Liraglutide and Sitagliptin in Type 2 Diabetes
Liraglutide and Sitagliptin in Type 2 Diabetes
We designed a non-interventional, retrospective, database study utilising EMR data from the UK Clinical Practice Research Datalink (CPRD) database.
The UK CPRD is the world's largest computerised database of anonymised longitudinal medical records from primary care. The CPRD is a well-validated cohort with robust, high-quality information on diagnoses, comorbidities, prescription medication, and laboratory data. Currently, data are being collected on 5.5 million active patients from around 672 primary care practices throughout the UK, accounting for approximately 8% coverage of the UK population.
The overall study design is illustrated in Figure 1A. The study population was defined as all those individuals with type 2 diabetes recorded in the CPRD, who started treatment with liraglutide or sitagliptin at the index date and continued on the index drug for a period of 6 months. The index date was defined as the first recorded prescription of either drug during the period 1 July 2009 [the day after the European Medicines Agency (EMA) license was issued for liraglutide] to 31 July 2012 (database closure).
(Enlarge Image)
Figure 1.
(A) Retrospective study design, (B) Eligible patient populations (liraglutide and sitagliptin) based on study inclusion and exclusion criteria
The inclusion criteria to select the patient cohorts reflected the prescribing conditions of liraglutide based on the EMA prescribing label. The criteria were as follows: (i) individuals treated with liraglutide or sitagliptin from 1 July 2009 to 31 July 2012; (ii) 18 years of age or older with a type 2 diabetes diagnosis (based on READ codes); (iii) a minimum of 6 months of medical history pre-index; (iv) at least one measurement of HbA1c and weight at index (+7/−180 days) to maximise sample size, while ensuring the capture of the latest HbA1c measurement that is not likely to be impacted by the index treatment; and (v) at least one HbA1c and weight measurement at 6 months (± 45 days) post-index to allow for a reasonable timeframe to capture at least one outcome measure for analyses. Resulting from the nature of EMR data, time windows for measurement collection pre- and post-index were designed to maximise the number of patients included for baseline and outcome variable capture, thereby allowing for statistically robust data analyses. Specifically, a baseline window of 6 months was defined to reflect UK guidance on frequency of HbA1c testing, which is undertaken at 2–6 month intervals, as well as to ensure capture of the maximum number of patients initiated on the treatments of interest given the prescription patterns in the UK (i.e. prescriptions are valid for up to 6 months).
Patients were excluded if they had any of the following: (i) previous or concomitant liraglutide or DPP-4 inhibitor use at index; (ii) prior or concomitant insulin use at index; or (iii) on fixed dose combinations with metformin. The decision to exclude patients taking insulin is explained by the highly variable dose of insulin in the context of discrete, fixed doses of liraglutide or sitagliptin. To assess the effectiveness of the two drugs of interest, it was decided that the effect on glycaemic control and body weight of variable doses of insulin would reduce the validity of the study. In addition, patients on fixed dose combination with metformin were also excluded to minimise the bias because of possible prescribing indication driven by adherence-related issues.
The study protocol was approved by the UK Independent Scientific and Advisory Committee. Power calculations were performed based on the number of patients identified for each treatment group, using a Pearson χ test for two proportions, and considering a 21% difference in achieving a change in HbA1c ≥ 1% (54% in liraglutide compared with 33% in sitagliptin). The power estimated for the study was > 0.99 at a level of significance of 0.05. Additional power calculations were performed for the subgroup of patients with baseline HbA1c >9 mmol/mol, considering a 12% difference in achieving a change in HbA1c ≥ 1% (72% in liraglutide compared with 60% in sitagliptin), with the power estimated being 0.75 at a level of significance of 0.05.
Data were extracted from CPRD as delimited text files. All post-extraction data analyses were conducted using SAS® Version 9.3 (SAS Institute, Cary, NC). The main study outcomes considered included: (i) the absolute change in HbA1c, weight and systolic blood pressure after 6 months of therapy; (ii) the percentage of patients achieving ≥ 1% HbA1c reduction (i.e. treatment continuation criteria when liraglutide is used as triple therapy); (iii) the percentage of patients with an HbA1c reduction ≥ 1% and a weight reduction ≥ 3% (i.e. treatment continuation criteria when liraglutide is used as dual therapy; and (iv) the percentage of patients achieving treatment target of HbA1c < 7% (< 53 mmol/mol), in line with American Diabetes Association recommendations at 6 months. The main exposure of interest was treatment with either liraglutide or sitagliptin. Other independent factors (covariates) assessed included age, gender, Charlson comorbidity index, baseline body mass index (BMI), time from type 2 diabetes diagnosis to treatment (either liraglutide or sitagliptin), baseline HbA1c and concomitant oral antidiabetic (OAD) use. The validated Charlson comorbidity index, which is the most widely used score to classify patients according to their disease burden, was included as a measure of severity for each patient to control for any confounding effect that comorbid conditions may have on treatment effectiveness.
Descriptive statistics were run for all variables. Chi-squared tests, Fisher's exact tests or t-tests were used to assess any statistically significant differences between the treatment groups based on their baseline characteristics and outcomes. A statistical significance level of 0.05 was adopted.
Independent samples t-tests were performed to compare the absolute change in outcomes after 6 months of treatment in liraglutide and sitagliptin patients, relative to the baseline data gathered −180/+7 days to the index date. Multiple logistic regression models were built to assess the relationship between exposures of interest and outcomes, and to control for potential confounding of key covariates. The model covariates were selected on the basis of their clinical plausibility. The trend in HbA1c over 1 year prior to index (i.e. trajectory) was also adjusted for in the multiple logistic regression models to control for its impact as a potential confounder of measured effectiveness.
The potential impact of any selection bias caused by population censoring (e.g. patients on index therapy < 6 months) was also explored by time-to-event analysis of the actual time on therapy, where event was defined as 'stop of treatment'.
Materials and Methods
We designed a non-interventional, retrospective, database study utilising EMR data from the UK Clinical Practice Research Datalink (CPRD) database.
The UK CPRD is the world's largest computerised database of anonymised longitudinal medical records from primary care. The CPRD is a well-validated cohort with robust, high-quality information on diagnoses, comorbidities, prescription medication, and laboratory data. Currently, data are being collected on 5.5 million active patients from around 672 primary care practices throughout the UK, accounting for approximately 8% coverage of the UK population.
The overall study design is illustrated in Figure 1A. The study population was defined as all those individuals with type 2 diabetes recorded in the CPRD, who started treatment with liraglutide or sitagliptin at the index date and continued on the index drug for a period of 6 months. The index date was defined as the first recorded prescription of either drug during the period 1 July 2009 [the day after the European Medicines Agency (EMA) license was issued for liraglutide] to 31 July 2012 (database closure).
(Enlarge Image)
Figure 1.
(A) Retrospective study design, (B) Eligible patient populations (liraglutide and sitagliptin) based on study inclusion and exclusion criteria
The inclusion criteria to select the patient cohorts reflected the prescribing conditions of liraglutide based on the EMA prescribing label. The criteria were as follows: (i) individuals treated with liraglutide or sitagliptin from 1 July 2009 to 31 July 2012; (ii) 18 years of age or older with a type 2 diabetes diagnosis (based on READ codes); (iii) a minimum of 6 months of medical history pre-index; (iv) at least one measurement of HbA1c and weight at index (+7/−180 days) to maximise sample size, while ensuring the capture of the latest HbA1c measurement that is not likely to be impacted by the index treatment; and (v) at least one HbA1c and weight measurement at 6 months (± 45 days) post-index to allow for a reasonable timeframe to capture at least one outcome measure for analyses. Resulting from the nature of EMR data, time windows for measurement collection pre- and post-index were designed to maximise the number of patients included for baseline and outcome variable capture, thereby allowing for statistically robust data analyses. Specifically, a baseline window of 6 months was defined to reflect UK guidance on frequency of HbA1c testing, which is undertaken at 2–6 month intervals, as well as to ensure capture of the maximum number of patients initiated on the treatments of interest given the prescription patterns in the UK (i.e. prescriptions are valid for up to 6 months).
Patients were excluded if they had any of the following: (i) previous or concomitant liraglutide or DPP-4 inhibitor use at index; (ii) prior or concomitant insulin use at index; or (iii) on fixed dose combinations with metformin. The decision to exclude patients taking insulin is explained by the highly variable dose of insulin in the context of discrete, fixed doses of liraglutide or sitagliptin. To assess the effectiveness of the two drugs of interest, it was decided that the effect on glycaemic control and body weight of variable doses of insulin would reduce the validity of the study. In addition, patients on fixed dose combination with metformin were also excluded to minimise the bias because of possible prescribing indication driven by adherence-related issues.
The study protocol was approved by the UK Independent Scientific and Advisory Committee. Power calculations were performed based on the number of patients identified for each treatment group, using a Pearson χ test for two proportions, and considering a 21% difference in achieving a change in HbA1c ≥ 1% (54% in liraglutide compared with 33% in sitagliptin). The power estimated for the study was > 0.99 at a level of significance of 0.05. Additional power calculations were performed for the subgroup of patients with baseline HbA1c >9 mmol/mol, considering a 12% difference in achieving a change in HbA1c ≥ 1% (72% in liraglutide compared with 60% in sitagliptin), with the power estimated being 0.75 at a level of significance of 0.05.
Data were extracted from CPRD as delimited text files. All post-extraction data analyses were conducted using SAS® Version 9.3 (SAS Institute, Cary, NC). The main study outcomes considered included: (i) the absolute change in HbA1c, weight and systolic blood pressure after 6 months of therapy; (ii) the percentage of patients achieving ≥ 1% HbA1c reduction (i.e. treatment continuation criteria when liraglutide is used as triple therapy); (iii) the percentage of patients with an HbA1c reduction ≥ 1% and a weight reduction ≥ 3% (i.e. treatment continuation criteria when liraglutide is used as dual therapy; and (iv) the percentage of patients achieving treatment target of HbA1c < 7% (< 53 mmol/mol), in line with American Diabetes Association recommendations at 6 months. The main exposure of interest was treatment with either liraglutide or sitagliptin. Other independent factors (covariates) assessed included age, gender, Charlson comorbidity index, baseline body mass index (BMI), time from type 2 diabetes diagnosis to treatment (either liraglutide or sitagliptin), baseline HbA1c and concomitant oral antidiabetic (OAD) use. The validated Charlson comorbidity index, which is the most widely used score to classify patients according to their disease burden, was included as a measure of severity for each patient to control for any confounding effect that comorbid conditions may have on treatment effectiveness.
Descriptive statistics were run for all variables. Chi-squared tests, Fisher's exact tests or t-tests were used to assess any statistically significant differences between the treatment groups based on their baseline characteristics and outcomes. A statistical significance level of 0.05 was adopted.
Independent samples t-tests were performed to compare the absolute change in outcomes after 6 months of treatment in liraglutide and sitagliptin patients, relative to the baseline data gathered −180/+7 days to the index date. Multiple logistic regression models were built to assess the relationship between exposures of interest and outcomes, and to control for potential confounding of key covariates. The model covariates were selected on the basis of their clinical plausibility. The trend in HbA1c over 1 year prior to index (i.e. trajectory) was also adjusted for in the multiple logistic regression models to control for its impact as a potential confounder of measured effectiveness.
The potential impact of any selection bias caused by population censoring (e.g. patients on index therapy < 6 months) was also explored by time-to-event analysis of the actual time on therapy, where event was defined as 'stop of treatment'.
Source...