FormalPara Key Summary Points

Why carry out this study?

In people with type 2 diabetes (PwT2D) who also have obesity, efforts targeting weight loss are recommended

We provide statistics on the prevalence of overweight/obesity and glycemic control among PwT2D in Europe and Australia and evaluated the association between glycemic control and obesity using recent real-world data and applying consistent methodology across countries

What was learned from the study?

Across countries/databases, proportion of patients with baseline glycated hemoglobin A1c (HbA1c) ≥ 7% ranged from 36.8–58.0% while proportion with obesity (body mass index [BMI] ≥ 30.0 kg/m2) ranged from 39.4–56.7%

Higher BMI was associated with poorer glycemic control, and for most countries/databases, patients with obesity were associated with substantially lower odds of post-index HbA1c below target

Therapies that both improve glycemic control and reduce weight could have substantial impact towards improving health outcomes in PwT2D

Introduction

Globally, 537 million adults were living with diabetes in 2021, and this figure is expected to increase to 783 million by 2045 [1]. An estimated 90% of people with diabetes have type 2 diabetes (T2D) [2].

Obesity and T2D share key multi-system pathophysiological mechanisms, and obesity is an important modifiable risk factor for the development of T2D [3, 4]. Weight loss appears to reverse the underlying metabolic impairments of T2D. In people with T2D (PwT2D) who also have obesity, efforts targeting weight loss, including lifestyle, healthy diet, regular physical activity, medication and surgical interventions, are recommended [5]. Modest and sustained weight loss has been shown to improve glycemic control and reduce the need for glucose-lowering agents (GLAs) [6]. The Look AHEAD (Action for Health in Diabetes) trial, conducted among PwT2D with overweight and obesity, found that loss of body weight was associated with improved overall fitness, reduced glycated hemoglobin A1c (HbA1c) levels, reduced cardiovascular risk factors [7] and reduced depression symptoms [8]. Guidelines recommend a target HbA1c < 7% for most adult PwT2D [5, 9].

A retrospective database study using linked claims and electronic medical records (EMR) data in the USA investigated the relationship between HbA1c and body mass index (BMI) among PwT2D. During the 2012–2019 study period, higher BMI was associated with higher HbA1c overall [10]. Another study using the same database and study period observed similar trends [11]. Compared to PwT2D with normal weight, those with class I to III obesity were 16–24% more likely to have HbA1c ≥ 7%. A literature review evaluated the association between glycemic control and obesity in PwT2D based on real-world studies and found greater levels of HbA1c in individuals with higher BMI, and vice versa [12]. In addition, those who had overweight or obesity were less likely to achieve glycemic control targets.

Despite recently published real-world data on the association between HbA1c and BMI in the US, there are limited published data specific to Europe or Australia. While general statistics on the prevalence of T2D and obesity are available in Europe and Australia, a thorough evaluation of the relationship between glycemic control and obesity is lacking. The objective of this study was to evaluate the association between glycemic control and obesity among PwT2D in Europe (France, Germany, Italy, Spain and UK) and Australia using recent real-world data and consistent methodology across countries. It was our hypothesis that higher BMI would be associated with worse glycemic outcomes. This study was also conducted to provide statistics on the prevalence of overweight/obesity and glycemic control among PwT2D in the countries of interest. Given the importance of weight management in the clinical treatment of T2D [4], the findings may add to the real-world evidence supporting the need for targeted interventions for glycemic control and weight reduction in PwT2D. The results of our study may aid in a better understanding of the PwT2D populations in these countries and, in particular, those that may benefit the most from such targeted interventions.

Methods

Study Design and Data Sources

This was a retrospective database study that utilized IQVIA EMR databases in France (Longitudinal Patient Database [LPD]), Spain (LPD), Italy (LPD), Germany (Disease Analyzer [DA]), UK (IQVIA Medical Research Data [IMRD] incorporating THIN, a Cegedim Database) and Australia (EMR). The databases provide EMR data sourced from physician offices. The group of physicians that contribute to the database are referred to as a panel. Depending on the country, different panels are available and separated based on physician specialty type (e.g., general practitioner [GP] panel or endocrinologist/diabetologist [E/D] panel). GP panels were used in Italy (n = 900 contributing GPs), UK (n = 833 GP practices) and Australia (n = 2500 GPs). GP and E/D panels were analyzed separately in France (n = 1210 GPs and n = 40 E/Ds) and Germany (n = 1540 GPs and n = 80 E/Ds). In the Spanish database, patients could be followed across physician specialties (including, but not limited to, n = 1300 GPs and n = 34 E/Ds). The databases have been previously described, and the patient populations in each country are generally representative of the respective country population according to age and sex distribution, as provided by national statistic authorities [13,14,15,16,17,18]. Prevalence of T2D in the databases are generally consistent with respective country prevalence. PwT2D were identified in each country/panel between 2014 and 2019. Each patient had ≥ 1 record of HbA1c and BMI measured within 90 days of each other between 2015 and 2018. This 90-day window was selected to best maximize the potential sample while considering the recommended frequency of HbA1c testing (every 3 months or greater) [9]. Patients were characterized using a 1-year baseline period. Post-index HbA1c targets (above or below HbA1c target: ≥ or < 7%) were assessed at the end of the 1-year follow-up. See Supplementary Fig. S1 in the electronic supplementary material for an overview of the study design.

Study Population

An initial population with ≥ 1 diagnosis for diabetes was identified during the study period (January 2014 to December 2019). Diabetes was identified through diagnosis codes for general diabetes or T2D, not including type 1 diabetes (T1D). Patients were then required to have ≥ 1 record of HbA1c and BMI measured within 90 days of each other in the selection window (January 2015 to December 2018). The ‘index date’ was defined as the date of the second record in the first observed HbA1c/BMI combination during the selection window.

Eligible patients had ≥ 1 diagnosis for diabetes in the 1-year pre-index period, were ≥ 18 years old at index, had ≥ 12 months of history in the database (based on activity prior to the 1-year pre-index period) and had ≥ 1 recorded HbA1c at 1-year post-index (index date + 359 [± 90 days]). Patients were excluded if they had diagnosis of gestational diabetes, pregnancy, T1D or cancer, or BMI < 18.5kg/m2 during the 1-year pre- or post-index periods.

Eligibility criteria were standardized across countries with a few differences following country-specific norms. In UK and Australia, due to the absence of International Classification of Diseases (ICD) diagnosis codes, Read codes were used in UK and search terms in Australia; patients with prediabetes in the 1-year pre-index period were also excluded. In Australia, diabetes was identified based on search terms for diabetes and/or GLA use because of under-recording of diagnoses.

Study Measures

Baseline Demographic and Clinical Characteristics

Demographics (e.g., age, sex) were collected as of the patient’s index date. Baseline comorbidity and GLA use were assessed over the 1-year pre-index period (not including the index date). Comorbidity was identified through ICD codes, Read codes or search terms, as applicable, and included Charlson Comorbidity Index (CCI) score and Diabetes Complications Severity Index (DCSI) score [19]. Baseline GLA use was evaluated based on written prescriptions.

Baseline HbA1c, Baseline BMI and Post-Index HbA1c

The HbA1c and BMI records on or closest (prior) to the index date were taken as the baseline values. Baseline HbA1c was reported categorically and by HbA1c target (above or below target; ≥ or < 7%). Baseline BMI was reported by classification (normal—18.5 to < 25.0; overweight—25.0 to < 30.0; Class I obesity—30.0 to < 35.0; Class II obesity—35.0 to < 40.0; Class III obesity— ≥ 40.0 kg/m2) and by obesity (with or without obesity; ≥ or < 30.0 kg/m2). The HbA1c records at 1-year post-index were taken from the record on or closest (absolute) to index date + 359 days (± 90-days). Baseline and post-index HbA1c (above or below target) were reported by baseline obesity.

Statistical Analyses

Descriptive statistics were reported using frequency and percentage distributions for categorical variables. Continuous and count variables were described using mean, standard deviation (SD) and median. As relevant, continuous variables were also categorized into intervals. Analyses were conducted separately by country/panel, except for the overall Spanish analysis where patients could be followed across physician specialties.

Logistic regression models were developed to evaluate the association between baseline obesity (BMI ≥ 30.0 kg/m2) and odds of having HbA1c below target (HbA1c < 7%) at the end of the 1-year follow-up while controlling for baseline covariates. The stepwise model selection procedure was used to select covariates to be included in a model. The stepwise selection process combines alternating forward selection and backward elimination steps. Forward selection steps add variables to the model, while backward elimination steps remove variables from the model. A variable had to be significant at the 0.2 level to be entered into the model and significant at the 0.05 level to stay (or remain) in the model.

Per protocol, baseline HbA1c and baseline obesity were included as covariates in the models. However, baseline HbA1c was found to dominate the models, and baseline BMI was dropped from the final model for seven of eight countries/panels. Baseline HbA1c < 7% accounted for so much of the variation in the dependent variable (post-index HbA1c < 7%) that we could not estimate the influence of other variables, in particular, baseline BMI. Given our study objective, post hoc models were developed excluding baseline HbA1c to understand the relationship between baseline obesity and post-index HbA1c. In addition, given the dominance of baseline HbA1c in the per-protocol models, post hoc logistic regression models were also developed to explore the association between baseline obesity and odds of baseline HbA1c below target. This allowed us to evaluate the relationship between HbA1c and BMI using more proximal values. Baseline obesity was forced if not retained in these post hoc models.

Other relevant baseline patient characteristics with a potential relationship with HbA1c were considered for entry in the models: age (continuous), sex, CCI score, DCSI score and baseline GLA use (mutually exclusive categories). Country-specific variables were also considered as available. Collinearity was evaluated as part of model development.

All analyses were conducted using SAS® Release 9.4 or above (SAS Institute Inc., Cary, NC). This study involved a retrospective analysis of de-identified data from several countries. Ethical review approval was obtained as required in Spain from the Clinical Research Ethics Committee of the Hospital Clinic de Barcelona (reference no. HCB/2022/0311) and in the UK from the Scientific Review Committee (SRC; reference no. 22SRC017). In France, the study was compliant with reference methodology MR-004 from the Commission Nationale Informatique & Libertés (CNIL), The French Data Protection Agency. Therefore, a submission to CNIL was not required, but a Privacy Impact Assessment was completed and registered. Ethics approval was not required for the analysis of the de-identified data in the other countries. IQVIA had permission to access and analyze the data in each country. This study was performed in accordance with the Helsinki Declaration of 1964 and its later amendments.

Results

Study Sample

A starting total of 1,456,898 patients were identified having evidence of diabetes during the study period (January 2014–December 2019). The attrition of study sample by country/panel can be found in Supplementary Table S1 in the electronic supplementary material. The requirement of ≥ 1 HbA1c and BMI record within 90 days of each other within the selection window (January 2015–December 2018) was the most common reason for exclusion across study countries and panels, resulting in exclusion of 34.6–88.0% of starting patients. The final sample comprised 194,729 patients or 13.4% of the starting sample. By country/panel, the final sample ranged from 719 (2.2% of the respective starting sample) patients in France E/D to 103,866 (22.9%) patients in UK.

Baseline Characteristics

Table 1 summarizes baseline demographic and clinical characteristics across countries/panels. Overall, mean (SD) age ranged from 63.9 (12.6) years in Australia to 69.5 (10.7) years in Italy, and median age ranged from 64 to 70 years, respectively. Patients were most often in the 60–69-year-old age group in France, Spain, UK and Australia (30.2–36.6%) whereas patients were most often in the 70–79-year-old age group in Germany and Italy (29.0–35.0%). Proportion male was higher across countries/panels (52.4–57.0%). Overall, most patients had index year of 2015 (44.5–79.8%).

Table 1 Baseline characteristics
Full size table

Demographic characteristics were generally similar by baseline HbA1c (below or above target, < 7% or ≥ 7%) (data not shown). With the exception of France E/D and Italy, mean age of patients with baseline HbA1c below target was generally higher (ranging from 65.5 to 68.6 years) than patients with baseline HbA1c above target (ranging from 62.6 to 66.5 years).

Mean (SD) CCI score ranged from 0.2 (0.5) in UK to 1.0 (1.3) in Germany GP, and mean DCSI (SD) score ranged from 0.2 (0.6) in UK and Australia to 1.3 (1.7) in Germany GP. Proportion with cardiovascular disease (CVD) ranged from 3.9% in UK to 36.1% in Germany GP. See Supplementary Table S2 in the electronic supplementary material for CVD diagnosis codes.

With the exception of Australia, oral GLA use ranged from 60.0% in Germany E/D to 87.5% in France GP, and biguanides were the most commonly used GLA class (40.9–67.3%). In Australia, 96.1% used baseline oral GLA (and 69.9% used biguanides, specifically). Note that patients with diabetes in Australia were identified based on diagnosis or GLA search terms. GLP-1 RA use was higher in E/D panels (10.6–16.3%) and was otherwise infrequent (< 5%). Similarly, insulin use was higher in E/D panels (38.1–50.8%) and otherwise less common (7.8–20.8%). GLA use was also evaluated in mutually exclusive categories. Use of oral GLA only was most common.

Baseline HbA1c and BMI

Mean (SD) baseline HbA1c ranged from 6.9 (1.3)% in Germany GP to 7.6 (1.6)% in UK and median ranged from 6.6% in Germany GP to 7.2% in France E/D and UK (Table 1). Proportion of patients with baseline HbA1c ≥ 7% ranged from 36.8% in Germany GP to 58.0% in UK. In France and Germany, proportion with baseline HbA1c above target was lower for GP compared to E/D panels (France 40.9% and 57.3%; Germany 36.8% and 44.7%). Mean (SD) baseline BMI ranged from 29.3 (5.3) kg/m2 in Italy to 31.6 (6.3) kg/m2 in UK and median ranged from 28.6 to 30.7 kg/m2, respectively. Proportion with obesity (BMI ≥ 30.0 kg/m2) ranged from 39.4% in Italy to 56.7% in Australia. In France, proportion with obesity was higher for the GP compared to E/D panel (49.2% and 43.1%).

Within each country/panel, patients with obesity had a higher proportion with baseline HbA1c above target (ranging from 40.8% in Germany GP to 64.2% in France E/D) compared to their counterparts without obesity (ranging from 32.2% in Germany GP to 52.4% in UK) (Fig. 1). Difference in proportion with baseline HbA1c above target between patients with and without baseline obesity was highest in France E/D (64.2% and 52.1%) followed by UK (62.6% and 52.4%).

Fig. 1
figure 1

Baseline HbA1c (above or below target) by baseline BMI (with or without obesity) and country/panel. BMI body mass index, E/D endocrinologist/diabetologist, GP general practitioner, HbA1c glycated hemoglobin A1c

Full size image

HbA1c at 1-Year Post-Index

Mean (SD) HbA1c at 1-year post-index ranged from 6.8 (1.0)% in Germany GP to 7.5 (1.5)% in UK and median ranged from 6.6% in Germany GP to 7.1% in France E/D and UK. Across countries/panels, mean 1-year post-index HbA1c values were the same or slightly lower as compared to baseline, with difference ranging from 0.0 to 0.2%. The proportion of patients with post-index HbA1c ≥ 7% at 1-year post-index ranged from 34.7% in Germany GP to 58.0% in France E/D. With the exception of France E/D where a marginal increase was observed, proportion of patients with post-index HbA1c above target slightly decreased from baseline. While small, the greatest decrease from baseline to post-index was seen in Spain (44.3–40.2%).

Within each country/panel, patients with obesity at baseline had a higher proportion with post-index HbA1c above target (ranging from 38.1% in Germany GP to 60.6% in France E/D) compared to their counterparts without obesity (ranging from 30.9% in Germany GP to 56.0% in France E/D). The greatest difference in proportion with post-index HbA1c ≥ 7% between patients with and without baseline obesity was seen in Germany E/D (47.0% vs. 38.6%) and the smallest difference in Spain (41.2% vs. 39.2%) (Fig. 2).

Fig. 2
figure 2

Post-index HbA1c (above or below target) by baseline BMI (with or without obesity) and country/panel. BMI body mass index, E/D endocrinologist/diabetologist, GP general practitioner, HbA1c glycated hemoglobin A1c

Full size image

Logistic Regression Models for Odds of Being Below Target HbA1c

Post hoc logistic regression models examined the association between baseline obesity and odds of HbA1c below target (< 7%) at 1-year post-index. With the exception of France E/D, Spain and Australia, patients with obesity were associated with substantially lower odds of post-index HbA1c below target compared to their counterparts without obesity, ranging from 12.4% in Italy (odds ratio [OR] = 0.876; 95% confidence intervals [CI] 0.826–0.929; p < 0.0001) to 16.7% in France GP (OR = 0.833; 95% CI 0.755–0.920; p = 0.0003). Post hoc logistic regression models also explored the association between baseline obesity and odds of baseline HbA1c below target. With the exception of Spain and Australia, patients with obesity had notably lower odds of having baseline HbA1c below target compared to their counterparts without obesity, ranging from 12.2% in Germany E/D (OR = 0.878; 95% CI 0.795–0.970, p = 0.0103) to 29.9% in France E/D (OR = 0.701; 95% CI 0.500–0.984; p = 0.0399) (Fig. 3).

Fig. 3
figure 3

Odds ratios for patients with vs. without obesity of having HbA1c below target at end of 1-year follow-up and at baseline by country/panel. a Post-index; b baseline. Post hoc logistic regression models. E/D endocrinologist/diabetologist, GP general practitioner, HbA1c glycated hemoglobin A1c

Full size image

GLA use was retained in all models. Other variables in the final models varied by country/panel, but age (continuous) and CCI (1 + vs. 0) were commonly retained (Supplementary Table S3 in the electronic supplementary material).

BMI Trends by Year

In a secondary analysis, BMI trends were reported over time by calendar year. At the patient level, a patient contributed a BMI measure (as available) to the calendar years included in their 1-year pre-index and 1-year follow-up periods. BMI was generally consistent by calendar year (2014–2018) (Fig. 4). Note that 2019 was not included because of limited sample size. Mean BMI in each year was ≥ 30.0 kg/m2 in all countries/panels except for France E/D and Italy.

Fig. 4
figure 4

BMI trends by calendar year by country/panel. Note that the N is for the overall study sample and does not reflect the patient n contributing a BMI at each year. BMI body mass index, E/D endocrinologist/diabetologist, GP general practitioner

Full size image

Discussion

This study evaluated the relationship between glycemic control and obesity among 194,729 PwT2D in Europe and Australia. The study sample across countries/panels represented an older population (mean age 63.9–69.5 years), which is consistent with European and Australian statistics on diabetes [20, 21]. At baseline, more than half of patients were above glycemic targets (HbA1c ≥ 7%) in France (E/D), UK and Australia, and more than half of patients had obesity (BMI ≥ 30.0 kg/m2) in Germany, UK and Australia. Descriptively, there was little change in proportion above target at baseline compared to at 1-year post-index. Similarly, BMI was consistent across calendar years. These findings suggest that there is room for improvement of glycemic control and weight management in T2D.

Across countries/panels, a higher proportion of patients with obesity at baseline had post index HbA1c above target compared to their counterparts without obesity. With the exception of France E/D, Spain and Australia, post hoc multivariable regression analysis found that patients with obesity were associated with sizably lower odds of post-index HbA1c below target compared to their counterparts without obesity, ranging from 12.4 to 16.7%. Given the unexpected findings in France E/D, Spain and Australia, additional research is needed to further investigate the association among these populations. A retrospective claims database study in the US specifically evaluated BMI class, and compared to PwT2D with normal weight, those with class I and class II obesity were both 24% more likely to have HbA1c ≥ 7%, while those with class III obesity were 16% more likely [11].

Study results also showed an association between baseline HbA1c and baseline BMI. Across countries/panels, proportion of patients with baseline HbA1c above target was higher among patients with baseline obesity compared to those without obesity. In the regression analysis, with the exception of Spain and Australia, patients with obesity were associated with considerably lower odds of having baseline HbA1c below target compared to their counterparts without obesity. Recent guidelines from the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD) have recognized the importance of weight management in T2D [22]. The guidelines emphasize weight reduction as a targeted intervention to improve HbA1c and reduce the risk for weight-related complications. A greater magnitude of weight loss results in better outcomes ranging from metabolic improvement to disease-modifying effect and remission of diabetes. Other benefits may include improvements in risk factors for cardiometabolic disease and increases in quality of life. The findings from our study support the influence of weight on HbA1c (even if measured at the same time).

The relationship between glycemic control and obesity has also been demonstrated in several other real-world studies. A retrospective claims database study of PwT2D in the US from 2012 to 2019 found higher BMI was associated with higher HbA1c, particularly among age group 18–44 years [10]. However, the US sample described in that study appeared to have higher HbA1c and BMI relative to the European and Australian populations in the current study. A literature review evaluated the association between glycemic control and obesity in PwT2D based on real-world studies and found most studies reported a lower chance of achieving glycemic control targets in those with overweight or obesity [12]. Most studies found a similar trend of positive correlation between HbA1c and BMI (e.g., HbA1c values were greater among individuals with higher BMI). Prior literature in Germany, Spain and Scotland has analyzed the relationship between HbA1c/glycemic control and BMI/obesity, but the data are more than a decade old and study designs and populations have varied [23,24,25]. A relatively recent population-based cohort study in Germany included healthy children and adolescents recruited through 2017 and found that obesity was associated with higher HbA1c values [26]. Prior research in the US has demonstrated that short-term weight loss is associated with increased odds of achieving HbA1c < 7% [27].

This study had some important strengths which are worth noting. We analyzed various EMR datasets providing a comprehensive overview of glycemic control and BMI statistics across Europe and Australia. Many datasets were relatively large. While GP data were used in all countries, the use of E/D panels (as available in France and Germany) and all specialties in Spain allowed for additional insights and a broader view. The association between baseline BMI and baseline/1-year post-index HbA1c was evaluated in multivariable logistic regression models controlling for relevant factors which may impact glycemic control.

There were several limitations of this study related to data sources and study design, which should be considered when interpreting the results. First, the study was limited to PwT2D who visit office-based physicians and are not representative of all the patients in each respective country. In particular, the France E/D sample size was notably small. The study sample may be biased towards capturing patients who more frequently have reported values for HbA1c and BMI. BMI may be more likely to be reported for patients that are overweight or with obesity and/or with some comorbidity where overweight or obesity is a risk factor. This study relied exclusively on BMI values, and our study sample did not specifically capture any additional patients with diagnosis codes indicating BMI or obesity. Second, data are available for a patient from one practice only (except for Spain). In countries where only GP data were available, specialist management was not captured. Nevertheless, it is anticipated that a small percentage of PwT2D are treated for diabetes in specialists’ settings only and only a minority of PwT2D with more complex disease are treated in specialists setting in addition to GP settings for their T2D [28, 29]. GPs are expected to be the primary caregiver for most PwT2D in Europe. Third, the information provided by the physicians in health records may be underreported as it is not collected for research purposes. Incident comorbidities occurring over the 1-year baseline period may be more likely to be recorded than prevalent comorbidities. Severe complications or comorbidities may not be captured in the outpatient physician data (or GP data, specifically). Finally, the logistic regression models could only adjust for available and measured patient factors. Duration of diabetes, insurance information and income were not available in the data, and availability of lifestyle risk factors, race/ethnicity and socioeconomic status data were unavailable or extremely limited even if available. Future research is warranted in data sources with availability of these important socioeconomic patient characteristics.

Conclusion

This study presents statistics on HbA1c and BMI among PwT2D in Europe and Australia. The uses of recent real-world data and consistent methodology across countries were key strengths of this study. Higher BMI was associated with poorer glycemic control. Therapies that both improve glycemic control and reduce weight could have substantial impact towards improving health outcomes in PwT2D.