o

r
i
g
i
n
a
l
a
r
t
i
c
l
e
Diabetes, Obesity and Metabolism 2014.
2014 John Wiley & Sons Ltd
original article
Association between rst-line monotherapy with
sulphonylurea versus metformin and risk of all-cause
mortality and cardiovascular events: a retrospective,
observational study

C. Ll. Morgan
1
, J. Mukherjee
2
, S. Jenkins-Jones
1
, S. E. Holden
1,3
& C. J. Currie
1,3
1
Global Epidemiology, Pharmatelligence, Cardiff, UK
2
Global Health Economics and Outcomes Research, Bristol-Myers Squibb, Wallingford, CT, USA
3
Cochrane Institute of Primary Care and Public Health, Cardiff University, Cardiff, UK
Aims: To evaluate the risk of all-cause mortality and major adverse cardiovascular events (MACE) for patients exposed to rst-line monotherapy
with sulphonylurea or metformin.
Methods: Data were from the Clinical Practice Research Datalink (CPRD). Patients with type 2 diabetes were selected if initiated with
metformin or sulphonylurea monotherapy as their rst-line glucose-lowering regimen 20002012. The primary endpoint was all-cause
mortality; the secondary endpoint was MACE (myocardial infarction or stroke). Times to endpoints were compared using Cox proportional
hazards models. Additional analyses were performed on subsets matched directly on key characteristics and by propensity score.
Results: In the main analysis, 76 811 patients were prescribed metformin monotherapy (mean follow-up 2.9 years) and 15 687 sulphonylurea
monotherapy (mean follow-up 3.1 years). A total of 2604 patients were included in each arm of the directly matched cohorts and 8836 in the
propensity-matched. With respect to all-cause mortality, using all three analytical approaches the hazard ratio (HR) was signicantly increased
for sulphonylurea compared with metformin: adjusted HR =1.580 (95% CI 1.4831.684) for the main analysis, 1.902 (1.7332.088) for those
matched on propensity score, and 1.272 (1.0211.584) for the directly matched cohort analysis. For MACE, the respective HRs were 1.196
(1.0901.313), 1.202 (1.0011.442) and 0.814 (0.5781.148), respectively.
Conclusions: All-cause mortality was signicantly increased in patients prescribed sulphonylurea compared with metformin monotherapy.
Whilst residual confounding and confounding by indication may remain, this study indicates that rst-line treatment with sulphonylurea
monotherapy should be reconsidered.
Keywords: rst-line therapy, MACE, metformin, mortality, myocardial infarction, stroke, sulphonylurea
Date submitted 10 February 2014; date of rst decision 3 March 2014; date of nal acceptance 4 April 2014
Introduction
Type 2 diabetes has an estimated prevalence of around 5%
in the UK [1]. The general focus of diabetes treatment is the
maintenance of normoglycaemia toreduce progressiontolong-
term microvascular and macrovascular complications. Initial
management may consist of diet and lifestyle interventions
prior to the introduction of oral hypoglycaemic agents
(OHAs). Current guidelines in the UK [2] recommend that,
in the absence of contraindications, metformin should be the
rst-line choice of OHA. As an alternative, treatment with
sulphonylureas is recommended. In the USA, the proportion
of patients initially treated with sulphonylurea monotherapy
was 18% in 2008 [3]. However, recent ndings from both
Correspondence to: Prof. Craig J. Currie, Cochrane Institute of Primary Care and Public
Health, Cardiff University, The Pharma Research Centre, Cardiff Medicentre, Cardiff CF14
4UJ, UK.
E-mail: currie@cardiff.ac.uk

Parts of this study were presented in abstract form at the 49th Annual Meeting of the
European Association for the Study of Diabetes, Barcelona, Spain, 2327 September 2013.
clinical and observational studies have raised concerns about
the safety of intensive blood glucose management in type
2 diabetes, especially treatment with either sulphonylurea
or insulin [46]. The principal purpose of this study was
to characterize all-cause mortality and progression to major
adverse cardiovascular events (MACE) in patients with type
2 diabetes treated with either metformin or sulphonylurea as
rst-line monotherapy.
Methods
Data Source
The study was conducted using data from the Clinical Practice
Research Datalink (CPRD), a longitudinal, anonymized
research database derived from more than 600 primary
care practices in the UK. CPRD contains clinical records
from over 13 million people (based on the July 2012
release), of whom approximately seven million are actively
registered and can be followed prospectively. Available data
original article DIABETES, OBESITY AND METABOLISM
included demographics, medical history (including diagnoses
and health contacts), clinical investigation results and drug
prescriptions. Diagnostic information in CPRD is recorded
using the Read code classication, a UK general practice
standard. A subset of CPRD practices based in England
is also linked to additional health-based datasets including
the Health Episode Statistics (HES) inpatient data and the
Ofce for National Statistics (ONS) mortality data. HES
provides details of all National Health Service (NHS) inpatient
admissions in England since 1997, including primary and
contributory causes coded using the ICD-10 classication.
ONS provides details of all deaths in England, with immediate
and antecedent causes coded using the ICD-9 and ICD-10
classications.
Cohort Selection
Patients were extracted with a Read code indicative of diabetes.
As not all Read codes for diabetes differentiate between type
1 and type 2 diabetes, and some patient histories erroneously
contain codes for both types, patients with type 2 diabetes were
dened by one or more of the following:
1 Read codes exclusively indicative of type 2 diabetes
2 AReadcode indicative of type 2 diabetes (regardless of others
indicative of type 1 or non-specic diabetes) and at least one
prescription for an OHA
3 Prescription of two or more classes of OHA
4 Read codes indicative of both type 1 and type 2 diabetes but
an age of diagnosis older than 35 years.
Patients dened as incident diabetes cases based on a wash-
in period of at least 180 days from practice registration to
diabetes diagnosis were selected; of these, patients subsequently
initiating on a sulphonylurea or metformin were extracted.
Index date was dened as the rst prescription date and
patients were followed to each endpoint or censorship.
Study Endpoints
The two endpoints were all-cause mortality (deduced from
death date appearing in the patient record) and progression
to MACE. MACE, dened here as either acute myocardial
infarction (MI) or stroke, was ascertained from Read codes.
Statistical Methods
Comparison of baseline characteristics associated with each
therapy regimen was presented using appropriate univariate
statistics (t-test or nonparametric equivalent for continuous
data; chi-squared test for categorical data). The following
a priori dened variables were included in the analyses:
age at the index date (limited to accuracy of one whole
year), gender, diabetes duration (time from incidence date
to index date), haemoglobin A1c (HbA1c), systolic blood
pressure (SBP), total cholesterol, body mass index (BMI) and
smoking status. Baseline HbA1c, SBP, cholesterol and BMI
were dened as the nearest value to index date, searching
in the following order (i) 30 days prior to baseline, (ii) 30
days following baseline and (iii) 1 year prior to baseline.
Therapy history at baseline was modelled as binary variables
for each of: antiplatelet agents, anti-hypertensives and lipid-
lowering drugs. Baseline morbidity was modelled as the
total number of contacts with the general practitioner in
the year prior to the index date and as the Charlson co-
morbidity index [7]. Prior renal disease was indicated by
clinical diagnosis or by baseline serum creatinine levels
above 132 mol/l for males or 123 mol/l for females.
Where appropriate, missing data were handled using multiple
imputation.
Patient follow-up time was dened as the period from
index date to the earliest of (i) the date of the target event,
(ii) censoring by the cases transition to a different treatment
cohort plus 90 days, (iii) the date of the patients leaving
the participating primary-care practice or (iv) the last data-
collection date for that particular practice.
In addition to an analysis of all patients adjusting for baseline
variables, two additional analyses were conducted to address
issues of confounding by indication:
1 Sulphonylurea patients matched to metformin patients.
Matched variables were age (2 years), gender, year of index,
diabetes duration (1 year), BMI (3 kg/m
2
), serum creati-
nine (10 mol/l), and HbA1c (1% [11 mmol/mol]).
2 Sulphonylurea patients matched to metformin patients by
propensity score. This method matches exposed subjects
(in this instance, those exposed to sulphonylurea) with
non-exposed subjects (those exposed to metformin) on the
probability of their being initiated on sulphonylurea based
upon their baseline covariates. Covariates included in the
propensity matching were age, gender, serum creatinine,
HbA1c, BMI, smoking status, systolic blood pressure, total
cholesterol, primary-care contacts in the preceding year,
prior antihypertensive scripts, prior lipid-lowering scripts,
prior antiplatelet scripts, Charlson index, prior cancer and
MACE status.
In both sensitivity analyses only patients with complete (i.e.
non-imputed) data for the matching criteria were considered
for matching.
Analysis
Crude rates for all-cause mortality and MACE by person-
time for metformin versus sulphonylurea monotherapies were
compared using the mid-P exact test. For all three analysis
groups, progression to each outcome was tested by Cox
proportional hazards modelling (CPHM). Adjusted hazard
ratios (aHRs) for each endpoint were calculated. Threshold
statistical signicance in all tests was p =0.05 and 95%
condence intervals were presented for aHRs. All candidate
covariates with p <0.20 were included in the nal model.
In addition, to explore issues of residual confounding, a
subset of patients who failed on monotherapy and switched to
dual therapy with metformin and sulphonylurea were followed
from date of dual therapy initiation until last date in the
dataset. Baseline characteristics were re-calculated for date of
dual therapy initiation.
Initial data processing was undertaken using Microsoft SQL
Server 2008. Statistical analysis was conducted using IBM SPSS
2 Morgan et al. 2014
DIABETES, OBESITY AND METABOLISM original article
Table 1. Baseline characteristics of patients initiating either metformin or sulphonylurea by study design: main analysis, propensity matched and directly
matched.
Main analysis Propensity-matched Directly matched
Metformin Sulphonylurea Metformin Sulphonylurea Metformin Sulphonylurea
Patients, n 76811 15687 8836 8836 2604 2604
Gender, n (%)
Male 43282 (56.3) 9121 (58.1) 3736 (60.5) 3747 (60.7) 931 (35.8) 931 (35.8)
Female 33529 (43.7) 6566 (41.9) 2435 (39.5) 2424 (39.3) 1673 (64.2) 1673 (64.2)
Age, years, mean (s.d.) 61.4 (12.8) 68.1 (12.9) 68.2 (11.4) 68.5 (12.7) 66.6 (10.4) 66.6 (10.4)
Diabetes duration, years, mean (s.d.) 1.5 (2.4) 1.8 (3.0) 2.0 (2.8) 2.3 (3.1) 1.4 (1.8) 1.5 (1.8)
Serum creatinine, mol/l, mean (s.d.) 85.4 (19.1) 98.1 (33.7) 95.0 (22.5) 95.1 (27.9) 89.6 (16.5) 89.8 (16.7)
HbA1c, %, mean (s.d.) 8.6 (1.8) 9.1 (2.1) 8.9 (2.0) 8.9 (1.9) 8.6 (1.5) 8.6 (1.6)
HbA1c, mmol/mol, mean (s.d.) 70.5 (19.7) 76.0 (23.0) 73.8 (21.9) 73.8 (20.8) 70.5 (16.4) 70.5 (17.5)
BMI, kg/m
2
, mean (s.d.) 32.3 (5.8) 27.1 (4.8) 27.8 (4.3) 27.4 (4.8) 28.4 (3.7) 27.9 (3.9)
Systolic BP, mmHg, mean (s.d.) 139.0 (17.0) 140.3 (19.4) 138.9 (17.5) 138.8 (18.8) 139.7 (17.4) 139.2 (18.0)
Total cholesterol, mmol/l, mean (s.d.) 5.0 (1.2) 5.1 (1.2) 5.0 (1.2) 5.0 (1.2) 5.1 (1.2) 5.0 (1.2)
HDL, mmol/l, mean (s.d.) 1.2 (0.3) 1.2 (0.4) 1.2 (0.3) 1.2 (0.4) 1.2 (0.3) 1.2 (0.3)
LDL, mmol/l, mean (s.d.) 2.9 (1.0) 2.9 (1.1) 2.8 (1.0) 2.9 (1.0) 2.9 (1.0) 2.9 (1.0)
Triglycerides, mmol/l, mean (s.d.) 2.3 (1.5) 2.3 (1.5) 2.2 (1.4) 2.2 (1.4) 2.3 (1.5) 2.1 (1.3)
GP contacts preceding year, mean (s.d.) 11.1 (9.7) 11.7 (10.5) 12.2 (10.8) 11.9 (10.1) 10.4 (8.2) 12.0 (9.8)
Charlson index, mean (s.d.) 1.9 (1.3) 2.2 (1.6) 2.4 (1.7) 2.3 (1.7) 2.0 (1.3) 2.1 (1.5)
Prior anti-hypertensive scripts, n (%) 51188 (66.6) 9773 (62.3) 4272 (69.2) 4195 (68.0) 1784 (68.5) 1726 (66.3)
Prior lipid-lowering scripts, n (%) 37834 (49.3) 5117 (32.6) 2940 (47.6) 2937 (47.6) 1266 (48.6) 1272 (48.8)
Prior antiplatelet scripts, n (%) 28327 (36.9) 5904 (37.6) 2721 (44.1) 2682 (43.5) 1066 (40.9) 1100 (42.2)
Smoking history, n (%)
Non-smoker 31199 (40.6) 7079 (45.1) 2435 (39.5) 2516 (40.8) 1047 (68.5) 1047 (68.5)
Ex-smoker 31902 (41.5) 5432 (34.6) 2687 (43.5) 2571 (41.7) 1215 (48.6) 1215 (48.6)
Current smoker 13614 (17.7) 2995 (19.1) 1049 (17.0) 1084 (17.6) 342 (40.9) 342 (40.9)
Missing data for metformin, sulphonylurea, respectively: serum creatinine 8653 (11.3%), 4715 (30.1%); HbA1c 14 558 (19%), 5229 (33.3%); BMI 10 886
(14.2%), 4717 (30.1%); systolic BP 3839 (5%), 2144 (13.7%); total cholesterol 8813 (11.5%), 5078 (32.4%); HDL 21 531 (28%), 8722 (55.6%); LDL 32 915
(42.9%), 10 576 (67.4%); triglycerides 20 269 (26.4%), 8126 (51.8%); smoking history 181 (0.2%), 96 (0.6%). HbA1c, haemoglobin A1c; BMI, body mass
index; BP, blood pressure; HDL, high-density lipoprotein; LDL, low-density lipoprotein.
Statistics 20. Studies using the CPRD are covered by ethics
approval granted by the Trent Multicentre Research Ethics
Committee (reference 05/MRE04/87). This study was granted
CPRD Independent Scientic Advisory Committee approval
(ISAC 12155).
Results
In the main analysis, 76 811 patients were prescribed
metformin monotherapy (mean follow-up 2.9 years) and
15 687 sulphonylureas (3.1 years). Of the patients prescribed
sulphonylurea the majority were initiated with gliclazide
(12 948; 82.5%). Other agents were glimepiride (1092; 7.0%),
glipizide (713; 4.5%), glibenclamide (629; 4.0%) and tolbu-
tamide (305; 1.9%). A total of 2604 patients were included
in each arm of the directly matched cohorts and 8836 in the
propensity-matched. Baseline characteristics for these groups
are shown in Table 1. For the overall cohort, there were sig-
nicant differences between those treated with metformin and
sulphonylurea, including age (61.4 vs. 68.1 years), creatinine
(85.4 vs. 98.1 mol/l), Charlson index (1.9 vs. 2.2), and HbA1c
[8.6% (70.5 mmol/mol) vs. 9.1% (76.0 mmol/mol)]. During
follow-up, 357 (2.3%) patients treated with sulphonylurea had
at least one recording of a hypoglycaemic event compared with
546 (0.7%) patients treated with metformin.
Main Analysis
There were with 2172 deaths in the sulphonylurea cohort
compared with 3209 in the metformin cohort, representing
crude rates of 44.6 and 13.6 events per 1000 person-years,
respectively, and a crude relative risk of 3.270 (95% CI
3.0963.453) (Table 2). After adjustment inthe CPHM, the HR
was 1.580 (1.4831.684), p <0.001. In sub-group analyses by
age, gender, glucose control, prior renal disease and morbidity
status, all HRs were signicantly greater for sulphonylurea
compared with metformin, ranging from 1.476 (1.3031.671)
for patients with a Charlson index of 1 to 2.280 (1.8572.799)
for those aged under 63 years (Figure 1). Notably, aHRs were
similar for patients with or without prior renal disease: 1.654
(1.4271.916) and 1.577 (1.4661.695), respectively.
There were 787 rst MACE events in the sulphonylurea
cohort compared with 2035 in the metformin cohort,
representing crude rates of 20.1 and 9.7 events per 1000
patient-years, respectively, and a crude relative risk of 2.079
(1.9142.257) (Table 3). After adjustment in the CHPM, there
was a signicantly increased HR for sulphonylurea relative to
metformin of 1.196 (1.0901.313), p <0.001.
Directly Matched Analysis
In the directly matched analysis, there were 174 and 154 deaths
in the sulphonylurea and metformin cohorts, representing
2014 doi:10.1111/dom.12302 3
original article DIABETES, OBESITY AND METABOLISM
Table 2. Events, crude rates, risk ratios and adjusted HRs for all-cause mortality for patients treated with sulphonylurea and metformin monotherapy.
Cohort n Events
Crude
rates (pkpy) Risk ratio (95% CI) HR (95% CI) p-Value
All subjects Sulphonylurea 15687 2172 44.6 3.270 (3.0963.453) 1.580 (1.4831.684)* <0.001
Metformin 76811 3209 13.6
Directly matched Sulphonylurea 2604 174 21.7 1.390 (1.1191.727) 1.272 (1.0211.584)* 0.032
Metformin 2604 154 15.6
Propensity-matched Sulphonylurea 8836 1121 41.3 1.806 (1.6461.982) 1.902 (1.7332.088)
1.504 (1.3671.654)*
<0.001
Metformin 8836 739 22.9
Pkpy, per thousand person-years; HR, hazard ratio.
*Adjusted.
Unadjusted.
aHR(95%CI) Analysis aHR 95%CI
Events/ Number
of patients
Overall 1.58 (1.48-1.58) (5381/92,498)
Gender
Males 1.61 (1.48-1.61) (2999/52,403)
Females 1.53 (1.39-1.53) (2,382/40,095)
Age (years)
62 2.28 (1.86-2.28) (591/44,752)
>63 1.51 (1.42-1.51) (4,790/47,746)
Charlson index
1 1.48 (1.30-1.48) (1,593/47251)
2 1.62 (1.51-1.62) (3788/45247)
HbA1c at baseline (%)
7.3 1.49 (1.35-1.49) (918/18,270)
>7.3 1.64 (1.48-1.64) (2835/55,081)
Prior renal impairment
Absent 1.58 (1.47-1.58) (4,290/81,017)
Present 1.65 (1.43-1.65) (1,091/11,481)
5
0.5 1.0 2.0
aHR
Figure 1. Adjusted hazard ratios for all-cause mortality by sub-groups for sulphonylurea versus metformin.
crude rates of 21.7 and 15.6 events per 1000 person years,
respectively, and a crude relative risk of 1.390 (1.1191.727)
(Table 2). After adjustment, the HR was 1.272 (1.0211.584).
There were 62 rst MACE events for patients treated with
sulphonylurea and 85 for those treated with metformin;
respective rates of 10.7 and11.7 per 1000 patient years (Table 3).
The crude relative risk was 0.913 (0.6591.265).
Propensity-matched Analysis
In the propensity-matched cohorts, there were 1121 deaths
in those treated with sulphonylurea and 739 deaths in those
treated with metformin; respective rates of 41.3 and 22.9
events per 1000 person-years and a crude relative risk of 1.806
(1.6461.982) (Table 2). In the CPHM, the HR was 1.902
(1.7332.088) prior to adjustment and 1.504 (1.3671.654)
following adjustment. For MACE, there were 249 events in
the sulphonylurea cohort and 216 in the metformin cohort;
respective rates of 15.3and12.7per 1000person-years (Table 3).
Survival Following Discontinuation
Following failure of rst-line monotherapy, 20 341 patients
switched to a dual therapy of metformin and sulphonylurea;
Table 3. Events, crude rates, risk ratios and adjusted HRs for major adverse cardiovascular events (MACE) for patients treated with sulphonylurea and
metformin monotherapy.
Cohort n Events
Crude
rates (pkpy) Crude risk ratio (95% CI) HR (95% CI) p-Value
All subjects Sulphonylurea 12860 787 20.1 2.079 (1.9142.257) 1.196 (1.0901.313)* <0.001
Metformin 68139 2035 9.7
Directly matched Sulphonylurea 1927 62 10.7 0.913 (0.6591.265) 0.814 (0.5781.148)* 0.241
Metformin 1927 85 11.7
Propensity-matched Sulphonylurea 5356 249 15.3 1.202 (1.0311.40) 1.202 (1.0011.442)
1.068 (0.8811.295)*
<0.001
0.503 Metformin 5356 216 12.7
Pkpy, per thousand person-years; HR, hazard ratio.
*Adjusted.
Unadjusted.
4 Morgan et al. 2014
DIABETES, OBESITY AND METABOLISM original article
14 621 (71.9%) were previously treated with metformin and
5720 (28.1%) with sulphonylurea. There were 987 deaths
(29.6 per 1000 person-years) in those initially treated with
sulphonylurea and1102 (17.5) inthose treatedwithmetformin.
The aHR for sulphonylurea versus metformin was 1.026 (95%
CI 0.9311.131).
Discussion
This study reports an increased risk of mortality associated with
sulphonylurea monotherapy when compared with patients
treatedwithmetformin; a nding observedusing three differing
analytical approaches. Comparable ndings fromsimilar retro-
spective cohort designs have been published previously [810].
For the secondary endpoint of MACE the results were
less clear, with signicantly increased HRs for sulphonylurea
relative to metformin observed in the main and unadjusted
propensity-matched analyses, but no signicant difference
foundinthe adjustedpropensity-matchedanddirectlymatched
approaches. The latter may have been underpowered for
the MACE endpoint. In the USA, a recent cohort study
reportedsignicantlyincreasedHRs of non-fatal cardiovascular
endpoints for sulphonylurea relative to metformin, which
were similar regardless of the type of sulphonylurea used
[9]. A meta-analysis of randomized controlled trials (RCTs)
and cohort studies [11] reported similar ndings, although
a signicant difference was not observed when the meta-
analysis was restricted to RCTs. However, comparative trials
of sulphonylurea and metformin have tended to have been
powered for efcacy rather than for longer-term safety
outcomes such as either mortality or cardiovascular events
[1214]. The United Kingdom Prospective Diabetes Study
(UKPDS), with a 10-year follow-up, reported signicant
reduction in both myocardial infarction and mortality for
obese patients initiated with metformin compared with those
treated with either sulphonylurea or insulin [15].
Metformin is contraindicated in patients with renal
impairment due to the risk of lactic acidosis resulting from the
inhibition of hepatic gluconeogenesis. However, it is known
that patients with a degree of renal impairment are treated
with metformin in real-world clinical practice [16] and that
consequent lactic acidosis is a rare event [17]. In our study,
patients with renal impairment or elevated serum creatinine
were represented equally in both cohorts (metformin 11.3%vs.
sulphonylurea 12.7%). In contrast to other studies [8,9], we did
not exclude these patients but adjusted for themin our analyses
and also performed a subgroup analysis for those with and
without a history of renal impairment and/or elevated serum
creatinine. It is of interest that, in this sub-group analysis, the
HR for all-cause mortality remained signicantly higher (aHR
=1.654) for patients treated with sulphonylurea. The decision
not to treat a patient with metformin due to contraindication
should therefore be balanced against other risks and benets as
discussed in two recent commentaries [17,18].
It is not clear from our study whether the difference in
the risk of all-cause mortality is due to the adverse impact
of sulphonylurea or to the benecial effect of metformin.
Sulphonylureas are known to have an increased risk of
hypoglycaemia and there was a threefold increase in our study.
There is evidence fromepidemiological studies [19] andclinical
trials [20] suggesting that hypoglycaemia is associated with
increased risk of all-cause mortality and cardiovascular events.
Sulphonylureas are also known to lead to weight gain [21]
and may inhibit ischaemic preconditioning [22]. In contrast,
metformin is associated with lower rates of cancer [23] and
is also believed to have an impact upon myocardial infarction
beyond its glucose-lowering properties [24,25].
This study has a number of limitations common to observa-
tional studies. Whilst observational studies cannot replace
randomized trials they can be considered complementary
as they do not omit patients on the basis of strict inclu-
sion/exclusioncriteria andwill usuallyhave increasedfollow-up
time. This may be of particular importance in evaluating out-
comes with long latencies.
As cases were not randomized to each cohort, systematic
baseline differences between the treatment cohorts existed.
Furthermore, treatments may have been dictated by individual
patient prole. For example, sulphonylureas may be used
in patients contraindicated to metformin due to renal
impairment, whilst metformin may be preferred in patients
prone to weight gain. Although the survival models accounted
for some of these differences as factors and covariates, there
is still potential for some unseen selection bias to remain. To
partly address this, we analysed two subsets of patients, the
rst matched directly on key characteristics and the second
by propensity score. Furthermore, following cessation of
monotherapy, we followed those patients who switched to a
regimen of metformin and sulphonylurea dual therapy to death
or last inclusion in the dataset. The HR for initial rst-line
therapy (sulphonylurea vs. metformin) for all-cause mortality
was unity (1.026), indicating that there was no underlying
difference between these two cohorts. It is accepted that, as
these patients were not contraindicated to either metformin or
sulphonylurea, they may represent a subgroupthat is not gener-
alizable to the cohort as a whole. However, it does indicate that,
at least for these patients, residual confounding is negligible.
In our study, exposure to either sulphonylurea or
metforminwas basedonprimary-care prescription. We have no
data as to whether the patient actually lled the prescription at
the pharmacy or, further, as to whether they took the medicine
at the recorded dosage. Patterns of adherence may have been
different between the two cohorts due to characteristics of both
the individual therapy and of the cohort itself at baseline.
The weight of scientic evidence so far shows that sulphony-
lurea represents an increased hazard of serious adverse events
over metformin [26]. This study conrms these ndings, with
sulphonylurea being associated with increased risk of all-cause
mortality compared with metformin. Guidelines recom-
mending metformin as rst-line monotherapy are therefore
appropriate. However, for those patients for whom metformin
is deemed unsuitable due to contraindication, there must be
some question as to whether sulphonylurea should still be con-
sidered the rst-choice alternative. An RCT may be required
to determine the appropriate alternative, as these data raise
questions about the safety of sulphonylurea more generally.
2014 doi:10.1111/dom.12302 5
original article DIABETES, OBESITY AND METABOLISM
Acknowledgements
This research was funded by AstraZeneca and Bristol-Myers
Squibb.
Conict of Interest
C. Ll. M. is a contractor of Pharmatelligence, S. J. J. and S.
E. H. are employed by Pharmatelligence and C. J. C. is a
director of Pharmatelligence, a research consultancy receiving
funding from pharmaceutical companies. J. M. is an employee
of Bristol-Myers Squibb.
C. Ll. M. was the principal investigator, performed the
statistical analysis, wrote the rst draft of the manuscript and
reviewed and edited subsequent versions. J. M. contributed to
the study design and reviewed the data and manuscript. S. J. J.
extracted and reviewed the data and contributed to and edited
the manuscript. S. E. H. contributed to the study design and
manuscript. C. J. C. conceived, designed and supervised the
study, and contributed to and reviewed the manuscript. C. J. C.
is the guarantor of this work and, as such, had full access to all
the data in the study and takes responsibility for the integrity
of the data and the accuracy of the data analysis.
References
1. Diabetes UK. Diabetes in the UK 2011/2012: Key Statistics
on Diabetes. London: Diabetes UK 2011. Available from URL:
http://www.diabetes.org.uk/Documents/Reports/Diabetes-in-the-UK-
2011-12.pdf. Accessed 7 February 2014.
2. The National Collaborating Centre for Chronic Conditions. Type 2
Diabetes: The Management of Type 2 Diabetes (Update) (CG66).
London: Royal College of Physicians 2008. Available from URL:
http://guidance.nice.org.uk/CG66. Accessed 7 February 2014.
3. Desai NR, Shrank WH, Fischer MA et al. Patterns of medication initiation
in newly diagnosed diabetes mellitus: quality and cost implications. Am J
Med 2012; 125: 302.e1302.e7.
4. Simpson SH, Majumdar SR, Tsuyuki RT, Eurich DT, Johnson JA.
Doseresponse relation between sulfonylurea drugs and mortality in
type 2 diabetes mellitus: a population-based cohort study. CMAJ 2006;
174: 169174.
5. Currie CJ, Poole CD, Gale EAM. The inuence of glucose-lowering therapies
on cancer risk in type 2 diabetes. Diabetologia 2009; 52: 17661777.
6. Garratt KN, Brady PA, Hassinger NL, Grill DE, Terzic A, Holmes DR Jr.
Sulfonylurea drugs increase early mortality in patients with diabetes
mellitus after direct angioplasty for acute myocardial infarction. J Am Coll
Cardiol 1999; 33: 119124.
7. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of
classifying prognostic comorbidity in longitudinal studies: development
and validation. J Chronic Dis 1987; 40: 373383.
8. Wheeler S, Moore K, Forsberg CW et al. Mortality among veterans
with type 2 diabetes initiating metformin, sulfonylurea or rosiglitazone
monotherapy. Diabetologia 2013; 56: 19341943.
9. Roumie CL, Hung AM, Greevy RA. Comparative effectiveness of
sulfonylurea and metformin monotherapy on cardiovascular events in
type 2 diabetes mellitus: a cohort study. Ann Intern Med 2012; 157:
601610.
10. Tzoulaki I, Molokhia M, Curcin V et al. Risk of cardiovascular disease and
all cause mortality among patients with type 2 diabetes prescribed oral
antidiabetes drugs: retrospective cohort study using UK general practice
research database. BMJ 2009; 339: b4731.
11. Phung OJ, Schwartzman E, Allen RW, Engel SS, Rajpathak SN.
Sulphonylureas and risk of cardiovascular disease: systematic review
and meta-analysis. Diabet Med 2013; 30: 11601171.
12. DeFronzo RA, Goodman AM. Efcacy of metformin in patients with non-
insulin-dependent diabetes mellitus. The Multicenter Metformin Study
Group. N Engl J Med 1995; 333: 541549.
13. Garber AJ, Donovan DS Jr, Dandona P, Bruce S, Park JS. Efcacy of
glyburide/metformin tablets compared with initial monotherapy in type
2 diabetes. J Clin Endocrinol Metab 2003; 88: 35983604.
14. Kahn SE, Haffner SM, Heisse MA, Herman WH, Holman RR. Glycemic
durability of rosiglitazone, metformin, or glyburide monotherapy. N Engl J
Med 2006; 355: 24272443.
15. UK Prospective Diabetes Study (UKPDS) Group. Effect of intensive blood-
glucose control with metformin on complications in overweight patients
with type 2 diabetes (UKPDS 34). Lancet 1998; 352: 854865.
16. Lipska KJ, Bailey CJ, Inzucchi SE. Use of metformin in the setting of mild-
to-moderate renal insufciency. Diabetes Care 2011; 34: 14311437.
17. Emslie-Smith AM, Boyle DIR, Evans JMM, Sullivan F, Morris AD.
Contraindications to metformin therapy in patients with Type 2
diabetesa population-based study of adherence to prescribing
guidelines. Diabet Med 2001; 18: 483488.
18. Kajbaf F, Arnouts P, de Broe M, Lalau J-D. Metformin therapy and kidney
disease: a review of guidelines and proposals for metformin withdrawal
around the world. Pharmacoepidemiol Drug Saf 2013; 22: 10271035.
19. Currie CJ, Peters JR, Tynan A et al. Survival as a function of HbA1c in people
with type 2 diabetes: a retrospective cohort study. Lancet 2010; 375:
481489.
20. Zoungas S, Patel A, Chalmers J et al. ADVANCE Collaborative Group. Severe
hypoglycemia and risks of vascular events and death. N Engl J Med 2010;
363: 14101418.
21. Morgan CL, Jenkins-Jones S, Evans M, Barnett AH, Poole CD, Currie CJ.
Weight change in people with type 2 diabetes: secular trends and the
impact of alternative antihyperglycaemic drugs. Diabetes Obes Metab
2012; 14: 424432.
22. Cleveland JC, Meldrum DR, Cain BS, Banerjee A, Harken AH. Oral
sulfonylurea hypoglycemic agents prevent ischemic preconditioning in
human myocardium. Two paradoxes revisited Circulation 1997; 96: 2932.
23. Soranna D, Scotti L, Zambon A et al. Cancer risk associated with use of
metformin and sulfonylurea in type 2 diabetes: a meta-analysis. Oncologist
2012; 17: 813822.
24. Kirpichnikov D, McFarlane SI, Sowers JR. Metformin: an update. Ann Intern
Med 2002; 137: 2533.
25. Roussel R, Travert F, Pasquet B et al. Reduction of Atherothrombosis
for Continued Health (REACH) Registry Investigators. Metformin use and
mortality among patients with diabetes and atherothrombosis. Arch Intern
Med 2010; 170: 18921899.
26. Holden SE, Currie CJ. Mortality risk with sulfonylureas compared to
metformin. Diabetes Obes Metab 2014; DOI: 10.1111/dom.12280 [Epub
ahead of print].
6 Morgan et al. 2014