Initiation of antidepressant medication in people with type 2 diabetes living in the United Kingdom—A retrospective cohort study

Abstract Introduction Depression is a common comorbidity in people with type 2 diabetes and it is associated with poorer outcomes. There is limited data on the treatments used for depression in this population. The aim of this study was to explore the rates of initiation of antidepressant prescriptions in people with type 2 diabetes in the UK and identify those most at risk of needing such treatment. Research Design and Methods This was a retrospective cohort study using data from IQVIA Medical Research Data (IMRD)‐UK data. Data from general practices in IMRD‐UK between January 2008 and December 2017 were used for this study. Results The overall rates of antidepressant prescribing were stable over the study period. The rate of initiation of antidepressant medication in people with type 2 diabetes was 22.93 per 1000 person years at risk (PYAR) with a 95%CI 22.48 to 23.39 compared to 16.89 per 1000 PYAR (95%CI 16.77 to 17.01) in an age and gender matched cohort. The risk of being prescribed antidepressant medication with age had a U‐shaped distribution with the lowest risk in the 65–69 age group. The peak age for antidepressant initiation in men and women was 40–44, with a rate in men of 32.78 per 1000 PYAR (95% CI 29.57 to 36.34) and a rate in women of 46.80 per 1000 PYAR (95% CI 41.90 to 52.26). People with type 2 diabetes with in the least deprived quintile had an initiation rate of 19.66 per 1000 PYAR (95%CI 18.67 to 20.70) compared to 27.19 per 1000 PYAR (95%CI 25.50 to 28.93) in the most deprived quintile, with a 32% increase in the risk of starting antidepressant medication (95%CI 1.22 to 1.43). Conclusions People with type 2 diabetes were 30% more likely to be started on antidepressant medication than people without type 2 diabetes. Women with type 2 diabetes were 35% more likely than men to be prescribed antidepressants and the risks increased with deprivation and in younger or older adults, with the lowest rates in the 65–69 year age band. The rates of antidepressant prescribing were broadly stable over the 10‐year period in this study. The antidepressant medications prescribed changed slightly over time with sertraline becoming more widely used and fewer prescriptions of citalopram.

• Sertraline is now the most commonly initiated antidepressant in people with type 2 diabetes in the UK.

Plain language summary
Depression is a common problem in people with type 2 diabetes and it is associated with poorer health outcomes. The aim of this study was to explore the rates of starting antidepressant medicines in people with type 2 diabetes in the United Kingdom and identify which patients with type 2 diabetes were most likely to be started on antidepressant medication. This study looked at historical data stored in electronic medical records in the IQVIA Medical Research Data (IMRD)-UK data between January 2008 and December 2017. People with type 2 diabetes were 30% more likely to be started on antidepressant medication than people without type 2 diabetes.
Women with type 2 diabetes were 35% more likely than men to be prescribed antidepressants and the risks increased with deprivation and in younger or older adults, with the lowest rates in the 65-69 year age band. The rates of antidepressant prescribing were broadly stable over the 10-year period in this study. The antidepressant medication that people with type 2 diabetes were started on changed over time, with sertraline becoming more widely used and fewer prescriptions of citalopram.

| INTRODUCTION
According to primary care registers, diabetes is the third most common long-term condition in England after hypertension and depression, and it affects around 6% of the population. Diabetes causes significant morbidity and mortality: it can shorten life expectancy by 8-10 years if diabetes is poorly controlled. 1 Around four million deaths worldwide are related to diabetes and direct healthcare costs range from 2.5% to 15% of annual heathcare budgets-around the world over 500 billion US dollars is spent on treating diabetes, with the majority of costs spent on treating diabetes related complications. 2,3 The presence of depression increases the risk of poorer outcomes in diabetes as it is associated with poor glycemic control and increased rates of complications. 4,5 Depression has been found to significantly increase the likelihood of being diagnosed with diabetes. 6,7 There are a number of studies that have examined this bidirectional relationship between diabetes and depression. The relationship between the two might be partly due to shared underlying pathophysiology driven by changes in hormones in the hypothalamus-pituitaryadrenal cortex axis and sympathetic nervous system. 8,9 Both conditions are also associated with subclinical inflammation. 10 There are also behavioural factors and complications associated with these conditions that link them with poorer self-care requiring healthier lifestyles and medication adherence. 11 The net result is a shared increase in vulnerability to these common chronic conditions and poorer outcomes (including increased mortality) where they co-exist. 12 Estimates of the prevalence of depression in people with type 2 diabetes vary widely, ranging from 0% to 39% in a systematic review. 13 However, there are not many studies on populations within the UK. A retrospective cohort study following 3225 people in Hertfordshire, UK estimated the prevalence of depression in people with diabetes as 8.5%. 14 Risk factors for developing depression include the presence of microvascular and macrovascular complications, female gender, being single, external locus of control and previous depression. 15,16 The prevalence of depression in the general population in UK has been estimated to be 3.3%. 17 When examining primary care records, depression recorded by general practitioners (GPs) has lower incidence rates that epidemiological studies as GPs increasingly use symptoms rather than labels to code the illness and the incidence rate from studies on electronic records has been estimated at 2.5%. 18 Looking at rates of antidepressant prescriptions avoids such risks of coding issues for common mental health problems and can help identify patients most at risk of needing to start antidepressant medication and likely to be higher risk of developing diabetesrelated complications.
The aim of this study was to explore the rates of initiation of antidepressant prescriptions in people with type 2 diabetes in the United Kingdom and identify those most at risk of needing such treatment.

| RESEARCH DESIGN AND METHODS
This was a retrospective cohort study using data from IQVIA Medical Research Data (IMRD)-UK data. The IMRD-UK database contains electronic primary care health records for approximately 12 million patients in the UK. The majority of diabetes and depression is usually treated and managed in primary care hence diagnoses, monitoring and treatments should be captured by IMRD-UK.
Multiple validation studies have shown IMRD-UK data to be generalizable to the wider UK population. 19,20 Data quality was ensured by using practices which had reached the standard for acceptable mortality rate (AMR) and acceptable computer usage (ACU). Acceptable mortality rates were reached when death rates recorded in the electronic database matched expected standardised mortality ratios and acceptable computer usage was reached when records included at least one medical record, additional health data and two therapy records per patient per year. The combination of AMR and ACU can be used as a marker of data quality. 21,22 Data from general practices in IMRD-UK between 1 January 2008 and 31 December 2017 were used for this study.
Definitions for key variables and inclusion criteria are described below: Eligible participants were people diagnosed with type 2 diabetes aged 35-99 years and who were permanently registered. People who had been registered for less than 9 months at the practice prior to first treatment were excluded for incidence calculations (but not prevalence calculations) as these cases were more likely to represent delayed coding of patients with pre-existing diabetes. Patients with diabetes diagnosed <35 and on insulin were excluded as they were likely to have type 1 diabetes.
Patients with diabetes were defined as people using a previously validated algorithm with at least two of the following records 1 : a diagnostic code for diabetes, 2 supporting evidence of diabetes, for example, screening for diabetic retinopathy or 3 treatment for diabetes. 23 The codes used to identify people living with type 2 diabetes can be found in Appendix S1.
A cohort of patients from the same practice, matched for age and gender, and without type 2 diabetes were identified for the comparison group. A sample size of up to five times the population with diabetes was used. 24 The matching was done in 5 year bands between

| Statistical analysis
To analyse the incidence data, we used a Poisson regression to examine the relationship between initiation of an antidepressant medication and gender, age and Townsend quintile. Townsend scores were used as a measure of social deprivation, where social deprivation was assigned quintiles with 1 being the least deprived and 5 being the most deprived. 26 Models were estimated separately for men and women and included a random effect to account for clustering by practice.
We calculated the ratio of and difference between incidence rates of new prescriptions of antidepressant use in people with and without type 2 diabetes with 95% confidence intervals. To examine the impact of duration of diabetes on the risk of starting on antidepressant medication, we looked at the time to first prescription from diagnosis of type 2 diabetes in the incident cohorts of patients newly diagnosed with type 2 diabetes during the study period and compared this with duration from a randomly generated index date (pseudo-diagnosis date) for the matched comparison cohort. We plotted a Kaplan-Meier curve of time to first prescription of antidepressant stratified by gender, age and Townsend score.
We used Poisson regression to examine the relationship between antidepressant initiation and gender, age, and Townsend score. There was a significant interaction between age and gender, so the results for age were stratified by sex. Models included a random effect to account for clustering by practice.
Analyses were conducted with Stata software version 16.0 (Stata Corporation, USA).

| Ethics
The data provider (IQVIA) obtained overall ethical approval for the  Figure S1).
In men with type 2 diabetes, the initial peak for starting on anti-  People with type 2 diabetes were 1.29 (95%CI 1.26-1.33) times more likely to be started on antidepressant medication than people in an age and gender matched cohort of people from the same practices (Table 1).   4 | CONCLUSIONS

| Main findings
People with type 2 diabetes were 30% more likely to be started on antidepressant medication than people without type 2 diabetes.
Women with type 2 diabetes were 35% more likely than men to be prescribed antidepressants and the risks increased with deprivation and in younger or older adults, with the lowest rates in the 65-69 year age band. The rates of antidepressant prescribing were broadly stable over the 10-year period in this study. The antidepressant medications prescribed changed slightly over time with sertraline becoming more widely used and fewer prescriptions of citalopram.

| Comparability with other findings
A recent systematic review of antidepressant use in type 2 diabetes highlighted a dearth of studies on the use of antidepressant medication in people with type 2 diabetes, with only seven included studies, of which six were in the USA and one in Australia. 28  with a dip in rates of prescriptions in the seventh decade before rising again later in life. 32 Brown et al. 30 did not find an association between diabetes and depression, however the comparison cohort was not matched by age and the mean age of the comparison cohort was significantly younger (and would have a higher risk of developing depression). Chen et al. 31 found the highest prevalence of depression in younger people but grouped together all patients over 65 so trends in older people were not described.
Trends in antidepressant medication used in people with type 2 diabetes are similar to trends in antidepressant prescribing in the general population, with sertraline becoming increasingly popular and citalopram less popular. 33

| Results by stratification
There is mixed evidence in the general population about the changes in prevalence of depression in older adults. 34 Reasons for increased risk of depression in older age could include physical disability, higher cognitive impairment, relative proportion of women and lower socioeconomic status. 35

| Strengths and weaknesses
This study has a number of strengths. The data were drawn from a large cohort of people with type 2 diabetes, followed up over 10 years, with detailed information about their diabetes care and prescriptions. Most people with T2DM and depression in the United Kingdom will receive their care, treatment and prescriptions within general practice and will have been well recorded in IMRD-UK and IMRD-UK data has also been shown to representative of the UK population.
This study also has some limitations that need to be borne in mind when considering the results. A very small minority of patients will receive care for these conditions away from general practices or receive care outside of NHS settings which might not be captured in IMRD-UK. These small numbers are unlikely to substantially affect our results. We have excluded low dose TCA and those prescribed duloxetine from our analysis as these are more likely to be used for neuropathic pain. There will also be differences between rates of diagnosed depression and rates of antidepressant medication prescribing.
Not all depression will be treated by medication and many cases will be treated with psychological therapy in combination with or instead of medical therapy. Antidepressant medication can also be used to treat other conditions like anxiety, obsessive-compulsive disorder, and panic attacks-therefore, antidepressant prescription records will most closely correlate with moderate to severe episodes of common mental health problems.
While the literature suggests that outcomes are poorer for people with type 2 diabetes with co-morbid depression, what is less clear is whether treating depression has reversed that effect. Further studies examining the impact on glycaemic control of treating depression and comparing the effects of different antidepressants would be beneficial in exploring the best treatment modality for co-morbid type 2 diabetes and depression.

| CONCLUSION
People with type 2 diabetes were 30% more likely to be started on antidepressant medication than people without type 2 diabetes.
Women with diabetes were 35% more likely to prescribed antidepressants than men and the risks increased with deprivation, extremes of age and treatment with insulin. The rates of antidepressant prescribing were broadly stable over the 10-year period in this study. The antidepressant medications prescribed changed slightly over time with Sertraline becoming more widely used and fewer prescriptions of Citalopram.