Cardiovascular Events and All-cause Mortality in T2D
A retrospective cohort study was performed using data from The Health Improvement Network (THIN), a computerized primary care database containing anonymized records for individuals currently registered with participating primary care practices in the UK. THIN is age, sex and geographically representative of the UK population and has been extensively validated for epidemiological studies. Anonymized data on patients are systematically recorded by participating primary care physicians (PCPs) as part of their routine patient care and regularly delivered to THIN for use in research projects. The computerized information includes demographics, details of PCP visits, diagnoses, referrals to specialists and hospital admissions, and a free-text section. Participating practices are required to record prescriptions and new courses of therapy. THIN also provides a standardized system for the reliable and comprehensive recording of additional health data such as results of laboratory tests (including serum creatinine concentration, when appropriate). The Read classification is used to code specific diagnoses, and a drug dictionary based on data from the Multilex classification is used to record prescriptions. The collection of data in THIN database was approved by a Multicentre Research Ethics Committee in the UK (MREC reference number: 08/H0305/49).
A cohort of patients with diagnosed type 2 diabetes who were aged 20–89 years between January 1, 2000 and December 31, 2005 was identified from THIN (n = 64,755). The wide age range was chosen to include the general adult population with prevalent type 2 diabetes. Eligible individuals were required to be registered for at least 3 years with their PCP, to have had at least one visit recorded in the past 3 years, and to have a recorded prescription history of 3 years or more. Patients were included in the study cohort if they had at least one creatinine measurement of 10–250 μmol/L recorded between 1 January 2000 and 31 December 2005. Patients with a record of hemodialysis (n = 109) or renal transplant (n = 60) before their start date were excluded, and patients with a recorded incidence of hemodialysis or renal transplant during follow-up were censored from the analysis (n = 108 for hemodialysis and n = 5 for renal transplant).
Among all individuals with type 2 diabetes meeting these criteria (n = 57,957), 56,693 (97.8%) had a first recorded creatinine measurement of 10–250 μmol/L. The date of this first recorded creatinine measurement was defined as their start date. The remaining 1264 individuals (2.2%) had a first creatinine measurement < 10 μmol/L (n = 1161) or > 250 μmol/L (n = 103), and a subsequent measurement within the range 10–250 μmol/L. The date of their first serum creatinine measurement between 10 and 250 μmol/L was defined as their start date. The mean and median times from their first recorded measurement to their start date were 341 days and 202 days, respectively. All patients were followed up from their start date to the first occurrence of either of the following endpoints in three different analyses based on the studied outcome: outcome of interest (death, MI or IS/TIA), reaching the age of 90 years, or end of the study period (December 31, 2010). It should be noted that 11 patients were excluded from the final cohort (seven individuals who had died at start date, and four who had no visits during follow-up), resulting in a final cohort of 57,946 patients.
Type 2 diabetes diagnosis was based on the Read classification codes assigned by the PCP or use of hypoglycemic drugs or insulin. For the majority of cases, the type of diabetes was specifically reported by the physician. If the physician used an unspecific diagnostic code (e.g., diabetes mellitus), we reviewed the patient's medical record back to one year before the diagnosis including any referral letters and physicians' free-text comments to assign the type of diabetes. If the age of onset was ≤ 35 years and the patient had one or more prescriptions for insulin and less than one year of oral hypoglycemic treatment, the case was classified as type 1 diabetes. Conversely, if the age of onset was ≥ 50 years and the patient used oral hypoglycemic treatment for at least 1 year, the case was classified as type 2 diabetes. A previous THIN study with a similar diabetes ascertainment algorithm estimated a diabetes prevalence that closely matched the prevalence in the Health Survey of England, which is a national population survey.
Duration of diabetes was defined as the time interval between the first ever recorded entry for type 2 diabetes in the database (including treatment for diabetes) and the start date (date of the first ever valid recorded serum creatinine measurement). Duration of diabetes was categorized into five groups: < 1 year, 1–4 years, 5–9 years, 10–14 years and ≥ 15 years.
The modification of diet in renal disease (MDRD) study formula and the Cockcroft–Gault formula are routinely used to calculate eGFR from serum creatinine concentration. In this study, the eGFR at baseline was calculated using the MDRD study formula (eGFR = 186 × Cr × age × 1.212 [if black] × 0.742 [if female], where Cr is the serum creatinine concentration in mg/dL). Ethnicity is not recorded in THIN, hence the same formula was used for all patients (eGFR = 186 × Cr × age × 0.742 [if female]) to classify them into five subgroups according to their baseline eGFR: < 15 mL/min (CDK stage 5), 15–29 mL/min (CKD stage 4), 30–44 mL/min (CKD stage 3B), 45–59 mL/min (CKD stage 3A) and ≥ 60 mL/min (CKD stages 1 and 2, or no CKD).
An automatic computer search for specific Read codes was used for the ascertainment of MI cases. Previous studies using this method have shown a very high specificity for MI, resulting in a confirmation rate greater than 90% when validated with the PCP via a questionnaire. Therefore, additional steps of validation of the ascertainment of MI cases, such as manual review of patients' profiles or validation with a questionnaire, were not carried out in the present study. A total of 3435 cases of MI were identified.
The predictive value of computer-detected IS/TIA is lower than that for other outcomes such as MI owing to the level of misclassification of diagnoses using Read codes. Therefore, we used a multistep approach to ascertain IS/TIA cases (see Additional file 1 http://www.cardiab.com/content/14/1/38/additional for a detailed description). Briefly, a computer search using Read codes suggestive of IS/TIA identified 4799 potential cases. Among these cases, 902 were matched to patients reviewed in other projects in which we looked at a diagnosis of IS/TIA in THIN; 653 were classified as non-cases and 249 as cases. For the remaining 3897 patients, the cases of IS/TIA were ascertained in a stepwise fashion by first searching for indicators of hospitalization or referral and then searching for indicators of symptoms, diagnostic procedures and new treatment related to stroke in the 30 days before and after the date of the computer-detected IS/TIA. Finally, the profiles (including free text) of sample patients from different subgroups were manually reviewed to validate the ascertainment of cases. Overall, we identified 3785 cases of IS/TIA.
Data on demographic variables including sex, age, smoking status, alcohol use, body mass index (BMI) and Townsend deprivation index (a measure of material deprivation within a population that takes into account four main variables: unemployment rate, car ownership, home ownership and household overcrowding) were collected any time before the start date. Exposure to drugs was collected before the start date and categorized as follows: current use, when the supply of the most recent prescription lasted until the start date or ended in the 90 days before the start date; recent use, when supply of the most recent prescription ended more than 90 days before the start date; and non-use, when there was no recorded use any time before the start date. Data on healthcare service use (PCP visits, referrals and hospitalizations) were collected for the year before the start date. Information on comorbidities was collected any time before the start date. Data on levels of glycated hemoglobin (HbA1c) were collected for the year before the start date. Patients were classified into subgroups according to the HbA1c data recorded closest to their start date: < 7.00%, 7.00–7.99%, 8.00–8.99%, 9.00–9.99%, 10.00–10.99% and ≥ 11.00%. Individuals without a recorded level of HbA1c in the year before their start date were included in the 'missing' category.
Incidence rates of death, MI and IS/TIA were calculated overall and by eGFR categories. Kaplan–Meier survival curves for all-cause mortality, MI and IS/TIA were calculated overall and according to eGFR category. Hazard ratios (HRs) and their 95% confidence intervals (CIs) were calculated using Cox proportional hazard models adjusted for sex, age, BMI, smoking status, hyperlipidemia, hypertension, history of MI, history of IS/TIA, history of ischemic heart disease (excluding MI), eGFR category, duration of diabetes, HbA1c category, and polypharmacy (in the month before the start date). A two-sided p value < 0.05 was considered to be statistically significant. Statistical analyses were performed using the Stata package version 12.0 (StataCorp LP, College Station, TX, USA).
Methods
Data Source
A retrospective cohort study was performed using data from The Health Improvement Network (THIN), a computerized primary care database containing anonymized records for individuals currently registered with participating primary care practices in the UK. THIN is age, sex and geographically representative of the UK population and has been extensively validated for epidemiological studies. Anonymized data on patients are systematically recorded by participating primary care physicians (PCPs) as part of their routine patient care and regularly delivered to THIN for use in research projects. The computerized information includes demographics, details of PCP visits, diagnoses, referrals to specialists and hospital admissions, and a free-text section. Participating practices are required to record prescriptions and new courses of therapy. THIN also provides a standardized system for the reliable and comprehensive recording of additional health data such as results of laboratory tests (including serum creatinine concentration, when appropriate). The Read classification is used to code specific diagnoses, and a drug dictionary based on data from the Multilex classification is used to record prescriptions. The collection of data in THIN database was approved by a Multicentre Research Ethics Committee in the UK (MREC reference number: 08/H0305/49).
Study Design
A cohort of patients with diagnosed type 2 diabetes who were aged 20–89 years between January 1, 2000 and December 31, 2005 was identified from THIN (n = 64,755). The wide age range was chosen to include the general adult population with prevalent type 2 diabetes. Eligible individuals were required to be registered for at least 3 years with their PCP, to have had at least one visit recorded in the past 3 years, and to have a recorded prescription history of 3 years or more. Patients were included in the study cohort if they had at least one creatinine measurement of 10–250 μmol/L recorded between 1 January 2000 and 31 December 2005. Patients with a record of hemodialysis (n = 109) or renal transplant (n = 60) before their start date were excluded, and patients with a recorded incidence of hemodialysis or renal transplant during follow-up were censored from the analysis (n = 108 for hemodialysis and n = 5 for renal transplant).
Among all individuals with type 2 diabetes meeting these criteria (n = 57,957), 56,693 (97.8%) had a first recorded creatinine measurement of 10–250 μmol/L. The date of this first recorded creatinine measurement was defined as their start date. The remaining 1264 individuals (2.2%) had a first creatinine measurement < 10 μmol/L (n = 1161) or > 250 μmol/L (n = 103), and a subsequent measurement within the range 10–250 μmol/L. The date of their first serum creatinine measurement between 10 and 250 μmol/L was defined as their start date. The mean and median times from their first recorded measurement to their start date were 341 days and 202 days, respectively. All patients were followed up from their start date to the first occurrence of either of the following endpoints in three different analyses based on the studied outcome: outcome of interest (death, MI or IS/TIA), reaching the age of 90 years, or end of the study period (December 31, 2010). It should be noted that 11 patients were excluded from the final cohort (seven individuals who had died at start date, and four who had no visits during follow-up), resulting in a final cohort of 57,946 patients.
Ascertainment and Duration of Type 2 Diabetes
Type 2 diabetes diagnosis was based on the Read classification codes assigned by the PCP or use of hypoglycemic drugs or insulin. For the majority of cases, the type of diabetes was specifically reported by the physician. If the physician used an unspecific diagnostic code (e.g., diabetes mellitus), we reviewed the patient's medical record back to one year before the diagnosis including any referral letters and physicians' free-text comments to assign the type of diabetes. If the age of onset was ≤ 35 years and the patient had one or more prescriptions for insulin and less than one year of oral hypoglycemic treatment, the case was classified as type 1 diabetes. Conversely, if the age of onset was ≥ 50 years and the patient used oral hypoglycemic treatment for at least 1 year, the case was classified as type 2 diabetes. A previous THIN study with a similar diabetes ascertainment algorithm estimated a diabetes prevalence that closely matched the prevalence in the Health Survey of England, which is a national population survey.
Duration of diabetes was defined as the time interval between the first ever recorded entry for type 2 diabetes in the database (including treatment for diabetes) and the start date (date of the first ever valid recorded serum creatinine measurement). Duration of diabetes was categorized into five groups: < 1 year, 1–4 years, 5–9 years, 10–14 years and ≥ 15 years.
Estimated Glomerular Filtration Rate
The modification of diet in renal disease (MDRD) study formula and the Cockcroft–Gault formula are routinely used to calculate eGFR from serum creatinine concentration. In this study, the eGFR at baseline was calculated using the MDRD study formula (eGFR = 186 × Cr × age × 1.212 [if black] × 0.742 [if female], where Cr is the serum creatinine concentration in mg/dL). Ethnicity is not recorded in THIN, hence the same formula was used for all patients (eGFR = 186 × Cr × age × 0.742 [if female]) to classify them into five subgroups according to their baseline eGFR: < 15 mL/min (CDK stage 5), 15–29 mL/min (CKD stage 4), 30–44 mL/min (CKD stage 3B), 45–59 mL/min (CKD stage 3A) and ≥ 60 mL/min (CKD stages 1 and 2, or no CKD).
Myocardial Infarction Ascertainment
An automatic computer search for specific Read codes was used for the ascertainment of MI cases. Previous studies using this method have shown a very high specificity for MI, resulting in a confirmation rate greater than 90% when validated with the PCP via a questionnaire. Therefore, additional steps of validation of the ascertainment of MI cases, such as manual review of patients' profiles or validation with a questionnaire, were not carried out in the present study. A total of 3435 cases of MI were identified.
Ischemic Stroke Ascertainment
The predictive value of computer-detected IS/TIA is lower than that for other outcomes such as MI owing to the level of misclassification of diagnoses using Read codes. Therefore, we used a multistep approach to ascertain IS/TIA cases (see Additional file 1 http://www.cardiab.com/content/14/1/38/additional for a detailed description). Briefly, a computer search using Read codes suggestive of IS/TIA identified 4799 potential cases. Among these cases, 902 were matched to patients reviewed in other projects in which we looked at a diagnosis of IS/TIA in THIN; 653 were classified as non-cases and 249 as cases. For the remaining 3897 patients, the cases of IS/TIA were ascertained in a stepwise fashion by first searching for indicators of hospitalization or referral and then searching for indicators of symptoms, diagnostic procedures and new treatment related to stroke in the 30 days before and after the date of the computer-detected IS/TIA. Finally, the profiles (including free text) of sample patients from different subgroups were manually reviewed to validate the ascertainment of cases. Overall, we identified 3785 cases of IS/TIA.
Data Collection
Data on demographic variables including sex, age, smoking status, alcohol use, body mass index (BMI) and Townsend deprivation index (a measure of material deprivation within a population that takes into account four main variables: unemployment rate, car ownership, home ownership and household overcrowding) were collected any time before the start date. Exposure to drugs was collected before the start date and categorized as follows: current use, when the supply of the most recent prescription lasted until the start date or ended in the 90 days before the start date; recent use, when supply of the most recent prescription ended more than 90 days before the start date; and non-use, when there was no recorded use any time before the start date. Data on healthcare service use (PCP visits, referrals and hospitalizations) were collected for the year before the start date. Information on comorbidities was collected any time before the start date. Data on levels of glycated hemoglobin (HbA1c) were collected for the year before the start date. Patients were classified into subgroups according to the HbA1c data recorded closest to their start date: < 7.00%, 7.00–7.99%, 8.00–8.99%, 9.00–9.99%, 10.00–10.99% and ≥ 11.00%. Individuals without a recorded level of HbA1c in the year before their start date were included in the 'missing' category.
Statistical Analysis
Incidence rates of death, MI and IS/TIA were calculated overall and by eGFR categories. Kaplan–Meier survival curves for all-cause mortality, MI and IS/TIA were calculated overall and according to eGFR category. Hazard ratios (HRs) and their 95% confidence intervals (CIs) were calculated using Cox proportional hazard models adjusted for sex, age, BMI, smoking status, hyperlipidemia, hypertension, history of MI, history of IS/TIA, history of ischemic heart disease (excluding MI), eGFR category, duration of diabetes, HbA1c category, and polypharmacy (in the month before the start date). A two-sided p value < 0.05 was considered to be statistically significant. Statistical analyses were performed using the Stata package version 12.0 (StataCorp LP, College Station, TX, USA).
SHARE
