Herpes Zoster Incidence and Burden in Adults With Type 2 Diabetes in the U.S.: A Retrospective Database Analysis

Herpes zoster (HZ) is a neurocutaneous disorder that develops due to the reactivation of the varicella zoster virus latent in the sensory ganglia following a previous infection (typically in childhood). Upon reactivation, a characteristic vesicular rash develops on the skin, with rash location corresponding to involved dermatomes (1). Varicella zoster virus reactivation can also induce neurologic responses, leading to several complications, notably, postherpetic neuralgia (PHN), chronic pain that persists for ≥3 months after the initial HZ episode, affecting 5–30% of patients (2,3).

Nearly one in every three individuals will contract HZ in their lifetime, with almost 1 million cases per year in the U.S. (4,5). The risk of HZ increases between 40 and 50 years of age and affects up to half of all people who live to 85 years of age (4). Several studies have suggested that type 2 diabetes (T2D) represents an important risk factor for HZ (6–10). In a recent meta-analysis, preexisting diabetes was associated with a 1.6-fold increased risk of HZ (9), and PHN may also be more frequent in patients with T2D with HZ (2,10).

While studies have reported on health care resource use (HCRU) and costs associated with HZ (and PHN and other HZ-related complications) in different populations in the U.S. (11–13), limited data exist for the real-world impact of HZ on HCRU and costs among U.S. patients with T2D. Understanding such an impact is important given the increasing prevalence of both T2D and HZ in the U.S. (5,14). The current study assessed the incidence and risk of HZ in patients with T2D compared with patients without T2D. In addition, we measured the impact of HZ on HCRU and evaluated incremental all-cause health care and HZ-related costs in patients with T2D.

This was an observational, retrospective study (study identifier: VxHO-000044) based on U.S. administrative health insurance claims databases. The study used medical data, pharmacy data, and enrollment and demographic information from the IBM Watson Health Analytics MarketScan suite of data, the Commercial Claims and Encounters (CCAE) database (which covers patients aged 18–64 years), the Medicare Supplemental database (Medicare, ≥65 years), and the Medicaid Multi-State (Medicaid) database (≥18 years). Patient data were analyzed from 1 August 2013 to 31 July 2018 for patients in the CCAE and Medicare databases, and from 1 January 2012 to 31 December 2016 for patients in the Medicaid database. All data were deidentified prior to acquisition. As a retrospective observational study using deidentified patient data, the current study was considered exempt from Institutional Review Board approval.

The study used data in two ways: firstly, to estimate HZ incidence and risk of developing HZ in patients with T2D, compared with patients without T2D (Supplementary Fig. 1); secondly, HCRU and associated costs were evaluated over a 12-month period in three propensity score-matched cohorts (patients with T2D with an HZ diagnosis, T2D patients without HZ, and patients without T2D with HZ), to estimate incremental HCRU and costs associated with HZ (Supplementary Fig. 1).

Eligible subjects were patients ≥18 years of age identified on the basis of HZ and T2D claims associated with relevant International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) or the ICD-10-CM diagnosis codes (15,16). Patients with T2D were identified based on one inpatient or two outpatient (≥30 days apart) primary or secondary diagnosis claims (ICD-9-CM diagnosis codes 250.x0 and 250.x2; ICD-10-CM diagnosis code E11.XX) (17). Patients with HZ were identified by the presence of a primary or secondary diagnosis of HZ (ICD-9-CM diagnosis code 053.XX; ICD-10-CM diagnosis code B02.XX). Patients with any prior receipt of HZ vaccination during the study period (identifiable by Current Procedural Terminology and National Drug Code claims codes) were excluded.

For incidence, we considered all patients with and without T2D, followed up for at least 12 months, or until censoring at the earlier of one of two events, incident HZ or HZ vaccination. For HCRU and cost analysis, we then selected a subset of patients with assignment into three mutually exclusive study cohorts; patients with T2D with an HZ diagnosis (cohort 1), patients with T2D and no HZ (cohort 2), and patients without T2D with HZ (cohort 3). In this approach, those patients with and without T2D with HZ (cohorts 1 and 3) required ≥12 months of continuous enrollment prior to an index date (defined as the date of the first HZ diagnosis) and ≥12 months of continuous enrollment after this date. Patients with T2D without HZ (cohort 2) were assigned an index date defined as the date occurring 364 days after the first observed T2D diagnosis date and required ≥12 months of continuous enrollment before and after this date. Specific eligibility criteria are listed in Supplementary Table 1, and patient selection is illustrated in Supplementary Fig. 2.

After applying eligibility criteria, propensity score matching was then used to adjust for covariates as follows. Covariates included in propensity score matching included age, sex, health plan type, modified Charlson-Quan comorbidity index (CCI) score (18,19), and baseline health care costs (as measured in the 12 months prior to the index date). Baseline T2D-related medication use and diabetes complication severity index (DCSI) score (20) were also included as covariates for T2D patients (in cohorts 1 and 2). Patients with T2D with HZ (cohort 1) were matched firstly with patients with T2D without HZ (cohort 2) and then secondly with patients without T2D with HZ (cohort 3). A 1:1 matching ratio was used (for every patient being matched in a specific cohort, one patient from the other cohort was selected). Matching was conducted using the greedy matching method (nearest-neighbor best-pair matches) (21,22), in which cases patients were first matched based on 5 digits of the propensity score. For those patients who did not match based on all 5 digits of the propensity score, patients were then matched based on 4 digits of the propensity score (and continued to 3, 2, and 1 digit, as necessary).

The primary outcomes of interest were 1) the incidence and comparative risk of HZ in patients with and without T2D, and 2) HCRU and all-cause and HZ-related costs in patients with T2D with and without HZ (and in patients without T2D with HZ). Additional outcomes of interest included HZ-related complications and diabetes medication use.

Demographic characteristics (age, sex, insurance type, and geographic region) were recorded at the HZ index date and at the first observed T2D diagnosis date. Clinical characteristics (e.g., CCI scores, specific comorbidities, DCSI score) were evaluated at baseline. T2D medication use in the matched cohorts was identified using relevant Current Procedural Terminology/National Drug Code codes, evaluated at baseline and across the 12-month follow-up period.

The incidence of HZ was calculated in patients with and without T2D before any enrollment criteria were applied. Patients were followed for at least 18 months or until censoring at the earlier of one of two events, incident HZ or HZ vaccination. Incident HZ was defined as at least one claim associated with an HZ diagnosis (as indicated above), with the exception of diagnosis/treatment claims associated with HZ-related complications. PHN (ICD-9-CM, 053.12; ICD-10-CM, B02.22) or postherpetic polyneuropathy (ICD-9-CM, 053.13; ICD-10-CM, B02.23) that indicate care for a previous rather than new HZ episode. PHN and other HZ-related complications (cutaneous, neurologic, and ophthalmic complications) were identified based on previously reported criteria (11,13) and evaluated in the prematched and postmatched populations.

The influence of HZ on diabetes treatment patterns was assessed by the proportion of patients receiving specific classes of antihyperglycemic medications in the T2D-matched cohorts (cohorts 1 and 2) in the baseline and follow-up periods. For the HCRU and cost analyzes, HCRU included hospital inpatient stays, emergency department visits, and hospital or physician office outpatient visits, pharmacy, and other outpatient and supplementary care. Cost outcomes included longitudinal total all-cause and HZ-related costs incurred over 12 months after the HZ episode. The incremental cost burden associated with HZ in patients with T2D was estimated as the difference in all-cause health care costs between the matched cohorts of patients with T2D with and without HZ (cohorts 1 and 2, respectively). Incremental HZ-related costs were also assessed in HZ patients with and without T2D (cohorts 1 and 3, respectively).

Descriptive analyses were performed for all study variables. For continuous variables, means, SDs, and medians were calculated; frequency counts and percentages were calculated for categorical variables. In the matched cohort cost analyzes, outcome measures (in the follow-up observation period) were analyzed descriptively for all three cohorts.

HZ incidence was estimated in the total patient population (i.e., patients with and without T2D) before any enrollment criteria were applied. Incidence rates were calculated as the number of patients with incident HZ divided by patient-years (PYs) of observation and expressed per 1,000 PYs. Incidence was calculated for populations with and without T2D, overall and stratified by age and health plan. Risk comparisons for incident HZ between patients with and without T2D were assessed using Poisson analysis, with adjustment for age and sex in the overall analysis and for sex in the age- and health plan-stratified analyses, to generate adjusted incidence rate ratios (aIRRs) with 95% CIs and P values. Cox proportional hazards analyses were performed to evaluate demographic and clinical factors associated with the development of HZ-related complications in a subset of patients with and without T2D with incident HZ (those fulfilling eligibility for inclusion in the matched cohort analysis before propensity matching). Covariates included in the Cox proportional hazards models were cohort (i.e., T2D and non-T2D), age category, sex, and baseline CCI score (excluding scoring for T2D-related conditions). Hazards model plots were generated, and hazard ratios (HR) were calculated to determine the relative risk of HZ-related complications.

In the matched cohort analyzes, costs were extracted for each individual and aggregated within respective strata and cohorts to generate mean (SD) and median values. All source cost data were adjusted to 2018 U.S. $ values using the medical care component of the U.S. Consumer Price Index (23). Longitudinal cumulative costs were reported for each of the matched cohorts across 12 months of follow-up and for shorter time horizons (i.e., 1 month, 1 quarter postindex date). Costs and incremental cost differences were reported overall, by age-group, and by insurance plan type. T2D medication use is presented as the percentage of patients (in cohorts 1 and 2) using specific agents in the 12-month baseline and follow-up periods. The percentage of patients experiencing HZ-related complications (in cohorts 1 and 3) was also reported. All statistical analyses were conducted using the statistical software SAS 9.4 or SAS Enterprise Guide 7.1 (SAS Institute Inc., Cary, NC).

The incidence study population included 91,319,189 patients; 5,928,052 patients with T2D and 85,391,137 patients without T2D, contributing a total of 12,748,009 and 231,082,154 PYs of observation time, respectively. There were 124,683 cases of HZ in the T2D population and 596,467 cased of HZ in the non-T2D cohort. The overall crude incidence rate for HZ was 9.8/1,000 PYs in patients with T2D and 2.6/1,000 PYs among all patients without T2D (Fig. 1). HZ incidence rates progressively increased with increasing age in populations both with and without T2D. After adjustment for age and sex, compared with patients without T2Ds, patients with T2D had a higher risk of HZ, with an overall aIRR of 1.84 (95% CI 1.82–1.85). Risk of HZ varied according to age. Patients with T2D aged 18–49 years had a 4.2-times higher risk of HZ (aIRR, 4.20; 95% CI 4.10–4.30), whereas the magnitude of risk declined in older age strata; in those patients with T2D aged ≥80 years, the aIRR was 1.35 (95% CI 1.32–1.38) (Fig. 1).

A total of 2,034,660 patients aged ≥18 years identified within the three contributing MarketScan databases (CCAE, Medicare, and Medicaid) fulfilled all inclusion criteria and were eligible for propensity score matching as follows: 26,071 patients with T2D and HZ (cohort 1), 1,812,747 patients with T2D without HZ (cohort 2), and 195,842 patients without T2D and with HZ (cohort 3) (Supplementary Fig. 2). After 1:1 matching, all but one of those patients with T2D and HZ in cohort 1 could be matched to a patient in cohort 2, and all could be matched to a patient in cohort 3. As such, the final cohort numbers were as follows: 26,070 patients were included in cohorts 1 and 2 for comparisons between patients with T2D with and without HZ, and 26,071 included in cohorts 1 and 3 for comparisons between patients with HZ with and without T2D.

Baseline characteristics in the matched populations are provided in Table 1. A more complete presentation of baseline characteristics in the prematched and matched populations is available in Supplementary Table 2. In these matched populations, the mean age was ∼62 years, with the proportion aged <50 years ranging from 14.4–15.2%; in each cohort, women accounted for ∼58% of patients. Most patients were covered by commercial or Medicare insurance (86.5%). Comorbidity, as measured by the mean CCI score, was comparable across all three cohorts. Prior to matching, baseline DCSI was higher in patients with T2D with HZ than those without HZ (mean [SD]: 2.6 [2.3] vs. 1.4 [1.9]) (Supplementary Table 2). After propensity matching, this difference was smaller, although baseline DCSI remained higher in patients with T2D with HZ than in patients without HZ (mean [SD]: 2.6 [2.3] vs. 2.1 [2.2]). Some differences in all-cause health care service costs in the 12 months prior to index date were observed between the cohorts (Table 1). In those patients with T2D developing HZ, mean (SD) costs were $20,305 ($53,023) compared with $16,538 ($44,248) in those patients with T2D not developing HZ. These higher costs may in part reflect higher baseline DCSI.

Multivariable Cox regression model analysis was performed in 221,913 patients with HZ in the matched cohort analysis before propensity matching (26,071 with T2D and 195,842 without T2D). Patient age was the strongest predictor of having any HZ-related complication, with HRs increasing with increasing patient age (Supplementary Fig. 3). Compared with patients aged 18–44 years, patients aged ≥75 years had a 1.8-times greater risk of developing complications (HR 1.80; 95% CI 1.75–1.86). Compared with patients without T2D, having T2D was also associated with some risk, although any such risk was minimal (HR 1.03; 95% CI 1.00–1.05), while conversely, risk of HZ-related complications was lower in women versus men (HR 0.90; 95% CI 0.88–0.91). In the matched-cohort population the proportion of patients experiencing any HZ-related complications was similar in patients with and without T2D (30.9% and 29.9%, respectively) (Supplementary Table 3). In both cohorts, PHN was the most commonly observed HZ-related complication (24.8% and 24.1% in patients with and without T2D, respectively).

Minor differences in the proportion of patients receiving specific T2D medications were observed in the matched cohorts of T2D patients with and without HZ (cohorts 1 and 2) at baseline and at follow-up. Although our objective here was not to try to ascertain risk on the basis of current medications, baseline medication treatment patterns were broadly comparable in both cohorts (Supplementary Fig. 4). Insulin use was reported for 28.9% and 29.2%, respectively, in patients with T2D with and without HZ (and metformin in 53.8% and 51.5% and sulfonylureas in 26.7% and 28.5%, respectively). Sodium–glucose cotransporter 2 (SGLT-2) inhibitors were more frequently reported in patients with T2D who developed HZ (7.1% vs. 3.4%). Medication use in the 12-month follow-up period was broadly comparable in both T2D cohorts (although some numerical differences were observed), with a slight increase observed in use of insulin and SGLT-2 inhibitors in both cohorts and a small decline in metformin and sulfonylurea use, although percentage changes were small. Some evolution in disease severity was observed in both T2D cohorts in the 12-month follow-up period, with the mean (SD) DCSI score increasing from 2.6 (2.3) at baseline to 3.3 (2.4) in patients with T2D with HZ and from 2.1 (2.2) to 2.5 (2.5) in patients with T2D without HZ.

Patients with HZ (with and without T2D) incurred a greater number of HCRU events than patients with T2D without HZ in the first month postindex date (Supplementary Table 4). At 1 month, 4.9% of patients with T2D with HZ and 4.0% of patients without T2D with HZ (cohort 1 and cohort 3) incurred inpatient hospitalizations compared with 2.1% of patients with T2D without HZ (cohort 2). The proportion incurring emergency department visits in the first month were 19.6% and 17.4%, respectively, for patients with and without T2D with HZ versus 4.9% of patients with T2D without HZ (Supplementary Table 4). Although the relative differences in HCRU use in these cohorts declined over the full 12 months of follow-up, cumulative HCRU at 12 months postindex date remained greatest in the patients with T2D with HZ for all visit categories.

Longitudinal all-cause and HZ-related health care costs and incremental cost differences over the 12-month follow-up period in the matched cohorts are presented in Figs. 2 and 3 and in Supplementary Table 5. Unadjusted all-cause health care costs in the first month were $1,231 higher for patients with T2D with HZ compared with those without HZ (mean [SD]: $3,497 [$11,772] vs. $2,265 [$8,903], respectively) (Fig. 2). Cumulative all-cause total health care cost differences were observed across the entire 12-month follow-up period, with an incremental cost difference of $1,868 after the first quarter postindex date. At 12 months, costs in patients with T2D with HZ were $5,216 higher than those in patients with T2D without HZ. All-cause health care costs were higher in older age-groups, with notable cost differences at 3 months in patients with T2D with HZ ≥65 years compared with those without HZ (Supplementary Fig. 5). Patients covered by commercial or Medicaid insurance had similar patterns in all-cause health care costs in the first quarter of the follow-up period (again highest in those patients with T2D with HZ), with higher costs observed in Medicare beneficiaries (Supplementary Fig. 6), which reflects the older population in this insurance plan.

HZ-related costs in the first month after the HZ index date were $306 higher for patients with T2D than for those without T2D (mean [SD]: $1,622 [$6,110] vs. $1,317 [$6,064], respectively) (Fig. 3 and Supplementary Table 5). Cumulative HZ-related costs remained higher in patients with T2D throughout the entire 1-year follow-up period. At 12 months, the mean (SD) costs were $9,077 ($23,040) in patients with T2D and $6,351 ($18,132) in patients without T2D, an incremental cost difference of $2,726.

In this retrospective claims database analysis, we evaluated HZ incidence in a large population of patients with and without T2D and found a higher HZ incidence in patients with T2D compared with patients without T2D (with crude HZ incidence rates of 9.8 and 2.6/1,000 PYs respectively). In a recent meta-analysis of cohort studies in patients with diabetes, Lai et al. (9) reported pooled incidence rates of 7.2 and 4.1/1,000 PYs in patients with and without T2D, respectively. In the current study, HZ incidence progressively increased with increasing age and was highest in older patients, evident in the populations both with and without T2D, again consistent with previous observations in other studies (6,9,10). However, when looking at incidence rates stratified by age, incidence was higher in patients with T2D in all age strata (and substantially higher in those patients aged 18–49 years). This is reflected in our risk estimations using Poisson analysis, where we found that patients with T2D have a 1.8-fold higher risk for HZ compared with patients without T2D. This is slightly higher than previously reported risk estimates for HZ in patients with diabetes. In previous meta-analyses, Kawai et al. (6) reported a pooled risk ratio of 1.30 (95% CI 1.17–1.45), and Lai et al. (9) reported an IRR of 1.60 (95% CI 1.33–1.93). In a previous claims-based analysis (using 2005–2009 data) in a U.S. cohort, Suaya et al. (10) reported an adjusted HR of 1.45 (95% CI 1.43–1.46).

Nearly one-third of patients with and without T2D who experienced HZ had HZ-related complications, with PHN being the most commonly observed complication. Multivariable regression analysis showed that increasing patient age was the strongest predictor of HZ-related complications; T2D had minimal impact on HZ-related complication risk. Previous studies evaluating the impact of diabetes on HZ complications are conflicting, and most focus on the risk of PHN. In their systematic review and meta-analysis, Forbes et al. (2) found that any significantly greater risk of PHN in patients with diabetes with HZ was reported in a few, but not all studies, one exception being a study from Taiwan, reporting a 1.35 relative risk of PHN in diabetes (24). In their U.S. study (not included in that meta-analysis), Suaya et al. (10) reported increased risk of persistent pain in patients with diabetes compared with patients without diabetes, with an adjusted odds ratio of 1.18 (95% CI, 1.13–1.24).

In our propensity-matched cohorts, minor changes were observed in diabetes medication use in both cohorts of patients with T2D (either with or without HZ) in the 12-month follow-up compared with the previous 12 months. These changes were small, however, broadly similar in both cohorts, and whether HZ had any direct impact on T2D treatment remains unclear.

We found that HZ episodes are associated with increased HCRU and all-cause health care costs. During the 1-year follow-up period, mean incremental all-cause health care costs for patients with T2D with HZ versus patients with T2D without HZ were $5,216. This is higher than that previously reported (in an immunocompetent population) where the annual incremental all-cause health care cost associated with HZ was $2,564 (13). We also found that mean incremental HZ-related health care costs for patients with T2D with HZ versus patients with HZ without T2D were $2,726. We should note that the costs we report represent only direct costs incurred through health plan paid amounts and patient copayments and do not account for additional out-of-pocket costs incurred, resulting in potential underestimation.

One point worth considering is the different patterns for HZ incidence, risk, and cost outcomes we observed in patients covered by different health plans. When considering our findings in this context, we find that the patterns chiefly reflect the age strata covered by these different plans. Hence, Medicare patients have a higher HZ incidence, lower relative HZ risk between cohorts (as reflected in the aIRR), and greater longitudinal health care costs, all of which are consistent with our general findings across the older population (indeed most of our older patients were derived from this database). In a similar manner, the lower HZ incidence and higher risk (notably for Medicaid patients) seen for patients covered by commercial insurance/Medicaid generally align with the findings for the younger or mixed-patient populations we report.

Given the increased risk of HZ in patients with T2D and the associated high HCRU and cost burden compared with patients without T2D, prevention of HZ in individuals with T2D might be particularly beneficial. The recombinant zoster vaccine (RZV) is recommended by the Advisory Committee on Immunization Practices (ACIP) for prevention of HZ in adults ≥50 years of age, where efficacy in preventing HZ is >90% (25). The ACIP has also recently recommended use of RZV in individuals aged ≥19 years who are or will be immunodeficient or immunosuppressed because of disease or therapy (26). Since its introduction in the U.S., RZV uptake has been increasing (with 4.28 million patients completing the two-dose series between October 2017 and September 2019), although the coronavirus disease 2019 pandemic has impacted uptake in more recent years (27,28). The American Diabetes Association highly recommends that adults ≥50 years of age with diabetes should receive HZ vaccination according to age-appropriate recommendations (29).

This study has several limitations. Firstly, it was conducted using administrative claims data with limitations common to all such retrospective database analyses, such as billing record miscoding, missing data, and lack of detailed clinical data. While we used validated ICD-9-CM and ICD-10-CM codes to identify patients with HZ and T2D, this may have failed to identify some eligible patients, including patients with less severe HZ not seeking medical care. Consequently, HZ case numbers and resultant incidence estimations may be underestimated.

Secondly, in our study, patients with T2D accounted for 6.4% of the overall population, while prevalence in the broader U.S. adult population is higher (estimated as 8.6% for 2016) (30). It seems likely that the lower prevalence of T2D in the current study population simply reflects the number of patients with T2D enrolled across the claims databases at the time of the study. It is well recognized that the prevalence of specific conditions in claims database populations do not fully align with broader population estimates. This is illustrated in another recent study evaluating diabetes patterns and costs using an alternative U.S. claims database for 2016, where the prevalence of T2D was ∼7.3% (31). As such we believe there was little if any potential for selection bias in our incidence estimations.

Thirdly, our HCRU and cost analysis was performed in propensity-matched populations, and while we included a robust range of covariates, we did not formally evaluate this via pairwise comparisons, and it remains that residual confounding may have influenced our results. Patients with T2D with HZ had higher disease severity (as measured by DCSI) at baseline and higher all-cause health care service costs in the 12 months prior to the index date. These patients may also have had T2D of longer duration than those without HZ, which may have influenced our HCRU and cost outcomes. In addition, the observed costs we report are unadjusted (so do not account for such potential confounding) and show substantial variation (with often large SD). Our findings should be considered in light of these limitations.

Finally, this study examined patients in select commercial, employer-sponsored, Medicare supplemental, and Medicaid plans, and may not be generalizable to other populations, including patients covered by other public plans or uninsured individuals.

In conclusion, we found that patients with T2D were at greater risk of developing HZ. Furthermore, patients with T2D with HZ accrued higher health care costs than those without HZ.

Acknowledgments. The authors would like to thank Elizabeth La (GSK) for assistance with manuscript preparation. The authors would also like to thank Business & Decision Life Sciences platform for editorial assistance and manuscript coordination, on behalf of GSK. Amrita Ostawal and Iain O’Neill, on behalf of GSK, provided medical writing support.

Duality of Interest. GlaxoSmithKline (GSK) Biologicals SA funded this study (Study identifier: VxHO-000044) and was involved in all stages of study conduct, including analysis of the data. GSK Biologicals SA also covered all costs associated with the development and publication of this manuscript. J.-E.P., B.J.P., and L.I.G. were employees of GSK and held shares in GSK. J.-E.P. is currently an employee of Parexel Belgium. B.J.P. is currently an employee of Janssen Global Services and holds shares in Janssen Global Services. L.I.G. is currently an employee of AstraZeneca and holds shares in AstraZeneca. S.P.N. and J.L.M. are employees of RTI Health Solutions, which was contracted by GSK for the conduct of this study. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. J.-E.P., J.L.M., S.P.N., B.J.P., L.I.G., and S.A.J. conceived and designed the study. J.L.M. and S.P.N. acquired or generated the data, and all authors analyzed and/or interpreted the data. All authors participated in the development of this manuscript and in its critical review with important intellectual contributions. All authors had full access to the data and gave approval before submission. All authors agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. J.-E.P. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

ICMJE Statement. The work described was performed in accordance to ICMJE recommendations for conduct, reporting, editing, and publishing of scholarly work in medical journals. The corresponding author had responsibility to submit the final manuscript for publication.

Prior Presentation. Parts of this study were presented in abstract form at the 80th Scientific Sessions of the American Diabetes Association, virtual meeting, 12–16 June 2020.