Approximate Bayesian Bootstrap procedures to estimate multilevel treatment effects in observational studies with application to type 2 diabetes treatment regimens
Restricted accessResearch articleFirst published online November, 2020
Approximate Bayesian Bootstrap procedures to estimate multilevel treatment effects in observational studies with application to type 2 diabetes treatment regimens
Randomized clinical trials are considered as the gold standard for estimating causal effects. Nevertheless, in studies that are aimed at examining adverse effects of interventions, randomized trials are often impractical because of ethical and financial considerations. In observational studies, matching on the generalized propensity scores was proposed as a possible solution to estimate the treatment effects of multiple interventions. However, the derivation of point and interval estimates for these matching procedures can become complex with non-continuous or censored outcomes. We propose a novel Approximate Bayesian Bootstrap algorithm that results in statistically valid point and interval estimates of the treatment effects with categorical outcomes. The procedure relies on the estimated generalized propensity scores and multiply imputes the unobserved potential outcomes for each unit. In addition, we describe a corresponding interpretable sensitivity analysis to examine the unconfoundedness assumption. We apply this approach to examine the cardiovascular safety of common, real-world anti-diabetic treatment regimens for type 2 diabetes mellitus in a large observational database.
PreisSRHwangSCoadyS, et al.
Trends in all-cause and cardiovascular disease mortality among women and men with and without diabetes mellitus in the Framingham Heart Study, 1950 to 2005. Circulation2009;
119: 1728–1735.
2.
Group UK Prospective Diabetes Study.
Effect of intensive blood-glucose control with metformin on complications in overweight patients with type 2 diabetes.Lancet1998;
352: 1558.
3.
GersteinHCMillerMEByingtonRP, et al.
Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med2008;
358: 2545–2559.
4.
FDA. FDA Drug Safety Communication: FDA adds warnings about heart failure risk to labels of type 2 diabetes medicines containing saxagliptin and alogliptin, www.fda.gov/drugs/drug-safety-and-availability (2018, accessed 28 September 2019).
InzucchiSEBergenstalRMBuseJB, et al.
Management of hyperglycemia in type 2 diabetes, 2015: a patient-centered approach: update to a position statement of the American diabetes association and the European association for the study of diabetes.Diabet Care2015;
38: 140–149.
7.
ChamberlainJJHermanWHLealS, et al.
Pharmacologic therapy for type 2 diabetes: synopsis of the 2017 American Diabetes Association standards of medical care in diabetes.Ann Intern Med2017;
166: 572–578.
8.
LeonardCEHanXBrensingerCM, et al.
Comparative risk of serious hypoglycemia with oral antidiabetic monotherapy: a retrospective cohort study. Pharmacoepidemiol Drug Safety2018;
27: 9–18.
9.
DormandyJA, et al.
Secondary prevention of macrovascular events in patients with type 2 diabetes in the PROactive Study (PROspective pioglitAzone Clinical Trial in macroVascular Events): a randomised controlled trial. Lancet2005;
366: 1279–1289.
10.
SeongJChoiNShinJ, et al.
Differential cardiovascular outcomes after dipeptidyl peptidase-4 inhibitor, sulfonylurea, and pioglitazone therapy, all in combination with metformin, for type 2 diabetes: a population-based cohort study. PLOS One2015;
10: 1–14.
11.
StrongmanH, et al.
Pioglitazone and cause-specific risk of mortality in patients with type 2 diabetes: extended analysis from a European multidatabase cohort study.BMJ Open Diabet Res Care2018;
6: e000481.
12.
ChanC, et al.
Glitazones and alpha-glucosidase inhibitors as the second-line oral anti-diabetic agents added to metformin reduce cardiovascular risk in type 2 diabetes patients: a nationwide cohort observational study. Cardiovasc Diabetol2018;
17: 1–13.
13.
PonikowskiP, et al.
Guidelines for the diagnosis and treatment of acute and chronic heart failure: the task force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC).Eur Heart J2016;
37: 2129–2200.
14.
SciricaBMBhattDLBraunwaldE, et al.
Saxagliptin and cardiovascular outcomes in patients with type 2 diabetes mellitus. N Engl J Med2013;
369: 1317–1326.
15.
WhiteWBCannonCPHellerSR, et al.
Alogliptin after acute coronary syndrome in patients with type 2 diabetes. N Engl J Med2013;
369: 1327–1335.
16.
GreenJBBethelMAArmstrongPW, et al.
Effect of sitagliptin on cardiovascular outcomes in type 2 diabetes. N Engl J Med2015;
373: 232–242.
17.
RosenstockJPerkovicVJohansenOE, et al.
Effect of linagliptin vs placebo on major cardiovascular events in adults with type 2 diabetes and high cardiovascular and renal risk: the CARMELINA randomized clinical trial.J Am Med Assoc2019;
321: 69–79.
18.
NissenSENichollsSJWolskiK, et al.
Comparison of pioglitazone vs glimepiride on progression of coronary atherosclerosis in patients with type 2 diabetes. J Am Med Assoc2008;
299: 1561–1573.
19.
RoumieCLHungAMGreevyRA, et al.
Comparative effectiveness of sulfonylurea and metformin monotherapy on cardiovascular events in type 2 diabetes mellitus: a cohort study. Ann Intern Med2012;
157(9).
20.
StuartEA.Matching methods for causal inference: a review and a look forward.Stat Sci2010;
25(1): 1–21.
21.
ImbensGWRubinDB.Causal inference for statistics, social, and biomedical sciences: an introduction.
Cambridge, UK:
Cambridge University Press, 2015.
22.
RosenbaumPRRubinDB.The central role of the propensity score in observational studies for causal effects. Biometrika1983;
70: 41–55.
23.
ImbensGW.The role of the propensity score in estimating dose-response functions. Biometrika2000;
87: 706–710.
24.
ImaiKvan DykDA.Causal inference with general treatment regimes: generalizing the propensity score. J Am Stat Assoc2004;
99: 854–866.
25.
McCaffreyDFGriffinBAAlmirallD, et al.
A tutorial on propensity score estimation for multiple treatments using generalized boosted models. Stat Med2013;
32: 3388–3414.
26.
YoshidaKHernandez-DiazSSolomonDH, et al.
Matching weights to simultaneously compare three treatment groups: comparison to three-way matching.Epidemiology2017;
28: 387–395.
27.
LiFLiF. Propensity score weighting for causal inference with multi-valued treatments. Ann Appl Stat2019. arXiv: 1808.05339.
28.
LopezMJGutmanR.Estimation of causal effects with multiple treatments: a review and new ideas. Stat Sci2017;
32: 432–454.
29.
RassenJASolomonDHGlynnRJ, et al.
Simultaneously assessing intended and unintended treatment effects of multiple treatment options: a pragmatic matrix design. Pharmacoepidemiol Drug Safety2011;
20: 675–683.
30.
YangSImbensGWCuiZ, et al.
Propensity score matching and subclassification in observational studies with multi-level treatment. Biometrics2016;
72: 1055–1065.
31.
ScotinaADGutmanR.Matching algorithms for causal inference with multiple treatments.Stat Med2019;
38: 3139–3167.
32.
Scotina ADBeaudoin FLGutmanR.Matching estimators for causal effects of multiple treatments.Stat Methods Med Res2020;
29: 1051–1066.
33.
RubinDB.Bayesian inference for causal effects: the role of randomization. Ann Stat1978;
6: 34–58.
34.
GutmanRRubinDB.Estimation of causal effects of binary treatments in unconfounded studies.Stat Med2015;
34: 3381–3398.
35.
RubinDB.Randomization analysis of experimental data: the Fisher randomization test comment. J Am Stat Assoc1980;
75: 591–593.
36.
FleissJLLevinBPaikMC.Statistical methods for rates and proportions. 3rd ed.
Hoboken, NJ:
Wiley, 2003.
HoDEImaiKKingG, et al.
Matching as nonparametric preprocessing for reducing model dependence in parametric causal inference. Political Anal2007;
15: 199–236.
39.
AustinPCSmallDS.The use of bootstrapping when using propensity-score matching without replacement: a simulation study.Stat Med2014;
33: 4306–4319.
40.
AbadieAImbensGW.On the failure of the bootstrap for matching estimators. Econometrica2008;
76: 1537–1557.
41.
FengPZhouXZouQ, et al.
Generalized propenstiy score for estimating the average treatment effect of multiple treatments. Stat Med2011;
31: 681–697.
42.
GutmanRRubinDB.Estimation of causal effects of binary treatments in unconfounded studies with one continuous covariate.Stat Meth Med Res2017;
26: 1199–1215.
43.
RubinDB.The design versus the analysis of observational studies for causal effects: parallels with the design of randomized trials.Stat Med2007;
26: 20–36.
44.
RubinDB.Statistical inference for causal effects, with emphasis on applications in epidemiology and medical statistics. Handb Stat2007;
27: 28–63.
45.
GelmanACarlinJBSternHS, et al. Bayesian Data Anal. 3rd ed.London, UK:
Chapman and Hall/CRC, 2013.
46.
AndridgeRRLittleRJA.A review of hot deck imputation for survey non-response.Int Stat Rev2010;
78: 40–64.
47.
TuCJiaoSKohW. Comparison of clustering algorithms on generalized propensity score in observational studies: a simulation study, 2013.
48.
RubinDBSchenkerN.Multiple imputation for interval estimation from simple random samples with ignorable nonresponse. J Am Stat Assoc1986;
81: 366–374.
49.
RubinDB.The Bayesian bootstrap. Ann Stat1981;
9: 130–134.
50.
LavoriPW.A multiple imputation strategy for clinical trials with truncations of patient data.Stat Med1995;
14: 1913–1925.
51.
AzzaliniACapitanioA.Distributions generated by perturbation of symmetry with emphasis on a multivariate skew t-distribution. J R Stat Soc: Ser B (Stat Methodol)2003;
65: 367–389.
52.
CangulMZChretienYRGutmanR, et al.
Testing treatment effects in unconfounded studies under model misspecification: logistic regression, discretization, and their combination.Stat Med2009;
28: 2531–2551.
53.
KhanNFHarrisonSERosePW.Validity of diagnostic coding within the general practice research database: a systematic review.Brit J Gen Pract2010;
60: e128–e136.
54.
EkströmNSvenssonAMiftarajM, et al.
Cardiovascular safety of glucose-lowering agents as add-on medication to metformin treatment in type 2 diabetes: report from the Swedish National Diabetes Register.Diabet Obes Metabol2016;
18: 990–998.
55.
SuYGelmanAHillJ, et al.
Multiple imputation with diagnostics (mi) in R: opening windows into the black box. J Stat Softw2011;
45: 1–31.
56.
RubinDB.Multiple imputation for nonresponse in surveys.
Hoboken, NJ:
Wiley, 1987.
57.
HarelO.Inferences on missing information under multiple imputation and two-stage multiple imputation. Stat Methodol2007;
4: 75–89.
58.
ChenHCohenPChenS.How big is a big odds ratio? Interpreting the magnitudes of odds ratios in epidemiological studies. Commun Stat – Simul Comput2010;
39: 860–864.
59.
R Core Team. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria, www.R-project.org/ (2018, accessed 16 November 2019).
60.
HastieTTibshiraniRFriedmanJ.The elements of statistical learning: data mining, inference, and prediction.
Berlin, Germany:
Springer Science & Business Media, 2009.
61.
FraleyCRafteryAE.Model-based clustering, discriminant analysis, and density estimation. J Am Stat Assoc2002;
97: 611–631.
62.
GutmanRGutmanH.Monitoring methods in the surgical arena: assumptions and potential pitfalls.Eur J Surg Oncol2013;
39: 575–583.
63.
American Diabetes Association.Standards of medical care in diabetes.Diabetes Care2014;
37: S14–S80.
64.
HiattWRKaulSSmithRJ.The cardiovascular safety of diabetes drugs: insights from the rosiglitazone experience.N Engl J Med2013;
369: 1285–1287.
65.
SmithRJHiattWR.Two new drugs for homozygous familial hypercholesterolemia: managing benefits and risks in a rare disorder.JAMA Intern Med2013;
173: 1491–1492.
66.
O’BrienMJKaramSLWalliaA, et al.
Association of second-line antidiabetic medications with cardiovascular events among insured adults with type 2 diabetes.JAMA Network Open2018;
1: e186125.
67.
ZinmanBWannerCLachinJM, et al.
Empagliflozin, cardiovascular outcomes, and mortality in type 2 diabetes.N Engl J Med2015;
373: 2117–2128.
68.
NealBPerkovicVMahaffeyKW, et al.
Canagliflozin and cardiovascular and renal events in type 2 diabetes. N Engl J Med2017;
377: 644–657.
