Temporal variation of renal function in people with type 2 diabetes mellitus: A retrospective UK clinical practice research datalink cohort study

Abstract Aim To characterize the longitudinal variability of estimated glomerular filtration rate (eGFR) in people with type 2 diabetes mellitus (T2DM), including variation between categories and individuals. Methods People with T2DM and sufficient recorded serum creatinine measurements were identified from the Clinical Practice Research Datalink (T2DM diagnosis from 1 January 2009 to 1 January 2011 with 5 years follow‐up); eGFR was calculated using the CKD‐EPI equation. Results In total, 7766 individuals were included; 32.8%, 50.2%, 12.4%, 4.0% and 0.6% were in glomerular filtration rate (GFR) categories G1, G2, G3a, G3b and G4, respectively. Overall, eGFR decreased by 0.44 mL/min/1.73 m2 per year; eGFR increased by 0.80 mL/min/1.73 m2 between index and year 1, then decreased by 0.75 mL/min/1.73 m2 annually up to year 5. Category G1 showed a steady decline in eGFR over time; G2, G3a and G3b showed an increase between index and year 1, followed by a decline. Category G4 showed a mean eGFR increase of 1.85 mL/min/1.73 m2 annually. People in categories G3‐G4 moved across a greater number of GFR categories than those in G1 and G2. Individual patients' eGFR showed a wide range of values (change from baseline at year 5 varied from −80 to +59 mL/min/1.73 m2). Conclusion Overall, eGFR declined over time, although there was considerable variation between GFR categories and individuals. This highlights the difficulty in prescribing many glucose‐lowering therapies, which require dose adjustment for renal function. The study also emphasizes the importance of regular monitoring of renal impairment in people with T2DM.


| INTRODUCTION
Diabetes is a leading cause of chronic kidney disease (CKD) 1 and it is expected that between 40% and 50% of people with type 2 diabetes mellitus (T2DM) will be affected by CKD in their lifetimes. [2][3][4] However, only a small number of glucose-lowering therapies can be used safely in people with renal impairment without requiring a dose adjustment. 5 Therefore, renal function is an important factor to consider when prescribing glucose-lowering medications in people with T2DM.
Previous research has showed that renal function, as measured by estimated glomerular filtration rate (eGFR), can vary considerably, especially among people with diabetes. [6][7][8][9][10][11][12][13][14] These studies have also suggested that eGFR improvement among people with T2DM is possible, 11 leading to increased complexity when considering optimal treatment. Published studies have tended to investigate renal variation at the population or category level, with one such study reporting eGFR trends in the UK. 11 There are no recent studies reporting patient-level variation in renal function in a T2DM population.
Using primary care clinical records, this study aims to further characterize the longitudinal variability of eGFR in a cohort of people with T2DM with availability of consistent eGFR measurements over a period of 5 years to further explore eGFR trends and patterns over a longer period, including analysis at the individual patient level.

| Data source
Patient records were obtained from the UK Clinical Practice Research Datalink (CPRD), a primary care database that includes data from general practices throughout the UK. As of November 2018, the database contained anonymized data for approximately 10 million people, with over 1 in 10 practices in the UK contributing data. 15 CPRD data have been used in over 2000 peer-reviewed publications 15 , and have been found to be broadly representative of the UK population in terms of age, sex, ethnicity and body mass index (BMI). 16 Medical records are updated monthly from participating practices, including complete clinical information, pathology tests, anthropometric data, referral and prescription records. CPRD is linked to Hospital Episode Statistics (HES), a database containing details of all hospital admissions, accident and emergency attendances and outpatient appointments, to improve ethnicity recording for glomerular filtration rate (GFR) estimation. 17

| Study population
Individuals were identified in CPRD based on their first diagnosis code of T2DM (codes are reported in the supporting information). Eligibility criteria included diagnosis of T2DM between 1 January 2009 and 1 January 2011; individuals also had to have a measure of serum creatinine after T2DM diagnosis (index measurement) and at least one measure of serum creatinine recorded in 5 yearly intervals post-first serum creatinine after diagnosis. In addition, the following inclusion criteria were applied: individuals must have at least 12 months' registration in practice prior to the index date; belong to an "up-to-standard" practice at the index date; have a record of ethnicity (identified through HES linkage, or CPRD if unavailable in HES). Individuals with a history of type 1 diabetes mellitus were excluded from the analysis.

| Renal function classification
Renal function was measured via eGFR using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation. To estimate GFR, the CKD-EPI equation requires data for serum creatinine, age, sex and ethnicity (see equation in the supporting information). 18 The CKD-EPI equation was selected as it is the recommended formula by the National Institute for Heath and Care Excellence (NICE). 19 Individuals were grouped into GFR categories, as adopted by NICE guidelines, according to their eGFR at baseline and follow-up. 19 These are Category G5 (<15 mL/min/1.73 m 2 ) was also considered, but none of the study population had an eGFR that fitted within this group.
Baseline characteristics, including age at T2DM diagnosis, age at the index date, BMI, HbA1c, systolic blood pressure, diastolic blood pressure and eGFR, were compared among individuals included and excluded from the analysis using Student's t-test. Renal function was described for each yearly interval based on the last recorded value per year and compared with baseline using mean values, counts and percentages to identify the raw change in eGFR as well as individual category changes. The analysis was performed using Stata version 14.
On average, there were no relevant differences in the baseline characteristics of those included and excluded from the analysis in terms of age, BMI, HbA1c, systolic and diastolic blood pressure and eGFR (Table S1).
During follow-up, patients changed GFR categories 1.5 times on average [standard deviation (SD) 1.6]. Those with reduced renal function below 60 mL/min/1.73 m 2 (G3 and higher categories) changed GFR categories more often compared with people with eGFR ≥60 mL/min/1.73 m 2 (G1 and G2) ( Table 2). In particular, people in categories G1 and G2 changed GFR categories 1.3 times on average (SD 1.6 and 1.5, respectively), and people in categories G3a, G3b and G4 changed GFR categories 2.6 (SD 1.7), 2.1 (SD 1.7) and 2.9 (SD 1.9) times, respectively.   Table S4). Change in renal function in the GFR categories and at the individual patient level showed the same trends as reported in the main analysis (Table S4 and Figure S4, respectively).

| Sensitivity analyses
Finally, we also looked at eGFR trends according to ACR. The results in both categories (A1 and A2) followed a similar trend to that observed in the main analysis ( Figure S5).  T2DM. This may also produce some bias in the results, as we are unable to identify how the practices included in our study perform against any clinical quality metrics. However, because all the practices included met the "up-to-standard" metric, it is probable that each possessed a reasonable level of quality and were suitable for research.