dialysis codes for the composite kidney outcome could capture cases of acute kidney injury, causing an overestimation of the composite outcome. However, these were limited in number and not expected to affect the conclusions. Additional studies are needed to evaluate the progression of diabetic kidney disease in adults with type 2 diabetes treated with DU.
These post hoc analyses suggest that the use of DU 1.5 mg for glycemic control in adults with type 2 diabetes at CV risk and a wide range of renal function may confer renal benefits.
Clinical trial reg. no. NCT01394952, clinicaltrials.gov
This article contains supplementary material online at https://doi.org/10.2337/figshare.22959695.
Article Information
Acknowledgments. The authors thank Karen Nunley, PhD, for providing medical writing assistance (Syneos Health) and Raena Fernandes, MSc, ELS, for providing editorial assistance (Syneos Health), Charles M. Atisso (retired from Eli Lilly and Company) for contributions to designing the analysis plan and data analysis, and all the participants of the REWIND study.
Duality of Interest. Eli Lilly and Company funded the study and was involved in the study design, data collection, data analysis, data interpretation, developing the manuscript, and decision to submit the article to the journal. F.T.B., R.M., C.N., A.H., and I.T. are employees of Eli Lilly. F.T.B., R.M., A.H., and I.T. own stock in Eli Lilly. F.T.B. received support for meetings/travel from Eli Lilly, reports patents planned, issued, or pending with Eli Lilly, and reports receipt of equipment, materials, drugs, medical writing, gifts, or other services from Eli Lilly. H.C.G. received grant funding from Eli Lilly for this project, and also reports grant funding from Sanofi, AstraZeneca, Novo Nordisk, Merck, Abbott, Boehringer Ingelheim, and Hanmi, and honoraria from Eli Lilly, Sanofi, AstraZeneca, Boehringer Ingelheim, Abbott, Novo Nordisk, DKSH, Zuellig, and Kowa Research Institute. H.M.C. received institutional fees for REWIND and grant funding from Eli Lilly, AstraZeneca, Pfizer, Novo Nordisk, and Regeneron, honoraria from Eli Lilly, Novartis, and Regeneron, served on advisory boards for Eli Lilly, Bayer, Sanofi, Novartis, Regeneron, and Novo Nordisk, and owns stock in Bayer AG and Roche Pharmaceuticals. J.E.S. received grant funding from Boehringer Ingelheim and AstraZeneca, consulting fees from Eli Lilly, Pfizer, AstraZeneca, Roche, and Sanofi, honoraria from AstraZeneca, Mylan, Sanofi, Boehringer Ingelheim, and Zeullig, and support for attending meetings/travel fees from Bayer. No other potential conflicts of interest relevant to this article were reported.
Author Contributions. F.T.B. contributed to the conception/design of the work, drafting the manuscript, and analysis and interpretation of the data. H.C.G. contributed to acquisition and analysis of the data and provided critical review of the manuscript. C.N. analyzed the data, interpreted the data, and provided critical review of the manuscript. I.T. contributed to drafting the manuscript and analyzing the data as well as providing critical review of the manuscript. H.M.C. contributed to conception/design of the work and acquisition of the data and provided critcal review of the manuscript. J.E.S. contributed to designing the work and drafting/critical review the manuscript. R.M. and A.H. contributed to interpreting the data and provided critical review/revision of the work. All authors approved the final version of this manuscript. F.T.B. is the guarantor of this work and, as such, had full access to all the data in the analyses and takes responsibility for the integrity of the data and the accuracy of the data analysis.