[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Figure 1.  Flowchart of Participant Exclusions and Eligibility
Flowchart of Participant Exclusions and Eligibility

Data were extracted from the electronic health record and administrative databases using virtual data warehouse databases at each study site. Data from January 1, 2005, through December 31, 2013, are included. ESRD indicates end-stage renal disease; HbA1c, hemoglobin A1c.

aIdentified based on diagnoses, procedure codes, or laboratory data in the 2 years before the index date, unless otherwise indicated.

bIndicates palliative care, hospice care, or stage IV cancer.

cDefined as estimated glomerular filtration rate of less than 15 mL/1.73 m2/min, dialysis, or transplant.

dIdentified using pharmacy codes.

Figure 2.  Crude Survival Curves for Mortality, Myocardial Infarction (MI), and Mortality Due to Cardiovascular Disease (CVD)
Crude Survival Curves for Mortality, Myocardial Infarction (MI), and Mortality Due to Cardiovascular Disease (CVD)

Unadjusted estimates of the survival curves by insulin therapy regimen are shown for 3 of the 5 outcomes studied. The P value on each plot is the 2-tailed P value of the statistical test that the area between the 2 survival curves is null (ie, the sum of the risk differences at each quarter is equal to 0). AI group indicates patients continuously receiving analogue insulin with or without human insulin; HI group, patients continuously receiving human insulin only.

Figure 3.  Adjusted Survival Curves for Mortality, Myocardial Infarction (MI), and Mortality Due to Cardiovascular Disease (CVD) Using Inverse Probability Weight Estimation to Fit a Saturated Marginal Structural Model
Adjusted Survival Curves for Mortality, Myocardial Infarction (MI), and Mortality Due to Cardiovascular Disease (CVD) Using Inverse Probability Weight Estimation to Fit a Saturated Marginal Structural Model

Adjusted estimates of the survival curves by insulin therapy regimen are shown for 3 of the 5 outcomes studied. The P value on each plot is the 2-tailed P value of the statistical test that the area between the 2 survival curves is null (ie, the sum of the risk differences at each quarter is equal to 0). AI group indicates patients continuously receiving analogue insulin with or without human insulin; HI group, patients continuously receiving human insulin only.

Table 1.  Baseline Clinical and Demographic Characteristics of Participantsa
Baseline Clinical and Demographic Characteristics of Participantsa
Table 2.  Event Rates and Reasons for End of Analytic Follow-up
Event Rates and Reasons for End of Analytic Follow-up
Table 3.  Primary Analysis Resultsa
Primary Analysis Resultsa
Supplement.

eMethods 1. Cohort Construction

eMethods 2. Data Structure and Notation

eMethods 3. Causal Estimands and Inverse Probability Estimator

eMethods 4. Denominator of the Inverse Probability Weights

eMethods 5. Standard Propensity Score Estimation With 3 Covariate Adjustment Sets

eMethods 6. Data-Adaptive Propensity Score Estimation

eMethods 7. Results

eTable 1. Sources of Data and Codes Used to Ascertain Major Cardiovascular Events and Mortality

eTable 2. Part I of II: Brief Description of All Attributes (L) in the Covariate Adjustment Sets

eTable 3. Part II of II: Brief Description of All Attributes (L) in the Covariate Adjustment Sets

eTable 4. Part I of II: List of Covariates Considered in the Various Analyses and Whether They Are Assumed to Impact Exposure Decisions, Censoring Events, or Outcomes

eTable 5. Part II of II: List of Covariates Considered in the Various Analyses and Whether They Are Assumed to Impact Exposure Decisions, Censoring Events, or Outcomes

eTable 6. Cutoffs Used to Discretize Continuous Covariates

eTable 7. PS Estimation Approach 1 in AMI Primary Analysis

eTable 8. Part I of II: PS Estimation Approach 2 in AMI Primary Analysis

eTable 9. Part II of II: PS Estimation Approach 2 in AMI Primary Analysis

eTable 10. Part I of II: PS Estimation Approach 3 in AMI Primary Analysis

eTable 11. Part II of II: PS Estimation Approach 3 in AMI Primary Analysis

eTable 12. PS Estimation Approach 4 in AMI Primary Analysis

eTable 13. Distribution of Type of Initial Insulin Therapy for Patients in the Main Cohort by Site and Year of Cohort Entry

eTable 14. Distribution of Follow-up Time (Expressed in 90-Day Intervals) for Patients Continuously Exposed to Analogue-Containing Insulin Therapy in the Primary AMI Analyses (All Sites Combined)

eTable 15. Distribution of Follow-up Time (Expressed in 90-Day Intervals) for Patients Continuously Exposed to Human-Only Insulin Therapy in the Primary AMI Analyses (All Sites Combined)

eTable 16. Summary Statistics of the Distribution of Follow-up Time (Expressed in 90-Day Intervals) by Exposure Regimen in the Primary AMI Analyses (All Sites Combined)

eTable 17. Distribution of Follow-up Time (Expressed in 90-Day Intervals) for Patients Continuously Exposed to Analogue-Containing Insulin Therapy in the Primary CHF Analyses (All Sites Combined)

eTable 18. Distribution of Follow-up Time (Expressed in 90-Day Intervals) for Patients Continuously Exposed to Human-Only Insulin Therapy in the Primary CHF Analyses (All Sites Combined)

eTable 19. Summary Statistics of the Distribution of Follow-up Time (Expressed in 90-Day Intervals) by Exposure Regimen in the Primary CHF Analyses (All Sites Combined)

eTable 20. Distribution of Follow-up Time (Expressed in 90-Day Intervals) for Patients Continuously Exposed to Analogue-Containing Insulin Therapy in the Primary CVA Analyses (All Sites Combined)

eTable 21. Distribution of Follow-up Time (Expressed in 90-Day Intervals) for Patients Continuously Exposed to Human-Only Insulin Therapy in the Primary CVA Analyses (All Sites Combined)

eTable 22. Summary Statistics of the Distribution of Follow-up Time (Expressed in 90-Day Intervals) by Exposure Regimen in the Primary CVA Analyses (All Sites Combined)

eTable 23. Distribution of Follow-up Time (Expressed in 90-Day Intervals) for Patients Continuously Exposed to Analogue-Containing Insulin Therapy in the Primary CVD Mortality Analyses (All Sites Combined)

eTable 24. Distribution of Follow-up Time (Expressed in 90-Day Intervals) for Patients Continuously Exposed to Human-Only Insulin Therapy in the Primary CVD Mortality Analyses (All Sites Combined)

eTable 25. Summary Statistics of the Distribution of Follow-up Time (Expressed in 90-Day Intervals) by Exposure Regimen in the Primary CVD Mortality Analyses (All Sites Combined)

eTable 26. Distribution of Follow-up Time (Expressed in 90-Day Intervals) for Patients Continuously Exposed to Analogue-Containing Insulin Therapy in the Primary All-Cause Mortality Analyses (All Sites Combined)

eTable 27. Distribution of Follow-up Time (Expressed in 90-Day Intervals) for Patients Continuously Exposed to Human-Only Insulin Therapy in the Primary All-Cause Mortality Analyses (All Sites Combined)

eTable 28. Summary Statistics of the Distribution of Follow-up Time (Expressed in 90-Day Intervals) by Exposure Regimen in the Primary All-Cause Mortality Analyses (All Sites Combined)

eTable 29. AMI Results

eTable 30. CHF Results

eTable 31. CVA Results

eTable 32. CVD Mortality Results

eTable 33. All-Cause Mortality Results

eTable 34. Summary Statistics of the Inverse Probability Weights (IPW) Involved in the AMI Analyses

eTable 35. Summary Statistics of the Inverse Probability Weights (IPW) Involved in the CHF Analyses

eTable 36. Summary Statistics of the Inverse Probability Weights (IPW) Involved in the CVA Analyses

eTable 37. Summary Statistics of the Inverse Probability Weights (IPW) Involved in the CVD Mortality Analyses

eTable 38. Summary Statistics of the Inverse Probability Weights (IPW) Involved in the All-Cause Mortality Analyses

eFigure 1. Survival Curve Estimates for AMI (Primary and Sensitivity Analyses)

eFigure 2. Survival Curve Estimates for CHF (Primary and Sensitivity Analyses)

eFigure 3. Survival Curve Estimates for CVA (Primary and Sensitivity Analyses)

eFigure 4. Survival Curve Estimates for CVD Mortality (Primary and Sensitivity Analyses)

eFigure 5. Survival Curve Estimates for All-Cause Mortality (Primary and Sensitivity Analyses)

eReferences.

1.
American Diabetes Association.  Introduction: Standards of Medical Care in Diabetes–2019.  Diabetes Care. 2019;42(suppl 1):S1-S2. doi:10.2337/dc19-Sint01PubMedGoogle ScholarCrossref
2.
Davies  MJ, D’Alessio  DA, Fradkin  J,  et al.  Management of hyperglycemia in type 2 diabetes, 2018: a consensus report by the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD).  Diabetes Care. 2018;41(12):2669-2701. doi:10.2337/dci18-0033PubMedGoogle ScholarCrossref
3.
Cefalu  WT, Kaul  S, Gerstein  HC,  et al.  Cardiovascular outcomes trials in type 2 diabetes: where do we go from here? reflections from a Diabetes Care editors’ expert forum.  Diabetes Care. 2018;41(1):14-31. doi:10.2337/dci17-0057PubMedGoogle ScholarCrossref
4.
Marso  SP, Daniels  GH, Brown-Frandsen  K,  et al; LEADER Steering Committee; LEADER Trial Investigators.  Liraglutide and cardiovascular outcomes in type 2 diabetes.  N Engl J Med. 2016;375(4):311-322. doi:10.1056/NEJMoa1603827PubMedGoogle ScholarCrossref
5.
Marso  SP, Bain  SC, Consoli  A,  et al; SUSTAIN-6 Investigators.  Semaglutide and cardiovascular outcomes in patients with type 2 diabetes.  N Engl J Med. 2016;375(19):1834-1844. doi:10.1056/NEJMoa1607141PubMedGoogle ScholarCrossref
6.
Zinman  B, Wanner  C, Lachin  JM,  et al; EMPA-REG OUTCOME Investigators.  Empagliflozin, cardiovascular outcomes, and mortality in type 2 diabetes.  N Engl J Med. 2015;373(22):2117-2128. doi:10.1056/NEJMoa1504720PubMedGoogle ScholarCrossref
7.
Neal  B, Perkovic  V, Mahaffey  KW,  et al; CANVAS Program Collaborative Group.  Canagliflozin and cardiovascular and renal events in type 2 diabetes.  N Engl J Med. 2017;377(7):644-657. doi:10.1056/NEJMoa1611925PubMedGoogle ScholarCrossref
8.
Rosenstock  J, Kahn  SE, Johansen  OE,  et al; CAROLINA Investigators.  Effect of linagliptin vs glimepiride on major adverse cardiovascular outcomes in patients with type 2 diabetes: the CAROLINA randomized clinical trial.  JAMA. 2019;322(12):1155-1166. doi:10.1001/jama.2019.13772PubMedGoogle ScholarCrossref
9.
Roumie  CL, Chipman  J, Min  JY,  et al.  Association of treatment with metformin vs sulfonylurea with major adverse cardiovascular events among patients with diabetes and reduced kidney function.  JAMA. 2019;322(12):1167-1177. doi:10.1001/jama.2019.13206PubMedGoogle ScholarCrossref
10.
Pathak  RD, Schroeder  EB, Seaquist  ER,  et al; SUPREME-DM Study Group.  Severe hypoglycemia requiring medical intervention in a large cohort of adults with diabetes receiving care in US integrated health care delivery systems: 2005-2011.  Diabetes Care. 2016;39(3):363-370. doi:10.2337/dc15-0858PubMedGoogle ScholarCrossref
11.
Lipska  KJ, Parker  MM, Moffet  HH, Huang  ES, Karter  AJ.  Association of initiation of basal insulin analogs vs neutral protamine hagedorn insulin with hypoglycemia-related emergency department visits or hospital admissions and with glycemic control in patients with type 2 diabetes.  JAMA. 2018;320(1):53-62. doi:10.1001/jama.2018.7993PubMedGoogle ScholarCrossref
12.
Nathan  DM, Genuth  S, Lachin  J,  et al; Diabetes Control and Complications Trial Research Group.  The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus.  N Engl J Med. 1993;329(14):977-986. doi:10.1056/NEJM199309303291401PubMedGoogle ScholarCrossref
13.
UK Prospective Diabetes Study (UKPDS) Group.  Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33).  Lancet. 1998;352(9131):837-853. doi:10.1016/S0140-6736(98)07019-6PubMedGoogle ScholarCrossref
14.
Gerstein  HC, Miller  ME, Byington  RP,  et al; Action to Control Cardiovascular Risk in Diabetes Study Group.  Effects of intensive glucose lowering in type 2 diabetes.  N Engl J Med. 2008;358(24):2545-2559. doi:10.1056/NEJMoa0802743PubMedGoogle ScholarCrossref
15.
Patel  A, MacMahon  S, Chalmers  J,  et al; ADVANCE Collaborative Group.  Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes.  N Engl J Med. 2008;358(24):2560-2572. doi:10.1056/NEJMoa0802987PubMedGoogle ScholarCrossref
16.
Gerstein  HC, Miller  ME, Genuth  S,  et al; ACCORD Study Group.  Long-term effects of intensive glucose lowering on cardiovascular outcomes.  N Engl J Med. 2011;364(9):818-828. doi:10.1056/NEJMoa1006524PubMedGoogle ScholarCrossref
17.
Holman  RR, Paul  SK, Bethel  MA, Matthews  DR, Neil  HA.  10-year follow-up of intensive glucose control in type 2 diabetes.  N Engl J Med. 2008;359(15):1577-1589. doi:10.1056/NEJMoa0806470PubMedGoogle ScholarCrossref
18.
Duckworth  W, Abraira  C, Moritz  T,  et al; VADT Investigators.  Glucose control and vascular complications in veterans with type 2 diabetes.  N Engl J Med. 2009;360(2):129-139. doi:10.1056/NEJMoa0808431PubMedGoogle ScholarCrossref
19.
Lipska  KJ, Hirsch  IB, Riddle  MC.  Human insulin for type 2 diabetes: an effective, less-expensive option.  JAMA. 2017;318(1):23-24. doi:10.1001/jama.2017.6939PubMedGoogle ScholarCrossref
20.
Fullerton  B, Siebenhofer  A, Jeitler  K,  et al.  Short-acting insulin analogues versus regular human insulin for adult, non-pregnant persons with type 2 diabetes mellitus.  Cochrane Database Syst Rev. 2018;12:CD013228. doi:10.1002/14651858.CD013228PubMedGoogle Scholar
21.
Roglic  G, Norris  SL.  Medicines for treatment intensification in type 2 diabetes and type of insulin in type 1 and type 2 diabetes in low-resource settings: synopsis of the World Health Organization guidelines on second- and third-line medicines and type of insulin for the control of blood glucose levels in nonpregnant adults with diabetes mellitus.  Ann Intern Med. 2018;169(6):394-397. doi:10.7326/M18-1149PubMedGoogle ScholarCrossref
22.
World Health Organization (WHO).  Guidelines on Second- and Third-Line Medicines and Type of Insulin for the Control of Blood Glucose Levels in Non-pregnant Adults With Diabetes Mellitus. Geneva, Switzerland: World Health Organization; 2018.
23.
Robins  JM. Marginal structural models. In: 1997 Proceedings of the Section on Bayesian Statistical Science; Alexandria, VA: American Statistical Association, 1998;1-10. https://cdn1.sph.harvard.edu/wp-content/uploads/sites/343/2013/03/msm-web.pdf. Accessed December 9, 2019.
24.
Hernán  MA, Robins  JM.  Using big data to emulate a target trial when a randomized trial is not available.  Am J Epidemiol. 2016;183(8):758-764. doi:10.1093/aje/kwv254PubMedGoogle ScholarCrossref
25.
Hernán  MA, Hernández-Díaz  S.  Beyond the intention-to-treat in comparative effectiveness research.  Clin Trials. 2012;9(1):48-55. doi:10.1177/1740774511420743PubMedGoogle ScholarCrossref
26.
Hernán  MA, Robins  JM.  Per-protocol analyses of pragmatic trials.  N Engl J Med. 2017;377(14):1391-1398. doi:10.1056/NEJMsm1605385PubMedGoogle ScholarCrossref
27.
Hernán  MA.  The hazards of hazard ratios.  Epidemiology. 2010;21(1):13-15. doi:10.1097/EDE.0b013e3181c1ea43PubMedGoogle ScholarCrossref
28.
Pirracchio  R, Petersen  ML, van der Laan  M.  Improving propensity score estimators’ robustness to model misspecification using super learner.  Am J Epidemiol. 2015;181(2):108-119. doi:10.1093/aje/kwu253PubMedGoogle ScholarCrossref
29.
Neugebauer  R, Schmittdiel  JA, van der Laan  MJ.  A case study of the impact of data-adaptive versus model-based estimation of the propensity scores on causal inferences from three inverse probability weighting estimators.  Int J Biostat. 2016;12(1):131-155. doi:10.1515/ijb-2015-0028PubMedGoogle ScholarCrossref
30.
Neugebauer  R, Fireman  B, Roy  JA, Raebel  MA, Nichols  GA, O’Connor  PJ.  Super learning to hedge against incorrect inference from arbitrary parametric assumptions in marginal structural modeling.  J Clin Epidemiol. 2013;66(8)(suppl):S99-S109. doi:10.1016/j.jclinepi.2013.01.016PubMedGoogle ScholarCrossref
31.
Steiner  JF, Paolino  AR, Thompson  EE, Larson  EB.  Sustaining research networks: the twenty-year experience of the HMO Research Network.  EGEMS (Wash DC). 2014;2(2):1067. doi:10.13063/2327-9214.1067PubMedGoogle Scholar
32.
Nichols  GA, Desai  J, Elston Lafata  J,  et al; SUPREME-DM Study Group.  Construction of a multisite DataLink using electronic health records for the identification, surveillance, prevention, and management of diabetes mellitus: the SUPREME-DM project.  Prev Chronic Dis. 2012;9:E110. doi:10.5888/pcd9.110311PubMedGoogle Scholar
33.
Kaiser Permanente Division of Research, Northern California; Leong  TK, Tabada  GH, Yang  JM, Zhu  Z, Neugebauer  RS. MSMstructure SAS macro version 1.0.4. https://divisionofresearch.kaiserpermanente.org/projects/biostatistics/causalinferencesoftware. Published March 16, 2017. Accessed November 22, 2019.
34.
Hernán  MA, Hernández-Díaz  S, Robins  JM.  A structural approach to selection bias.  Epidemiology. 2004;15(5):615-625. doi:10.1097/01.ede.0000135174.63482.43PubMedGoogle ScholarCrossref
35.
Robins  JM.  Association, causation, and marginal structural models.  Synthese. 1999;121(1-2):151-179. doi:10.1023/A:1005285815569Google ScholarCrossref
36.
Hernán  MA, Brumback  B, Robins  JM.  Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men.  Epidemiology. 2000;11(5):561-570. doi:10.1097/00001648-200009000-00012PubMedGoogle ScholarCrossref
37.
Neugebauer  R, Schmittdiel  JA, van der Laan  MJ.  Targeted learning in real-world comparative effectiveness research with time-varying interventions.  Stat Med. 2014;33(14):2480-2520. doi:10.1002/sim.6099PubMedGoogle ScholarCrossref
38.
Sauer  BC, Brookhart  MA, Roy  J, VanderWeele  T.  A review of covariate selection for non-experimental comparative effectiveness research.  Pharmacoepidemiol Drug Saf. 2013;22(11):1139-1145. doi:10.1002/pds.3506PubMedGoogle ScholarCrossref
39.
VanderWeele  TJ, Shpitser  I.  A new criterion for confounder selection.  Biometrics. 2011;67(4):1406-1413. doi:10.1111/j.1541-0420.2011.01619.xPubMedGoogle ScholarCrossref
40.
van der Laan  MJ, Polley  EC, Hubbard  AE.  Super learner.  Stat Appl Genet Mol Biol. 2007;6(Article 25):e25. doi:10.2202/1544-6115.1309PubMedGoogle Scholar
41.
Joffe  MM.  Exhaustion, automation, theory, and confounding.  Epidemiology. 2009;20(4):523-524. doi:10.1097/EDE.0b013e3181a82501PubMedGoogle ScholarCrossref
42.
Neugebauer  R, Schmittdiel  JA, Zhu  Z, Rassen  JA, Seeger  JD, Schneeweiss  S.  High-dimensional propensity score algorithm in comparative effectiveness research with time-varying interventions.  Stat Med. 2015;34(5):753-781. doi:10.1002/sim.6377PubMedGoogle ScholarCrossref
43.
Cole  SR, Hernán  MA.  Constructing inverse probability weights for marginal structural models.  Am J Epidemiol. 2008;168(6):656-664. doi:10.1093/aje/kwn164PubMedGoogle ScholarCrossref
44.
Petersen  ML, Porter  KE, Gruber  S, Wang  Y, van der Laan  MJ.  Diagnosing and responding to violations in the positivity assumption.  Stat Methods Med Res. 2012;21(1):31-54. doi:10.1177/0962280210386207PubMedGoogle ScholarCrossref
45.
Price  HI, Agnew  MD, Gamble  JM.  Comparative cardiovascular morbidity and mortality in patients taking different insulin regimens for type 2 diabetes: a systematic review.  BMJ Open. 2015;5(3):e006341. doi:10.1136/bmjopen-2014-006341PubMedGoogle Scholar
46.
Strandberg  AY, Hoti  FJ, Strandberg  TE, Christopher  S, Haukka  J, Korhonen  P.  All-cause and cause-specific mortality among users of basal insulins NPH, detemir, and glargine.  PLoS One. 2016;11(3):e0151910. doi:10.1371/journal.pone.0151910PubMedGoogle Scholar
47.
Rathmann  W, Schloot  NC, Kostev  K, Reaney  M, Zagar  AJ, Haupt  A.  Macro- and microvascular outcomes in patients with type 2 diabetes treated with rapid-acting insulin analogues or human regular insulin: a retrospective database analysis.  Exp Clin Endocrinol Diabetes. 2014;122(2):92-99. doi:10.1055/s-0033-1363684PubMedGoogle ScholarCrossref
48.
Horvath  K, Jeitler  K, Berghold  A,  et al.  Long-acting insulin analogues versus NPH insulin (human isophane insulin) for type 2 diabetes mellitus.  Cochrane Database Syst Rev. 2007;(2):CD005613. doi:10.1002/14651858.CD005613.pub3PubMedGoogle Scholar
49.
Cammarota  S, Falconio  LM, Bruzzese  D,  et al.  Lower rate of cardiovascular complications in patients on bolus insulin analogues: a retrospective population-based cohort study.  PLoS One. 2013;8(11):e79762. doi:10.1371/journal.pone.0079762PubMedGoogle Scholar
50.
Glymour  MM, Weuve  J, Chen  JT.  Methodological challenges in causal research on racial and ethnic patterns of cognitive trajectories: measurement, selection, and bias.  Neuropsychol Rev. 2008;18(3):194-213. doi:10.1007/s11065-008-9066-xPubMedGoogle ScholarCrossref
51.
Brookhart  MA, Schneeweiss  S, Rothman  KJ, Glynn  RJ, Avorn  J, Stürmer  T.  Variable selection for propensity score models.  Am J Epidemiol. 2006;163(12):1149-1156. doi:10.1093/aje/kwj149PubMedGoogle ScholarCrossref
52.
Schneeweiss  S, Rassen  JA, Glynn  RJ, Avorn  J, Mogun  H, Brookhart  MA.  High-dimensional propensity score adjustment in studies of treatment effects using health care claims data.  Epidemiology. 2009;20(4):512-522. doi:10.1097/EDE.0b013e3181a663ccPubMedGoogle ScholarCrossref
53.
Luo  J, Khan  NF, Manetti  T,  et al.  Implementation of a health plan program for switching from analogue to human insulin and glycemic control among Medicare beneficiaries with type 2 diabetes.  JAMA. 2019;321(4):374-384. doi:10.1001/jama.2018.21364PubMedGoogle ScholarCrossref
54.
Lipska  KJ.  Insulin analogues for type 2 diabetes.  JAMA. 2019;321(4):350-351. doi:10.1001/jama.2018.21356PubMedGoogle ScholarCrossref
55.
Lipska  KJ, Yao  X, Herrin  J,  et al.  Trends in drug utilization, glycemic control, and rates of severe hypoglycemia, 2006-2013.  Diabetes Care. 2017;40(4):468-475. doi:10.2337/dc16-0985PubMedGoogle ScholarCrossref
56.
Madenidou  AV, Paschos  P, Karagiannis  T,  et al.  Comparative benefits and harms of basal insulin analogues for type 2 diabetes: a systematic review and network meta-analysis.  Ann Intern Med. 2018;169(3):165-174. doi:10.7326/M18-0443PubMedGoogle ScholarCrossref
57.
Insulins for type 2 diabetes. The Medical Letter on Drugs and Therapeutics, issue 1571. https://secure.medicalletter.org/article-share?a=1571a&p=tml&title=Insulins%20for%20Type%202%20Diabetes&cannotaccesstitle=1. Published May 6, 2019. Accessed December 3, 2019.
58.
Marso  SP, McGuire  DK, Zinman  B,  et al; DEVOTE Study Group.  Efficacy and safety of degludec versus glargine in type 2 diabetes.  N Engl J Med. 2017;377(8):723-732. doi:10.1056/NEJMoa1615692PubMedGoogle ScholarCrossref
59.
Freemantle  N, Chou  E, Frois  C,  et al.  Safety and efficacy of insulin glargine 300 μ/mL compared with other basal insulin therapies in patients with type 2 diabetes mellitus: a network meta-analysis.  BMJ Open. 2016;6(2):e009421. doi:10.1136/bmjopen-2015-009421PubMedGoogle Scholar
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    Original Investigation
    Diabetes and Endocrinology
    January 24, 2020

    Comparison of Mortality and Major Cardiovascular Events Among Adults With Type 2 Diabetes Using Human vs Analogue Insulins

    Author Affiliations
    • 1Division of Research, Kaiser Permanente Northern California, Oakland
    • 2Kaiser Permanente Colorado Institute for Health Research, Aurora
    • 3Department of Research and Evaluation, Kaiser Permanente Southern California, Pasadena
    • 4HealthPartners Institute, Minneapolis, Minnesota
    • 5Rocky Mountain Regional Veterans Affairs and University of Colorado (Anschutz) Medical Center, Denver
    • 6HealthPartners Center for Chronic Care Innovation, Minneapolis, Minnesota
    JAMA Netw Open. 2020;3(1):e1918554. doi:10.1001/jamanetworkopen.2019.18554
    Key Points español 中文 (chinese)

    Question  Are there significant differences in cardiovascular outcomes in adults with type 2 diabetes who use human insulin compared with those who use analogue insulin?

    Findings  In this cohort study of 127 600 adults with type 2 diabetes, no differences were found in overall mortality, cardiovascular mortality, myocardial infarction, stroke, and hospitalization for congestive heart failure.

    Meaning  In this study, insulin-naive adults with type 2 diabetes treated with human vs analogue insulin had similar rates of major cardiovascular events, mortality due to cardiovascular disease, and overall mortality.

    Abstract

    Importance  The comparative cardiovascular safety of analogue and human insulins in adults with type 2 diabetes who initiate insulin therapy in usual care settings has not been carefully evaluated using machine learning and other rigorous analytic methods.

    Objective  To examine the association of analogue vs human insulin use with mortality and major cardiovascular events.

    Design, Setting, and Participants  This retrospective cohort study included 127 600 adults aged 21 to 89 years with type 2 diabetes at 4 health care delivery systems who initiated insulin therapy from January 1, 2000, through December 31, 2013. Machine learning and rigorous inference methods with time-varying exposures were used to evaluate associations of continuous exposure to analogue vs human insulins with mortality and major cardiovascular events. Data were analyzed from September 1, 2017, through June 30, 2018.

    Exposures  On the index date (first insulin dispensing), participants were classified as using analogue insulin with or without human insulin or human insulin only.

    Main Outcomes and Measures  Overall mortality, mortality due to cardiovascular disease (CVD), myocardial infarction (MI), stroke or cerebrovascular accident (CVA), and hospitalization for congestive heart failure (CHF) were evaluated. Marginal structural modeling (MSM) with inverse probability weighting was used to compare event-free survival in separate per-protocol analyses. Adjusted and unadjusted hazard ratios and cumulative risk differences were based on logistic MSM parameterizations for counterfactual hazards. Propensity scores were estimated using a data-adaptive approach (machine learning) based on 3 nested covariate adjustment sets. Sensitivity analyses were conducted to address potential residual confounding from unmeasured differences in risk factors across delivery systems.

    Results  The 127 600 participants (mean [SD] age, 59.4 [12.6] years; 68 588 men [53.8%]; mean [SD] body mass index, 32.3 [7.1]) had a median follow-up of 4 quarters (interquartile range, 3-9 quarters) and experienced 5464 deaths overall (4.3%), 1729 MIs (1.4%), 1301 CVAs (1.0%), and 3082 CHF hospitalizations (2.4%). There were no differences in adjusted hazard ratios for continuous analogue vs human insulin exposure during 10 quarters for overall mortality (1.15; 95% CI, 0.97-1.34), CVD mortality (1.26; 95% CI, 0.86-1.66), MI (1.11; 95% CI, 0.77-1.45), CVA (1.30; 95% CI, 0.81-1.78), or CHF hospitalization (0.93; 95% CI, 0.75-1.11).

    Conclusions and Relevance  Insulin-naive adults with type 2 diabetes who initiate and continue treatment with human vs analogue insulins had similar observed rates of major cardiovascular events, CVD mortality, and overall mortality.

    Introduction

    Because major cardiovascular events and mortality due to cardiovascular disease (CVD) are the principal causes of excess mortality and health care costs in adults with type 2 diabetes, the selection of agents to lower glucose levels for those with type 2 diabetes is necessarily informed by the effects of various agents on major cardiovascular events and mortality, as well as by other factors such as rates of hypoglycemia, convenience of use, and medication costs.1,2 The US Food and Drug Administration has required randomized cardiovascular outcome trials for all agents to lower glucose levels approved since 2008. The results of these trials demonstrate that some agents confer substantive cardiovascular-related benefits among some individuals with type 2 diabetes, whereas other agents do not.3-9

    Currently, approximately 90% of insulin users in the United States use analogue insulins, which were first introduced to the US market in 1996 and rapidly became widely used despite higher costs, because of effective marketing and studies indicating a lower rate of mild hypoglycemia.1,10,11 Although hypoglycemia rates related to various types of insulin have been extensively investigated, the effects of human compared with analogue insulins on cardiovascular events and mortality have received much less attention. Human insulin and the most commonly used analogue insulins were introduced before 2008 and thus were not evaluated in US Food and Drug Administration–mandated cardiovascular outcome trials. Some major clinical trials, such as the Diabetes Control and Complications Trial12 and United Kingdom Prospective Diabetes Study,13 did not include use of analogue insulins. Although more recent trials, such as ACCORD (Action to Control Cardiovascular Risk in Diabetes)14 and ADVANCE (Action in Diabetes and Vascular Disease: Preterax and Diamicron MR Controlled Evaluation),15 included the use of analogue insulins, the study designs preclude inferences about the cardiovascular safety of specific agents to lower glucose levels and direct comparison of the cardiovascular effects of human and analogue insulins.12,15-18

    The price differential between human and analogue insulins and the lack of significant differences in rates of serious hypoglycemia in recent reports have sparked new interest in the use of human insulin as a way to make health care more affordable to patients with diabetes.11,19 However, a recent meta-analysis20 and a recent World Health Organization position paper on diabetes care21,22 note that few methodologically rigorous studies have compared the relative effect of human vs analogue insulins on rates of major cardiovascular events and mortality in adults with type 2 diabetes. Herein we report the results of a large, multisite, National Institutes of Health–funded retrospective cohort study designed to assess the occurrence of mortality, CVD mortality, acute myocardial infarction (MI), stroke or cerebrovascular accident (CVA), and hospitalization for congestive heart failure (CHF) in adults with type 2 diabetes who initiated and adhered to a regimen of human vs analogue insulin. The present study differs from prior investigations of this topic by including a large number of US participants receiving care in community-based clinics, having relatively complete clinical and clinical outcome data, and applying current guidelines for machine learning and other modern statistical techniques that accommodate time-varying exposures and large health care databases under explicit assumptions such as those identified in this report.23-30

    Methods
    Study Design, Study Sites, and Data Sources

    This retrospective cohort study aimed to emulate24 a randomized experiment in which insulin-naive adults with type 2 diabetes would have been randomized at the time of dispensing the first insulin prescription to a continuous regimen of human insulin only (HI group) or analogue insulin with or without human insulin (AI group). The study sites included 4 integrated health care delivery systems from the Health Care Services Research Network: HealthPartners in Minnesota, Kaiser Permanente Colorado, Kaiser Permanente Northern California, and Kaiser Permanente Southern California.31 Health system electronic medical records, administrative claims data, 2010 census data, and mortality data were used to identify eligible participants, insulin type and use, demographics, clinical values, outcome variables, and covariates. The institutional review board of HealthPartners reviewed, approved, and monitored the progression of this study and approved our request to waive written informed consent for participants in this retrospective cohort study. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

    Overall, the combined membership of the 4 participating organizations was approximately 17 million members, of whom approximately 1.1 million individuals met criteria for diabetes from January 1, 2000, through December 31, 2013 (Figure 1).32 Operational definitions of type 2 diabetes and detailed definitions of other inclusion and exclusion criteria are presented in eMethods 1 in the Supplement.

    Participants

    We searched the entire adult membership of the 4 participating Health Care Services Research Network health plans to identify those who filled a first insulin prescription from January 1, 2005, through December 31, 2013, and who met all eligibility criteria detailed in eMethods 1 in the Supplement on the date when insulin was first dispensed (index date). Following prospective study enrollment principles, participants were not excluded from the study based on information collected after the first insulin fill.

    Participants were followed up from their index date until the earliest of (1) December 31, 2013 (administrative end of the study), (2) plan disenrollment (defined as a health insurance coverage gap of >90 days), or (3) death. For CVD mortality, the administrative end of study was December 31, 2011, owing to a 2-year lag of state death records. All participants with a first insulin prescription after that date were thus excluded from the analysis of CVD mortality.

    Exposures

    Two exposure groups defined by continuous treatment with the same insulin therapy (HI and AI groups) were compared. In the primary analyses, participants were considered exposed to a given insulin therapy from its dispensing date until the earliest of 180 days after dispensing or the date of a new prescription fill. If a new insulin prescription was filled before the 180th day after a prior insulin prescription was dispensed, we assumed that the patient expended the prior insulin supply (ie, stockpiling of insulin was assumed to be null). In sensitivity analyses, continuous treatment with the same insulin therapy was determined based on the assumption that each prescription could last as long as 365 days instead of 180 days.

    Clinical Outcomes

    Five clinical time-to-event outcomes were examined (eTable 1 in the Supplement). Acute MI (International Statistical Classification of Diseases, Tenth Revision, Clinical Modification [ICD-9-CM] code 410.xx), stroke/CVA (ICD-9-CM codes 430.xx, 431.xx, 433.x1, and 434.x1), and heart failure (ICD-9-CM codes 402.01, 402.11, 402.91, 404.01, 404.03, 404.11, 404.13, 404.91, 404.93, and 428.xx) were based on the inpatient principal discharge diagnosis. All-cause and CVD mortality were based on health system, state, and national vital statistics data. Mortality due to CVD included coronary heart disease, heart failure, cerebrovascular disease, peripheral artery disease, and atherosclerosis as defined by the primary cause of death.

    Covariates

    Based on the current medical literature or consensus medical judgment, we identified a comprehensive list of covariates (eTable 2 and eTable 3 in the Supplement) potentially affecting the exposures, outcomes, and censoring events (plan disenrollment, adherence to the initial insulin regimen, and death). These included patient demographics, clinical values, comorbid conditions, concomitant medications, smoking, neighborhood-level socioeconomic variables, and clinician and site characteristics (eTables 4-6 in the Supplement).

    Statistical Analysis

    Data were analyzed from September 1, 2017, through June 30, 2018. A separate analytic data set was constructed33 for each of the 5 clinical outcomes to conduct per-protocol analyses. Time-dependent variables (eMethods 2 in the Supplement) were updated every 90 days from the index date to the analytic end of follow-up, defined as the earliest of failure occurrence or one of the following right-censoring events: interruption of insulin therapy, switch in therapy type, start of pregnancy, or the administrative end of the study follow-up period.

    To account for baseline confounding and time-dependent sources of bias from informative censoring,34 we used inverse probability weight (IPW) estimation to evaluate the counterfactual cumulative risks of failure if all participants were continuously exposed to human-only or analogue-containing insulin therapy.23,35 For each outcome, IPW was used to fit 2 logistic marginal structural models (MSM) for the discrete-time counterfactual hazards (eMethods 3 in the Supplement) during the first 2.5 years of follow-up: an MSM that relies on the proportionality assumption27,36 to provide a single summary effect measure estimate (hazard ratio) and a saturated MSM29,37 to provide estimates of differences in cumulative risks (at 1 and 2 years) between the 2 exposure regimens without reliance on the proportionality assumption.

    Four approaches for estimating the propensity scores that define the IPW were considered. The first 3 approaches were based on logistic modeling (eMethods 4 in the Supplement) with different covariate adjustment sets38,39 (eMethods 5 in the Supplement). The fourth approach was based on data-adaptive propensity score estimation with a machine learning method known as Super Learning.40 Super Learning was used for adapting the covariate adjustment set that best predicts each propensity score outcome based on a logistic model (eMethods 6 in the Supplement).41,42 All IPWs were stabilized and truncated43,44 at 20. Adjusted effect measure estimates from the 2 MSMs and 4 propensity score estimation approaches considered were also compared with their unadjusted effect measure estimates.

    Motivated by results from the evaluation of patterns of first insulin use across sites, we conducted post hoc site-specific sensitivity analyses restricted to data from the 3 sites that used similar proportions of human vs analogue insulin (sites 2-4) and data from the site (site 4) that had the most variability in the type of first insulin use during the years of study. Sensitivity analyses as well as the primary analyses include interaction terms between site and year of study entry variables in the propensity score logistic models used to predict the initial insulin therapy prescribed at the index date. Statistical details of the main and sensitivity analyses are provided in eMethods 7 in the Supplement. For each outcome analysis, we computed a 2-tailed P value for the statistical test that the area between the 2 survival curves is null (ie, the sum of the risk differences at each quarter is equal to 0). P < .05 indicated statistical significance.

    Results

    Of the 1 084 052 participants with diabetes in the 4 health care systems, 223 908 had a first fill of insulin from January 1, 2005, through December 31, 2013. Of these, 127 600 participants with type 2 diabetes initiating insulin therapy were included in the main study cohort (mean [SD] age, 59.4 [12.6] years; 68 588 men [53.8%] and 59 012 women [46.2%]), with 108 672 (85.2%) in the HI group and 18 928 (14.8%) in the AI group (Figure 1). Only 95 300 of the 127 600 insulin-using participants were evaluated in the CVD mortality analyses for the reasons discussed earlier.

    Table 1 describes selected demographic and clinical characteristics of participants at the index date in the main cohort by type of insulin initiated. The AI group was slightly younger (mean [SD] age, 58.8 [13.2] vs 59.5 [12.5] years) and had slightly higher rate of comorbidities (including coronary artery disease [18.7% vs 16.1%], CVA [2.2% vs 1.6%], and CHF [7.6% vs 6.5%]). Mean hemoglobin A1c levels, blood pressure, and smoking rates were similar between the groups. Of the 127 600 patients, 98 965 (77.6%) initiated only long-acting insulin therapy at the index date, 10 379 (8.1%) initiated only short-acting insulin therapy, and 18 256 (14.3%) initiated a combination of both insulin types. Change in hemoglobin A1c values after the index date and during the 2.5-year follow-up showed no systematic major differences between the AI and HI groups.

    The median time from the index date to the analytic end of follow-up (eg, owing to a switch in insulin therapy) was 4 quarters in the main and CVD cohorts, with an interquartile range of 3 to 9 quarters in the main cohort and 3 to 8 quarters in the CVD cohort. We thus restricted the evaluation of hazards to the first 10 quarters of follow-up. Table 2 quantifies exposure time to analogue or human insulin and the number of events for each of the 5 outcomes. Overall, participants experienced 5464 deaths (4.3%), 1729 MIs (1.4%), 1301 CVAs (1.0%), and 3082 CHF hospitalizations (2.4%). Table 2 also displays distribution of reasons for end of follow-up in all primary analyses. Interruption of initial insulin therapy was the primary source of right censoring (57.1% for MI; 57.7% for mortality; 57.2% for CVA; 56.6% for CHF; and 54.6% for CVD mortality) and occurred for the following 4 reasons in order of decreasing frequency: a gap of more than 180 days between 2 consecutive insulin prescriptions dispensed, discontinuation of the initial insulin therapy with no subsequent insulin refill before the study end of follow-up, switching from analogue to human insulin therapy or vice versa, or dispensing of inhaled or animal insulins. For all primary analyses, more than half (>57%) of participants whose analytic end of follow-up was due to a gap in or discontinuation of the initial insulin therapy were right censored in the third quarter of follow-up because they did not refill a prescription for insulin within 180 days from the index date. This amounts to more than 26% of participants in the main and CVD cohorts being right censored owing to not filling a second insulin prescription within 180 days from their first insulin fill.

    Figure 2 displays the results of the unadjusted (crude) primary per-protocol analyses for mortality, MI, and CVD mortality and includes data on the number of participants at risk for each outcome in each quarter during the analysis period. Figure 3 displays the adjusted (IPW) primary per-protocol analyses based on the saturated MSM and Super Learning for propensity score estimation. Additional results and results for CHF and CVA are presented in eFigures 1 to 5 in the Supplement. Table 3 shows the adjusted hazard ratios and risk differences at 1 and 2 years along with their 95% CIs from the primary per-protocol analyses of each outcome using data-adaptive estimation of the propensity score. Adjusted estimates from the primary per-protocol analyses provide no evidence of a statistically significant difference in cumulative risks between the HI and AI groups during the first 10 quarters of follow-up for any outcome irrespective of the propensity score estimation approach used. Adjusted hazard ratios and 95% CIs for continuous analogue vs human insulin exposure also demonstrated no statistically significant associations with overall mortality (1.15; 95% CI, 0.97-1.34), CVD mortality (1.26; 95% CI, 0.86-1.66), MI (1.11; 95% CI, 0.77-1.45), CVA (1.30; 95% CI, 0.81-1.78), and CHF hospitalization (0.93; 95% CI, 0.75-1.11). Estimations of risk differences at 1 and 2 years (Table 3) were consistent with these results.

    eMethods 7 in the Supplement provides the results of all primary and sensitivity analyses based on the 2 MSM and 4 propensity score estimation approaches considered, including details of propensity score estimation (eTables 7-12 in the Supplement), patterns of first insulin use across sites (eTable 13 in the Supplement), and follow-up time by quarter (eTables 14-28 in the Supplement). Null findings from the primary per-protocol analyses (Figure 2, Figure 3, and eTables 29-33 in the Supplement) are generally supported by the adjusted estimates from sensitivity per protocol analyses (eFigures 1-5 in the Supplement). Inverse probability weights are provided in eTables 34 to 38 in the Supplement.

    Discussion

    The results of this cohort study show no consistent statistically significant differences in rates of MI, CVA, CHF hospitalizations, CVD mortality, or overall mortality between adults with type 2 diabetes initiating human vs analogue insulin treatment. Unadjusted analyses showed some results that favored the HI over the AI groups, but analyses adjusted for possible baseline or time-varying confounding and informative right censoring using standard and machine learning estimates of propensity scores based on 3 covariate adjustment sets showed no consistent differences in outcomes across exposure groups.

    To our knowledge, no randomized clinical trials have compared the relative effects of human and analogue insulin on major cardiovascular events or mortality.20,45 Prior cohort studies of the relative effects of human vs analogue insulin on major cardiovascular outcomes have reported mixed results, with most studies having significant limitations related to selection of participants, sample size, or description of analytic details.45-48 Several cohort studies focused exclusively on bolus insulin preparations, included relatively small numbers of participants, were limited to US veterans, or may have included in the analysis cardiovascular events that occurred before the initiation of insulin therapy.45,47,49

    Our results address many of the limitations of prior studies, including the inclusion of bolus and basal preparations of analogue and human insulin, inclusion of a large number of eligible participants from multiple US health care systems, systematic ascertainment of cardiovascular events and CVD mortality, inclusion of a broad representation of adults with type 2 diabetes being treated predominantly in primary care settings, and use of sophisticated analytic methods including but not limited to machine learning. The validity of results with the per-protocol analyses based on IPW estimation used in this study relies on the usual assumption of no unmeasured confounding35 or sources of selection bias,34 that is, the sequential randomization assumption. Upholding this assumption relies on the selection of an adequate set of covariates for bias adjustment. Several alternate pragmatic criteria for covariate selection, including machine learning, have been proposed,39,41,42,50-52 and we selected 4 such alternatives herein to evaluate consistency of results across analytic methods.

    The extensive sensitivity analyses conducted in this study (eMethods 7 in the Supplement) led to multiple comparisons when analyzing the association of human vs analogue insulin with 5 clinical outcomes. Although multiple comparisons do not change point estimates and corresponding (pointwise) CIs, there is no correction to our P values to compensate for multiple hypothesis testing. This observation further supports the main finding of no significant differences in mortality or major cardiovascular events between the AI and HI groups.

    Our results suggest that cardiovascular outcomes and mortality should not be a motivating factor in the decision to start human vs analogue insulin therapy in insulin-naive adults with type 2 diabetes. Other relevant factors to consider include hypoglycemia, glycemic control, cost, and ease of use. Recent reports have shown similar effects of human and analogue insulins on control of glucose levels53,54 and serious hypoglycemic events in primary care practice,11 which suggest that human and analogue insulins are safe and effective treatments in type 2 diabetes.11,19,53-56 The wholesale acquisition cost or list price of a 10-mL vial of human insulin was $24.90 for short- or intermediate-acting insulin, compared with $283 for a 10-mL vial of long-acting analogue insulin glargine and $289 for a 10-mL vial of rapid-acting insulin aspart analogue, according to data published in May 2019.57

    Limitations

    Several factors constrain the interpretation of our results. First, even with the use of rigorous statistical methods and a large sample size, the cohort study design is a limitation. However, owing to the high cost of conducting large randomized trials in a rapidly evolving insulin market, there is little chance that a large randomized trial will address cardiovascular outcomes of human vs other insulins, although manufacturers have compared newer and older analogue insulins. Results show few differences in cardiovascular events, suggesting that newer analogue insulins are unlikely to have better cardiovascular outcomes than the analogue insulins we evaluated.58,59 Second, human insulin was the initial treatment for 5.0%, 82.5%, 90.4%, and 96.1% of participants at our 4 study sites owing to formulary preferences. We exploited temporal variability in prescription rates at one of the sites in a sensitivity analysis to address potential concerns over residual bias from unobserved differences in risk factors between human and analogue insulin users at the other 3 sites that consistently favored 1 insulin type during the years of the study. Third, we elected to include short-acting and long-acting analogue and human insulin preparations in this analysis because the chief clinical choice patients and physicians make on a daily basis is between analogue and human insulin. Additional research to compare the cardiovascular safety of short- and rapid-acting vs long-acting insulins is warranted. It is challenging to precisely detect interruption in insulin exposure solely from pharmacy-dispensing data. To address this concern, we demonstrated that inferences were not sensitive to our assumption about the maximum duration of each insulin prescription (ie, 180 vs 365 days).

    Conclusions

    This detailed analysis of a large data set using rigorous modern statistical methods and machine learning suggests that adults with type 2 diabetes who are new users of human or analogue insulin have similar rates of mortality, CVD mortality, and major cardiovascular events during 2.5 years of follow-up. These results suggest that cardiovascular outcomes and mortality should not be a motivating factor in the decision to start human vs analogue insulin therapy in insulin-naive adults with type 2 diabetes. Other relevant factors to consider include hypoglycemia, glycemic control, cost, and ease of use. These results contribute important new clinical information that can help inform insulin-related treatment decisions made by adults with type 2 diabetes and their clinicians.

    Back to top
    Article Information

    Accepted for Publication: October 15, 2019.

    Published: January 24, 2020. doi:10.1001/jamanetworkopen.2019.18554

    Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2020 Neugebauer R et al. JAMA Network Open.

    Corresponding Author: Patrick J. O’Connor, MD, MA, MPH, HealthPartners Institute, 8170 33rd Ave S, Mail Stop 23301A, Minneapolis, MN 55425 (patrick.j.oconnor@healthpartners.com).

    Author Contributions: Dr Neugebauer 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.

    Concept and design: Neugebauer, Schroeder, Reynolds, Loes, Desai, Vazquez-Benitez, O’Connor.

    Acquisition, analysis, or interpretation of data: Neugebauer, Schroeder, Reynolds, Schmittdiel, Dyer, Desai, Vazquez-Benitez, Ho, Anderson, Pimentel, O’Connor.

    Drafting of the manuscript: Neugebauer, Schroeder, Desai, Pimentel, O’Connor.

    Critical revision of the manuscript for important intellectual content: Neugebauer, Schroeder, Reynolds, Schmittdiel, Loes, Dyer, Vazquez-Benitez, Ho, Anderson, Pimentel, O’Connor.

    Statistical analysis: Neugebauer, Schmittdiel, Dyer, Vazquez-Benitez, Pimentel, O’Connor.

    Obtained funding: Neugebauer, Schmittdiel, O’Connor.

    Administrative, technical, or material support: Neugebauer, Schroeder, Reynolds, Loes, Desai, Pimentel, O’Connor.

    Supervision: Neugebauer, O’Connor.

    Conflict of Interest Disclosures: Dr Neugebauer reported receiving grants from the National Institutes of Health (NIH) during the conduct of the study. Dr Schroeder reported grants from the National Heart, Lung, and Blood Institute (NHLBI) and the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) during the conduct of the study and grants from the NIDDK, the Garfield Foundation, Centers for Disease Control and Prevention, the American Heart Association, and the Agency for Healthcare Research and Quality outside the submitted work. Dr Reynolds reported receiving grants from Merck & Co outside the submitted work. Dr Schmittdiel reported receiving grants from the NIH during the conduct of the study. Dr Loes reported receiving grants from the NIH during the conduct of the study. Ms Dyer reported receiving grants from the NHLBI during the conduct of the study. Dr Desai reported receiving grants from the NIH during the conduct of the study. Dr Vazquez-Benitez reported receiving grants from the NIH during the conduct of the study. Dr Ho reported receiving grants from NHLBI and the Department of Veterans Affairs Health Services Research and Development and serving as the deputy editor for Circulation: Cardiovascular Quality and Outcomes. Dr Anderson reported receiving grants from the NIH during the conduct of the study and internal grants from HealthPartners outside the submitted work. Dr O’Connor reported receiving grants from the NIH and Patient-Centered Outcomes Research Institute during the conduct of the study. No other disclosures were reported.

    Funding/Support: This study was supported by grant R01HL124461 from the NHLBI of the NIH and by grants K23DK099237 (Dr Schroeder) and P30DK092924 (Drs Schmittdiel, Desai, and O’Connor) from the NIDDK.

    Role of the Funder/Sponsor: The NIH approved the study design but had no role in the conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.

    References
    1.
    American Diabetes Association.  Introduction: Standards of Medical Care in Diabetes–2019.  Diabetes Care. 2019;42(suppl 1):S1-S2. doi:10.2337/dc19-Sint01PubMedGoogle ScholarCrossref
    2.
    Davies  MJ, D’Alessio  DA, Fradkin  J,  et al.  Management of hyperglycemia in type 2 diabetes, 2018: a consensus report by the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD).  Diabetes Care. 2018;41(12):2669-2701. doi:10.2337/dci18-0033PubMedGoogle ScholarCrossref
    3.
    Cefalu  WT, Kaul  S, Gerstein  HC,  et al.  Cardiovascular outcomes trials in type 2 diabetes: where do we go from here? reflections from a Diabetes Care editors’ expert forum.  Diabetes Care. 2018;41(1):14-31. doi:10.2337/dci17-0057PubMedGoogle ScholarCrossref
    4.
    Marso  SP, Daniels  GH, Brown-Frandsen  K,  et al; LEADER Steering Committee; LEADER Trial Investigators.  Liraglutide and cardiovascular outcomes in type 2 diabetes.  N Engl J Med. 2016;375(4):311-322. doi:10.1056/NEJMoa1603827PubMedGoogle ScholarCrossref
    5.
    Marso  SP, Bain  SC, Consoli  A,  et al; SUSTAIN-6 Investigators.  Semaglutide and cardiovascular outcomes in patients with type 2 diabetes.  N Engl J Med. 2016;375(19):1834-1844. doi:10.1056/NEJMoa1607141PubMedGoogle ScholarCrossref
    6.
    Zinman  B, Wanner  C, Lachin  JM,  et al; EMPA-REG OUTCOME Investigators.  Empagliflozin, cardiovascular outcomes, and mortality in type 2 diabetes.  N Engl J Med. 2015;373(22):2117-2128. doi:10.1056/NEJMoa1504720PubMedGoogle ScholarCrossref
    7.
    Neal  B, Perkovic  V, Mahaffey  KW,  et al; CANVAS Program Collaborative Group.  Canagliflozin and cardiovascular and renal events in type 2 diabetes.  N Engl J Med. 2017;377(7):644-657. doi:10.1056/NEJMoa1611925PubMedGoogle ScholarCrossref
    8.
    Rosenstock  J, Kahn  SE, Johansen  OE,  et al; CAROLINA Investigators.  Effect of linagliptin vs glimepiride on major adverse cardiovascular outcomes in patients with type 2 diabetes: the CAROLINA randomized clinical trial.  JAMA. 2019;322(12):1155-1166. doi:10.1001/jama.2019.13772PubMedGoogle ScholarCrossref
    9.
    Roumie  CL, Chipman  J, Min  JY,  et al.  Association of treatment with metformin vs sulfonylurea with major adverse cardiovascular events among patients with diabetes and reduced kidney function.  JAMA. 2019;322(12):1167-1177. doi:10.1001/jama.2019.13206PubMedGoogle ScholarCrossref
    10.
    Pathak  RD, Schroeder  EB, Seaquist  ER,  et al; SUPREME-DM Study Group.  Severe hypoglycemia requiring medical intervention in a large cohort of adults with diabetes receiving care in US integrated health care delivery systems: 2005-2011.  Diabetes Care. 2016;39(3):363-370. doi:10.2337/dc15-0858PubMedGoogle ScholarCrossref
    11.
    Lipska  KJ, Parker  MM, Moffet  HH, Huang  ES, Karter  AJ.  Association of initiation of basal insulin analogs vs neutral protamine hagedorn insulin with hypoglycemia-related emergency department visits or hospital admissions and with glycemic control in patients with type 2 diabetes.  JAMA. 2018;320(1):53-62. doi:10.1001/jama.2018.7993PubMedGoogle ScholarCrossref
    12.
    Nathan  DM, Genuth  S, Lachin  J,  et al; Diabetes Control and Complications Trial Research Group.  The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus.  N Engl J Med. 1993;329(14):977-986. doi:10.1056/NEJM199309303291401PubMedGoogle ScholarCrossref
    13.
    UK Prospective Diabetes Study (UKPDS) Group.  Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33).  Lancet. 1998;352(9131):837-853. doi:10.1016/S0140-6736(98)07019-6PubMedGoogle ScholarCrossref
    14.
    Gerstein  HC, Miller  ME, Byington  RP,  et al; Action to Control Cardiovascular Risk in Diabetes Study Group.  Effects of intensive glucose lowering in type 2 diabetes.  N Engl J Med. 2008;358(24):2545-2559. doi:10.1056/NEJMoa0802743PubMedGoogle ScholarCrossref
    15.
    Patel  A, MacMahon  S, Chalmers  J,  et al; ADVANCE Collaborative Group.  Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes.  N Engl J Med. 2008;358(24):2560-2572. doi:10.1056/NEJMoa0802987PubMedGoogle ScholarCrossref
    16.
    Gerstein  HC, Miller  ME, Genuth  S,  et al; ACCORD Study Group.  Long-term effects of intensive glucose lowering on cardiovascular outcomes.  N Engl J Med. 2011;364(9):818-828. doi:10.1056/NEJMoa1006524PubMedGoogle ScholarCrossref
    17.
    Holman  RR, Paul  SK, Bethel  MA, Matthews  DR, Neil  HA.  10-year follow-up of intensive glucose control in type 2 diabetes.  N Engl J Med. 2008;359(15):1577-1589. doi:10.1056/NEJMoa0806470PubMedGoogle ScholarCrossref
    18.
    Duckworth  W, Abraira  C, Moritz  T,  et al; VADT Investigators.  Glucose control and vascular complications in veterans with type 2 diabetes.  N Engl J Med. 2009;360(2):129-139. doi:10.1056/NEJMoa0808431PubMedGoogle ScholarCrossref
    19.
    Lipska  KJ, Hirsch  IB, Riddle  MC.  Human insulin for type 2 diabetes: an effective, less-expensive option.  JAMA. 2017;318(1):23-24. doi:10.1001/jama.2017.6939PubMedGoogle ScholarCrossref
    20.
    Fullerton  B, Siebenhofer  A, Jeitler  K,  et al.  Short-acting insulin analogues versus regular human insulin for adult, non-pregnant persons with type 2 diabetes mellitus.  Cochrane Database Syst Rev. 2018;12:CD013228. doi:10.1002/14651858.CD013228PubMedGoogle Scholar
    21.
    Roglic  G, Norris  SL.  Medicines for treatment intensification in type 2 diabetes and type of insulin in type 1 and type 2 diabetes in low-resource settings: synopsis of the World Health Organization guidelines on second- and third-line medicines and type of insulin for the control of blood glucose levels in nonpregnant adults with diabetes mellitus.  Ann Intern Med. 2018;169(6):394-397. doi:10.7326/M18-1149PubMedGoogle ScholarCrossref
    22.
    World Health Organization (WHO).  Guidelines on Second- and Third-Line Medicines and Type of Insulin for the Control of Blood Glucose Levels in Non-pregnant Adults With Diabetes Mellitus. Geneva, Switzerland: World Health Organization; 2018.
    23.
    Robins  JM. Marginal structural models. In: 1997 Proceedings of the Section on Bayesian Statistical Science; Alexandria, VA: American Statistical Association, 1998;1-10. https://cdn1.sph.harvard.edu/wp-content/uploads/sites/343/2013/03/msm-web.pdf. Accessed December 9, 2019.
    24.
    Hernán  MA, Robins  JM.  Using big data to emulate a target trial when a randomized trial is not available.  Am J Epidemiol. 2016;183(8):758-764. doi:10.1093/aje/kwv254PubMedGoogle ScholarCrossref
    25.
    Hernán  MA, Hernández-Díaz  S.  Beyond the intention-to-treat in comparative effectiveness research.  Clin Trials. 2012;9(1):48-55. doi:10.1177/1740774511420743PubMedGoogle ScholarCrossref
    26.
    Hernán  MA, Robins  JM.  Per-protocol analyses of pragmatic trials.  N Engl J Med. 2017;377(14):1391-1398. doi:10.1056/NEJMsm1605385PubMedGoogle ScholarCrossref
    27.
    Hernán  MA.  The hazards of hazard ratios.  Epidemiology. 2010;21(1):13-15. doi:10.1097/EDE.0b013e3181c1ea43PubMedGoogle ScholarCrossref
    28.
    Pirracchio  R, Petersen  ML, van der Laan  M.  Improving propensity score estimators’ robustness to model misspecification using super learner.  Am J Epidemiol. 2015;181(2):108-119. doi:10.1093/aje/kwu253PubMedGoogle ScholarCrossref
    29.
    Neugebauer  R, Schmittdiel  JA, van der Laan  MJ.  A case study of the impact of data-adaptive versus model-based estimation of the propensity scores on causal inferences from three inverse probability weighting estimators.  Int J Biostat. 2016;12(1):131-155. doi:10.1515/ijb-2015-0028PubMedGoogle ScholarCrossref
    30.
    Neugebauer  R, Fireman  B, Roy  JA, Raebel  MA, Nichols  GA, O’Connor  PJ.  Super learning to hedge against incorrect inference from arbitrary parametric assumptions in marginal structural modeling.  J Clin Epidemiol. 2013;66(8)(suppl):S99-S109. doi:10.1016/j.jclinepi.2013.01.016PubMedGoogle ScholarCrossref
    31.
    Steiner  JF, Paolino  AR, Thompson  EE, Larson  EB.  Sustaining research networks: the twenty-year experience of the HMO Research Network.  EGEMS (Wash DC). 2014;2(2):1067. doi:10.13063/2327-9214.1067PubMedGoogle Scholar
    32.
    Nichols  GA, Desai  J, Elston Lafata  J,  et al; SUPREME-DM Study Group.  Construction of a multisite DataLink using electronic health records for the identification, surveillance, prevention, and management of diabetes mellitus: the SUPREME-DM project.  Prev Chronic Dis. 2012;9:E110. doi:10.5888/pcd9.110311PubMedGoogle Scholar
    33.
    Kaiser Permanente Division of Research, Northern California; Leong  TK, Tabada  GH, Yang  JM, Zhu  Z, Neugebauer  RS. MSMstructure SAS macro version 1.0.4. https://divisionofresearch.kaiserpermanente.org/projects/biostatistics/causalinferencesoftware. Published March 16, 2017. Accessed November 22, 2019.
    34.
    Hernán  MA, Hernández-Díaz  S, Robins  JM.  A structural approach to selection bias.  Epidemiology. 2004;15(5):615-625. doi:10.1097/01.ede.0000135174.63482.43PubMedGoogle ScholarCrossref
    35.
    Robins  JM.  Association, causation, and marginal structural models.  Synthese. 1999;121(1-2):151-179. doi:10.1023/A:1005285815569Google ScholarCrossref
    36.
    Hernán  MA, Brumback  B, Robins  JM.  Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men.  Epidemiology. 2000;11(5):561-570. doi:10.1097/00001648-200009000-00012PubMedGoogle ScholarCrossref
    37.
    Neugebauer  R, Schmittdiel  JA, van der Laan  MJ.  Targeted learning in real-world comparative effectiveness research with time-varying interventions.  Stat Med. 2014;33(14):2480-2520. doi:10.1002/sim.6099PubMedGoogle ScholarCrossref
    38.
    Sauer  BC, Brookhart  MA, Roy  J, VanderWeele  T.  A review of covariate selection for non-experimental comparative effectiveness research.  Pharmacoepidemiol Drug Saf. 2013;22(11):1139-1145. doi:10.1002/pds.3506PubMedGoogle ScholarCrossref
    39.
    VanderWeele  TJ, Shpitser  I.  A new criterion for confounder selection.  Biometrics. 2011;67(4):1406-1413. doi:10.1111/j.1541-0420.2011.01619.xPubMedGoogle ScholarCrossref
    40.
    van der Laan  MJ, Polley  EC, Hubbard  AE.  Super learner.  Stat Appl Genet Mol Biol. 2007;6(Article 25):e25. doi:10.2202/1544-6115.1309PubMedGoogle Scholar
    41.
    Joffe  MM.  Exhaustion, automation, theory, and confounding.  Epidemiology. 2009;20(4):523-524. doi:10.1097/EDE.0b013e3181a82501PubMedGoogle ScholarCrossref
    42.
    Neugebauer  R, Schmittdiel  JA, Zhu  Z, Rassen  JA, Seeger  JD, Schneeweiss  S.  High-dimensional propensity score algorithm in comparative effectiveness research with time-varying interventions.  Stat Med. 2015;34(5):753-781. doi:10.1002/sim.6377PubMedGoogle ScholarCrossref
    43.
    Cole  SR, Hernán  MA.  Constructing inverse probability weights for marginal structural models.  Am J Epidemiol. 2008;168(6):656-664. doi:10.1093/aje/kwn164PubMedGoogle ScholarCrossref
    44.
    Petersen  ML, Porter  KE, Gruber  S, Wang  Y, van der Laan  MJ.  Diagnosing and responding to violations in the positivity assumption.  Stat Methods Med Res. 2012;21(1):31-54. doi:10.1177/0962280210386207PubMedGoogle ScholarCrossref
    45.
    Price  HI, Agnew  MD, Gamble  JM.  Comparative cardiovascular morbidity and mortality in patients taking different insulin regimens for type 2 diabetes: a systematic review.  BMJ Open. 2015;5(3):e006341. doi:10.1136/bmjopen-2014-006341PubMedGoogle Scholar
    46.
    Strandberg  AY, Hoti  FJ, Strandberg  TE, Christopher  S, Haukka  J, Korhonen  P.  All-cause and cause-specific mortality among users of basal insulins NPH, detemir, and glargine.  PLoS One. 2016;11(3):e0151910. doi:10.1371/journal.pone.0151910PubMedGoogle Scholar
    47.
    Rathmann  W, Schloot  NC, Kostev  K, Reaney  M, Zagar  AJ, Haupt  A.  Macro- and microvascular outcomes in patients with type 2 diabetes treated with rapid-acting insulin analogues or human regular insulin: a retrospective database analysis.  Exp Clin Endocrinol Diabetes. 2014;122(2):92-99. doi:10.1055/s-0033-1363684PubMedGoogle ScholarCrossref
    48.
    Horvath  K, Jeitler  K, Berghold  A,  et al.  Long-acting insulin analogues versus NPH insulin (human isophane insulin) for type 2 diabetes mellitus.  Cochrane Database Syst Rev. 2007;(2):CD005613. doi:10.1002/14651858.CD005613.pub3PubMedGoogle Scholar
    49.
    Cammarota  S, Falconio  LM, Bruzzese  D,  et al.  Lower rate of cardiovascular complications in patients on bolus insulin analogues: a retrospective population-based cohort study.  PLoS One. 2013;8(11):e79762. doi:10.1371/journal.pone.0079762PubMedGoogle Scholar
    50.
    Glymour  MM, Weuve  J, Chen  JT.  Methodological challenges in causal research on racial and ethnic patterns of cognitive trajectories: measurement, selection, and bias.  Neuropsychol Rev. 2008;18(3):194-213. doi:10.1007/s11065-008-9066-xPubMedGoogle ScholarCrossref
    51.
    Brookhart  MA, Schneeweiss  S, Rothman  KJ, Glynn  RJ, Avorn  J, Stürmer  T.  Variable selection for propensity score models.  Am J Epidemiol. 2006;163(12):1149-1156. doi:10.1093/aje/kwj149PubMedGoogle ScholarCrossref
    52.
    Schneeweiss  S, Rassen  JA, Glynn  RJ, Avorn  J, Mogun  H, Brookhart  MA.  High-dimensional propensity score adjustment in studies of treatment effects using health care claims data.  Epidemiology. 2009;20(4):512-522. doi:10.1097/EDE.0b013e3181a663ccPubMedGoogle ScholarCrossref
    53.
    Luo  J, Khan  NF, Manetti  T,  et al.  Implementation of a health plan program for switching from analogue to human insulin and glycemic control among Medicare beneficiaries with type 2 diabetes.  JAMA. 2019;321(4):374-384. doi:10.1001/jama.2018.21364PubMedGoogle ScholarCrossref
    54.
    Lipska  KJ.  Insulin analogues for type 2 diabetes.  JAMA. 2019;321(4):350-351. doi:10.1001/jama.2018.21356PubMedGoogle ScholarCrossref
    55.
    Lipska  KJ, Yao  X, Herrin  J,  et al.  Trends in drug utilization, glycemic control, and rates of severe hypoglycemia, 2006-2013.  Diabetes Care. 2017;40(4):468-475. doi:10.2337/dc16-0985PubMedGoogle ScholarCrossref
    56.
    Madenidou  AV, Paschos  P, Karagiannis  T,  et al.  Comparative benefits and harms of basal insulin analogues for type 2 diabetes: a systematic review and network meta-analysis.  Ann Intern Med. 2018;169(3):165-174. doi:10.7326/M18-0443PubMedGoogle ScholarCrossref
    57.
    Insulins for type 2 diabetes. The Medical Letter on Drugs and Therapeutics, issue 1571. https://secure.medicalletter.org/article-share?a=1571a&p=tml&title=Insulins%20for%20Type%202%20Diabetes&cannotaccesstitle=1. Published May 6, 2019. Accessed December 3, 2019.
    58.
    Marso  SP, McGuire  DK, Zinman  B,  et al; DEVOTE Study Group.  Efficacy and safety of degludec versus glargine in type 2 diabetes.  N Engl J Med. 2017;377(8):723-732. doi:10.1056/NEJMoa1615692PubMedGoogle ScholarCrossref
    59.
    Freemantle  N, Chou  E, Frois  C,  et al.  Safety and efficacy of insulin glargine 300 μ/mL compared with other basal insulin therapies in patients with type 2 diabetes mellitus: a network meta-analysis.  BMJ Open. 2016;6(2):e009421. doi:10.1136/bmjopen-2015-009421PubMedGoogle Scholar
    ×