Impact of Smoking and Drinking on Complications After Revision Total Joint Arthroplasty

A Matched Cohort Analysis

Venkatsaiakhil Tirumala, MS; Georges Bounajem, MD; Christian Klemt, PhD; Ameen Barghi, MD; Young-Min Kwon, MD, PhD


J Am Acad Orthop Surg. 2021;29(15):e769-e781. 

In This Article



We retrospectively evaluated a total of 4,302 consecutive patients who underwent either revision THA or TKA at our tertiary referral institution between 2010 and 2017 after approval from the Institutional Review Board. We classified revision THA as those that included revisions of (1) both acetabular and femoral implants, (2) either acetabular or femoral implant, and (3) acetabular liner and/or femoral head only. Revision TKAs were classified as those that included revisions of (1) both femoral and tibial implants, (2) either femoral or tibial implant, and (3) isolated tibial polyethylene implant exchanges. From this total, we excluded patients if they had previous revisions or manipulations of the same joint, any operation within three months before their revision TJA, incomplete perioperative data and/or social history, or were lost to a minimum follow-up of 2 years for reasons besides mortality or re-revision. These exclusions align with those in the existing TJA literature.[19,20] After these exclusions, a total of 3,336 patients remained in the cohort.

As obtained from chart review, patients were divided into four cohorts based on their social history of substance usage: (1) nonusers (patients who neither smoke nor drink), (2) smokers (smokers only), (3) drinkers (drinkers only), and (4) concurrent users (simultaneous smokers and drinkers). Drinkers were classified as those who, on average, consume at least a "moderate" amount of alcohol per day (1 drink/d for women and 2 drinks/d for men) per the US Department of Health and Human Services guidelines. Smokers were considered those who participated in daily light (1 to 4 cigarettes/d) to heavy (≥23 cigarettes/d) smoking.[21] Active users were those who smoked and/or drank within at least 12 months of their revision TJA. Nonusers were (1) those who abstained from smoking and/or drinking for at least 12 months before their revision surgery or (2) those with no history of smoking and/or drinking.[17] Drinking and smoking status at the time of presentation for revision TJA was captured from the electronic medical record (retrospective chart review), which is reconciled on the day of presentation for revision surgery. We then retrospectively collected demographics including age, sex, race, comorbid conditions, the American Society of Anesthesiologists (ASA) score, and other predictors of outcomes related to TJA. The Charlson comorbidity index was calculated for each patient.

Propensity Score Cohort Matching

Propensity score matching is an effective tool to reduce selection bias in nonrandomized studies,[22] especially when the number of potential confounders is large and dimensionality issues arise. For the multicohort matching in this study, generalized propensity scores were determined for each patient, which was defined relative to the conditional probability of the patient being in one of the four cohorts (dependent variable). These generalized propensity scores were obtained via multinomial regression analysis on the set of observed covariates (the independent variables) which included (1) age, (2) sex, (3) body mass index, (4) ASA score, and (5) indications for revision. These covariates were assumed to be related to both the cohort membership and the outcomes of interest. The regression analysis resulted in four different propensity scores, which were the estimated predicted probabilities of assignment to one of the four cohorts. A generalized overlap weighting scheme was then applied to the distribution of independent propensity scores to check and ensure that patients after matching have approximately the same probabilities of being assigned to all other cohorts.[22] Patients with extreme propensity scores or those who could not be matched into the four treatment groups were trimmed. This ensured balanced baseline characteristics among cohorts.


The perioperative outcomes included (1) inhospital complications, (2) length of stay (LOS), and (3) discharge disposition. An extended LOS was classified as being greater than the 90th percentile of all revision TJAs found in our database (11.2 days). The postdischarge outcomes of interest were (1) 30-, 60-, and 90-day readmissions; (2) 90-day medical complications; and (3) re-revisions for aseptic and septic failures. Minor complications included pneumonia, urinary tract infection, renal insufficiency, and superficial surgical site infection. Major complications included unplanned intubation, wound dehiscence, hematoma/seroma, deep surgical site infection, myocardial infarction, shock, sepsis, stroke, deep vein thrombosis, pulmonary embolism, revision surgery, and mortality. Re-revisions for septic and aseptic failures were evaluated until the latest follow-up. An occurrence of periprosthetic joint infection was considered a septic failure.

Statistical Analysis

Variables that were not charted for more than 80% of the cohort were excluded from this analysis. Propensity score matching was performed in R (R Core Team, 2017) with the use of the PSWeight package.[23] After propensity analysis, significance testing of all covariates and noncovariates was then performed with an analysis of covariance for continuous variables and logistic regression analysis for dichotomous variables. Follow-up Tukey tests were used to correct for the matching of cohorts. Multivariate regression analyses which were adjusted for propensity scores were then applied to the total matched cohort to obtain adjusted odds ratios (ORs) and beta coefficients. Statistical Package for Social Sciences version 25.0 (IBM Corp. Released 2019; IBM SPSS Statistics for Windows, Version 25.0) was used for the aforementioned significance testing and regression analyses.[24] The level of significance was P < 0.05 for all analyses.