Association of type 2 diabetes remission and risk of cardiovascular disease in pre‐defined subgroups

Abstract Aim To quantify the association between type 2 diabetes remission and 5‐year incidence of cardiovascular disease outcomes, overall and in pre‐defined subgroups. Methods Retrospective cohort analysis of 60,287 adults with type 2 diabetes from the Care and Health Information Analytics (CHIA) database. Multivariable Cox models were used to assess the association between remission within the first two years of follow‐up and incidence of cardiovascular disease (CVD) outcomes including events, microvascular and macrovascular complications at 7‐year follow‐up. Effect modification by age, sex, diabetes duration, pre‐existing CVD, baseline body mass index (BMI) and HbA1c level was assessed. Results 7489 (12.4%) people achieved remission during the first two years of follow‐up. Overall, remission was associated with lower risk of CVD outcomes. Remission was associated with lower risk of microvascular complications for younger compared with older age groups (eg aHR: 0.59 (0.41–0.84) and aHR: 0.78 (0.67–0.92) for those aged <45 years and 75–84 years, respectively). Amongst those achieving remission, those with no or 1–2 comorbidities had lower risk of microvascular complications (aHR: 0.65 (0.56–0.75)) compared to those with more than three comorbidities (aHR: 0.83 (0.69–0.99), respectively). There were no significant interactions in the remaining subgroups or for models assessing CVD events or macrovascular complications. Conclusions Achieving remission of type 2 diabetes is associated with a lower risk of microvascular complications, particularly for younger groups and those with fewer comorbidities. Targeted interventions that focus on promoting remission in these groups may reduce the impact of microvascular complications and associated health costs.


| INTRODUC TI ON
Type 2 diabetes is a progressive chronic condition affecting over 460 million adults globally and is associated with significantly increased risk of cardiovascular disease (CVD) and mortality. 1,2 CVD costs the National Health Service (NHS) approximately £9 billion a year, with much higher costs to wider UK economy. 3 A growing body of evidence from clinical trials has shown that biochemical remission of diabetes, defined as a level of glycaemia below a diagnostic threshold (HbA 1c < 6.5% or 48 mmol/mol) in the absence of pharmacological or surgical intervention, is achievable through lifestyle management. [4][5][6][7][8][9] In the Look AHEAD study, participants who achieved weight loss during the intensive lifestyle intervention experienced reduced cardiovascular events over a median follow-up of 9.6 years. 10 However, 15% of participants with well-controlled diabetes and poor self-reported general health experienced increased risk of CVD events and CVD-related mortality. Averaged across 4 years, intensive lifestyle intervention participants had a greater percentage of weight loss than diabetes support and education participants; however, in participants without CVD at baseline, CVD events were reduced by 14% (274 vs 240; 1.42 vs 1.23%) while those with CVD (14%) experienced an increase in CVD events of 13% (144 vs 163; 5.92 vs 6.56%), although this difference was statistically insignificant. 11 Recently, the Association of British Clinical Diabetologists (ABCD) and the Primary Care Diabetes Society (PCDS) have acknowledged that any weight loss-including unintentional weight loss-may contribute to remission. 12 It is likely that remission may be associated with reduced risk of cardiovascular events for some people; however, for others (eg those with pre-existing CVD or cancers) weight loss and remission may be harmful and associated with adverse events.
There is currently a lack of studies exploring whether the association between remission and CVD risk varies by certain subgroups.
It could be valuable to understand determinants that predict CVD events in people who achieve remission to allow interventions to be targeted and personalised. In the present study, we therefore sought to examine the remission-CVD link overall and in pre-defined subgroups using a large population-based cohort with established type 2 diabetes from routine clinical care.

| Study population
We used data from the Care and Health Information Analytics database (CHIA), a pseudo-anonymised live electronic database linking routinely collected primary care data for approximately 1.5 million people from 150 general practitioner (GP) practices across Hampshire and Isle of Wight (Southern England, UK) with clinical biochemistry data from local hospital laboratories. We identified a cohort of 60,715 adults (aged 18-85 years) over 7 years who were coded for type 2 diabetes based on the Quality and Outcomes Framework (QOF) Read code criteria prior to 1 January 2013, and who had continuously recorded electronic records over seven years from the 1 January 2013 to 1 April 2020.

| Exposure definition
Remission (at any point during the first two years of the study period: 1 January 2013-2031 December 2015) was defined as two consecutive HbA 1c level <48 mmol/mol (6.5%) measurements separated by a minimum period of 6 months in the absence of diabetes medications (assessed in 6-month intervals) or bariatric surgery. 9

| Outcomes
The outcomes of interest were CVD outcomes including events (defined as a composite of myocardial infarction (MI), amputation and stroke), microvascular complications (comprising peripheral neuropathy, retinopathy and nephropathy) and macrovascular complications (comprising stroke, MI, coronary heart disease (CHD), peripheral arterial disease (PAD) or amputation) occurring between years 2 and 7 of follow-up. QoF definitions for each diagnosis were used.

| Statistical analysis
Missing data on biochemistry variables, remission status, baseline weight and IMD were assumed to be missing at random and multiply imputed using chained equations using STATA SE 16.0. Ten cycles of imputation were used. All imputed data after patient death were recoded as missing. Missing data on ethnicity were recoded as white ethnicity.
Descriptive statistics were used to compare baseline sociodemographic and clinical characteristics for individuals who did and those who did not achieve remission during the first 2 years of follow-up. Cox proportional hazards models were fitted to estimate the association between remission in the first 2 years of follow-up and 5-year incidence of CVD events, macrovascular complications and microvascular complications. These models included only those alive at the end of the first 2 years of follow-up. The mid-point of the quarter of death was used as exact date of death was not available. Multivariable models were adjusted based on a priori reasoning for age, sex, ethnicity, IMD, pre-existing CVD, BMI, diabetes duration, number of comorbidities and clustering within practices (all at baseline). We additionally adjusted for last observed BMI and hypertension status. We then modelled interaction terms between remission and age (<45, 45-54,55-64, 75-84, 85+), sex (male, female), diabetes duration (<5, 5-<10, 10-<20, 20+ years), pre-existing CVD (no/yes; defined as a com-

| Study cohort characteristics
The study cohort included 60,287 people with type 2 diabetes for whom remission status could be assessed (ie had HbA 1c data at baseline and at least one follow-up measurement). The cohort was followed for a mean of 6.8 (SD = 1.3) years. Table 1  Those achieving remission were more likely to be older, female, nonsmokers, living in less deprived areas and with a shorter duration of diabetes (Table 1). Univariate-and multivariate-adjusted associations are presented in Table 2. Similar results were observed amongst those who did not have pre-existing CVD, microvascular or macrovascular complications at baseline. Table 3 presents the results of remission status-subgroup variable interactions. There were no significant interaction effects for models on CVD events or macrovascular complications. The association between remission and microvascular complications was  Table 4. Gender, BMI, pre-existing CVD, duration of diabetes and HbA 1c level at baseline did not modify the association between remission and microvascular complications. Similar results were observed amongst those who did not have pre-existing CVD, microvascular or macrovascular complications at baseline.

| DISCUSS ION
In this large population-based cohort of 60,287 people with type 2 diabetes, we found that remission was associated with a substantially lower risk of CVD outcomes including events, microvascular and macrovascular complications. The effect on microvascular complications was modified by age and number of comorbidities.
Younger age and fewer comorbidities amongst those achieving remission were associated with a lower risk of microvascular complications. Sex, diabetes duration, pre-existing CVD, BMI and HbA 1c level did not modify the remission-CVD association.
This study extends the literature by examining the association between remission (in the absence of medication or surgery) of type 2 diabetes and CVD outcomes. Although the literature around glucose control and CVD outcomes (especially microvascular and macrovascular complications) is mixed, [15][16][17] our findings are consistent with a few meta-analyses from clinical trials examining the effect of intensive glucose control on CVD that reported glucose control was associated with a lower risk of CVD. 18,19 Remission is defined biochemically by lower glucose levels so it is plausible that these findings would be comparable.
In contrast with some previous studies, we do not find evidence of effect modification by pre-existing CVD. [19][20][21] Other studies have also reported no modification effect of pre-existing CVD on the association between glucose control and CVD events. 22 Although biochemical remission is comparable to tighter glucose control, differences in results could be explained by hypotheses in the literature, suggesting that the state of remission for at least a few months goes beyond biochemical glucose control to include reduced inflammation, decreased insulin resistance and enhanced gut hormone release, which may contribute to lower CVD risk. 23,24 It is thought that remission is linked to a reduction in adipokines such as leptin and inflammatory cytokines such as TNFα and several interleukins, as well as an increase in adiponectin concentrations, which reduces several cardio-metabolic risk factors. 25 Other explanations for these differences in the findings may be due to variations in study population such as greater diversity in age and comorbidities in the present study. Our findings on the effect of baseline BMI levels are in line with previous study that reported BMI level did not modify the association between HbA 1c and cardiovascular events in a Dutch population. 26 However, this previous study looked at a much smaller population, of whom many had vascular disease at baseline, which may have explained the absence of significant associations.
Strengths of the study include the use of a large observational cohort with extended follow-up, which allowed interaction effects to be examined, as well as the availability of data on comorbidities, medication and all HbA 1c measurements over the seven-year duration of the study period. Limitations include the presence of missing data on variables which may reflect undertesting or under-reporting in routine data. Although we were able to use multiple imputation techniques which provide valid TA B L E 2 Association between remission and incidence of CVD outcomes over five-year follow-up in the CHIA type 2 diabetes cohort Given the limited NHS resources and growing prevalence of type 2 diabetes, prioritising subgroups for targeted and personalised interventions is challenging. Our findings suggest that a focus on younger age groups and those with fewer comorbidities could be one approach with potential to reduce the impact of microvascular complications, as they are more likely to benefit most. Targeted interventions that focus on promoting remission as an achievable outcome in these groups could be justified given the potential to reduce the burden of microvascular complications and associated costs.

ACK N OWLED G EM ENT
The authors would like to thank Simon Griffin for his helpful comments at the early stages of the study.

CO N FLI C T O F I NTE R E S T S
None to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
We do not have governance permissions to share individual-level data on which these analyses were conducted since they derive from clinical record data. However, direct data requests can be made to the database (CHIA). a Cox regression models included the specified interaction term and additionally adjusted for age group, sex, diabetes duration, pre-existing CVD, total number of comorbidities, last observed BMI, last observed hypertension status, baseline BMI and clustering within practices.