Study design and data source
We conducted a nested case–control analysis within a cohort of patients with incident T2DM between 01. January 1995 and 31. December 2019 in the UK-based primary care Clinical Practice Research Datalink (CPRD) GOLD.
CPRD GOLD contains anonymized medical records of over 11.3 million patients from more than 600 general practices in the UK. It is a governmental, non-profit database; the enrolled patients account for approximately 6.9% of the UK population. Patients within CPRD GOLD are representative of the UK general population with respect to age, sex, and ethnicity . The database was established in 1987 and is a collaborative project between the National Institute for Health Research (NIHR) and the Medicines and Healthcare Regulatory Agency (MHRA). The information in the database comes from participating general practitioners (GPs), who are trained on recording medical information using standard software and coding systems. Medical diagnoses, referrals to specialists and secondary care settings, prescriptions, diagnostic testing, lifestyle information, and demographic data are all part of the recorded information . Many validation studies have been performed that demonstrate the high quality of CPRD GOLD data [28,29,30]. The validity of the diagnoses of T2DM and VTE has been shown previously [31,32,33].
In order to ensure that we only included incident DM cases in the study population, patients had to have a minimum of 3 years of DM-free history in the database prior to onset. We identified patients based on specific codes for T2DM. We also included patients with an unspecific code for DM (e.g. general code for “diabetes”) if they were older than 30 years at diagnosis and received an oral antidiabetic drug (OAD). Independently of age, if DM patients never received insulin, we classified them as T2DM patients. We used the onset of DM as the study entry date, defined as the date of the first recorded DM code or the date of the first prescription for a DM medication. If the prescription occurred more than 365 days prior to the first recording of a DM diagnosis code, we excluded the patient.
We excluded patients with a diagnosis of cancer (except non-melanoma skin cancer), alcoholism, or HIV at any time in the patient record to avoid substantial bias and confounding.
We excluded patients with a history of VTE (at any time prior to the diagnosis of T2DM), or a code for surgery, immobilization, trauma, paralysis and paresis, or use of HRT or the contraceptive pill within 3 months prior to the index date. We further excluded patients with a code for pregnancy or puerperium within 12 months prior to the index date.
Case and control definition
We defined cases as patients with a first-time recording of VTE during the study period, who received at least one prescription for an antithrombotic drug within 7 days prior until 90 days after the VTE [5, 34, 35], including vitamin K antagonists, heparins, direct factor Xa inhibitors, direct thrombin inhibitors, fibrinolytic enzymes, or the synthetic penta-saccharide factor Xa inhibitor fondaparinux. The index date for each case was the date of the first recorded VTE. Since we excluded patients with known risk factors for a VTE prior to the outcome, we regard the VTE cases included in this study as having an unprovoked or idiopathic VTE .
We used risk set sampling to match each case to 4 controls from the study population, i.e. patients who did not experience a VTE between the onset of DM and the index date of their matched case. We matched controls to cases on age (± 3 years), sex, general practice, index date (same index date as the case, and the control had to be present in the database on the index date), and T2DM duration (± 365 days assessed by counting the days between the study entry date and the index date).
The exposure of interest in this study was glycemic control after the onset of DM defined by HbA1c levels. We used the last recorded HbA1c value before the index date for our analyses. We assessed HbA1c levels in 7 categories: ≤ 6.5% (≤ 48 mmol/mol), > 6.5–7.0% (> 48–53 mmol/mol, reference group), > 7.0–7.5% (> 53–58 mmol/mol), > 7.5–8.0% (> 58–64 mmol/mol), > 8.0–9.0% (> 64–75 mmol/mol), > 9.0% (> 75 mmol/mol), and no HbA1c measurement. Results for patients with missing values were presented in a separate category.
We used conditional logistic regression to assess the association between levels of glycemic control (expressed as HbA1c levels) with HbA1c levels of > 6.5–7.0% (> 48–53 mmol/mol) as the reference group and the risk of VTE, expressed as odds ratios (ORs) or adjusted ORs (aORs) with 95% confidence intervals (CI). We also assessed the association between HbA1c level and the risk of VTE according to the patients’ number of GP visits during the study period. Lastly, we conducted analyses in men and women separately.
We adjusted for the following comorbidities and co-medications (recorded at any time in the patient record before the index date) in the final model based on previous clinical knowledge: BMI (categorical variable), smoking status (current, past, non-smokers, and unknown), CVD (including congestive heart failure, ischemic heart disease, myocardial infarction, hypertension, and stroke), osteoarthritis, use of insulin, bisphosphonates, systemic corticosteroids, low-dose acetylsalicylic acid, and current (last prescription within 30 before the index date) or past (last prescription > 30 days prior to index date) use of metformin or sulfonylureas. We additionally tested for effect modification by obesity status (non-obese versus obese, defined as BMI levels < 30 and ≥ 30) of the association between level of HbA1c and risk of VTE.
In sensitivity analyses, we 1) restricted the sample to patients whose last HbA1c measurement was recorded within less than 90 days prior to the index date, 2) analyzed the risk of VTE separately for patients with a previous CVD diagnosis, and 3) conducted separate analyses of the risk of VTE by HbA1c levels for patient groups of different T2DM durations.
We conducted analyses using SAS software version 9.4 (SAS Institute, Inc., Cary, NC, USA).
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.