Introduction

Immunosuppressive medications inhibit activity of the immune system and include steroids, cytostatics, biologic modifiers, and other drugs used to suppress the body’s immune response [1]. Notably, synthetic steroids are potent anti-inflammatory drugs widely used to attenuate the general immune response [2–4]. Steroids along with other immunosuppressive medications are used to treat chronic inflammatory conditions such as asthma, chronic obstructive pulmonary disease, sarcoidosis, and osteoarthritis in addition to autoimmune diseases like rheumatoid arthritis and systemic lupus erythematosus.

Despite their benefit, chronic immunosuppressant and steroid medications have been associated with severe complications including infections, cardiac and thromboembolic events, gastrointestinal bleeding, wound healing, and anastomotic leakage following a variety of surgical procedures [5–8]. A large multicenter study of 635,265 patients undergoing a variety of operations reported a two- to threefold higher postoperative infection rate and a fourfold increase in mortality rate among steroid users compared with nonusers [8]. Unfortunately for many patients, discontinuation of such medications is not feasible due to risk of disease relapse or exacerbation.

Given the rise in the incidence of severe obesity worldwide, there has been an increase in bariatric procedures due to its effectiveness in achieving weight loss and improvement of obesity-related comorbidities [9–11]. In addition, more patients on chronic immunosuppressants are being referred for and undergoing bariatric surgery, especially in the transplant-recipient or candidate population who are severely obese [12, 13]. However, the literature on the postoperative outcomes in patients on chronic steroid or other immunosuppressants following primary bariatric surgery is limited [14–16]. We thus aimed to compare the 30-day major morbidity and mortality among steroid/immunosuppressant-dependent (SD) and non-dependent (ND) patients following primary bariatric surgery.

Methods

Data Collection and Study Cohort

This retrospective cohort study was based on the data from the American College of Surgeons-National Surgical Quality Improvement Program (ACS-NSQIP) database. The ACS-NSQIP prospectively collects information on over 300 variables, including demographic variables, comorbidities, laboratory values, and 30-day postoperative mortality and morbidity outcomes for patients undergoing major surgical procedures in over 430 participating centers in North America by the end of 2013. The ACS-NSQIP utilizes several mechanisms to ensure that the data collected are of the highest consistency and reliability [17].

We identified patients ≥18 years old who had a body mass index (BMI) ≥35 kg/m2 and underwent adjustable gastric banding (AGB), sleeve gastrectomy (SG), Roux-en-Y gastric bypass (RYGB), or biliopancreatic diversion-duodenal switch (BPD-DS) using their respective Current Procedural Terminology codes (AGB: 43770; SG: 43775; RYGB: 43644, 43645; BPD-DS: 43845). All operations were performed laparoscopically except for a small fraction of the BPD-DS procedures. We excluded patients who underwent a revisional bariatric procedure and procedures coded as emergent. Patients who had another surgery in the 30 days prior to the index bariatric surgery were also excluded. In order to further identify those patients that were at risk for a worse postoperative outcome, we excluded those with any evidence of preoperative sepsis, disseminated cancer, chronic dependence on dialysis, and American Society of Anesthesiology (ASA) class 5 (Moribund).

Covariates

The main exposure variable was chronic use of steroid or immunosuppressant medications. According to ACS-NSQIP variable definitions, we identified SD patients as those who were on a long course of steroid or immunosuppressant medications (defined by a duration >10 days) for a chronic medical condition (e.g., chronic obstructive pulmonary disease [COPD], asthma, rheumatologic disease, rheumatoid arthritis, inflammatory bowel disease), and within the 30 days prior to the principal operative procedure.

Independent demographic variables were age, sex, weight, BMI, race, and ethnicity. Examined comorbidities included diabetes (and insulin usage), hypertension, dyspnea, congestive heart failure (CHF), history of coronary artery disease ([CAD] binary variable representing angina, myocardial infarction, cardiac interventions, and cardiac surgeries), COPD, history of cerebrovascular diseases (binary variable representing transient ischemic attack and stroke), history of peripheral vascular disease (binary variable representing rest pain, gangrene, or need for revascularization or amputation), and bleeding disorder. Other baseline characteristics included were active smoking and functional status. Preoperative laboratory variables included serum hematocrit, albumin, bilirubin, and international normalized ratio. All variables are clearly defined in the ACS-NSQIP database user guide [17]. Operative variables considered were the ASA classification, type of bariatric surgery, and operation time.

End Points

The primary end points were 30-day postoperative mortality and major morbidity, which was defined as the presence of any of 14 major adverse events including organ/space surgical site infection, sepsis, septic shock, pneumonia, unplanned reintubation, failure to wean from mechanical ventilation beyond 48 hours from index operation, myocardial infarction, cardiac arrest, stroke, coma, venous thromboembolic event (VTE) including deep vein thrombosis (DVT) or pulmonary embolism (PE), need for transfusion, acute kidney injury, and wound disruption. We also considered any or infectious morbidities, VTE, postoperative bleeding requiring blood transfusion, 30-day return to operating room, unplanned 30-day readmissions, and length of stay (LOS) beyond 3 days as secondary outcomes.

Statistical Analysis

The estimates on the study parameters are expressed as mean ± standard deviation (SD) and number (%). Baseline, operative, and 30-day postoperative parameters between SD and ND patients were compared using the Pearson’s chi-square or Fisher’s exact tests for the categorical variables and Student’s t test or Wilcoxon rank-sum test for the continuous variables. Missing data is not tabulated since its consideration in the analyses did not change significance of the univariate analyses.

Prognostic impact of chronic dependence on steroid or other immunosuppressant medications was examined using multivariate logistic regression. In order to adequately control for confounding, we included all potential and empiric confounders of the relation between steroid/immunosuppressant use and 30-day postoperative outcome using a stepwise approach. The confounders were included in the regression model if they exerted a minimum of a 2 % change in either direction for the odds ratio (OR) of the main exposure variable [18]. A separate multivariate logistic regression was carried out within the subgroup of SD patients comparing the two most common bariatric procedures (SG and RYGB) in terms of 30-day postoperative outcomes. OR (and 95 % confidence intervals [CI]) was used as a measure of magnitude of association. Statistical analyses were carried out using the STATA software (Stata Corp, Texas, USA) version 12. Inference is based on a two-sided 5 % level.

Results

The study cohort consisted of 114,169 patients who underwent primary bariatric surgeries between 2005 and 2013. We identified 1277 (1.12 %) SD patients. The proportion of SD patients to all patients who had primary bariatric surgery increased from 1.44 % (N = 15) in 2005 to 1.76 % (N = 359) in 2013.

Compared to ND patients, the mean age of the SD patients was higher (44.76 versus 48.96 years old; P < 0.001). Mean BMI was 47.02 kg/m2 in SD patients and was significantly higher than a mean of 46.37 kg/m2 in ND patients (P = 0.003). Diabetes mellitus, hypertension, COPD, cardiac, cerebrovascular, and bleeding disorders were significantly more prevalent among SD patients (P < 0.001). SD patients had significantly lower mean preoperative hematocrit and serum albumin levels compared to ND patients (P < 0.001). The complete list of baseline characteristics of the study cohort is described in Table 1.

Table 1 Baseline characteristics of the study cohort
Full size table

The frequencies of AGB placement were similar between the SD (21.14 %) and ND patients (22.87 %; P = 0.144). The frequency of SG was higher in the SD (26.47 %) compared to ND patients (18.34 %; P < 0.001) and the frequency of RYGB was higher in the ND (57.46 %) compared to SD patients (50.35 %; P < 0.001). Higher proportion of SD patients underwent BPD-DS (2.04 %) compared to ND patients (1.33 %; P = 0.028). Operation times for AGB, SG, and RYGB were significantly longer among SD patients (P ≤ 0.013). Table 2 shows the operative characteristics of the study cohort.

Table 2 Operative characteristics of the study cohort
Full size table

Table 3 demonstrates the 30-day postoperative outcomes stratified on the chronic use of steroid/immunosuppressant medications. Thirty-day mortality rate for SD patients was significantly higher (0.55 %) compared to the ND patients (0.11 %; P < 0.001). SD patients had a higher 30-day major composite morbidity compared to ND patients (5.01 versus 2.54 %; P < 0.001). Similar frequencies for 30-day return to the operating room were found for SD patients (2.58 %) and ND patients (2.02 %; P = 0.152). However, unplanned 30-day readmissions were higher among SD patients (8.58 %) as compared to ND patients (4.77 %; P < 0.001). SD patients also had a longer mean length of stay compared to ND patients (2.41 versus 1.99 days; P < 0.001).

Table 3 Thirty-day postoperative outcomes stratified on the chronic use of steroid/immunosuppression
Full size table

Table 4 lists the adjusted OR for the chronic use of steroid/immunosuppressant medications corresponding to the primary and secondary end points. With respect to 30-day mortality, SD patients had an adjusted OR of 6.85 (95 % CI 1.95–24.12; c-statistic = 0.89). Chronic steroid/immunosuppressant use was also an independent prognostic factor for 30-day major morbidity (OR = 2.21; 95 % CI 1.29–3.79; c-statistic = 0.68). Chronic use of steroid/immunosuppressant drugs was not found to be an independent predictor of 30-day postoperative bleeding requiring transfusion, return to the operating room, and unplanned readmission.

Table 4 Adjusted odds ratios of preoperative chronic use of steroid/immunosuppressant on 30-day postoperative outcomes
Full size table

Table 5 shows the subgroup multivariate regression analysis comparing RYGB to SG in the SD patients only. Due to small number of events within the SD subgroup, no OR were reported for mortality and unplanned readmission. After adjustment, there was no significant difference in 30-day major morbidity after RYGB compared to SG (OR = 0.36; 95 % CI 0.08–1.66). LOS beyond 3 days was found to be significantly higher in patients who underwent RYGB compared to SG (OR = 5.43; 95 % CI 1.05–28.07).

Table 5 Multivariate analysis comparing SG and RYGB in the SD patients
Full size table

Discussion

We found that the 30-day postoperative mortality rate among patients dependent on chronic use of steroid/immunosuppressant after primary bariatric surgery was near sevenfold higher than that for ND patients. The occurrences of any, any major, and any infectious morbidity in the 30-day postoperative period were also over twofold higher in the SD compared to the ND patients. Diabetes mellitus, hypertension, COPD, cardiac, cerebrovascular, and bleeding disorders were significantly more prevalent at baseline among SD patients. However, the substantial prognostic effect of the chronic steroid use was observed after accounting for several confounders including the comorbidities at baseline.

To our knowledge, this is the largest multi-center study to show a highly negative prognostic impact of the chronic use of steroid/immunosuppressant medications on the 30-day postoperative morbidity and mortality following primary bariatric surgery. Our results are consistent with other large multi-center studies showing that chronic steroid use is associated with significant postoperative adverse events following a variety of surgical procedures mainly colorectal surgeries [4, 8, 19].

Our results following primary bariatric surgery are in contrast to a recent cohort study by Gribsholt et al. using the Danish nationwide medical database which evaluated the postoperative infection and bleeding rates in 325 current and 365 recent steroid users following RYGB from 2006 to 2012 [15]. Gribsholt et al. showed that current use of steroid is associated with a slightly increased but not significant risk of postoperative bleeding (2.8 versus 1.6 %) translating into an adjusted OR of 1.5 and virtually no difference in risk of infection (1.9 versus 1.7 %) [15]. This difference in rate of infectious complications with our results could partly be explained by the small sample size of the SD patients in the Danish study with a study cohort unique to a single Scandinavian country. But the difference could also be in part due to different definitions used in various studies and the lack of information on baseline comorbidities that could necessitate chronic steroid/immunosuppressant therapy in the ACS-NSQIP database such as systemic lupus erythematosus, asthma, or inflammatory bowel disease. Adjusting for such chronic diseases could likely blunt the magnitude of the effect measures in our study.

We also found that chronic steroid/immunosuppressant use is not an independent predictor of postoperative bleeding requiring transfusion, return to operating room, or unplanned readmission. Yet we showed that chronic steroid use is an independent predictor of 30-day postoperative DVT, and despite the lack of statistical significance, there was high trend toward a higher occurrence of 30-day VTE in the SD patients. This finding is consistent with another large ACS-NSQIP study on risk factors of postoperative symptomatic thromboembolism after nine common general, vascular, and orthopedic surgeries that after adjustment showed a 1.5- to 2-fold increase in 30-day VTE and PE directly attributed to chronic corticosteroid use [20]. Furthermore, we found that chronic steroid use was also an independent predictor of prolonged LOS defined as >3 days for a primary bariatric surgery which, given the associated increase in 30-day postoperative morbidity is not surprising.

Our study was not designed to identify the mechanism by which steroids or other immunosuppressant medications lead to such adverse early postoperative outcomes. Nonetheless, corticosteroids are involved in the regulation of various physiological processes maintaining homeostasis [21]. At supra-physiologic levels with exogenous use, they may suppress immune activity, proliferation and protein synthesis, and alter metabolism [5, 22] which can subsequently interfere with postoperative healing processes and recovery. In addition, steroids may mask the early symptoms of postoperative complications leading to a delay in diagnosis and treatment.

Regardless of the mechanism, given the strong negative prognostic effect of chronic steroid/immunosuppressant use on the 30-day postoperative morbidity and mortality after primary bariatric surgery, and the elective nature of weight loss procedures, all efforts must be made to help make these procedures safer. One such effort could be the cessation of the steroid/immunosuppressive therapy when clinically possible prior to the bariatric surgery in a supervised setting and to resume it after a short convalescent time interval. Unfortunately, no clear consensus exists regarding the need and optimum time for withholding therapy before surgery; likewise, clinicians are often uncertain of the appropriate time to resume therapy after the procedure [5]. Perhaps if the stability and course of the chronic disease permits, a 30 to 90-day preoperative window would be a reasonable interval to withhold the steroid/immunosuppressive therapy in a supervised setting. Support for the preoperative timing of drug cessation comes from a large population-level study by Ostenfeltd et al. which showed that, among 3966 SD patients from a total of 34,641 who underwent colorectal surgery for cancer, active SD patients had a 30 % higher 30-day mortality rate compared to non-users. Former users who did not take steroids in the 90-day period prior to the index operation did not have an increased mortality compared to the non-users [4]. Moreover, the study by Gribsholt et al. also showed no increased risk of 30-day postoperative infection or bleeding in the recent steroid users who did not receive any steroids in the 60-day period leading up to the RYGB [15].

Even though the literature investigating the impact of chronic steroid/immunosuppressant use on postoperative outcomes after bariatric surgery is lacking, most surgeons assume SD patients to be higher-risk candidates. In addition to consulting with the respective specialists to withhold such medications preoperatively, surgeons may be more inclined to offer such patients a SG as a potentially less morbid operation. Since SG and RYGB are currently the two most commonly performed primary bariatric surgeries in North America, we carried out a subgroup multivariate analysis in the SD patients only to compare the 30-day postoperative adverse outcomes between these two procedures. We found that, after accounting for various baseline and operative confounders, RYGB did not pose a higher risk than SG in terms of 30-day postoperative morbidity. This is an important finding since the ultimate goal of bariatric surgery is to achieve weight loss along with improvement of metabolic and obesity-related comorbidities [9, 11, 23, 24] and RGYB may be more suitable than SG in some SD patients.

In addition, understanding the prevalence and pattern of early postoperative adverse events in this population will help surgeons to have a heightened awareness for early signs of complications in these patients and address them in a timely fashion when they occur. Once the patient is beyond the early postoperative phase, the bariatric and metabolic benefits of the surgery may result in improvements in the underlying indication for immunosuppressant medication and may even allow decrease in steroid/immunosuppressant use [13, 14]. A recent study on bariatric surgery outcomes in patients with systemic lupus erythematosus with a mean follow-up time of 3 years reported that, despite a higher prevalence of early postoperative complications among those patients on immunosuppressive therapy, 42 % had a reduction in their medication dose and over 19 % were completely off their immunosuppressants by the end of the follow-up period [25].

Certain limitations need to be taken into consideration when interpreting our results. First, these analyses are limited by the retrospective study design. Another limitation is that we were unable to differentiate between steroid and other classes of immunosuppressants. Furthermore, the worsened outcomes in the SD group may be a consequence of the underlying disease requiring the chronic steroid/immunosuppressant use. However, the chronic use of steroid/immunosuppressant could be a proxy for such unknown baseline comorbidities.

There are also inherent limitations when the ACS-NSQIP database is used that can introduce both selection and misclassification bias. ACS-NSQIP captures only a sample of all procedures from each participating center and does not collect data unique to bariatric surgery patients (preoperative comorbidities such as obstructive sleep apnea, pulmonary hypertension, and history of venous thromboembolism) or unique bariatric postoperative complications such as anastomotic leaks. Furthermore, not all baseline comorbidities that could necessitate steroid/immunosuppressant therapy are captured in the ACS-NSQIP database such as systemic lupus erythematosus, rheumatoid arthritis, or asthma. In addition, the database only includes immediate postoperative outcomes up to day 30, which can lead to under-estimation of real risk of early adverse outcomes and has been suggested to be set at 90 days for other general surgical procedures [26]. In addition, long-term postoperative adverse events such as decrease in bone density and osteoporosis in patients with chronic steroid use after bariatric surgery may be missed [27].

Conclusions

In conclusion, we demonstrated that chronic and active use of steroid or immunosuppressant medications is a strong predictor of 30-day postoperative morbidity and mortality following primary bariatric surgery. Among the steroid/immunosuppressant users, RYGB and SG had a similar risk profile. Further studies are needed to help guide the management of such medications in the perioperative period.