[Skip to Navigation]
Sign In
Figure 1.  Flow Diagram of Patients in the Propensity Score–Matched Cohorts
Flow Diagram of Patients in the Propensity Score–Matched Cohorts

aExclusion criteria were not mutually exclusive.

bNo preoperative estimated glomerular filtration rate (eGFR) available in the 180 days before the operation was automatically abstracted from within the electronic health record.

cPatients for whom metformin therapy was not currently recommended.

Figure 2.  Kaplan-Meier Survival Curves for Postoperative Day 90 (A) and at Year 5 (B) in the Propensity Score–Matched Cohorts
Kaplan-Meier Survival Curves for Postoperative Day 90 (A) and at Year 5 (B) in the Propensity Score–Matched Cohorts

Patients who did not die and were lost to follow-up within the postoperative window were censored at their last contact date. Those censored at time 0 had no postoperative encounter with the health care system after hospital discharge.

Figure 3.  Hazard of 90-Day Mortality in Prespecified Subgroups in the Propensity Score–Matched Cohorts
Hazard of 90-Day Mortality in Prespecified Subgroups in the Propensity Score–Matched Cohorts

mARR indicates matched cohort absolute risk reduction; mHR, matched cohort hazard ratios (HRs).

aTotal number (%) of deaths and patients at 90 days and associated absolute risk reduction in the prespecified subgroups within the propensity score–matched cohorts.

bHRs compared those with preoperative metformin prescriptions to those without prescriptions in the propensity score–matched cohorts.

Table 1.  Baseline Characteristics of Patients With Diabetes Before and After Propensity Score Matchinga
Baseline Characteristics of Patients With Diabetes Before and After Propensity Score Matchinga
Table 2.  Absolute Risk Reduction in Mortality and Readmission in the Matched Cohorts
Absolute Risk Reduction in Mortality and Readmission in the Matched Cohorts
1.
Desborough  JP.  The stress response to trauma and surgery.   Br J Anaesth. 2000;85(1):109-117. doi:10.1093/bja/85.1.109 PubMedGoogle ScholarCrossref
2.
World Health Organization; US National Institute of Aging, eds.  Global Health and Ageing. World Health Organization; 2011.
3.
Cowie  CC, Rust  KF, Ford  ES,  et al.  Full accounting of diabetes and pre-diabetes in the U.S. population in 1988-1994 and 2005-2006.   Diabetes Care. 2009;32(2):287-294. doi:10.2337/dc08-1296PubMedGoogle ScholarCrossref
4.
Etzioni  DA, Liu  JH, Maggard  MA, Ko  CY.  The aging population and its impact on the surgery workforce.   Ann Surg. 2003;238(2):170-177. doi:10.1097/01.SLA.0000081085.98792.3d PubMedGoogle Scholar
5.
Kalyani  RR, Saudek  CD, Brancati  FL, Selvin  E.  Association of diabetes, comorbidities, and A1C with functional disability in older adults: results from the National Health and Nutrition Examination Survey (NHANES), 1999-2006.   Diabetes Care. 2010;33(5):1055-1060. doi:10.2337/dc09-1597 PubMedGoogle ScholarCrossref
6.
US Department of Health and Human Services; Administration for Community Living.  Administration of Aging. 2018 Profile of Older Americans. US Department of Health and Human Services; 2018.
7.
Joseph  B, Zangbar  B, Pandit  V,  et al.  Emergency general surgery in the elderly: too old or too frail?   J Am Coll Surg. 2016;222(5):805-813. doi:10.1016/j.jamcollsurg.2016.01.063 PubMedGoogle ScholarCrossref
8.
Franceschi  C, Garagnani  P, Parini  P, Giuliani  C, Santoro  A.  Inflammaging: a new immune-metabolic viewpoint for age-related diseases.   Nat Rev Endocrinol. 2018;14(10):576-590. doi:10.1038/s41574-018-0059-4 PubMedGoogle ScholarCrossref
9.
Jencks  SF, Williams  MV, Coleman  EA.  Rehospitalizations among patients in the Medicare fee-for-service program.   N Engl J Med. 2009;360(14):1418-1428. doi:10.1056/NEJMsa0803563 PubMedGoogle ScholarCrossref
10.
Bhutta  H, Agha  R, Wong  J, Tang  TY, Wilson  YG, Walsh  SR.  Neutrophil-lymphocyte ratio predicts medium-term survival following elective major vascular surgery: a cross-sectional study.   Vasc Endovascular Surg. 2011;45(3):227-231. doi:10.1177/1538574410396590 PubMedGoogle ScholarCrossref
11.
Glance  LG, Kellermann  AL, Osler  TM,  et al.  Hospital readmission after noncardiac surgery: the role of major complications.   JAMA Surg. 2014;149(5):439-445. doi:10.1001/jamasurg.2014.4 PubMedGoogle ScholarCrossref
12.
Hall  DE, Arya  S, Schmid  KK,  et al.  Association of a frailty screening initiative with postoperative survival at 30, 180, and 365 days.   JAMA Surg. 2017;152(3):233-240. doi:10.1001/jamasurg.2016.4219 PubMedGoogle ScholarCrossref
13.
Aguilar  D, Nambi  V.  Taking the air out of oxygen supplementation in individuals with diabetes and acute coronary syndromes. Standards of Medical Care in Diabetes—2019.   Diabetes Care. 2019;42(11):2019-2021. doi:10.2337/dci19-0035 PubMedGoogle ScholarCrossref
14.
Flory  J, Lipska  K.  Metformin in 2019.   JAMA. 2019;321(19):1926-1927. doi:10.1001/jama.2019.3805 PubMedGoogle ScholarCrossref
15.
Cameron  AR, Morrison  VL, Levin  D,  et al.  Anti-inflammatory effects of metformin irrespective of diabetes status.   Circ Res. 2016;119(5):652-665. doi:10.1161/CIRCRESAHA.116.308445 PubMedGoogle ScholarCrossref
16.
Campbell  JM, Bellman  SM, Stephenson  MD, Lisy  K.  Metformin reduces all-cause mortality and diseases of ageing independent of its effect on diabetes control: a systematic review and meta-analysis.   Ageing Res Rev. 2017;40:31-44. doi:10.1016/j.arr.2017.08.003 PubMedGoogle ScholarCrossref
17.
Sardu  C, Paolisso  P, Sacra  C,  et al.  Effects of metformin therapy on coronary endothelial dysfunction in patients with prediabetes with stable angina and nonobstructive coronary artery stenosis: the CODYCE multicenter prospective study.   Diabetes Care. 2019;42(10):1946-1955. doi:10.2337/dc18-2356 PubMedGoogle ScholarCrossref
18.
Miles  JM, Rule  AD, Borlaug  BA.  Use of metformin in diseases of aging.   Curr Diab Rep. 2014;14(6):490. doi:10.1007/s11892-014-0490-4 PubMedGoogle ScholarCrossref
19.
Bannister  CA, Holden  SE, Jenkins-Jones  S,  et al.  Can people with type 2 diabetes live longer than those without? A comparison of mortality in people initiated with metformin or sulphonylurea monotherapy and matched, non-diabetic controls.   Diabetes Obes Metab. 2014;16(11):1165-1173. doi:10.1111/dom.12354 PubMedGoogle ScholarCrossref
20.
Anisimov  VN, Berstein  LM, Popovich  IG,  et al.  If started early in life, metformin treatment increases life span and postpones tumors in female SHR mice.   Aging (Albany NY). 2011;3(2):148-157. doi:10.18632/aging.100273 PubMedGoogle ScholarCrossref
21.
Cabreiro  F, Au  C, Leung  KY,  et al.  Metformin retards aging in C. elegans by altering microbial folate and methionine metabolism.   Cell. 2013;153(1):228-239. doi:10.1016/j.cell.2013.02.035 PubMedGoogle ScholarCrossref
22.
Morris  MS, Deierhoi  RJ, Richman  JS, Altom  LK, Hawn  MT.  The relationship between timing of surgical complications and hospital readmission.   JAMA Surg. 2014;149(4):348-354. doi:10.1001/jamasurg.2013.4064 PubMedGoogle ScholarCrossref
23.
Gani  F, Lucas  DJ, Kim  Y, Schneider  EB, Pawlik  TM.  Understanding Variation in 30-Day Surgical Readmission in the Era of Accountable Care: Effect of the Patient, Surgeon, and Surgical Subspecialties.   JAMA Surg. 2015;150(11):1042-1049. doi:10.1001/jamasurg.2015.2215 PubMedGoogle ScholarCrossref
24.
Fisher  A, Srikusalanukul  W, Fisher  L, Smith  P.  The neutrophil to lymphocyte ratio on admission and short-term outcomes in orthogeriatric patients.   Int J Med Sci. 2016;13(8):588-602. doi:10.7150/ijms.15445 PubMedGoogle ScholarCrossref
25.
von Elm  E, Altman  DG, Egger  M, Pocock  SJ, Gøtzsche  PC, Vandenbroucke  JP; STROBE Initiative.  The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.   Int J Surg. 2014;12(12):1495-1499. doi:10.1016/j.ijsu.2014.07.013 PubMedGoogle ScholarCrossref
26.
Mirza  SK, Deyo  RA, Heagerty  PJ,  et al.  Development of an index to characterize the “invasiveness” of spine surgery: validation by comparison to blood loss and operative time.   Spine (Phila Pa 1976). 2008;33(24):2651-2661. doi:10.1097/BRS.0b013e31818dad07 PubMedGoogle ScholarCrossref
27.
Irons  BK, Minze  MG.  Drug treatment of type 2 diabetes mellitus in patients for whom metformin is contraindicated.   Diabetes Metab Syndr Obes. 2014;7:15-24. doi:10.2147/DMSO.S38753 PubMedGoogle ScholarCrossref
28.
American Diabetes Association.  Standards of Medical Care in Diabetes—2018.   J Clin Appl Res Education. 2018;41(suppl 1).Google Scholar
29.
Navar  AM, Peterson  ED, Steen  DL,  et al.  Evaluation of mortality data from the Social Security Administration death master file for clinical research.   JAMA Cardiol. 2019;4(4):375-379. doi:10.1001/jamacardio.2019.0198 PubMedGoogle ScholarCrossref
30.
Hill  ME, Rosenwaike  I.  The Social Security Administration’s death master file: the completeness of death reporting at older ages.   Soc Secur Bull. 2001-2002;64(1):45-51.PubMedGoogle Scholar
31.
Haukoos  JS, Lewis  RJ.  The propensity score.   JAMA. 2015;314(15):1637-1638. doi:10.1001/jama.2015.13480 PubMedGoogle ScholarCrossref
32.
Shrank  WH, Patrick  AR, Brookhart  MA.  Healthy user and related biases in observational studies of preventive interventions: a primer for physicians.   J Gen Intern Med. 2011;26(5):546-550. doi:10.1007/s11606-010-1609-1 PubMedGoogle ScholarCrossref
33.
Dormuth  CR, Patrick  AR, Shrank  WH,  et al.  Statin adherence and risk of accidents: a cautionary tale.   Circulation. 2009;119(15):2051-2057. doi:10.1161/CIRCULATIONAHA.108.824151 PubMedGoogle ScholarCrossref
34.
Eskin  M, Simpson  SH, Eurich  DT.  Evaluation of healthy user effects with metformin and other oral antihyperglycemia medication users in adult patients with type 2 diabetes.   Can J Diabetes. 2019;43(5):322-328. doi:10.1016/j.jcjd.2018.12.001 PubMedGoogle ScholarCrossref
35.
Austin  PC.  Some methods of propensity-score matching had superior performance to others: results of an empirical investigation and Monte Carlo simulations.   Biom J. 2009;51(1):171-184. doi:10.1002/bimj.200810488 PubMedGoogle ScholarCrossref
36.
Leuven  E, Sianesi  B. PSMATCH2: Stata module to perform full Mahalanobis and propensity score matching, common support graphing, and covariate imbalance testing. Statistical software components S432001. Boston College Department of Economics. Revised February 2018. https://ideas.repec.org/c/boc/bocode/s432001.html
37.
Evans  IVR, Phillips  GS, Alpern  ER,  et al.  Association between the New York sepsis care mandate and in-hospital mortality for pediatric sepsis.   JAMA. 2018;320(4):358-367. doi:10.1001/jama.2018.9071PubMedGoogle ScholarCrossref
38.
Gillespie  BW, Chen  Q, Reichert  H,  et al.  Estimating population distributions when some data are below a limit of detection by using a reverse Kaplan-Meier estimator.   Epidemiology. 2010;21(suppl 4):S64-S70. doi:10.1097/EDE.0b013e3181ce9f08PubMedGoogle ScholarCrossref
39.
Bernaudin  F, Verlhac  S, Peffault de Latour  R,  et al; DREPAGREFFE Trial Investigators.  Association of matched sibling donor hematopoietic stem cell transplantation with transcranial Doppler velocities in children with sickle cell anemia.   JAMA. 2019;321(3):266-276. doi:10.1001/jama.2018.20059 PubMedGoogle ScholarCrossref
40.
Austin  PC.  The use of propensity score methods with survival or time-to-event outcomes: reporting measures of effect similar to those used in randomized experiments.   Stat Med. 2014;33(7):1242-1258. doi:10.1002/sim.5984 PubMedGoogle ScholarCrossref
41.
Wang  R, Lagakos  SW, Ware  JH, Hunter  DJ, Drazen  JM.  Statistics in medicine–reporting of subgroup analyses in clinical trials.   N Engl J Med. 2007;357(21):2189-2194. doi:10.1056/NEJMsr077003PubMedGoogle ScholarCrossref
42.
Zhang  Z, Reinikainen  J, Adeleke  KA, Pieterse  ME, Groothuis-Oudshoorn  CGM.  Time-varying covariates and coefficients in Cox regression models.   Ann Transl Med. 2018;6(7):121-121. doi:10.21037/atm.2018.02.12 PubMedGoogle ScholarCrossref
43.
Tahrani  AA, Varughese  GI, Scarpello  JH, Hanna  FWF.  Metformin, heart failure, and lactic acidosis: is metformin absolutely contraindicated?   BMJ. 2007;335(7618):508-512. doi:10.1136/bmj.39255.669444.AE PubMedGoogle ScholarCrossref
44.
Imam  TH.  Changes in metformin use in chronic kidney disease.   Clin Kidney J. 2017;10(3):301-304. doi:10.1093/ckj/sfx017 PubMedGoogle ScholarCrossref
45.
Horlen  C, Malone  R, Bryant  B,  et al.  Frequency of inappropriate metformin prescriptions.   JAMA. 2002;287(19):2504-2505. doi:10.1001/jama.287.19.2504-a PubMedGoogle ScholarCrossref
46.
Haneuse  S, VanderWeele  TJ, Arterburn  D.  Using the e-value to assess the potential effect of unmeasured confounding in observational studies.   JAMA. 2019;321(6):602-603. doi:10.1001/jama.2018.21554 PubMedGoogle ScholarCrossref
47.
Austin  PC, Stuart  EA.  Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies.   Stat Med. 2015;34(28):3661-3679. doi:10.1002/sim.6607 PubMedGoogle ScholarCrossref
48.
Austin  PC.  Variance estimation when using inverse probability of treatment weighting (IPTW) with survival analysis.   Stat Med. 2016;35(30):5642-5655. doi:10.1002/sim.7084 PubMedGoogle ScholarCrossref
49.
Nathan  DM, Buse  JB, Davidson  MB,  et al; American Diabetes Association; European Association for Study of Diabetes.  Medical management of hyperglycemia in type 2 diabetes: a consensus algorithm for the initiation and adjustment of therapy: a consensus statement of the American Diabetes Association and the European Association for the Study of Diabetes.   Diabetes Care. 2009;32(1):193-203. doi:10.2337/dc08-9025 PubMedGoogle ScholarCrossref
50.
Desai  NR, Shrank  WH, Fischer  MA,  et al.  Patterns of medication initiation in newly diagnosed diabetes mellitus: quality and cost implications.   Am J Med. 2012;125(3):302.e1-302.e7. doi:10.1016/j.amjmed.2011.07.033 PubMedGoogle ScholarCrossref
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
    April 8, 2020

    Association Between Preoperative Metformin Exposure and Postoperative Outcomes in Adults With Type 2 Diabetes

    Author Affiliations
    • 1Department of Surgery, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania
    • 2Clinical Research, Investigation, and Systems Modeling of Acute Illness Center, Pittsburgh, Pennsylvania
    • 3Clinical Analytics, UPMC Health Services Division, Pittsburgh, Pennsylvania
    • 4Department of Epidemiology, University of Pittsburgh School of Public Health, Pittsburgh, Pennsylvania
    • 5Department of Critical Care Medicine, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania
    • 6Department of Endocrinology and Metabolism, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania
    • 7Veterans Affairs Pittsburgh Healthcare System, Pittsburgh, Pennsylvania
    • 8Division of Pulmonary, Allergy and Critical Care Medicine, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania
    • 9Department of Emergency Medicine, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania
    JAMA Surg. 2020;155(6):e200416. doi:10.1001/jamasurg.2020.0416
    Key Points

    Question  Are preoperative metformin prescriptions associated with a lower risk of postoperative mortality and readmission among patients with type 2 diabetes who underwent a major surgical intervention?

    Findings  In this cohort study of 5460 patients with diabetes who had a major surgical procedure, preoperative prescriptions of metformin were associated with a statistically significant decrease in the risk of 90-day mortality as well as 30- and 90-day readmission compared with no such prescriptions.

    Meaning  Findings from this study suggest that preoperative metformin prescriptions may be associated with decreased postoperative mortality and readmission compared with no preoperative exposure to this medication, but further research is needed to ascertain if this relationship is causal.

    Abstract

    Importance  Adults with comorbidity have less physiological reserve and an increased rate of postoperative mortality and readmission after the stress of a major surgical intervention.

    Objective  To assess postoperative mortality and readmission among individuals with diabetes with or without preoperative prescriptions for metformin.

    Design, Setting, and Participants  This cohort study obtained data from the electronic health record of a multicenter, single health care system in Pennsylvania. Included were adults with diabetes who underwent a major operation with hospital admission from January 1, 2010, to January 1, 2016, at 15 community and academic hospitals within the system. Individuals without a clinical indication for metformin therapy were excluded. Follow-up continued until December 18, 2018.

    Exposures  Preoperative metformin exposure was defined as 1 or more prescriptions for metformin in the 180 days before the surgical procedure.

    Main Outcomes and Measures  All-cause postoperative mortality, hospital readmission within 90 days of discharge, and preoperative inflammation measured by the neutrophil to leukocyte ratio were compared between those with and without preoperative prescriptions for metformin. The corresponding absolute risk reduction (ARR) and adjusted hazard ratio (HR) with 95% CI were calculated in a propensity score–matched cohort.

    Results  Among the 10 088 individuals with diabetes who underwent a major surgical intervention, 5962 (59%) had preoperative metformin prescriptions. A total of 5460 patients were propensity score–matched, among whom the mean (SD) age was 67.7 (12.2) years, and 2866 (53%) were women. In the propensity score–matched cohort, preoperative metformin prescriptions were associated with a reduced hazard for 90-day mortality (adjusted HR, 0.72 [95% CI, 0.55-0.95]; ARR, 1.28% [95% CI, 0.26-2.31]) and hazard of readmission, with mortality as a competing risk at both 30 days (ARR, 2.09% [95% CI, 0.35-3.82]; sub-HR, 0.84 [95% CI, 0.72-0.98]) and 90 days (ARR, 2.78% [95% CI, 0.62-4.95]; sub-HR, 0.86 [95% CI, 0.77-0.97]). Preoperative inflammation was reduced in those with metformin prescriptions compared with those without (mean neutrophil to leukocyte ratio, 4.5 [95% CI, 4.3-4.6] vs 5.0 [95% CI, 4.8-5.3]; P < .001). E-value analysis suggested robustness to unmeasured confounding.

    Conclusions and Relevance  This study found an association between metformin prescriptions provided to individuals with type 2 diabetes before a major surgical procedure and reduced risk-adjusted mortality and readmission after the operation. This association warrants further investigation.

    Introduction

    A surgical procedure imparts a complex, systematic stress to the body, initiating an inflammatory response that alters homeostatic function.1 Type 2 diabetes affects nearly 1 in 3 Americans older than 65 years,2,3 and nearly 50% of individuals with type 2 diabetes require a surgical intervention throughout their lifetime.4 Many people with diabetes also have kidney and/or cardiovascular disease,5 which can reduce physiological reserve,6,7 impair inflammatory resolution and restorative mechanisms,7-10 and increase postoperative morbidity and mortality.11,12

    Metformin, the most commonly prescribed noninsulin medication for type 2 diabetes,13,14 has anti-inflammatory properties independent of its role in glycemic control15 and reduces cardiovascular disease in individuals with diabetes or prediabetes.16,17 Metformin is associated with a decrease in chronic inflammation16,18 and may be associated with an increase in lifespan as demonstrated in multiple preclinical and large retrospective clinical studies.19-21 The association between preoperative metformin exposure and postsurgical outcomes for people with diabetes remains unexplored.

    Therefore, we sought to understand the association between metformin and risk-adjusted surgical outcomes. We hypothesized that preoperative metformin ameliorates the physiological stress of a surgical intervention and improves postoperative mortality and readmission in this patient population.

    Methods

    Metformin has pleiotropic and anti-inflammatory properties that may decrease the cellular inflammation and organ damage that accumulate overtime, independent of its effect on blood glucose.8,15 Comorbid illness may be reduced and lifespan extended in individuals with diabetes who use metformin.16,18 Thus, we studied a cohort of patients with diabetes, with and without preoperative exposure to metformin, who were at an increased risk for hospital readmission and mortality after a major surgical intervention. We excluded those without a clinical indication for metformin therapy (eg, kidney insufficiency, type 1 diabetes). To account for indication bias, we used propensity score matching to determine the risk-adjusted association between preoperative metformin prescription and both postoperative mortality and readmission, a surrogate for postoperative complications.9,22,23 We compared the preoperative neutrophil to leukocyte ratio (NLR), a simple index of systemic inflammation, between those with and without metformin prescriptions.10,24 The University of Pittsburgh Institutional Review Board approved this study and waived informed consent because data were deidentified. All reporting followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.25

    Study Design, Population, and Setting

    This retrospective cohort study included adults (aged ≥18 years) with diabetes who underwent a major surgical intervention from January 1, 2010, to January 1, 2016, at 15 community and academic hospitals within a single health care system in Pennsylvania. Follow-up lasted until December 18, 2018. Inpatient data were obtained from the electronic health record (EHR) of the health care system and were linked to its outpatient electronic records.

    Clinical Data Collection

    Outpatient-encounter data included medications; comorbidities with International Classification of Diseases, Ninth Revision, Clinical Modification or International Classification of Diseases, Tenth Revision, Clinical Modification codes, including diabetes; smoking status; laboratory values (eg, estimated glomerular filtration rate [eGFR], hemoglobin, hemoglobin A1c, and NLR); provider appointments; and postdischarge mortality (eAppendix and eTable 1 in the Supplement). Inpatient-encounter data included patient demographics (eg, age, sex, and race/ethnicity), insurance coverage, index surgical intervention, American Society of Anesthesiologists score and associated emergent status for the index operation, surrogate markers of operative complexity (ie, estimated intraoperative blood loss and total procedure time),26 length of stay, and discharge disposition.

    We defined a major surgical procedure as one that required both general anesthesia and a postoperative hospital admission. Using Current Procedural Terminology codes, we categorized surgical procedures into 1 of 6 high-volume specialties (eg, general, cardiothoracic, orthopedic, neurologic, vascular, or other). General and orthopedic surgical procedures (ie, laparoscopy, robotic, and arthroscopy) were identified as minimally invasive. A manual medical record review was performed to minimize the missing data and to validate patient characteristics associated with automated EHR data extraction (eAppendix in the Supplement).

    Metformin Exposure

    We defined preoperative metformin exposure as 1 or more prescriptions for metformin in the 180 days before the surgical procedure and as inclusion of metformin on the active medication list during the most recent preoperative encounter before the surgical procedure. Patients with no previous metformin prescription and no evidence of including metformin on their medication lists were considered not exposed. More than 1 prescription for the drug was required to strengthen the assumption that prescription status represented active treatment with metformin. We excluded patients without a clear indication for metformin (ie, those with type 1 diabetes). We also excluded patients with no documented eGFR in the 180 days before the surgical procedure or with a most recent eGFR of less than 30, which is a relative contraindication to metformin therapy.27,28 We defined postoperative metformin exposure as a documented prescription or as both active and physician-reviewed medication for any health care system encounter in the 365 days after the surgical procedure.

    Outcomes

    The primary outcome was 90-day postoperative mortality. The secondary outcomes were 30-day mortality, 5-year survival, 30- and 90-day postdischarge readmission, and the preoperative inflammatory marker NLR.10,24

    We assessed vitality status after the initiation of the operative intervention using 2 files29: (1) the Social Security Death Index, which is updated in the EHR monthly,30 and (2) mandated EHR documentation of deaths that occur within the health care system (eg, nursing or rehabilitation facilities, emergency departments, and acute care hospitals) (eAppendix in the Supplement). Outcomes were abstracted through December 18, 2018.

    Statistical Analysis

    We conducted the analysis with the following steps: (1) comparison of descriptive data from patients with and without metformin prescriptions; (2) a priori identification of preoperative variables that could confound outcomes; (3) determination of risk-adjusted estimates for outcomes using propensity score matching; (4) performance of sensitivity analyses to test the heterogeneity of treatment effect, understand potential unmeasured confounding, and assess the robustness of the findings in the primary analysis.

    All normally distributed and skewed continuous variables were expressed as mean (SD) or median (interquartile range [IQR]). Categorical variables were expressed as frequencies (%).

    To minimize the potential bias of treatment allocation and confounding,31 we generated a propensity score to estimate by logistic regression the likelihood that patients had preoperative metformin prescriptions. The initial model included all available variables (eg, preoperative demographic characteristics, risk factors, and laboratory values; operative American Society of Anesthesiologists score, specialty, year, and hospital; and perioperative medications). These variables included those representing the health habits of patients who received metformin prescriptions that may capture a so-called healthy user effect, such as colonoscopy compliance,32,33 primary care physician appointments, and statin and aspirin therapy (eAppendix in the Supplement).32-34

    To generate a matched population among those who did and did not have preoperative metformin prescriptions, we used a greedy 1:1 matching algorithm. We used calipers of 0.01 on the propensity score scale and 1:1 sampling without replacement.35,36 All patient characteristics included in generating and distributing the propensity scores were compared before and after propensity score–matching using a standardized mean difference. A threshold less than 0.1 was considered acceptable.31,35,37

    On the propensity score–matched cohort, we used a 2-sided t test to compare preoperative NLR, estimated intraoperative blood loss, total procedure time, and 1-year postoperative metformin prescription number. χ2 Testing compared postoperative prescription status (yes or no). We applied the Kaplan-Meier estimator with a reversed scale to quantify the median duration of postoperative follow-up.38

    The absolute risk reduction (ARR) in mortality and readmission was calculated. Differences between treatment groups were compared with McNemar testing.39 In addition, we generated Kaplan-Meier survival and freedom from readmission curves compared with log-rank testing, stratified by matched pairs.40 We used a univariable Cox proportional hazards regression model with the robust variance estimator to calculate the hazard ratio (HR) for mortality and sub-HR for readmission, with mortality as a competing risk.40 The heterogeneity of the treatment effect of preoperative metformin prescriptions on 90-day mortality among prespecified subpopulations was assessed according to the interaction between metformin and surgical specialty (eg, general, cardiothoracic, orthopedic, neurologic, vascular, or other), emergent operative status (yes or no), and operative hospital location (eAppendix in the Supplement).41

    A 2-sided P < .05 was considered statistically significant. Proportional hazards assumptions were tested with Schoenfeld residuals over time.42 All statistical analyses were conducted in Stata, version 15.0 (StataCorp LLC), on November 1, 2018.

    Sensitivity Analysis

    The robustness of these findings was assessed in multiple sensitivity analyses. First, we minimized the confounding by indication by selecting a group of patients who were equally as likely to receive as not receive preoperative metformin prescriptions. We restricted the cohort to patients for whom metformin was no longer contraindicated owing to recent US Food and Drug Administration and societal guideline changes: those with congestive heart failure and kidney dysfunction (ie, eGFR of 30-60).13,43-45 We then calculated the 90-day mortality ARR on the basis of metformin prescriptions. Second, we explored the potential for unmeasured confounding between preoperative prescriptions and 90-day mortality by calculating E-values.46 The E-value quantifies the required magnitude of an unmeasured confounder that could negate the observed association between metformin and 90-day mortality.

    To ascertain whether the results were sensitive to the matching method, we performed 2 additional sensitivity analyses using the full cohort. First, multivariable modeling of the association between preoperative metformin prescriptions and both mortality and readmission with mortality as a competing risk was conducted using Cox proportional hazards regression model. Second, we calculated the adjusted 5-year survival average treatment effect of both the treated and untreated patients using inverse probability of treatment weighting of the propensity scores.47,48

    Results
    Patient Population

    We identified 15 894 individuals with diabetes who underwent a major surgical procedure. Of these patients, 8583 (54%) had preoperative metformin prescriptions. The full and final cohort included 10 088 patients who met all inclusion and no exclusion criteria, of whom 5962 (59%) had preoperative metformin prescriptions. A total of 5460 patients (54%) were successfully matched (Figure 1). After propensity score matching, all baseline patient characteristics were similar across groups, propensity scores were equally distributed, and percent bias was minimized (Table 1; eFigures 1 and 2 in the Supplement). The extent of missing laboratory values and the comparison between automated and clinically adjudicated covariates are presented in the eAppendix and eTables 2 and 3 in the Supplement.

    Metformin Exposure, Operative Characteristics, and Diabetes Management

    Among the 5460 matched patients, the mean (SD) age was 67.7 (12.2) years, 2866 (53%) were women, 4867 (89%) were white individuals, and the mean (SD) body mass index (calculated as weight in kilograms divided by height in meters squared) was 33.5 (7.7). More than one-third of patients (1893 [35%]) were treated with insulin and had adequate glucose control (mean [SD] hemoglobin A1c of 7.1 [1.7] mmol/mol). Unmatched patients (4628) were younger, had a higher mean body mass index, and had less insulin use (eTable 4 in the Supplement).

    Most of the surgical interventions were general (1618 of 5460 [30%]) or orthopedic (1598 [29%]), of which 774 (24%) used a minimally invasive approach. Across all surgical specialties, no differences were observed in mean (SD) estimated intraoperative blood loss beween patients with or without preoperative metformin prescriptions (122 [288] mL vs 119 [292] mL; P = .67). Patients with preoperative metformin prescriptions had a longer mean (SD) procedural duration than those without (175 [107] minutes vs 169 [101] minutes; P = .02).

    Each treatment group had a median follow-up of more than 4.5 years (4.7 [95% CI, 4.6-4.9] vs 4.6 [95% CI, 4.5-4.7]). A total of 2502 patients with preoperative metformin prescriptions (95%) had evidence of receiving prescriptions in postoperative year 1. Of these patients, 1708 (72%) had at least 1 EHR-derived prescription with a mean (SD) of 4.4 (1.9) in postoperative year 1. Of 2727 patients without preoperative prescriptions, 218 (8%) had evidence of new prescriptions after the surgical procedure.

    Association Between Preoperative Prescriptions and Outcomes

    Among 2729 patients with preoperative metformin prescriptions, 49 (2%) deaths had occurred by postoperative day 30; 89 (3%) and 360 (13%) deaths had occurred by day 90 and year 5, respectively, whereas 56 deaths (2%) by day 30, 124 (5%) by day 90, and 471 (17%) by year 5 were noted among 2727 patients without prescriptions. Metformin prescriptions were associated with a statistically significant ARR in mortality by day 90 but not day 30 (ARR, 1.28%; 95% CI, 0.26-2.31) (Table 2). Patients with preoperative metformin prescriptions had a lower hazard for mortality by day 90 (HR, 0.72; 95% CI, 0.55-0.95; P = .02) and year 5 (HR, 0.74; 95% CI, 0.65-0.85; P = .001) (Table 2; Figure 2). The association between metformin and the reduced risk of 90-day mortality was statistically significant for those who underwent a general surgical intervention (matched ARR, 2.02; 95% CI, –0.18 to 3.85) and was similar across other surgical specialties, excluding neurosurgery (matched ARR, –0.89; 95% CI, –3.83 to 2.05), and for both emergent (matched ARR, 0.38; 95% CI, –2.25 to 9.81) and elective (matched ARR, 1.00; 95% CI, 0.02-1.93) operations (Figure 3). Rates of 90-day mortality per operative hospital are included in the eAppendix in the Supplement (eTable 5 in the Supplement).

    Hospital readmissions among those with preoperative metformin prescriptions (n = 2729) were observed by postdischarge days 30 and 90 (304 [11%] and 538 [20%]), whereas among those without prescriptions (n = 2727), 361 readmissions (13%) occurred by day 30 and 614 (23%) by day 90. Statistically significant ARRs in readmission by both days 30 (ARR, 2.09% [95% CI, 0.35-3.82]) and 90 (ARR, 2.78% [95% CI, 0.62-4.95]) were also observed (Table 2; eFigure 3 in the Supplement). Preoperative metformin was also associated with a reduced subhazard of readmission by day 30 (sub-HR, 0.84; 95% CI, 0.72-0.98) and day 90 (sub-HR, 0.86; 95% CI, 0.77-0.97), with death as a competing risk (Table 2).

    Before the operation, patients with metformin prescriptions had statistically significantly lower mean preoperative NLR compared with those without metformin prescriptions (4.5 [95% CI, 4.3-4.6] vs 5.0 [95% CI, 4.8-5.3]; P < .001).

    Sensitivity Analysis

    The association between preoperative metformin prescriptions and reduced risk of 90-day mortality continued in patients who were potentially equally as likely to receive as not receive prescriptions (ie, congestive heart failure ARR, 2.3% [95% CI, 0.02-1.32; P = .32]; eGFR, 30-60; ARR, 0.6% [95% CI, 0.46-0.89; P = .006]). We generated an E-value to assess the sensitivity to unmeasured confounding. The primary findings were robust, unless an unmeasured confounder existed with a lower relative risk of 90-day mortality, with an ARR greater than 2.4% or an HR lower than 0.47 (eAppendix in the Supplement).

    In the full cohort (N = 10 088), multivariable regression analysis similarly demonstrated that metformin was associated with a reduced hazard for both 90-day (adjusted HR, 0.77; 95% CI, 0.59-0.99) (eTable 6 in the Supplement) and 5-year (adjusted HR, 0.80; 95% CI, 0.70-0.91) mortality and for both 30-day (adjusted HR, 0.83; 95% CI, 0.73-0.97) and 90-day (adjusted HR, 0.86; 95% CI, 0.78-0.95) readmission, with mortality as a competing risk (Table 2). Using inverse probability weighting, metformin prescriptions were associated with a 22% increase in 5-year survival, compared with no prescriptions, without generating a mean treatment effect of 0.34 years (95% CI, 0.12-0.56; P = .003).

    Discussion

    In this retrospective, propensity score–matched cohort study, preoperative metformin prescriptions provided to individuals with diabetes were associated with a lower risk-adjusted 90-day mortality after a major surgical procedure, compared with no metformin prescription. This association was similarly found for the secondary outcomes of postoperative 30- and 90-day hospital readmissions and 5-year mortality and in additional models to control for indication bias.

    Previous studies demonstrated that treatment with metformin was associated with a decrease in all-cause mortality.16,18,19 Individuals with diabetes treated with metformin showed fewer age-related chronic diseases such as cardiovascular disease, kidney dysfunction, and cancer.16-18 These associations may reflect the antiaging properties of metformin against the onset of disease or diabetes-associated complications. This study extends these findings by demonstrating that preoperative metformin prescriptions were associated with a reduction in postoperative mortality and readmission, a surrogate for postoperative complications,9,22 and with long-term mortality.

    These findings have several explanations. First, propensity score matching created a consistent incidence of measured preoperative comorbidities, independent of the treatment group. This process potentially mitigated the true effects of metformin, yet preoperative metformin prescriptions were associated with an ARR in postoperative 90-day mortality as well as 30- and 90-day readmission. In this immediate perioperative period, the development of new, clinically significant disease states was unlikely. It suggests that metformin has ongoing physiological implications beyond its known association with chronic diseases, health care use, and glycemic control. Second, consistent with the findings in previous research, metformin was associated with lower NLR before the operation, a marker of systemic inflammation.15,17 As in other cohorts, lower preoperative inflammation can estimate superior perioperative morbidity and mortality.10,24 Therefore, the pleiotropic advantages of metformin in the present cohort may positively modulate the inflammatory response to surgical procedures to improve outcomes, although this potential remains unproven.

    These associations could be derived from the spurious implications of confounding. In particular, metformin is the first-line pharmacologic therapy recommended by the American Diabetes Association.28 Yet, metformin therapy is not appropriate for all patients with diabetes.27,49 The therapeutic association between metformin and postoperative outcomes may exist because those given a metformin prescription are both mutually exclusive and inherently different from those without prescriptions. The contraindications of metformin therapy, however, have evolved, resulting in prescribing variability among clinicians.27,28,49,50 Specifically, individuals with congestive heart failure and kidney disease are potential subpopulations for whom clinicians may be as likely to prescribe metformin as they would an alternative therapy.45 We, therefore, verified that, in this restricted cohort, the reduction in postoperative mortality supported by metformin, as opposed to unmeasured confounders, was associated with improved outcomes. In the subgroup analysis, metformin-associated postoperative mortality improvements did not differ across surgical specialty or the emergent status of the operation. Therefore, the advantages are likely not disease specific, but instead the pleiotropic properties may modulate the stress response generated by a major surgical intervention or confer consistently good outcomes, regardless of the surgical procedure.

    Limitations

    This study has multiple limitations. First, despite these precautions and an excellent balance in the propensity score–matched cohorts in this study, the potential for residual confounding exists, as with all retrospective analyses. Specifically, the surgical approach (ie, laparoscopic vs open) was distributed evenly across the treatment groups, yet the surgical stress experienced by each individual could not be quantified. We used the E-value sensitivity analysis to quantify the potential implications of unmeasured confounders and found that an unmeasured confounder was unlikely to explain the entirety of the treatment effect. Second, patients who received the standard of care for established disease states may also interact with the health care system differently compared with those who received alternative therapies, which we described as the healthy user effect and quantified and controlled for in this analysis.32,33 Third, all variables were extracted retrospectively from the EHR of a multicenter, single health care system. Consequently, those who received out-of-system prescriptions were at an increased risk for both differential misclassification and underreported information. Fourth, death and readmission rates were likely underrepresented. Vitality status is captured by the Social Security Death Index and supported by EHR documentation of deaths occurring within the health care system. Therefore, both vitality status and readmission are dependent on health care system use, creating a potential bias for underreporting in those who obtain a large portion of their health care services outside of the system.

    Fifth, although we found little evidence of treatment crossover after the surgical intervention, the preoperative duration and dose of metformin were unknown. Without this information, neither a temporal nor dose response relationship can support the causal inference. Sixth, the matched cohort consisted of primarily older individuals with diabetes who had substantial preoperative comorbidities, limiting the potential generalizability to all individuals with diabetes. Seventh, this analysis of preoperative metformin exposure and postoperative outcomes included a diverse selection of many surgical interventions. The study was not powered to adequately evaluate this association for individual procedures. Therefore, the effectiveness of perioperative metformin in reducing postoperative morbidity and mortality in individuals without diabetes warrants an investigation with a prospective randomized clinical trial.

    Conclusions

    This cohort study suggested that preoperative metformin prescriptions may be associated with reduced risk-adjusted postoperative mortality and readmission among individuals with diabetes who underwent a major surgical procedure.

    Back to top
    Article Information

    Accepted for Publication: February 7, 2020.

    Corresponding Author: Christopher W. Seymour, MD, MSc, Clinical Research, Investigation, and Systems Modeling of Acute Illness Center, 3520 Fifth Ave, Keystone Building, Ste 100, Pittsburgh, PA 15213 (seymourc@pitt.edu).

    Published Online: April 8, 2020. doi:10.1001/jamasurg.2020.0416

    Author Contributions: Drs Reitz and Neal had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

    Concept and design: Reitz, Marroquin, Zenati, Tzeng, Billiar, Zuckerbraun, Shapiro, Seymour, Angus, Neal.

    Acquisition, analysis, or interpretation of data: Reitz, Marroquin, Zenati, Kennedy, Korytkowski, Koscumb, Newhouse, Martinez Garcia, Vates, Billiar, Zuckerbraun, Simmons, Seymour, Angus, Rosengart, Neal.

    Drafting of the manuscript: Reitz, Koscumb, Newhouse, Angus, Rosengart, Neal.

    Critical revision of the manuscript for important intellectual content: Reitz, Marroquin, Zenati, Kennedy, Korytkowski, Tzeng, Newhouse, Martinez Garcia, Vates, Billiar, Zuckerbraun, Simmons, Shapiro, Seymour, Angus, Rosengart, Neal.

    Statistical analysis: Reitz, Zenati, Kennedy, Newhouse, Martinez Garcia, Seymour, Angus, Rosengart.

    Obtained funding: Shapiro, Neal.

    Administrative, technical, or material support: Marroquin, Koscumb, Newhouse, Vates, Billiar, Seymour, Angus, Neal.

    Supervision: Marroquin, Zenati, Korytkowski, Tzeng, Simmons, Shapiro, Seymour, Angus, Rosengart, Neal.

    Conflict of Interest Disclosures: Dr Seymour reported receiving grants from the NIH during the conduct of the study. Dr Angus reported receiving personal fees from Bristol-Myers Squibb, Bayer AG, and Ferring Pharmaceuticals Inc; receiving stock options from Alung Technologies Inc outside the submitted work; and holding a pending patent to Selepressin—Compounds, Compositions, and Methods for Treating Sepsis and a pending patent to Proteomic Biomarkers of Sepsis in Elderly Patients. Dr Neal reported receiving grants and personal fees from Janssen Pharmaceuticals; grants, personal fees, and nonfinancial support from Haemonetics; personal fees from CSL Behring; and grants from Noveome and Accriva Diagnostics outside the submitted work. No other disclosures were reported.

    Funding/Support: This study was funded internally by the UPMC and was supported in part by grant 5T32HL0098036 from the National Heart, Lung, and Bone Institute (Dr Reitz) and grant R35GM119519 from the NIH (Dr Seymour).

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

    Additional Contributions: We acknowledge the contributions of the patients, families, researchers, clinical staff, and sponsors of the cohort data in this study. The Clinical Analytics department in the UPMC Health Services Division prepared this data set with the support of Biostatistics and Data Management Core at the CRISMA Center in the Department of Critical Care Medicine at the University of Pittsburgh. The Strategies to Promote ResiliencY (SPRY) team contributed to this work. These contributors received no additional compensation, outside of their usual salary, for their work.

    References
    1.
    Desborough  JP.  The stress response to trauma and surgery.   Br J Anaesth. 2000;85(1):109-117. doi:10.1093/bja/85.1.109 PubMedGoogle ScholarCrossref
    2.
    World Health Organization; US National Institute of Aging, eds.  Global Health and Ageing. World Health Organization; 2011.
    3.
    Cowie  CC, Rust  KF, Ford  ES,  et al.  Full accounting of diabetes and pre-diabetes in the U.S. population in 1988-1994 and 2005-2006.   Diabetes Care. 2009;32(2):287-294. doi:10.2337/dc08-1296PubMedGoogle ScholarCrossref
    4.
    Etzioni  DA, Liu  JH, Maggard  MA, Ko  CY.  The aging population and its impact on the surgery workforce.   Ann Surg. 2003;238(2):170-177. doi:10.1097/01.SLA.0000081085.98792.3d PubMedGoogle Scholar
    5.
    Kalyani  RR, Saudek  CD, Brancati  FL, Selvin  E.  Association of diabetes, comorbidities, and A1C with functional disability in older adults: results from the National Health and Nutrition Examination Survey (NHANES), 1999-2006.   Diabetes Care. 2010;33(5):1055-1060. doi:10.2337/dc09-1597 PubMedGoogle ScholarCrossref
    6.
    US Department of Health and Human Services; Administration for Community Living.  Administration of Aging. 2018 Profile of Older Americans. US Department of Health and Human Services; 2018.
    7.
    Joseph  B, Zangbar  B, Pandit  V,  et al.  Emergency general surgery in the elderly: too old or too frail?   J Am Coll Surg. 2016;222(5):805-813. doi:10.1016/j.jamcollsurg.2016.01.063 PubMedGoogle ScholarCrossref
    8.
    Franceschi  C, Garagnani  P, Parini  P, Giuliani  C, Santoro  A.  Inflammaging: a new immune-metabolic viewpoint for age-related diseases.   Nat Rev Endocrinol. 2018;14(10):576-590. doi:10.1038/s41574-018-0059-4 PubMedGoogle ScholarCrossref
    9.
    Jencks  SF, Williams  MV, Coleman  EA.  Rehospitalizations among patients in the Medicare fee-for-service program.   N Engl J Med. 2009;360(14):1418-1428. doi:10.1056/NEJMsa0803563 PubMedGoogle ScholarCrossref
    10.
    Bhutta  H, Agha  R, Wong  J, Tang  TY, Wilson  YG, Walsh  SR.  Neutrophil-lymphocyte ratio predicts medium-term survival following elective major vascular surgery: a cross-sectional study.   Vasc Endovascular Surg. 2011;45(3):227-231. doi:10.1177/1538574410396590 PubMedGoogle ScholarCrossref
    11.
    Glance  LG, Kellermann  AL, Osler  TM,  et al.  Hospital readmission after noncardiac surgery: the role of major complications.   JAMA Surg. 2014;149(5):439-445. doi:10.1001/jamasurg.2014.4 PubMedGoogle ScholarCrossref
    12.
    Hall  DE, Arya  S, Schmid  KK,  et al.  Association of a frailty screening initiative with postoperative survival at 30, 180, and 365 days.   JAMA Surg. 2017;152(3):233-240. doi:10.1001/jamasurg.2016.4219 PubMedGoogle ScholarCrossref
    13.
    Aguilar  D, Nambi  V.  Taking the air out of oxygen supplementation in individuals with diabetes and acute coronary syndromes. Standards of Medical Care in Diabetes—2019.   Diabetes Care. 2019;42(11):2019-2021. doi:10.2337/dci19-0035 PubMedGoogle ScholarCrossref
    14.
    Flory  J, Lipska  K.  Metformin in 2019.   JAMA. 2019;321(19):1926-1927. doi:10.1001/jama.2019.3805 PubMedGoogle ScholarCrossref
    15.
    Cameron  AR, Morrison  VL, Levin  D,  et al.  Anti-inflammatory effects of metformin irrespective of diabetes status.   Circ Res. 2016;119(5):652-665. doi:10.1161/CIRCRESAHA.116.308445 PubMedGoogle ScholarCrossref
    16.
    Campbell  JM, Bellman  SM, Stephenson  MD, Lisy  K.  Metformin reduces all-cause mortality and diseases of ageing independent of its effect on diabetes control: a systematic review and meta-analysis.   Ageing Res Rev. 2017;40:31-44. doi:10.1016/j.arr.2017.08.003 PubMedGoogle ScholarCrossref
    17.
    Sardu  C, Paolisso  P, Sacra  C,  et al.  Effects of metformin therapy on coronary endothelial dysfunction in patients with prediabetes with stable angina and nonobstructive coronary artery stenosis: the CODYCE multicenter prospective study.   Diabetes Care. 2019;42(10):1946-1955. doi:10.2337/dc18-2356 PubMedGoogle ScholarCrossref
    18.
    Miles  JM, Rule  AD, Borlaug  BA.  Use of metformin in diseases of aging.   Curr Diab Rep. 2014;14(6):490. doi:10.1007/s11892-014-0490-4 PubMedGoogle ScholarCrossref
    19.
    Bannister  CA, Holden  SE, Jenkins-Jones  S,  et al.  Can people with type 2 diabetes live longer than those without? A comparison of mortality in people initiated with metformin or sulphonylurea monotherapy and matched, non-diabetic controls.   Diabetes Obes Metab. 2014;16(11):1165-1173. doi:10.1111/dom.12354 PubMedGoogle ScholarCrossref
    20.
    Anisimov  VN, Berstein  LM, Popovich  IG,  et al.  If started early in life, metformin treatment increases life span and postpones tumors in female SHR mice.   Aging (Albany NY). 2011;3(2):148-157. doi:10.18632/aging.100273 PubMedGoogle ScholarCrossref
    21.
    Cabreiro  F, Au  C, Leung  KY,  et al.  Metformin retards aging in C. elegans by altering microbial folate and methionine metabolism.   Cell. 2013;153(1):228-239. doi:10.1016/j.cell.2013.02.035 PubMedGoogle ScholarCrossref
    22.
    Morris  MS, Deierhoi  RJ, Richman  JS, Altom  LK, Hawn  MT.  The relationship between timing of surgical complications and hospital readmission.   JAMA Surg. 2014;149(4):348-354. doi:10.1001/jamasurg.2013.4064 PubMedGoogle ScholarCrossref
    23.
    Gani  F, Lucas  DJ, Kim  Y, Schneider  EB, Pawlik  TM.  Understanding Variation in 30-Day Surgical Readmission in the Era of Accountable Care: Effect of the Patient, Surgeon, and Surgical Subspecialties.   JAMA Surg. 2015;150(11):1042-1049. doi:10.1001/jamasurg.2015.2215 PubMedGoogle ScholarCrossref
    24.
    Fisher  A, Srikusalanukul  W, Fisher  L, Smith  P.  The neutrophil to lymphocyte ratio on admission and short-term outcomes in orthogeriatric patients.   Int J Med Sci. 2016;13(8):588-602. doi:10.7150/ijms.15445 PubMedGoogle ScholarCrossref
    25.
    von Elm  E, Altman  DG, Egger  M, Pocock  SJ, Gøtzsche  PC, Vandenbroucke  JP; STROBE Initiative.  The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.   Int J Surg. 2014;12(12):1495-1499. doi:10.1016/j.ijsu.2014.07.013 PubMedGoogle ScholarCrossref
    26.
    Mirza  SK, Deyo  RA, Heagerty  PJ,  et al.  Development of an index to characterize the “invasiveness” of spine surgery: validation by comparison to blood loss and operative time.   Spine (Phila Pa 1976). 2008;33(24):2651-2661. doi:10.1097/BRS.0b013e31818dad07 PubMedGoogle ScholarCrossref
    27.
    Irons  BK, Minze  MG.  Drug treatment of type 2 diabetes mellitus in patients for whom metformin is contraindicated.   Diabetes Metab Syndr Obes. 2014;7:15-24. doi:10.2147/DMSO.S38753 PubMedGoogle ScholarCrossref
    28.
    American Diabetes Association.  Standards of Medical Care in Diabetes—2018.   J Clin Appl Res Education. 2018;41(suppl 1).Google Scholar
    29.
    Navar  AM, Peterson  ED, Steen  DL,  et al.  Evaluation of mortality data from the Social Security Administration death master file for clinical research.   JAMA Cardiol. 2019;4(4):375-379. doi:10.1001/jamacardio.2019.0198 PubMedGoogle ScholarCrossref
    30.
    Hill  ME, Rosenwaike  I.  The Social Security Administration’s death master file: the completeness of death reporting at older ages.   Soc Secur Bull. 2001-2002;64(1):45-51.PubMedGoogle Scholar
    31.
    Haukoos  JS, Lewis  RJ.  The propensity score.   JAMA. 2015;314(15):1637-1638. doi:10.1001/jama.2015.13480 PubMedGoogle ScholarCrossref
    32.
    Shrank  WH, Patrick  AR, Brookhart  MA.  Healthy user and related biases in observational studies of preventive interventions: a primer for physicians.   J Gen Intern Med. 2011;26(5):546-550. doi:10.1007/s11606-010-1609-1 PubMedGoogle ScholarCrossref
    33.
    Dormuth  CR, Patrick  AR, Shrank  WH,  et al.  Statin adherence and risk of accidents: a cautionary tale.   Circulation. 2009;119(15):2051-2057. doi:10.1161/CIRCULATIONAHA.108.824151 PubMedGoogle ScholarCrossref
    34.
    Eskin  M, Simpson  SH, Eurich  DT.  Evaluation of healthy user effects with metformin and other oral antihyperglycemia medication users in adult patients with type 2 diabetes.   Can J Diabetes. 2019;43(5):322-328. doi:10.1016/j.jcjd.2018.12.001 PubMedGoogle ScholarCrossref
    35.
    Austin  PC.  Some methods of propensity-score matching had superior performance to others: results of an empirical investigation and Monte Carlo simulations.   Biom J. 2009;51(1):171-184. doi:10.1002/bimj.200810488 PubMedGoogle ScholarCrossref
    36.
    Leuven  E, Sianesi  B. PSMATCH2: Stata module to perform full Mahalanobis and propensity score matching, common support graphing, and covariate imbalance testing. Statistical software components S432001. Boston College Department of Economics. Revised February 2018. https://ideas.repec.org/c/boc/bocode/s432001.html
    37.
    Evans  IVR, Phillips  GS, Alpern  ER,  et al.  Association between the New York sepsis care mandate and in-hospital mortality for pediatric sepsis.   JAMA. 2018;320(4):358-367. doi:10.1001/jama.2018.9071PubMedGoogle ScholarCrossref
    38.
    Gillespie  BW, Chen  Q, Reichert  H,  et al.  Estimating population distributions when some data are below a limit of detection by using a reverse Kaplan-Meier estimator.   Epidemiology. 2010;21(suppl 4):S64-S70. doi:10.1097/EDE.0b013e3181ce9f08PubMedGoogle ScholarCrossref
    39.
    Bernaudin  F, Verlhac  S, Peffault de Latour  R,  et al; DREPAGREFFE Trial Investigators.  Association of matched sibling donor hematopoietic stem cell transplantation with transcranial Doppler velocities in children with sickle cell anemia.   JAMA. 2019;321(3):266-276. doi:10.1001/jama.2018.20059 PubMedGoogle ScholarCrossref
    40.
    Austin  PC.  The use of propensity score methods with survival or time-to-event outcomes: reporting measures of effect similar to those used in randomized experiments.   Stat Med. 2014;33(7):1242-1258. doi:10.1002/sim.5984 PubMedGoogle ScholarCrossref
    41.
    Wang  R, Lagakos  SW, Ware  JH, Hunter  DJ, Drazen  JM.  Statistics in medicine–reporting of subgroup analyses in clinical trials.   N Engl J Med. 2007;357(21):2189-2194. doi:10.1056/NEJMsr077003PubMedGoogle ScholarCrossref
    42.
    Zhang  Z, Reinikainen  J, Adeleke  KA, Pieterse  ME, Groothuis-Oudshoorn  CGM.  Time-varying covariates and coefficients in Cox regression models.   Ann Transl Med. 2018;6(7):121-121. doi:10.21037/atm.2018.02.12 PubMedGoogle ScholarCrossref
    43.
    Tahrani  AA, Varughese  GI, Scarpello  JH, Hanna  FWF.  Metformin, heart failure, and lactic acidosis: is metformin absolutely contraindicated?   BMJ. 2007;335(7618):508-512. doi:10.1136/bmj.39255.669444.AE PubMedGoogle ScholarCrossref
    44.
    Imam  TH.  Changes in metformin use in chronic kidney disease.   Clin Kidney J. 2017;10(3):301-304. doi:10.1093/ckj/sfx017 PubMedGoogle ScholarCrossref
    45.
    Horlen  C, Malone  R, Bryant  B,  et al.  Frequency of inappropriate metformin prescriptions.   JAMA. 2002;287(19):2504-2505. doi:10.1001/jama.287.19.2504-a PubMedGoogle ScholarCrossref
    46.
    Haneuse  S, VanderWeele  TJ, Arterburn  D.  Using the e-value to assess the potential effect of unmeasured confounding in observational studies.   JAMA. 2019;321(6):602-603. doi:10.1001/jama.2018.21554 PubMedGoogle ScholarCrossref
    47.
    Austin  PC, Stuart  EA.  Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies.   Stat Med. 2015;34(28):3661-3679. doi:10.1002/sim.6607 PubMedGoogle ScholarCrossref
    48.
    Austin  PC.  Variance estimation when using inverse probability of treatment weighting (IPTW) with survival analysis.   Stat Med. 2016;35(30):5642-5655. doi:10.1002/sim.7084 PubMedGoogle ScholarCrossref
    49.
    Nathan  DM, Buse  JB, Davidson  MB,  et al; American Diabetes Association; European Association for Study of Diabetes.  Medical management of hyperglycemia in type 2 diabetes: a consensus algorithm for the initiation and adjustment of therapy: a consensus statement of the American Diabetes Association and the European Association for the Study of Diabetes.   Diabetes Care. 2009;32(1):193-203. doi:10.2337/dc08-9025 PubMedGoogle ScholarCrossref
    50.
    Desai  NR, Shrank  WH, Fischer  MA,  et al.  Patterns of medication initiation in newly diagnosed diabetes mellitus: quality and cost implications.   Am J Med. 2012;125(3):302.e1-302.e7. doi:10.1016/j.amjmed.2011.07.033 PubMedGoogle ScholarCrossref
    ×