Determinants of treatment modification before and after implementation of the updated 2015 NICE guideline on type 2 diabetes: A retrospective cohort study

  • Judith van Dalem
    Affiliations
    Department of Clinical Pharmacy, Maastricht University Medical Centre+, Maastricht, the Netherlands

    Cardiovascular Research Institute Maastricht, Maastricht University Medical Centre+, Maastricht, the Netherlands
    Search for articles by this author
  • Martijn C.G.J. Brouwers
    Affiliations
    Department of Internal Medicine, Division of Endocrinology and Metabolic Disease, Cardiovascular Research Institute Maastricht (CARIM), Maastricht University Medical Centre+, Maastricht, the Netherlands
    Search for articles by this author
  • Andrea M. Burden
    Affiliations
    Department of Chemistry and Applied Biosciences, Institute of Pharmaceutical Sciences, ETH Zurich, Switzerland
    Search for articles by this author
  • Coen D.A. Stehouwer
    Affiliations
    Department of Internal Medicine and Cardiovascular Research Institute Maastricht (CARIM), Maastricht University Medical Centre+, Maastricht, the Netherlands
    Search for articles by this author
  • Olaf H. Klungel
    Affiliations
    Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht Institute of Pharmaceutical Sciences, Utrecht, the Netherlands
    Search for articles by this author
  • Frank de Vries
    Correspondence
    Corresponding author at: Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht Institute for Pharmaceutical Sciences, Utrecht, the Netherlands.
    Affiliations
    Department of Clinical Pharmacy, Maastricht University Medical Centre+, Maastricht, the Netherlands

    Cardiovascular Research Institute Maastricht, Maastricht University Medical Centre+, Maastricht, the Netherlands

    Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht Institute of Pharmaceutical Sciences, Utrecht, the Netherlands
    Search for articles by this author
  • Johanna H.M. Driessen
    Affiliations
    Department of Clinical Pharmacy, Maastricht University Medical Centre+, Maastricht, the Netherlands

    Cardiovascular Research Institute Maastricht, Maastricht University Medical Centre+, Maastricht, the Netherlands

    Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht Institute of Pharmaceutical Sciences, Utrecht, the Netherlands

    NUTRIM School for Nutrition and Translational Research in Metabolism, Maastricht University, Maastricht, the Netherlands
    Search for articles by this author
Open AccessPublished:April 21, 2021DOI:https://doi.org/10.1016/j.diabres.2021.108828

      Highlights

      • More diverse prescribing of second-line therapies was observed after 2015.
      • A first step towards individually tailoring prescribing has been made.
      • Not all noteworthy determinants influenced general practitioners’ prescribing.

      Abstract

      Aims

      To identify patient-specific factors associated with early metformin treatment modification among type 2 diabetes patients before and after implementation of the updated 2015 NICE (National Institute for Health and Care Excellence) guideline.

      Methods

      We conducted a population-based cohort study using data from the Clinical Practice Research Datalink GOLD database (2009–2016). Patients ≥ 18 years, newly treated with metformin only, during the period of valid data collection were included. The first prescription defined start of follow-up. Determinants of treatment modification in two cohorts (before and after implementation of the updated guideline) were studied by time-dependent Cox proportional hazards regression.

      Results

      After implementation of the updated guideline, patients were less likely to receive sulphonylureas (62.3% vs 41.3%) or thiazolidediones (4.7% vs 2.2%) and more likely to receive dipeptidyl peptidase-4 inhibitors (15.8% vs 27.1%) or sodium-glucose cotransporter-2 inhibitors (0.8% vs 9.9%). Some determinants influenced general practitioners’ prescribing differently after implementation of the updated guideline compared to before, including a high body mass index and heart failure.

      Conclusions

      Our results indicate that a first step towards tailored prescribing has been made. However, not all determinants that are important to consider when prescribing second-line glucose-lowering agents were of influence on general practitioners’ prescribing.

      Keywords

      1. Introduction

      For decades, metformin has been the recommended first-line treatment for patients with type 2 diabetes as it lowers fasting blood glucose levels by 20 percent and glycated haemoglobin (HbA1c) levels by 1.5 percent points [

      McCulloch DK. Metformin in the treatment of adults with type 2 diabetes mellitus. UpToDate; 2013 20 December 2013.

      ]. However, not all patients achieve adequate glucose control with metformin alone and therefore require stepped-up therapy.
      Although sulphonylureas have been the second-line therapy for many years, the arrival of several new therapies (e.g. dipeptidyl peptidase-4 [DPP-4] inhibitors, sodium-glucose cotransporter-2 [SGLT-2] inhibitors and glucagon-like peptide-1 [GLP-1] receptor agonists) has enabled tailoring of treatment to individual patient characteristics. This has led to substantial changes in type 2 diabetes management guidelines in many countries, including the UK NICE (National Institute for Health and Care Excellence) guideline in 2015. While sulphonylureas were the preferred second-line therapy in the 2009 guideline [

      National Institute for Health and Care Excellence. Type 2 diabetes: The management of type 2 diabetes. Published May 2009.

      ], the new guideline (published in December 2015) recommends to choose the second-line treatment based on patient characteristics, risk factors, treatment efficacy, safety and tolerability, costs and patient preferences [

      National Institute for Health and Care Excellence. Type 2 diabetes in adults: management Published December 2015, Last updated August 2019.

      ].
      Recent studies have shown that patient characteristics and risk factors are associated with treatment choices [

      Maguire A, Mitchell BD, Ruzafa JC. Antihyperglycaemic treatment patterns, observed glycaemic control and determinants of treatment change among patients with type 2 diabetes in the United Kingdom primary care: a retrospective cohort study. BMC Endocrine Disorders 2014 Aug 27; 14: 73. PubMed PMID: 25163796. Pubmed Central PMCID: 4161267.

      ,
      • Heintjes E.M.
      • Overbeek J.A.
      • Hall G.C.
      • Prieto-Alhambra D.
      • Lapi F.
      • Hammar N.
      • et al.
      Factors associated with type 2 diabetes mellitus treatment choice across four European countries.
      ]. In particular, body mass index (BMI), HbA1c, age, cardiovascular risk and renal function have been identified as significant determinants of general practitioners’ prescribing [

      Maguire A, Mitchell BD, Ruzafa JC. Antihyperglycaemic treatment patterns, observed glycaemic control and determinants of treatment change among patients with type 2 diabetes in the United Kingdom primary care: a retrospective cohort study. BMC Endocrine Disorders 2014 Aug 27; 14: 73. PubMed PMID: 25163796. Pubmed Central PMCID: 4161267.

      ,
      • Heintjes E.M.
      • Overbeek J.A.
      • Hall G.C.
      • Prieto-Alhambra D.
      • Lapi F.
      • Hammar N.
      • et al.
      Factors associated with type 2 diabetes mellitus treatment choice across four European countries.
      ]. However, these studies were performed with British data prior to 2015, with sulphonylureas as the recommended second-line therapy. Importantly, these studies did not account for important risk factors such as ethnicity, drug-related side-effects, contraindications or complications. Moreover, it is of great interest to investigate whether the implementation of the updated guideline has resulted in more individualised prescribing.
      Therefore, the objective of this study was to identify patient-specific determinants of early treatment modification (addition of or switching to second-line therapy within one year) in patients with type 2 diabetes before and after implementation of the updated NICE guideline in the UK.

      2. Subjects, materials and methods

      2.1 Data sources

      A cohort study was conducted using the Clinical Practice Research Datalink (CPRD) GOLD. The CPRD GOLD contains prospectively collected data of 674 primary care practices in the UK including approximately 7% of the British population. It comprises a wide range of information including demographics, ethnicity, diagnoses, referrals to secondary care, test results, prescription details and health-related behaviours. Data in CPRD GOLD have been shown to be valid and of high quality for a wide range of diseases [

      Herrett E, Gallagher AM, Bhaskaran K, Forbes H, Mathur R, van Staa T, et al. Data Resource Profile: Clinical Practice Research Datalink (CPRD). Int J Epidemiol 2015 Jun; 44(3): 827-3PubMed PMID: 26050254. Pubmed Central PMCID: 4521131.

      ].

      2.2 Study population

      Patients who received a first ever prescription of metformin (no other glucose-lowering agent) and aged ≥ 18 years during the period of valid CPRD GOLD data collection were included. All patients were required to have at least 1-year of eligible data collection to meet eligibility for our study. Therefore, we can ensure that all patients in our study had a minimum of a 1-year period of non-use of glucose-lowering agents prior to their first metformin prescription. For this study, two cohorts were created: one cohort with data from May 2009 (introduction of the old NICE guideline) - December 2014 and one cohort with data from January 2016 - December 2016. New users of metformin in the period January 2015 - December 2015 were not included in cohort 1 to exclude influence of the upcoming updated guideline (December 2015). Cohort entry (index date) was defined as the date of the first ever metformin prescription after start of valid data collection.

      2.3 Study outcomes

      To investigate determinants of prescribing of second-line therapies, the main outcome of interest was a prescription of a glucose-lowering agent other than metformin, i.e. second-line therapy. This was defined as a prescription of sulphonylureas, thiazolidinediones, DPP-4 inhibitors, GLP-1 receptor agonists, SGLT-2 inhibitors, insulin and/or other glucose-lowering agents (e.g. α-glucosidase inhibitors and repaglinide) after metformin only therapy. As a secondary outcome of interest we studied the potential determinants of a prescription of sulphonylureas only (and no other second-line therapies) after metformin only therapy. Sulphonylureas were chosen as we hypothesized that the greatest change would be observed in this group since sulphonylureas would no longer be the only second-line option after the implementation of the updated guideline.
      All patients were followed for a maximum of one year after cohort entry, or until patients received a prescription of a second-line therapy, the date of transfer out of the practice area, the date on which the practice stopped delivering data, or the date of death in the CPRD; whichever came first. All patients with a history of a glucose-lowering drug prescriptions, including insulin, were excluded.

      2.4 Determinants

      Determinants of treatment modification were largely based on diabetes-related complications, side-effects and contraindications of the different glucose-lowering agents and other patient- and prescribing-related factors. All determinants were selected based on a review of literature and reviewed by clinical experts to ensure clinical relevance prior to analysis. Potential determinants of treatment modification were assessed at index date or as time-dependent determinants. Total follow-up was divided into 90-day intervals in order to assess time-dependent determinants.
      The following determinants were examined at the index date: age, sex, BMI, alcohol use and ethnicity [

      Mathur R, Bhaskaran K, Chaturvedi N, Leon DA, vanStaa T, Grundy E, et al. Completeness and usability of ethnicity data in UK-based primary care and hospital databases. J Public Health 2014 Dec; 36(4) :684-92. PubMed PMID: 24323951. Pubmed Central PMCID: 4245896.

      ]. Other determinants considered in this study, including drug-related side-effects, contraindications or complications, were identified time-dependently at the start of each new interval. These included the following most recently recorded values in the past 6 months: HbA1c, fasting plasma glucose and estimated glomerular filtration rate (eGFR); chronic liver disease; a history of a hypoglycaemic event; cardiovascular disease; hypertension; heart failure; ischaemic heart disease; cerebrovascular disease; oedema; haematuria; microalbuminuria; bladder cancer; pancreatic cancer; pancreatitis; gastro-intestinal complications; (proxy indicators of) osteoporosis, schizophrenia/psychosis; Alzheimeŕs diseases/dementia; cognitive impairment; or a profession as a driver.

      2.5 Data analysis

      We investigated the percentage of patients with a treatment modification during the first year of metformin therapy, before and after implementation of the updated guideline. The proportional contribution of each individual glucose lowering treatment group was calculated.
      To evaluate trends of prescribing over time, the proportion of prescriptions for each second-line glucose-lowering agent was calculated annually for the years 2009–2014 and 2016. We used chi-square tests to evaluate differences before and after implementation of the guideline. In addition, we investigated if the treatment modification could be considered as a treatment intensification or a treatment switch. Stop- and start-dates of corresponding therapies were compared to define intensifications or switches. Patients were considered a switcher if they had not received metformin during the 90 days following a prescription of a second-line therapy. Otherwise the treatment modification was considered an intensification.
      Cox proportional hazard models were used to identify determinants of early treatment modification. Crude and adjusted Hazard Ratios (HRs) were calculated. HRs were adjusted for all other determinants using multivariate regression. Missing values were included in the models as a separate category. A test of interaction was performed to investigate if HRs before and after the updated guideline were statistically significantly different [

      Altman DG, Bland JM. Interaction revisited: the difference between two estimates. Bmj. 2003 Jan 25; 326(7382): 219. PubMed PMID: 12543843. Pubmed Central PMCID: 1125071.

      ]. In a sensitivity analysis we included only new metformin users in 2014 as the pre-intervention group (instead of 2009–2014). All other methods were identical to the primary analysis. All statistical analyses were performed using SAS statistical software, version 9.4 (SAS Institute, Inc., Cary, NC, USA).

      3. Results

      3.1 Patients’ characteristics before and after implementation of the updated guideline

      A total of 111,789 patients starting metformin therapy during the study period were included, 100,313 (89.7%) before and 11,476 (10.3%) after the implementation of the updated guideline (Table 1). The mean age of metformin users was 58.0 years before and 58.5 years after implementation of the guideline in 2015. The proportion of female users increased slightly after guideline implementation, from 47.8% to 48.8%. No substantial differences were observed for clinical characteristics.
      Table 1Baseline characteristics of all metformin users before and after implementation of the updated guideline (December 2015).
      Before the updated guideline (May 2009 – December 2014)After the updated guideline (2016)
      N = 100,313%N = 11,476%
      Follow-up, years (mean [SD])0.96(0.2)0.96(0.2)
      Number of females47,962(47.8)5,606(48.8)
      Age, years (mean [SD])58.5(15.8)58.0(16.0)
        Median (IQR)60.0(22.0)59.0(23.0)
      Age Category
      18 – 29 years5,025(5.0)587(5.1)
      30 – 39 years8,284(8.3)1,087(9.5)
      40 – 49 years14,405(14.4)1,606(14.0)
      50 – 59 years21,908(21.8)2,652(23.1)
      60 – 69 years24,647(24.6)2,629(22.9)
      70 – 79 years17,340(17.3)1,903(16.6)
      80 + years8704(8.7)1,012(8.8)
      Alcohol use
      Yes63,468(63.3)6,823(59.5)
      No30,394(30.3)3,634(31.7)
      Unknown6,451(6.4)1,019(8.9)
      Ethnicity
      White48,630(48.5)5,360(46.7)
      South Asian2,555(2.5)294(2.6)
      Black1,641(1.6)201(1.8)
      Mixed1,844(1.8)236(2.1)
      Other2,826(2.8)275(2.4)
      Missing42,817(42.7)5,110(44.5)
      Geographic region
      North East1,314(1.3)108(0.9)
      North West10,888(10.9)858(7.5)
      Yorkshire and the Humber1,983(2.0)161(1.4)
      East Midlands1,653(1.6)0(0.0)
      West Midlands10,377(10.3)830(7.2)
      East of England7,013(7.0)402(3.5)
      South West8,723(8.7)757(6.6)
      South Central11,002(11.0)836(7.3)
      London12,742(12.7)1,407(12.3)
      South East Coast10,318(10.3)1,716(15.0)
      Northern Ireland2,947(2.9)565(4.9)
      Scotland8,338(8.3)1,631(14.2)
      Wales13,015(13.0)2,205(19.2)
      BMI, kg/m2 (mean [SD])32.3(6.9)32.6(7.1)
      <201,102(1.1)118(1.0)
      20–24.910,065(10.0)1,054(9.2)
      25–29.928,326(28.2)3,024(26.4)
      30–34.928,327(28.2)3,091(26.9)
      ≥ 3528,940(28.8)3,395(29.6)
      Unknown3,553(3.5)794(6.9)
      Most recent eGFR measurement (mL/min/1.73 m2) 6 months prior to index date
      <30106(0.1)<6(0.0)
      30–597,605(7.6)721(6.3)
      ≥6061,727(61.5)7,025(61.2)
      Unknown30,875(30.8)3,725(32.5)
      Most recent HbA1c measurement 6 months prior to index date (mean [SD])8.3(1.8)8.2(1.7)
      <6.5% (48 mmol/mol)4,400(4.4)605(5.3)
      6.5–7.4% (48–57 mmol/mol)18,054(18.0)2,757(24.0)
      7.5–8.5% (58–69 mmol/mol)15,901(15.9)1,781(15.5)
      >8.5% (69 mmol/mol)20,625(20.6)2,512(21.9)
      Unknown41,333(41.2)3,821(33.3)
      Most recent fasting plasma glucose level 6 months prior to index date (mean [SD])9.9(4.0)9.6(4.0)
      <6.0 mmol/L1,855(1.9)203(1.8)
      6.0–7.4 mmol/L6,796(6.8)468(4.1)
      7.5–8.9 mmol/L6,507(6.5)372(3.2)
      ≥9 mmol/L12,067(12.0)756(6.6)
      Unknown73,088(72.9)9,677(84.3)
      History of disease
      Cardiovascular disease6,200(6.2)652(5.7)
      Heart failure2,425(2.4)289(2.5)
      Ischemic heart disease12,493(12.5)1,192(10.4)
      Cerebrovascular disease6,167(6.1)649(5.7)
      Hypertension36,848(36.7)3,895(33.9)
      Gastro-intestinal complications
       Nausea5,633(5.6)708(6.2)
       Diarrhoea14,919(14.9)1,743(15.2)
       Vomiting7,573(7.5)931(8.1)
       Flatulence1,196(1.2)134(1.2)
      (Proxies of) osteoporosis
       Osteoporosis2,053(2.0)270(2.4)
       History of fracture22,017(21.9)2,738(23.9)
       Use of bisphosphonates2,095(2.1)222(1.9)
      Chronic liver disease1,702(1.7)237(2.1)
      History of a hypoglycaemic event430(0.4)51(0.4)
      Oedema10,760(10.7)1,030(9.0)
      Haematuria5,094(5.1)541(4.7)
      Microalbuminuria2,130(2.1)174(1.5)
      Bladder cancer450(0.4)43(0.4)
      Pancreas carcinoma50(0.0)10(0.1)
      Pancreatitis1,143(1.1)165(1.4)
      Schizophrenia/psychosis1,604(1.6)189(1.6)
      Alzheimer/Dementia1,754(1.7)237(2.1)
      Cognitive impairment328(0.3)61(0.5)
      Profession as driver756(0.8)113(1.0)
      Abbreviations: BMI, Body Mass Index; eGFR, estimated glomerular filtration rate; HbA1c, glycated haemoglobin.

      3.2 Initiation of second-line treatment before and after updated guideline

      Table 2 shows the distribution of the second-line therapies before and after implementation of the updated guideline. In the first year of treatment, 23,334 (23.3%) and 2,523 (22.0%) users of metformin received one treatment modification, before and after implementation of the updated guideline, respectively. Patients who switched to or received ≥ 2 new agents are not shown (6.6% and 6.1% of all treatment modifications). Following the new guideline, patients were less likely to receive a prescription of sulphonylureas (62.3% vs 41.3%) or thiazolidinediones (4.7% vs 2.2%) and more likely to receive a prescription of DPP-4 inhibitors (15.8% vs 27.1%) or SGLT-2 inhibitors (0.8% vs 9.9%). Treatment modifications were the result of a treatment intensification (i.e. second-line therapy was added to metformin) in 83% of the patients, both before and after implementation of the updated guideline (data not shown). In the remaining patients (17%), treatment modification was suggestive of a switch from metformin to another therapy.
      Table 2Switch to or addition of one second-line therapy in the first year of treatment.
      Before the updated guideline
      Distribution of second-line therapies is statistically significant different before and after implementation of the updated guideline (Chi-square test, p < 0.01).
      After the updated guideline
      Distribution of second-line therapies is statistically significant different before and after implementation of the updated guideline (Chi-square test, p < 0.01).
      Glucose-lowering drugN%N%
      All second-line therapies23,334(23.3)2,523(22.0)
      Sulphonylureas14,536(62.3)1,043(41.3)
      DPP-4 inhibitors3,676(15.8)683(27.1)
      Thiazolidinediones1,103(4.7)56(2.2)
      Insulin3,254(13.9)432(17.1)
      GLP-1 receptor agonists465(2.0)53(2.1)
      SGLT-2 inhibitors186(0.8)251(9.9)
      other114(0.5)5(0.2)
      Abbreviations: DPP-4, dipeptidyl peptidase 4; GLP-1 glucagon-like peptide-1; SGLT-2 sodium-glucose co-transporter 2.
      Distribution of second-line therapies is statistically significant different before and after implementation of the updated guideline (Chi-square test, p < 0.01).
      Prescribing trends from 2009 to 2014 (supplemental Fig. S1) show that DPP-4 inhibitors were already prescribed more often from 2009 to 2014, regardless of the introduction of the new guideline. However, a substantial shift in the distribution of second-line therapies was observed after implementation of the new guideline, in particular a rapid decline in sulphonylureas.

      3.3 Determinants of initiation of second-line treatment before and after updated guideline

      Overall, initiation of a second-line therapy was associated with a similar set of determinants before and after the guideline change (Table 3). For example, male sex, older age, lower eGFR, higher HbA1c, higher fasting plasma glucose level, history of hypoglycaemic events, pancreas carcinoma and pancreatitis were associated with an increased likelihood of prescribing second-line therapy. Conversely, the presence of gastro-intestinal complications, oedema, haematuria and a profession as driver were negatively associated with the initiation of second-line therapy.
      Table 3Influence of determinants on GP's prescribing of second line therapy before (May 2009 – December 2014) and after implementation (2016) of the new guideline among new users of metformin only in the first year of therapy.
      Before the updated guideline (May 2009 – December 2014)After the updated guideline (2016)
      Crude HR95% CIAdj. HR
      Results were corrected for all other determinants.
      95% CICrude HR95% CIAdj. HR
      Results were corrected for all other determinants.
      CI
      Sex
      FemaleReferenceReference
      Male1.19(1.16–1.22)1.09(1.07–1.13)1.18(1.10–1.27)1.11(1.02–1.21)
      Age Category
      18 – 29 yearsReferenceReference
      30 – 39 years2.27(2.06–2.50)2.52(2.29–2.78)1.91(1.46–2.49)2.18(1.66–2.85)
      40 – 49 years3.21(2.94–3.52)4.43(4.03–4.86)2.58(2.00–3.33)3.65(2.81–4.73)
      50 – 59 years2.90(2.65–3.17)4.46(4.07–4.90)2.33(1.82–2.99)3.68(2.85–4.76)
      60 – 69 years2.58(2.36–2.82)4.40(4.01–4.83)2.25(1.75–2.88)3.84(2.97–4.97)
      70 – 79 years2.54(2.32–2.78)4.49(4.08–4.94)1.96(1.52–2.52)3.42(2.62–4.47)
      80 + years2.74(2.49–3.02)3.90(3.52–4.32)2.31(1.77–3.02)3.36(2.52–4.48)
      Alcohol Use
      YesReferenceReference
      No1.29(1.26–1.33)1.30(1.26–1.33)1.23(1.13–1.33)1.25(1.14–1.36)
      Unknown1.62(1.54–1.70)1.38(1.31–1.46)1.81(1.61–2.04)1.36(1.19–1.56)
      Ethnicity
      WhiteReferenceReference
      South Asian1.19(1.10–1.27)1.09(1.01–1.18)1.02(0.80–1.40)1.05(0.82–1.34)
      Black1.22(1.11–1.33)1.03(0.94–1.13)1.09(0.82–1.44)0.98(0.74–1.31)
      Mixed0.97(0.88–1.06)0.93(0.84–1.02)1.05(0.80–1.37)1.13(0.86–1.48)
      Test of interaction showed a statistically significant difference before and after implementation of the new guideline [8].
      Other1.19(1.10–1.28)1.05(0.98–1.13)1.10(0.87–1.29)0.91(0.71–1.16)
      Missing0.86(0.84–0.89)0.85(0.82–0.87)0.97(0.90–1.05)0.90(0.83–0.97)
      BMI, kg/m2
      <201.21(1.07–1.35)1.18(1.05–1.33)1.40(0.99–1.98)1.31(0.92–1.87)
      20–24.9ReferenceReference
      25–29.90.86(0.82–0.89)0.86(0.82–0.90)0.92(0.80–1.06)0.95(0.82–1.10)
      30–34.90.79(0.76–0.83)0.81(0.77–0.85)0.86(0.75–1.00)0.92(0.79–1.06)
      >= 350.79(0.76–0.83)0.84(0.80–0.88)0.87(0.75–1.00)0.99(0.85–1.14)
      Test of interaction showed a statistically significant difference before and after implementation of the new guideline [8].
      Unknown1.69(1.58–1.81)1.19(1.10–1.28)1.91(1.61–2.26)1.36(1.13–1.62)
      Most recent eGFR (mL/min/1.73 m2) within 6 months prior to index date
      <302.30(1.77–2.97)2.61(2.01–3.38)3.45(1.11–10.65)4.87(1.55–15.29)
      30–591.17(1.11–1.23)1.28(1.21–1.36)1.15(0.97–1.35)1.39(1.15–1.67)
      ≥60ReferenceReference
      Unknown1.66(1.62–1.71)1.31(1.27–1.35)2.00(1.85–2.16)1.30(1.17–1.45)
      Most recent HbA1c within 6 months prior to index date
      <6.5% (48 mmol/mol)ReferenceReference
      6.5–7.4% (48–57 mmol/mol)1.79(1.64–1.96)1.75(1.60–1.91)1.65(1.29–2.10)1.63(1.27–2.08)
      7.5–8.5% (58–69 mmol/mol)4.63(4.25–5.04)4.42(4.06–4.81)3.43(2.69–4.37)3.38(2.64–4.31)
      >8.5% (69 mmol/mol)10.26(9.46–11.13)9.37(8.63–10.18)6.87(5.46–8.65)6.60(5.23–8.33)
      Test of interaction showed a statistically significant difference before and after implementation of the new guideline [8].
      Unknown7.39(6.82–8.01)6.71(6.18–7.28)7.01(5.61–8.76)5.99(4.74–7.57)
      Most recent fasting plasma glucose within 6 months prior to index date
      <6.0 mmol/LReferenceReference
      6.0–7.4 mmol/L1.18(0.99–1.40)1.06(0.89–1.27)1.37(0.74–2.56)1.51(0.81–2.82)
      7.5–8.9 mmol/L1.89(1.60–2.24)1.33(1.12–1.57)1.47(0.78–2.80)1.22(0.64–2.34)
      ≥9 mmol/L5.85(5.02–6.83)2.36(2.02–2.75)4.16(2.41–7.20)1.98(1.14–3.45)
      Unknown3.85(3.31–4.47)2.13(1.84–2.48)4.40(2.60–7.44)2.54(1.50–4.31)
      History of Disease
      Cardiovascular disease1.13(1.08–1.19)1.04(0.84–1.30)1.18(1.01–1.37)1.16(0.73–1.85)
      Heart failure1.24(1.15–1.33)1.20(1.11–1.29)0.90(0.71–1.14)0.90(0.70–1.15)
      Test of interaction showed a statistically significant difference before and after implementation of the new guideline [8].
      Ischemic heart disease1.06(1.02–1.09)1.10(1.06–1.15)0.95(0.84–1.07)1.01(0.89–1.15)
      Cerebrovascular disease1.13(1.08–1.19)1.04(0.84–1.30)1.16(1.00–1.35)1.02(0.64–1.63)
      Hypertension0.91(0.89–0.94)0.97(0.94–0.99)0.90(0.83–0.97)0.97(0.89–1.05)
      Gastro-intestinal complications
       Nausea0.75(0.71–0.80)0.95(0.89–1.01)0.71(0.60–0.84)1.00(0.84–1.19)
       Diarrhoea0.74(0.71–0.77)0.87(0.84–0.90)0.63(0.56–0.70)0.81(0.71–0.91)
       Vomiting0.74(0.70–0.78)0.91(0.86–0.96)0.69(0.60–0.81)0.92(0.79–1.09)
       Flatulence0.68(0.59–0.77)0.84(0.74–0.96)0.37(0.22–0.62)0.50(0.30–0.83)
      (Proxies) of osteoporosis
       Osteoporosis0.93(0.85–1.02)1.03(0.93–1.14)0.89(0.70–1.14)1.03(0.79–1.36)
       History of fracture1.02(0.99–1.06)1.05(1.02–1.08)0.89(0.81–0.97)0.93(0.85–1.02)
      Test of interaction showed a statistically significant difference before and after implementation of the new guideline [8].
       Use of bisphosphonates0.78(0.71–0.85)0.82(0.74–0.91)0.76(0.57–1.01)0.83(0.61–1.13)
      Chronic liver disease1.01(0.92–1.11)1.07(0.97–1.17)0.98(0.77–1.25)1.07(0.84–1.37)
      History of a hypoglycaemic event2.03(1.77–2.33)1.82(1.59–2.09)2.43(1.70–3.46)2.27(1.58–3.26)
      Oedema0.67(0.64–0.70)0.75(0.71–0.78)0.57(0.49–0.66)0.68(0.58–0.80)
      Haematuria0.80(0.76–0.85)0.87(0.82–0.93)0.63(0.51–0.77)0.77(0.63–0.95)
      Microalbuminuria0.88(0.81–0.95)0.91(0.84–0.99)0.92(0.70–1.22)1.04(0.78–1.38)
      Bladder cancer1.00(0.83–1.20)1.12(0.93–1.34)1.09(0.62–1.91)1.07(0.60–1.90)
      Pancreas carcinoma2.83(1.93–4.16)2.12(1.44–3.12)3.31(1.49–7.36)3.84(1.71–8.60)
      Pancreatitis1.58(1.43–1.74)1.48(1.35–1.64)2.01(1.59–2.55)1.89(1.48–2.40)
      Schizophrenia/psychosis1.30(1.19–1.42)1.17(1.07–1.28)0.72(0.52–1.00)0.64(0.46–0.89)
      Test of interaction showed a statistically significant difference before and after implementation of the new guideline [8].
      Alzheimer/Dementia1.41(1.30–1.54)0.96(0.88–1.05)1.60(1.29–2.00)1.29(1.01–1.64)
      Test of interaction showed a statistically significant difference before and after implementation of the new guideline [8].
      Cognitive impairment1.34(1.11–1.62)1.10(0.91–1.34)0.79(0.46–1.36)0.55(0.31–0.97)
      Test of interaction showed a statistically significant difference before and after implementation of the new guideline [8].
      Profession as driver0.67(0.58–0.78)0.64(0.55–0.75)0.55(0.36–0.86)0.61(0.39–0.94)
      Abbreviations: BMI, Body Mass Index; eGFR, estimated glomerular filtration rate; HbA1c, glycated haemoglobin.
      Results were corrected for all other determinants.
      # Test of interaction showed a statistically significant difference before and after implementation of the new guideline

      Altman DG, Bland JM. Interaction revisited: the difference between two estimates. Bmj. 2003 Jan 25; 326(7382): 219. PubMed PMID: 12543843. Pubmed Central PMCID: 1125071.

      .
      However, we found eight determinants of second-line therapy prescribing with a statistically significant different adjusted HR (aHR) before and after implementation of the updated guideline. The statistically significant, negative association with a BMI ≥ 35 kg/m2 that was observed before the implementation of the updated guideline (aHR 0.84; 95%CI 0.80–0.88), was not observed after implementation (aHR 0.99; 95%CI 0.85–1.14). The association of a mixed ethnicity with the likelihood of a second-line therapy prescription changed after the implementation of the updated guideline, but was not statistically significant at either time interval (aHR 1.13; 95%CI 0.86–1.48 after versus aHR 0.93; 95%CI 0.84–1.02 before). After the guideline change, a positive association with Alzheimer’s disease/dementia was observed (aHR 1.29; 95%CI 1.01–1.64), which was not observed before (aHR 0.96; 95%CI 0.88–1.05). In contrast, after the guideline change, a negative association with cognitive impairment was observed (aHR 0.55; 95%CI 0.31–0.97), which was not observed before (aHR 1.10; 95%CI 0.91–1.34). The increased likelihood of prescribing a second-line therapy in patients with heart failure before the implementation of the updated guideline (aHR 1.20; 95%CI 1.11–1.29), was not observed after (aHR 0.90; 95%CI 0.70–1.15). A similar result was observed for a history of fracture (aHR 0.93; 95%CI 0.85–1.02 after versus aHR 1.05; 95%CI 1.02–1.08 before). Schizophrenia/psychosis was associated with an reduced likelihood of a second-line therapy prescription after the updated guideline (aHR 0.64; 95%CI 0.46–0.89), but an increased likelihood before (aHR 1.17; 95%CI 1.07–1.28). Finally, a HbA1c measurement > 8.5% (69 mmol/mol) was less strongly associated with a prescription of a second-line therapy after implementation of the updated guideline (aHR 6.60; 95%CI 5.23–8.33) than before (aHR 9.37; 95%CI 8.63–10.18). Overall, the sensitivity analysis with new metformin users (N = 15,690) in 2014 as the pre-intervention group showed similar results (supplemental Table S1) as compared to our primary analysis with new metformin users in May 2009 through December 2014 as the pre-intervention group (Table 3). However, some determinants no longer showed a statistically significant difference. A BMI ≥ 35 kg/m2, a history of fracture, heart failure, and a HbA1c measurement > 8.5% were no longer statistically significantly different between the pre- and post-intervention periods, although the determinants BMI ≥ 35 kg/m2 and heart failure showed a similar trend as in the primary analysis.

      3.4 Determinants of initiation of sulphonylureas before and after the updated guideline

      We found five determinants of sulphonylurea prescribing, after metformin therapy, with a statically significant different aHR before and after implementation of the updated guideline (Table 4). Patients with a BMI 30–34.9 kg/m2 were less likely to receive a sulphonylurea after implementation of the updated guideline (aHR 0.54; 95%CI 0.55–0.68 after versus aHR 0.70; 95%CI 0.66–0.74 before). Similarly, prior to the guideline change, a stronger negative association with flatulence was observed (aHR 0.30; 95%CI 0.11–0.79 after versus aHR 0.83; 95%CI 0.70–0.98 before). In contrast, pancreatitis was more strongly associated with a prescription of a sulphonylurea after (aHR 2.69; 95%CI 1.94–3.73) than before (aHR 1.38; 95%CI 1.21–1.57) implementation of the updated guideline. The increased likelihood of a sulphonylurea prescription in patients with heart failure before the implementation of the updated guideline (aHR 1.27; 95%CI 1.16–1.40), was not observed after (aHR 0.82; 95%CI 0.55–1.23). A similar result was observed for schizophrenia/psychosis (aHR 0.57; 95%CI 0.32–1.01 after versus aHR 1.23; 95%CI 1.10–1.39 before).
      Table 4Influence of determinants on GP's prescribing of sulphonylureas before (May 2009 – December 2014) and after implementation (2016) of the new guideline among new users of metformin only in the first year of therapy.
      Before the updated guideline (May 2009 – December 2014)After the updated guideline (2016)
      Crude HR95% CIAdj. HR
      Results were corrected for all other determinants.
      95% CICrude HR95% CIAdj. HR
      Results were corrected for all other determinants.
      CI
      Sex
      FemaleReferenceReference
      Male1.24(1.20–1.28)1.09(1.05–1.13)1.38(1.22–1.56)1.18(1.03–1.35)
      Age Category
      18 – 29 yearsReferenceReference
      30 – 39 years3.19(2.69–3.77)3.46(2.92–4.09)2.55(1.33–4.90)2.74(1.43–5.28)
      40 – 49 years5.62(4.79–6.58)7.40(6.30–8.69)6.23(3.39–11.45)7.96(4.29–14.78)
      50 – 59 years5.27(4.51–6.17)7.74(6.60–9.08)5.39(2.95–9.86)7.69(4.16–14.23)
      60 – 69 years4.92(4.20–5.75)7.95(6.77–9.33)5.11(2.79–9.34)7.95(4.29–14.73)
      70 – 79 years5.15(4.40–6.03)8.49(7.22–9.98)4.85(2.62–8.91)7.57(4.06–14.14)
      80 + years6.19(5.27–7.27)8.04(6.79–9.51)6.91(3.73–12.81)8.96(4.72–17.00)
      Alcohol Use
      YesReferenceReference
      No1.26(1.21–1.30)1.25(1.20–1.29)1.15(1.01–1.31)1.15(1.00–1.33)
      Unknown1.51(1.42–1.62)1.33(1.24–1.43)1.45(1.17–1.79)1.18(0.93–1.50)
      Ethnicity
      WhiteReferenceReference
      South Asian1.10(0.99–1.22)1.06(0.95–1.17)0.92(0.61–1.39)1.05(0.69–1.60)
      Black1.26(1.12–1.42)1.13(1.00–1.28)1.34(0.87–2.05)1.22(0.79–1.88)
      Mixed1.03(0.91–1.16)1.02(0.90–1.16)1.20(0.79–1.82)1.30(0.85–1.99)
      Other1.28(1.17–1.40)1.17(1.07–1.28)1.01(0.67–1.54)0.82(0.54–1.26)
      Missing0.93(0.90–0.96)0.89(0.86–0.93)1.00(0.88–1.14)0.92(0.81–1.05)
      BMI, kg/m2
      <201.26(1.10–1.44)1.25(1.08–1.43)1.01(0.58–1.74)0.92(0.53–1.61)
      20–24.9ReferenceReference
      25–29.90.80(0.76–0.85)0.80(0.75–0.84)0.75(0.62–0.92)0.75(0.61–0.91)
      30–34.90.69(0.65–0.73)0.70(0.66–0.74)0.53(0.43–0.65)0.54(0.44–0.68)
      Test of interaction showed a statistically significant difference before and after implementation of the new guideline [8].
      >= 350.58(0.55–0.62)0.63(0.59–0.67)0.48(0.39–0.60)0.55(0.44–0.69)
      Unknown1.41(1.28–1.54)1.00(0.91–1.11)1.09(0.83–1.42)0.80(0.60–1.06)
      Most recent eGFR (mL/min/1.73 m2) within 6 months prior to index date
      <302.69(1.99–3.64)2.72(2.00–3.70)n.a.
      No patients with a renal function < 30 ml/min/1.73 m2 received a sulphonylurea after implementation of the updated guideline.
      n.a.
      No patients with a renal function < 30 ml/min/1.73 m2 received a sulphonylurea after implementation of the updated guideline.
      30–591.23(1.16–1.31)1.22(1.14–1.31)1.30(1.02–1.66)1.49(1.13–1.98)
      ≥60ReferenceReference
      Unknown1.41(1.36–1.46)1.16(1.11–1.21)1.59(1.40–1.81)1.03(0.86–1.23)
      Most recent HbA1c within 6 months prior to index date
      <6.5% (48 mmol/mol)ReferenceReference
      6.5–7.4% (48–57 mmol/mol)1.88(1.67–2.11)1.80(1.60–2.02)1.86(1.18–2.93)1.74(1.11–2.75)
      7.5–8.5% (58–69 mmol/mol)5.13(4.57–5.75)4.77(4.25–5.35)4.15(2.64–6.50)3.87(2.74–6.08)
      >8.5% (69 mmol/mol)11.81(10.59–13.18)10.54(9.43–11.77)10.56(6.91–16.15)9.53(6.21–14.62)
      Test of interaction showed a statistically significant difference before and after implementation of the new guideline [8].
      Unknown7.35(6.60–8.19)7.37(6.60–8.23)8.33(5.48–12.67)8.62(5.58–13.32)
      Most recent fasting plasma glucose within 6 months prior to index date
      <6.0 mmol/LReferenceReference
      6.0–7.4 mmol/L1.67(1.28–2.17)1.35(1.04–1.76)1.84(0.51–6.69)1.63(0.45–5.95)
      7.5–8.9 mmol/L2.83(2.19–3.65)1.77(1.37–2.29)2.39(0.66–8.66)1.59(0.44–5.83)
      ≥9 mmol/L9.64(7.59–12.23)3.43(2.70–4.37)9.56(3.02–30.23)3.25(1.02–10.38)
      Unknown5.37(4.24–6.79)2.82(2.23–3.57)7.44(2.40–23.07)3.49(1.12–10.90)
      History of Disease
      Cardiovascular disease1.21(1.14–1.29)0.90(0.68–1.19)1.35(1.07–1.70)0.85(0.40–1.77)
      Heart failure1.41(1.29–1.55)1.27(1.16–1.40)0.89(0.61–1.32)0.82(0.55–1.23)
      Test of interaction showed a statistically significant difference before and after implementation of the new guideline [8].
      Ischemic heart disease1.06(1.01–1.11)1.02(0.97–1.07)0.86(0.70–1.06)0.83(0.67–1.03)
      Cerebrovascular disease1.22(1.14–1.30)1.20(0.90–1.59)1.38(1.09–1.74)1.48(0.71–3.08)
      Hypertension0.93(0.90–0.97)0.94(0.91–0.98)0.94(0.83–1.07)0.96(0.84–1.10)
      Gastro-intestinal complications
       Nausea0.83(0.77–0.89)1.02(0.94–1.10)0.81(0.62–1.06)1.20(0.91–1.59)
       Diarrhoea0.83(0.79–0.87)0.95(0.91–1.00)0.67(0.56–0.81)0.87(0.72–1.07)
       Vomiting0.79(0.74–0.84)0.94(0.87–1.00)0.68(0.53–0.88)0.92(0.70–1.20)
       Flatulence0.74(0.62–0.87)0.83(0.70–0.98)0.27(0.10–0.72)0.30(0.11–0.79)
      Test of interaction showed a statistically significant difference before and after implementation of the new guideline [8].
      (Proxies of) osteoporosis
       Osteoporosis1.10(0.99–1.23)1.02(0.90–1.15)1.03(0.71–1.50)0.94(0.62–1.44)
       History of fracture1.06(1.02–1.10)1.08(1.03–1.12)0.94(0.82–1.09)0.98(0.84–1.13)
       Use of bisphosphonates1.00(0.90–1.11)0.91(0.81–1.03)1.15(0.78–1.70)1.14(0.74–1.76)
      Chronic liver disease1.03(0.92–1.16)1.09(0.97–1.23)1.09(0.75–1.60)1.27(0.86–1.87)
      History of a hypoglycaemic event1.46(1.18–1.81)1.25(1.01–1.55)1.70(0.85–3.42)1.47(0.73–3.00)
      Oedema0.77(0.73–0.82)0.82(0.78–0.87)0.55(0.43–0.72)0.65(0.50–0.84)
      Haematuria0.90(0.83–0.97)0.92(0.85–0.99)0.72(0.52–0.98)0.78(0.57–1.09)
      Microalbuminuria0.88(0.80–0.98)0.88(0.79–0.97)0.80(0.49–1.31)0.79(0.48–1.31)
      Bladder cancer1.17(0.94–1.45)1.17(0.93–1.46)2.25(1.17–4.33)1.83(0.93–3.60)
      Pancreas carcinoma2.47(1.43–4.24)1.56(0.91–2.70)2.97(0.74–11.89)2.43(0.60–9.85)
      Pancreatitis1.56(1.37–1.77)1.38(1.21–1.57)3.06(2.22–4.21)2.69(1.94–3.73)
      Test of interaction showed a statistically significant difference before and after implementation of the new guideline [8].
      Schizophrenia/psychosis1.33(1.18–1.50)1.23(1.10–1.39)0.65(0.37–1.14)0.57(0.32–1.01)
      Test of interaction showed a statistically significant difference before and after implementation of the new guideline [8].
      Alzheimer/Dementia1.61(1.45–1.79)0.98(0.87–1.09)1.78(1.26–2.52)1.29(0.88–1.89)
      Cognitive impairment1.27(0.98–1.65)0.93(0.71–1.20)1.00(0.45–2.22)0.61(0.26–1.42)
      Profession as driver0.64(0.52–0.78)0.61(0.50–0.75)0.30(0.11–0.81)0.32(0.12–0.85)
      Abbreviations: BMI, Body Mass Index; eGFR, estimated glomerular filtration rate; HbA1c, glycated haemoglobin.
      Results were corrected for all other determinants.
      No patients with a renal function < 30 ml/min/1.73 m2 received a sulphonylurea after implementation of the updated guideline.
      # Test of interaction showed a statistically significant difference before and after implementation of the new guideline

      Altman DG, Bland JM. Interaction revisited: the difference between two estimates. Bmj. 2003 Jan 25; 326(7382): 219. PubMed PMID: 12543843. Pubmed Central PMCID: 1125071.

      .

      4. Discussion

      In this large, population-based study we identified patient-specific determinants related to early treatment modification in patients with type 2 diabetes before and after the implementation of the updated NICE guideline in the UK. Results show that after introduction of the updated NICE guideline, DPP-4 inhibitors and SGLT-2 inhibitors were more often prescribed as second-line therapy in the first year of treatment. Prescriptions of sulphonylureas and thiazolidediones decreased. Overall, initiation of a second-line therapy was associated with similar determinants before and after implementation of the updated guideline. However, we identified several patient-specific determinants that were significantly different before and after implementation of the updated guideline.
      The changes in treatment patterns, as observed in the present study, are generally consistent with results of studies investigating type 2 diabetes treatment patterns in the UK over time [
      • Overbeek J.A.
      • Heintjes E.M.
      • Prieto-Alhambra D.
      • Blin P.
      • Lassalle R.
      • Hall G.C.
      • et al.
      Type 2 diabetes mellitus treatment patterns across Europe: A population-based multi-database study.
      ,

      Curtis HJ, Dennis JM, Shields BM, Walker AJ, Bacon S, Hattersley AT, et al. Time trends and geographical variation in prescribing of drugs for diabetes in England from 1998 to 2017. Diabetes, Obesity Metabolism 2018 Sep;20(9):2159-68. PubMed PMID: 29732725. Pubmed Central PMCID: 6099452.

      ]. However, in our study sulphonylureas were still preferred as second-line treatment option after implementation of the updated guideline (December 2015). The study by Curtis and colleagues reported DPP-4 inhibitors as preferred second-line therapy since 2016 [

      Curtis HJ, Dennis JM, Shields BM, Walker AJ, Bacon S, Hattersley AT, et al. Time trends and geographical variation in prescribing of drugs for diabetes in England from 1998 to 2017. Diabetes, Obesity Metabolism 2018 Sep;20(9):2159-68. PubMed PMID: 29732725. Pubmed Central PMCID: 6099452.

      ]. This difference might be explained by different methods used, e.g. other definitions of first line therapy and treatment cessation. Of note, the prescribing trends from 2009 to 2014 showed that the prescription of DDP-4 inhibitors already changed proportionally before the introduction of the new guideline. In contrast, the introduction of the new guideline appeared to accelerate the decline in the prescription of sulphonylureas. These results should, however be interpreted with caution, since the observational nature of this study hampers causal inference. For instance, the continuing increase in prescribing SGLT-2 inhibitors after the implementation of the new guideline could also be explained by the publication of landmark studies on SGLT-2 inhibition that appeared in the same period [
      • Zinman B.
      • Wanner C.
      • Lachin J.M.
      • Fitchett D.
      • Bluhmki E.
      • Hantel S.
      • et al.
      Empagliflozin, Cardiovascular Outcomes, and Mortality in Type 2 Diabetes.
      ]. Nonetheless, while sulfonylureas continue to play an important role in the management of type 2 diabetes, it is clear that DPP-4 inhibitors have become a cornerstone of second-line therapy in the UK.
      Several other studies identified clinical features associated with prescribing of second-line glucose-lowering agents. Elevated HbA1c levels, BMI, age, cardiovascular comorbidities, eGFR, duration of diabetes and diabetes related comorbidities are known factors associated with prescribing of second-line therapies [

      Maguire A, Mitchell BD, Ruzafa JC. Antihyperglycaemic treatment patterns, observed glycaemic control and determinants of treatment change among patients with type 2 diabetes in the United Kingdom primary care: a retrospective cohort study. BMC Endocrine Disorders 2014 Aug 27; 14: 73. PubMed PMID: 25163796. Pubmed Central PMCID: 4161267.

      ,
      • Heintjes E.M.
      • Overbeek J.A.
      • Hall G.C.
      • Prieto-Alhambra D.
      • Lapi F.
      • Hammar N.
      • et al.
      Factors associated with type 2 diabetes mellitus treatment choice across four European countries.
      ,
      • Montvida O.
      • Shaw J.
      • Atherton J.J.
      • Stringer F.
      • Paul S.K.
      Long-term trends in antidiabetes drug usage in the U.S.: real-world evidence in patients newly diagnosed with type 2 diabetes.
      ,

      Fujihara K, Igarashi R, Matsunaga S, Matsubayashi Y, Yamada T, Yokoyama H, et al. Comparison of baseline characteristics and clinical course in Japanese patients with type 2 diabetes among whom different types of oral hypoglycemic agents were chosen by diabetes specialists as initial monotherapy (JDDM 42). Medicine 2017 Feb;96(7):e6122. PubMed PMID: 28207538. Pubmed Central PMCID: 5319527.

      ,

      Kanatsuka A, Sato Y, Kawai K, Hirao K, Kobayashi M, Kashiwagi A, et al. Relationship between the efficacy of oral antidiabetic drugs and clinical features in type 2 diabetic patients (JDDM38). J Diabetes Investigation 2016 May;7(3):386-95. PubMed PMID: 27330726. Pubmed Central PMCID: 4847894.

      ,
      • Monami M.
      • Ragghianti B.
      • Zannoni S.
      • Vitale V.
      • Nreu B.
      • Mannucci E.
      Identification of predictors of response to basal insulin and DPP4 inhibitors in patients with type 2 diabetes failing to other therapies.
      ,
      • Grabner M.
      • Peng X.
      • Geremakis C.
      • Bae J.
      Demographic and Clinical Profiles of Type 2 Diabetes Mellitus Patients Initiating Canagliflozin Versus DPP-4 Inhibitors in a Large U.S. Managed Care Population.
      ,
      • Bihan H.
      • Ng W.L.
      • Magliano D.J.
      • Shaw J.E.
      Predictors of efficacy of GLP-1 agonists and DPP-4 inhibitors: A systematic review.
      ]. We also found that a higher fasting plasma glucose, history of a hypoglycaemic event, pancreas carcinoma and pancreatitis were positively associated with a prescription of a second-line glucose-lowering agent both before and after implementation of the updated guideline. Most of the patients with one of the above clinical features are probably in need of more intensive therapy. However, with regard to a history of a hypoglycaemic event, we did not expect to find a positive association as hypoglycaemia is not related to metformin use, yet is for sulphonylureas [

      van Dalem J, Brouwers MC, Stehouwer CD, Krings A, Leufkens HG, Driessen JH, et al. Risk of hypoglycaemia in users of sulphonylureas compared with metformin in relation to renal function and sulphonylurea metabolite group: population based cohort study. Bmj. 2016 Jul 13;354:i3625. PubMed PMID: 27413017. Pubmed Central PMCID: 4948031.

      ]. In contrast, gastro-intestinal complications, oedema, haematuria and a report of being a driver as a profession were negatively associated with prescribing second-line therapies. We can partially explain these results as thiazolidinediones are associated with oedema [

      Mudaliar S, Chang AR, Henry RR. Thiazolidinediones, peripheral edema, and type 2 diabetes: incidence, pathophysiology, and clinical implications. Endocrine practice : official journal of the American College of Endocrinology and the American Association of Clinical Endocrinologists. 2003 Sep-Oct;9(5):406-16. PubMed PMID: 14583425.

      ] and sulphonylureas/insulin are not optimal treatment options for patients with a profession as driver due to the risk of hypoglycaemia. However, we cannot explain why gastro-intestinal complications, a common side effect of metformin, and haematuria were negatively associated with prescribing second-line therapies.
      In our primary analysis examining all second-line therapies, we identified eight determinants with statistically different aHRs. Since the determinants of a treatment modification to all second-line therapies combined are difficult to interpret, e.g. we cannot explain why Alzheimer was positively and cognitive impairment negatively associated with the prescription of second-line therapies, it is of great interest to study the differences in prescribing individual second-line therapies, i.e. sulphonylureas. In the analysis examining the prescribing of sulphonylureas only, we identified five key determinants that had a statistically significant different aHR before and after implementation of the updated guideline. Most of the changes are likely a result of more options to consider as second-line therapy after the 2015 guideline. First, patients with a high BMI are more likely to receive GLP-1 receptor agonists as they are known to have more favourable effects on body weight than sulphonylureas [

      National Institute for Health and Care Excellence. Type 2 diabetes in adults: management Published December 2015, Last updated August 2019.

      ,

      Cheng V, Kashyap SR. Weight considerations in pharmacotherapy for type 2 diabetes. J Obesity. 2011;2011. PubMed PMID: 20885921. Pubmed Central PMCID: 2946585.

      ]. Surprisingly, we did not find a statistically significant different aHR in patients with a BMI ≥ 35 kg/m2. Second, patients with heart failure are more likely to receive SGLT-2 inhibitors, as more evidence has become available regarding the beneficial effects of SGLT-2 inhibitors on heart failure [
      • Zinman B.
      • Wanner C.
      • Lachin J.M.
      • Fitchett D.
      • Bluhmki E.
      • Hantel S.
      • et al.
      Empagliflozin, Cardiovascular Outcomes, and Mortality in Type 2 Diabetes.
      ,
      • Neal B.
      • Perkovic V.
      • Mahaffey K.W.
      • de Zeeuw D.
      • Fulcher G.
      • Erondu N.
      • et al.
      Canagliflozin and Cardiovascular and Renal Events in Type 2 Diabetes.
      ]. Third, metformin users with gastrointestinal side-effects are more likely to switch to second-line therapies [
      • Bytzer P.
      • Talley N.J.
      • Jones M.P.
      • Horowitz M.
      Oral hypoglycaemic drugs and gastrointestinal symptoms in diabetes mellitus.
      ]. A stronger negative association after implementation of the new guideline with a sulphonylurea prescription was probably the result of the availability of more second-line therapies. Similarly, schizophrenia/psychosis was negatively associated with a sulphonylurea prescription after implementation of the updated guideline. Due to the risk of hypoglycaemia and severe consequences when not taken correctly/overdosed, sulphonylureas are not the best option for patients with schizophrenia/psychosis. In contrast, pancreatitis was more strongly associated with a sulphonylurea prescription after implementation of the updated guideline, as compared to the period before. With the on-going debate regarding the risk of pancreatitis in patients using incretin-based agents [

      Saisho Y. Incretin-based therapy and pancreatitis: accumulating evidence and unresolved questions. Annals of translational medicine. 2018 Apr;6(7):131. PubMed PMID: 29955591. Pubmed Central PMCID: 6015951 Company, Boehlinger Ingelheim and AstraZeneca, and research funding from AstraZeneca.

      ], this is not surprising.
      Although we found some statistically significant differences following the implementation of the updated guideline, we did expect to identify more shifts in determinants. For instance, we expected that a history of hypoglycaemia, alcohol use and older age would have been negatively associated with a sulphonylurea prescription after the implementation of the updated guideline as the new therapies are not associated with hypoglycaemia. However, it is possible that we did not find such results as use of metformin, the starting point in our study, is not contraindicated in patients with these determinants.
      A key strength of this study is the availability of detailed information for various potentially relevant determinants in the CPRD. Second, prescriptions for glucose-lowering agents are most of the time issued by a GP. Therefore, our prescribing data can be considered accurate and representative. Third, most of the results of our primary analysis remained consistent in a sensitivity analysis. The loss of power might explain the loss of significance for some determinants, e.g. heart failure. Finally, the separate analysis of data before and after implementation of the updated guideline provides important data regarding the uptake of the guideline.
      Our study also has some potential limitations. With limited data after implementation of the updated guideline, we could only investigate treatment modifications in the first year of therapy. It is possible that implementation of the updated guideline requires more time than investigated. Second, the CPRD contains information on prescription and not dispensed data, which might result in misclassification of exposure. However, it is unlikely that this misclassification would be differential before and after implementation of the updated guideline. Third, the patients’ electronic records are not collected for research purposes. This can result in missing or incomplete information regarding patients’ characteristics, risk factors and lifestyle factors, e.g. profession as a driver, schizophrenia and diet. Fourth, the shift in GP’s prescribing can also be the result of sales marketing by pharmaceutical companies for new medications, the influence of (social) media, policies and/or new clinical evidence. Fifth, there is geographical variation in the prescribing of second-line therapy in the UK [

      Curtis HJ, Dennis JM, Shields BM, Walker AJ, Bacon S, Hattersley AT, et al. Time trends and geographical variation in prescribing of drugs for diabetes in England from 1998 to 2017. Diabetes, Obesity Metabolism 2018 Sep;20(9):2159-68. PubMed PMID: 29732725. Pubmed Central PMCID: 6099452.

      ]. The geographical shift in patients registered in the database (shown in Table 1) may, therefore, have affected our results. Sixth, as previously mentioned, due to the design of our study we could only investigate associations and not causality. Seventh, although the CPRD contains information of a wide range of determinants, the influence of possible important determinants, as well as patient and provider preferences, could not be investigated as this information is not captured in the CPRD GOLD database. Moreover, our analysis did not correct for the potential influence of patients’ current non-diabetes medication. Finally, all treatment modifications (switch and intensification) were included in our primary analyses. We showed that treatment modifications were mainly the result of a treatment intensification.
      In conclusion, our results show increased diversity in the prescribing of second-line glucose-lowering agents after the implementation of the updated NICE guideline. While many of the patient characteristics associated with prescribing a second-line therapy remained stable before and after implementation of the new guideline, we recognise that more time may be needed to observe the optimal implementation of the updated guideline. We believe our results suggest that a first step towards individually tailoring prescribing to patient-specific characteristics has already been made. However, not all determinants that are important to consider when prescribing second-line glucose-lowering agents, e.g. older age and a history of hypoglycaemia, were of influence on GP’s prescribing.

      5. Contribution Statement

      JvD, initiated the study, did the literature review and wrote the first draft of the paper. JD analysed the data. JvD, MB, FV, AMB and JD and were responsible for the study concept and design and participated in the interpretation of the data. All authors had full access to all of the data in the study, critically revised the paper for intellectual content, approved the final version to be published and can take responsibility for the integrity of the data and accuracy of the data analyses.

      6. Funding

      This research did not receive any specific grant from funding agencies in the public, commercial or not-for-profit sectors.

      7. Scientific approval

      This protocol was approved by the Interdisciplinary Scientific Advisory Committee (ISAC) for Medicines and. Healthcare products Regulatory Agency. (MHRA) database research, protocol no. 18_126R.

      Declaration of Competing Interest

      The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

      Appendix A. Supplementary material

      The following are the Supplementary data to this article:

      References

      1. McCulloch DK. Metformin in the treatment of adults with type 2 diabetes mellitus. UpToDate; 2013 20 December 2013.

      2. National Institute for Health and Care Excellence. Type 2 diabetes: The management of type 2 diabetes. Published May 2009.

      3. National Institute for Health and Care Excellence. Type 2 diabetes in adults: management Published December 2015, Last updated August 2019.

      4. Maguire A, Mitchell BD, Ruzafa JC. Antihyperglycaemic treatment patterns, observed glycaemic control and determinants of treatment change among patients with type 2 diabetes in the United Kingdom primary care: a retrospective cohort study. BMC Endocrine Disorders 2014 Aug 27; 14: 73. PubMed PMID: 25163796. Pubmed Central PMCID: 4161267.

        • Heintjes E.M.
        • Overbeek J.A.
        • Hall G.C.
        • Prieto-Alhambra D.
        • Lapi F.
        • Hammar N.
        • et al.
        Factors associated with type 2 diabetes mellitus treatment choice across four European countries.
        Clin Ther. 2017; (PubMed PMID: 29108837)
      5. Herrett E, Gallagher AM, Bhaskaran K, Forbes H, Mathur R, van Staa T, et al. Data Resource Profile: Clinical Practice Research Datalink (CPRD). Int J Epidemiol 2015 Jun; 44(3): 827-3PubMed PMID: 26050254. Pubmed Central PMCID: 4521131.

      6. Mathur R, Bhaskaran K, Chaturvedi N, Leon DA, vanStaa T, Grundy E, et al. Completeness and usability of ethnicity data in UK-based primary care and hospital databases. J Public Health 2014 Dec; 36(4) :684-92. PubMed PMID: 24323951. Pubmed Central PMCID: 4245896.

      7. Altman DG, Bland JM. Interaction revisited: the difference between two estimates. Bmj. 2003 Jan 25; 326(7382): 219. PubMed PMID: 12543843. Pubmed Central PMCID: 1125071.

        • Overbeek J.A.
        • Heintjes E.M.
        • Prieto-Alhambra D.
        • Blin P.
        • Lassalle R.
        • Hall G.C.
        • et al.
        Type 2 diabetes mellitus treatment patterns across Europe: A population-based multi-database study.
        Clin Ther. 2017 Apr; 39 (PubMed PMID: 28342563): 759-770
      8. Curtis HJ, Dennis JM, Shields BM, Walker AJ, Bacon S, Hattersley AT, et al. Time trends and geographical variation in prescribing of drugs for diabetes in England from 1998 to 2017. Diabetes, Obesity Metabolism 2018 Sep;20(9):2159-68. PubMed PMID: 29732725. Pubmed Central PMCID: 6099452.

        • Zinman B.
        • Wanner C.
        • Lachin J.M.
        • Fitchett D.
        • Bluhmki E.
        • Hantel S.
        • et al.
        Empagliflozin, Cardiovascular Outcomes, and Mortality in Type 2 Diabetes.
        New England J Med. 2015 Nov 26; 373 (PubMed PMID: 26378978): 2117-2128
        • Montvida O.
        • Shaw J.
        • Atherton J.J.
        • Stringer F.
        • Paul S.K.
        Long-term trends in antidiabetes drug usage in the U.S.: real-world evidence in patients newly diagnosed with type 2 diabetes.
        Diabetes Care. 2018 Jan; 41 (PubMed PMID: 29109299): 69-78
      9. Fujihara K, Igarashi R, Matsunaga S, Matsubayashi Y, Yamada T, Yokoyama H, et al. Comparison of baseline characteristics and clinical course in Japanese patients with type 2 diabetes among whom different types of oral hypoglycemic agents were chosen by diabetes specialists as initial monotherapy (JDDM 42). Medicine 2017 Feb;96(7):e6122. PubMed PMID: 28207538. Pubmed Central PMCID: 5319527.

      10. Kanatsuka A, Sato Y, Kawai K, Hirao K, Kobayashi M, Kashiwagi A, et al. Relationship between the efficacy of oral antidiabetic drugs and clinical features in type 2 diabetic patients (JDDM38). J Diabetes Investigation 2016 May;7(3):386-95. PubMed PMID: 27330726. Pubmed Central PMCID: 4847894.

        • Monami M.
        • Ragghianti B.
        • Zannoni S.
        • Vitale V.
        • Nreu B.
        • Mannucci E.
        Identification of predictors of response to basal insulin and DPP4 inhibitors in patients with type 2 diabetes failing to other therapies.
        Acta Diabetol. 2016 Feb; 53 (PubMed PMID: 25805649): 35-40
        • Grabner M.
        • Peng X.
        • Geremakis C.
        • Bae J.
        Demographic and Clinical Profiles of Type 2 Diabetes Mellitus Patients Initiating Canagliflozin Versus DPP-4 Inhibitors in a Large U.S. Managed Care Population.
        J Managed Care Specialty Pharm. 2015 Dec; 21 (PubMed PMID: 26679969): 1204-1212
        • Bihan H.
        • Ng W.L.
        • Magliano D.J.
        • Shaw J.E.
        Predictors of efficacy of GLP-1 agonists and DPP-4 inhibitors: A systematic review.
        Diabetes Res Clin Pract. 2016 Nov; 121 (PubMed PMID: 27622682): 27-34
      11. van Dalem J, Brouwers MC, Stehouwer CD, Krings A, Leufkens HG, Driessen JH, et al. Risk of hypoglycaemia in users of sulphonylureas compared with metformin in relation to renal function and sulphonylurea metabolite group: population based cohort study. Bmj. 2016 Jul 13;354:i3625. PubMed PMID: 27413017. Pubmed Central PMCID: 4948031.

      12. Mudaliar S, Chang AR, Henry RR. Thiazolidinediones, peripheral edema, and type 2 diabetes: incidence, pathophysiology, and clinical implications. Endocrine practice : official journal of the American College of Endocrinology and the American Association of Clinical Endocrinologists. 2003 Sep-Oct;9(5):406-16. PubMed PMID: 14583425.

      13. Cheng V, Kashyap SR. Weight considerations in pharmacotherapy for type 2 diabetes. J Obesity. 2011;2011. PubMed PMID: 20885921. Pubmed Central PMCID: 2946585.

        • Neal B.
        • Perkovic V.
        • Mahaffey K.W.
        • de Zeeuw D.
        • Fulcher G.
        • Erondu N.
        • et al.
        Canagliflozin and Cardiovascular and Renal Events in Type 2 Diabetes.
        New England J Med. 2017 Aug 17; 377 (PubMed PMID: 28605608): 644-657
        • Bytzer P.
        • Talley N.J.
        • Jones M.P.
        • Horowitz M.
        Oral hypoglycaemic drugs and gastrointestinal symptoms in diabetes mellitus.
        Aliment Pharmacol Ther. 2001 Jan; 15 (PubMed PMID: 11136287): 137-142
      14. Saisho Y. Incretin-based therapy and pancreatitis: accumulating evidence and unresolved questions. Annals of translational medicine. 2018 Apr;6(7):131. PubMed PMID: 29955591. Pubmed Central PMCID: 6015951 Company, Boehlinger Ingelheim and AstraZeneca, and research funding from AstraZeneca.