Association Between Statin Use at the Time of Intra-abdominal Surgery and Postoperative Adhesion-Related Complications and Small-Bowel Obstruction

Key Points Question Are statins associated with a reduced risk of postoperative adhesion-related complications? Findings In these 2 separate retrospective cohort studies using population-representative data sets including more than 1.3 million individuals, statin use at the time of intra-abdominal surgery was significantly associated with an 8% to 19% reduction in adhesion-related complications and a 12% to 20% reduction in small-bowel obstruction after intra-abdominal surgery. Meaning The findings of this cohort study suggest that statin use at the time of intra-abdominal surgery may be associated with a decreased risk of postoperative adhesion-related complications.


Small Bowel Obstruction analyses
A major clinical sequela of adhesive-related complications is small bowel obstruction (SBO). Further, approximately 70% of SBO admissions are thought to be secondary to adhesion formation. While SBOs were included in the overall adhesion-related complication used in the primary analyses, a secondary analysis was performed examining this specific outcome alone. All other components of these analyses were identical to those described for the primary analysis. These analyses were first conducted in The Health Improvement Network (THIN), and then repeated in Optum's Clinformatics ® Data Mart (Optum).

Propensity score analyses
In order to ensure we were comparing individuals of similar baseline characteristics at the time of surgery, we performed a propensity score analysis using the pscore and psmatch2 routines in Stata within the THIN cohort 1,2 , as per our previously described methods 3 . Variables included in propensity score specification included sex, age at the time of surgery, previously recorded comorbidities associated with statin use (obesity, diabetes, hyperlipidemia, and hypertension), smoking, surgical location, and history of malignancy. Multiple methods of matching were ascertained to assess which specification resulted in greatest balance of these factors, including 1:1 and 1:5 caliper matching and kernel weighting. Balance was assessed graphically and statistically using pstest.
After exploring these three different matching/weight methods, 1:1 matching was employed given its degree of balance (see Supplemental Figure 1). Within the matched cohort, there were 17,997 individuals. Adjusting for the propensity score in this cohort, statin use at the time of surgery was more strongly associated with a reduction in adhesion-related complications than in our base analyses (adjusted Hazard Ratio 0.41, 95% CI 0.23, 0.74). To ensure this effect estimate was not an artifact induced through matching, we performed 100 iterations of matching, re-sorting the patient population using a randomly generated new patient identifier with each iteration. The average of these 100 hazard ratios was 0.42, the 2.5th percentile 0.24, and the 97.5th percentile 0.62. Only two of 100 iterations were non-significant based on 95% confidence intervals.
Assessing outcomes within specific time windows after the incident surgery In order to assess if the association between statin use at the time of intra-abdominal surgery and ARCs or SBOs varied depending on the time interval between the surgery of interest and outcomes, we repeated our analyses using our parsimonious model within THIN and only considering outcomes that occurred within specific time-from-surgery intervals as valid. We assessed outcomes occurring within 5, 4, 3, 2, or 1 year of the index surgical event for each of these analyses. The association between statins and ARCs or SBOs remained similar when only considering outcomes within each of these time windows as true outcomes (See Supplemental Table 10). E-value estimation for the hazard ratio of ARCs with statin use.
E-values have been a suggested means for estimating the potential effect estimate an unmeasured confounder would need to have with both the exposure of interest (statins) and outcome (ARCs) to bias the measured effect estimate to 1.0 4 . These were calculated using Stata's evalue package, using the methods described by VanderWeele and Ding. In THIN, the e-value was 1.77, and in Optum it was 1.40 eTable 1. Diagnostic codes used to identify surgical events within the abdomen in The Health Improvement Network READ