Effectiveness of sulphonylureas in the therapy of diabetes mellitus type 2 patients: an observational cohort study

Background We compared all-cause mortality, major macrovascular events (MACE) and diabetes-related hospitalizations in T2DM-incident patients newly treated with metformin (MET) versus sulphonylureas (SU) monotherapy and in T2DM-prevalent patients newly treated with MET+SU versus MET+DPP4-inhibitor combination therapy. Methods We analysed anonymized data obtained from a German health fund. Patients were included when they had started MET versus SU therapy or MET+SU versus MET+DPP4 therapy between 01/07/2010 and 31/12/2011. Observation started with the first MET/SU prescription or the first prescription of the second agent of a MET+SU/MET+DPP4 combination therapy. Follow-up time lasted until the end of data availability (a minimum of 12 months), death or therapy discontinuation. Results In total, 434,291 T2DM-prevalent and 35,661 T2DM-incident patients were identified. Of the identified T2DM-incident patients, 904/7,874 started SU/MET monotherapy, respectively, with a mean age of 70.1/61.4 years (54.6/50.3 % female; Charlson Comorbidity Index (CCI) 1.4/2.2; 933/7,350 observed SU/MET patient years). 4,157/1,793 SU+MET/DPP4+MET therapy starters had a mean age of 68.1/62.2 years (53.4/50.8 % female; CCI 2.8/2.6; 4,556/1,752 observed SU+MET/ DPP4+MET patient years). In a propensity score matched (PSM) comparison, the HRs (95 % CIs) associated with SU monotherapy compared to MET monotherapy exposure were 1.4 (0.9–2.3) for mortality, 1.4 (0.9–2.2) for MACE, 4.1 (1.5–10.9) for T2DM hospitalizations and 1.6 (1.2–2.3) for composite event risk. In a multivariable Cox regression model, SU monotherapy was associated with higher mortality (aHR 2.0; 1.5–2.6), higher MACE (aHR 1.3; 1.0–1.7) and higher T2DM hospitalizations (aHR 2.8; 1.8–4.4), which corresponded with a higher composite event risk (aHR 1.8; 1.5–2.1). No significant differences in event rates were observed in the PSM comparison between DPP4+MET/SU+MET combination therapy starters and in the multivariable Cox regression analysis. Conclusions Our results show that SU monotherapy may be associated with increased mortality, MACE and T2DM hospitalizations, compared to MET monotherapy. When considering SU therapy, the associated cardiovascular risk should also be taken into account. Electronic supplementary material The online version of this article (doi:10.1186/s40200-016-0251-9) contains supplementary material, which is available to authorized users.


Background
Amongst the most common chronic diseases, type 2 diabetes mellitus (T2DM) presents some of the greatest clinical and health economic challenges [1]. In addition to burdens directly associated with the underlying disease, T2DM patients have an increased frequency of microand macrovascular complications and hospitalizations as well as increased mortality rates [2][3][4][5][6][7].
Previous observational studies have shown that a substantial number of T2DM patients receive SUs [10,11]. In fact, in countries like Germany, public agencies frequently see SUs as a main comparator therapy when assessing the potential value and reimbursement price of new second-line T2DM treatment agents such as DPP4s or GLP1s [12][13][14]. That being said, findings from clinical trials and observational studies have also raised concerns about the effectiveness and safety of SU treatment, especially in terms of its association with risks of hypoglycaemic as well as macrovascular events [11,[15][16][17][18][19]. Specifically, a recent UK analysis concluded that both SU monotherapy (compared to MET monotherapy) and SU combination therapy with MET (compared to MET+DPP4 combination therapy) are associated with an increased macrovascular/mortality event risk [11,19].
In this study, we assessed all-cause mortality, major macrovascular events (MACE) and diabetes-related hospitalizations in T2DM-incident patients newly treated with MET versus SU monotherapy and in T2DMprevalent patients newly treated with MET+DPP4 versus MET+SU combination therapy.

T2DM samples
We used an anonymized dataset obtained from the German health fund AOK PLUS (2010-2012) which initially included all T2DM-prevalent patients [at least one outpatient or inpatient T2DM diagnosis (ICD-10 codes: E11.-) in 01/07/2010-31/12/2011] who were insured by this health fund for the entire study period. The dataset contained information on patient sociodemographics, outpatient prescriptions, diagnosisassociated outpatient visits to GPs and specialists, and finally inpatient treatment in hospitals.
All patients were followed from the moment they were enrolled in the study until the occurrence of the outcomes of interest or until the end of the study period (whichever came first). By applying additional inclusion criteria, T2DM-incident patients were identified as a subgroup of all T2DM-prevalent patients. These patients had at least one outpatient/inpatient T2DM diagnosis recorded in 01/07/2010-31/12/2011 without any previous T2DM diagnosis and without any prescriptions of an antidiabetic agent (ATC groups: A10*) in the preceding 6 months.

SU monotherapy versus MET monotherapy
The study included T2DM-incident patients who started either MET or SU monotherapy between 01/07/2010 and 31/12/2011 without having received any prior antidiabetic medication during the preceding 180 days (Figs. 1 and 2). Observation started with the date of the first observed MET/SU prescription; follow-up time for each patient was at least 12 months (with death as an exception) and lasted until the first observed event, death, therapy discontinuation (treatment gap >180 days or prescription of another agent) or the end of 2012, whichever came first. All patients were followed with regard to the following events: In order to reliably differentiate between acute events and treatment for previous diagnoses/events, this analysis only considered ICD-10 diagnoses or documented procedures (i.e. documented by means of German OPS codes) to represent an event if they were the main motivation for acute hospitalization. The main outcome used in this study was a composite outcome (occurrence of any of the above events); in secondary analyses, the three event types were analysed separately.

SU+MET combination therapy versus DPP4+MET combination therapy
Our analyses of SU+MET combination therapy versus DPP4+MET combination therapy exclusively included T2DM-prevalent patients who had been prescribed MET monotherapy before and who started either MET-SU or MET-DPP4 combination therapy (combination therapy starters; first prescriptions needed to overlap within 30 days) between 01/07/2010 and 31/12/2011 without having received any prior SU/DPP4 medication (for the preceding 180 days). Data are presented in Figs. 1 and 3. Follow-up started with the first observed prescription of the second dual combination agent. All patients were followed with respect to the events as defined above. The follow-up period ended at therapy discontinuation (treatment gap >180 days or prescription of another agent), at death/first observed event or at the end of data availability (31/12/2012).

Statistical analysis
Differences in event risk for patients who received MET/SU monotherapy or SU+MET/DPP4+MET combination therapy were reported as unadjusted hazard ratios (HRs) in a Cox regression model censoring for death in the analyses addressing time to first MACE and time to first T2DM-related hospitalization and, additionally, censoring for therapy discontinuation/ end of follow-up period for all outcome categories including death. Furthermore, the percentage of eventfree patients over time was depicted by means of Kaplan Meier (KM) curves, and log-rank tests were used for testing statistical significance of differences.
To address the issue of confounding, two additional analyses were conducted: an analysis of event rates in propensity score matched patient samples and a multivariate Cox regression analysis using time to event as the dependent variable and reporting adjusted HRs (aHRs).
In the propensity score matching (PSM) procedure, SU-exposed patients (either mono or in combination with MET) were matched to SU non-exposed patients (MET mono or DPP4+MET) by propensity score. Only patients with complete (non-imputed) data were included in the analyses. Propensity scores were calculated using logistic regression estimation (with group affiliation as the dependent variable) including age, gender, age-adjusted Charlson Comorbidity Index (CCI; Additional file 1: Table S2) and adapted Diabetes Complications Severity Index (aDCSI; Additional file 2: Table S1) [4] as general independent variables, even if a certain overlap existed between some of these variables. Furthermore, the following variables related to the six months prior to the index prescription were included as independent variables in case these variables significantly influenced group exposition: the number of general practitioner visits, any previous   micro-/macrovascular complications and prescription of antithrombotic, antihypertensive or lipid lowering medication. A backward elimination approach was used to eliminate any variables that did not reach significance in explaining group exposition; in such cases, these variables were excluded from the PSM model. In any case, models included age, gender and age-adjusted CCI. For the PSM matched cohorts, separate estimates of HRs were calculated following the methodology as described above.
In order to analyse independent factors associated with the observed event risk, additional multivariable Cox regression analyses were conducted covering MET/SU monotherapy patients (Model 1) and SU+MET/DPP4 +MET combination therapy patients (Model 2); results were reported as aHRs. In addition to the exposure to either MET/SU monotherapy or SU+MET/DPP4+MET combination therapy, age (as dichotomous variable with a cut-off point at 65 years), gender, age-adjusted CCI and aDCSI were included in these models as independent variables.
All reported p-values were two-sided, and 95 % CIs were calculated for HRs/aHRs. All descriptive analyses were performed with Microsoft SQL Server 2008 and Microsoft Excel 2010. All other statistical analyses were performed with SPSS 17.0.

T2DM patient characteristics
In our study population, a total of 434,291 T2DMprevalent and a subgroup of 35,661 T2DM-incident patients were identified (Table 1, Figs. 2 and 3). Of the T2DM-prevalent patients, 56.2 % were female and their mean age was 70.2 years. We also observed a high number of comorbidities per patient in this sample, expressed as a mean CCI (without age factor) of 3.7, which indicates a significant burden in terms of comorbidities experienced per patient.

SU monotherapy versus MET monotherapy
Of the T2DM-incident patients in our study, 904 patients who were new initiators of SU monotherapy were significantly older (mean age of 70.1 years), were more likely to be female (54.6 %) and had a significantly higher mean age-adjusted CCI (2.23) than the 7,874 therapy-naïve users of MET monotherapy [mean age of 61.4 years (p <0.001), 50.3 % female (p <0.050), mean age-adjusted CCI of 1.44 (p <0.001); Table 1]. We observed 933 patient years of SU monotherapy exposure (mean follow-up period 376.9 days) and 7,850 patient years of MET monotherapy exposure (mean follow-up period 363.9 days).
In the unmatched patient sample comparisons (Table 2; supplemental KM curves in Additional file 3: Figure S3  In the multivariable Cox regression models ( Table 2; Additional file 5: Figure S5), older age, higher age-adjusted CCI and higher aDCSI were associated with increased MACE/death rates. With respect to hospitalization rates, female gender was associated with lower event rates, while a higher aDCSI was associated with higher event rates. SU monotherapy was associated with higher mortality rates (aHR 2.0; 1.5-2.6), higher MACE rates (aHR 1.3; 1.0-1.7) and higher T2DM-related hospitalization rates (aHR 2.8; 95 % CI: 1. 8-4.4). This corresponded with higher composite event rates (aHR 1.8; 1.5-2.1).

SU+MET combination therapy versus DPP4+MET combination therapy
Among the T2DM-prevalent patients, 4,157 patients who were newly prescribed with a SU+MET combination therapy were significantly older (mean age of 68.1 years), were more likely to be female (53.4 %) and had a significantly higher mean age-adjusted CCI (2.79) than the 1,793 patients with newly prescribed DPP4+MET combination therapy [mean age of 62.2 years (p <0.001); 50.8 % female (p <0.050) and a mean age-adjusted CCI of 2.56 (p <0.001); Table 1]. We observed 4,556 patient years of SU+MET exposure (mean follow-up period of 400.0 days) and 1,752 patient years of DPP4+MET exposure (mean follow-up period of 356.6 days).
In the unmatched patient sample comparisons (Table 3; Additional file 6: Figure S4 (Table 3; Additional file 8: Figure S8). In the multivariable Cox regression models (Table 3; Additional file 9: Figure S6), older age, higher age-adjusted CCI, higher aDSCI and male gender were associated with an increased risk of all-cause events (including MACE, deaths and T2DMrelated hospitalizations). When we compared SU+MET combination therapy to DPP4+MET combination therapy, as was done in the PSM analysis, no statistically significant results were found (Table 3; Additional file 10: Figure S7).

Discussion
The results of this study indicate that SU monotherapy may be associated with an increased risk of death, MACE and hospitalizations for T2DM patients compared to MET monotherapy, taking into account the differences in patient characteristics. This was seen in crude as well as multivariate Cox regression analyses, but due to small sample sizes this could not be confirmed for all observed outcomes in the PSM comparison. However, point estimates indicated similar associations, and we also observed  The higher SU-associated T2DM hospitalization risk may underpin the disadvantage of higher rates of hypoglycaemia associated with SU therapy [20], something which has also been confirmed by earlier studies and reported in a recently published review [21][22][23]. In addition, another systematic review and meta-analysis as well as another study found that patients receiving SU treatment had an increased all-cause mortality risk [16,24]; this, however, could not be confirmed in every study [25]. Furthermore, a UK-based study which was very similar to the one reported here compared MACE/mortality risk among T2DM-incident patients treated with either SU or MET monotherapy; this study did not include T2DM-related hospitalizations as an event type. It concluded that SU monotherapy was associated with increased MACE/mortality risk [11]. Another study which examined SU monotherapy in T2DM patients in comparison to MET monotherapy reported that SU users experienced treatment failure (defined as progression to a combination of oral anti-hyperglycaemia drug therapy, insulin use or an HbA1C >7.5 %) significantly earlier and more frequently than MET monotherapy users [26].
While further examinations of potential risk factors related to an increased mortality/MACE/hospitalization risk associated with SU monotherapy are not available in this current study, evidence from previous studies indicates that several factors may contribute to the underlying risks, including weight gain [27][28][29][30], links to cancer [31,32], increased insulin resistance and the underliying SU mechanism of action [33][34][35][36].
A German analysis covering data provided by 1,201 GPs reported a lower macrovascular event frequency under DPP4 treatment in comparison to SU treatment. This could not be confirmed in our study. However, in this study, events were identified through GP diagnoses only; these may have described more existing co-morbidities in T2DM patients than incident events in our definition, which identified events through acute hospitalizations only. Moreover, we observed patients who received either SU or MET monotherapy or SU+MET or DPP4+MET combination therapy only, whereas this analysis only excluded concomitant insulin therapy but allowed for all other antidiabetic agents [24]. Furthermore, our analysis covered prescriptions and outpatient treatment by a larger number of physicians (12,419 outpatient physicians, with 5,055 different GPs and outpatient specialists involved).
In contrast to our study, a similar analysis based on a retrospective sample of UK patients found all-cause mortality to be lower in the DPP4+MET group; a similar trend was also observed for MACE risk [24]. This UK analysis was based on a significantly larger sample size of 33,983 MET+SU and 7,864 MET +DPP4 patients in the unmatched comparison and 13,802 patients in the PSM comparison. In addition, median follow-up time was also longer in the UK study when compared to our study. Furthermore, the patient characteristics in our study also differed significantly from the UK analysis: whereas mean age in our PSM cohorts was 64.6-64.8 years, mean age in the UK-PSM cohorts was 59.8-60.4 years. Results similar to the abovementioned UK analysis were found in another large study [24]; our results were confirmed in several other retrospective database studies [24,37].
There may be specific clinical reasons why SU/MET +SU patients received this specific type of therapy (e.g. low risk of hypoglycaemia). In choosing SU therapy, MET contraindications may have played a major role. We observed MET contraindications in 44 % of the patients contained in our database. This may also explain event/mortality rate differences in the unmatched comparisons between MET/SU and MET +SU/MET+DPP4 groups. Other reasons for choosing a specific antidiabetic therapy were unknown to us, but could have confounded the results. Furthermore, we observed comparatively old/comorbid T2DM patients. This is due to the characteristics of those insured in the health care fund which provided the data. This means that T2DM patients with higher comorbidity levels are over-represented in our study.
Our data show that patients receiving SU therapy (mono or combo) differ significantly from other T2DM patients treated with MET in any combination: they tend to be older, have greater comorbidity and are more often female. So, for example, the mean ages of patients who received MET mono, GLP-1+OAD, GLP-1+OAD+insulin or MET+DPP4 combination therapy were 69.0, 57.5, 58.0 and 66.8 years, respectively. In comparison, the mean ages of patients who received SU mono or SU +MET combination therapy were 76.8 and 72.2 years, respectively. This makes a real-world comparison of SU with GLP-1s/DPP4s a challenging task because, obviously, new antidiabetic agents address completely different T2DM patient cohorts in real-life practice than SUs. Consequently, a substantial number of patients were excluded in the PSM comparisons. To reduce the bias risk for those patients included in the PSM cohorts, we used all available variables that significantly influenced group exposition, even if there was a certain overlap between these variables, as was the case with CCI and aDCSI.

Limitations
The current study is an observational cohort study with several limitations commonly associated with observational studies. First of all, it is limited with regard to its sample size and, more importantly, to the relatively short duration of follow-up. In addition, a significant number of patients was lost to follow-up, with 42.8 % of SU monotherapy patients and 47.8 % of SU+MET combination therapy patients discontinuing their treatment (treatment gap >180 days or prescription of another agent) or suffering a fatal event within the first 12 months after initiation of therapy. Thus, these patients had a much shorter follow-up than the planned minimum period of 12 months, and they were censored prematurely before the end of the study period.
Finally, certain information concerning several risk factors known from patient demographics and clinical characteristics associated with event risk were not available in our claims data. This included information about HbA1C [9,[38][39][40] and blood pressure [41,42], which may predict MACE/mortality based on a U-curve pattern [3]. It also included preclinical atherosclerosis [43], specific GFR values [44], level of physical activity [45] and total or low density lipoprotein (LDL)-cholesterol values, which have been found to be independent cardiovascular risk factors in other T2DM studies [38,46].

Conclusions
Our study suggests that SU monotherapy may be associated with an increased risk of mortality, MACE, T2DM hospitalizations and/or all-cause events, compared to MET monotherapy. Current German and European guidelines mostly recommend the use of SU as secondline therapy or, in case of MET contraindications, the use of SU as first-line therapy, and SU therapy is still prescribed in an important part of T2DM patients in Germany [8,9]. Our results indicate that in considering SU therapy, the associated cardiovascular risk should also be taken into account.

Additional files
Additional file 1: Table S1. Abbreviations AD medication, antidiabetic medication; aDCSI, adapted diabetes complications severity index; aHR, adjusted Hazard ratio; ATC, anatomical therapeutic chemical; CCI, Charlson comorbidity index; DMP, disease management programme; DPP-4, dipeptidyl peptidase-4; GLP-1, glucagon-like peptide-1; HR, Hazard ratio; ICD, International statistical classification of diseases; IRR, incidence rate ratio; KM, Kaplan Meier; MACE, macrovascular event; MET, metformin; OAD, oral antidiabetic drugs; PS, propensity score; SU, sulfonyl urea; T2DM, type 2 diabetes mellitus involved in the statistical analysis, in validating the database, in interpreting the results in the discussion section and in the conception/design of the study as well as in writing the introduction and methodology section. NH also took part in the clinical interpretation of the results, in the design of multivariate analyses and in writing the discussion part of the paper. The author KT took part in the statistical analysis and in validating the database as well as in the interpretation of the results in the discussion section.

Availability of data materials
In view of German data protection law (SGB X), we are not allowed to distribute the dataset which was analysed. Individuals interested in the dataset are invited to send an application to the dataset owner (statutory health insurance fund AOK PLUS; Dr. Ulf Maywald, ulfdr.maywald@plus.aok.de).