Introduction

Obesity is among the most common health conditions affecting orthopaedic patients. The current prevalence of obesity in the United States is approximately 35 %, and this figure is projected to increase to 45–50 % of the population by 2030 [1]. Rising obesity rates are estimated to result in additional obesity-related healthcare costs of $50 billion each year [2]. Obesity affects nearly every organ system and is associated with significant medical comorbidities [3].

The impact of body mass index (BMI) on surgical outcomes and costs has been studied extensively in the elective orthopaedic literature. A meta-analysis and systematic review exploring the impact of obesity on outcomes following total knee arthroplasty (TKA) demonstrated that obese patients had increased odds of postoperative infection (OR 1.90), deep infection requiring surgical debridement (OR 2.38), and all-cause revision surgery (OR 1.30) compared to patients of normal weight [4]. Following elective total hip arthroplasty (THA), Batsis et al. showed that morbidly obese patients were more likely to be transferred to a nursing facility than normal weight, overweight, or obese patients [5]. Following multilevel spinal arthrodesis, BMI >30 is associated with increased hospital length-of-stay and complications at 1- and 2-year follow-up [6]. Morbidly obese patients have complication rates three times greater than rates in underweight patients and eight times greater than rates in patients of normal weight [6].

Relatively little is known about the impact of obesity on outcomes following orthopaedic trauma. Increased complication rates have been reported in obese patients following specific injury patterns including acetabular fractures [7], pelvic ring injuries [8], and spine trauma [9]. However, very few studies have explored the impact of BMI on outcomes in the general orthopaedic trauma population. Utilizing the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) database, we sought to investigate the relationship between BMI and perioperative complications in orthopaedic trauma patients.

Materials and methods

Data extraction

Access to the NSQIP dataset collected between 2005 and 2013 was granted by the American College of Surgeons. The 135 patient variables reported within this multi-centre database include preoperative risk factors, intraoperative variables, and 30-day postoperative mortality and morbidity outcomes for patients undergoing major surgical procedures in both inpatient and outpatient settings. At each participating institution, two risk-assessment nurses trained as surgical clinical reviewers (SCR) were appointed to collect data directly from patients’ medical records. Inter-rater reliability disagreement of <5 % per site was considered acceptable. Audit reports of NSQIP data collection have identified disagreement rates of <1.8 % [10].

Patient selection

All patients who underwent an orthopaedic trauma procedure during the study period were identified from the NSQIP dataset using current procedural terminology (CPT) codes for orthopaedic trauma (n = 89). A description for each CPT code used is provided in the Appendix. Patient demographics including age, gender, and race were recorded, along with preoperative comorbidities including body mass index (BMI), recent weight loss (greater than 10 % in the last 6 months), insulin-dependent diabetes mellitus, smoking status, alcohol use, functional status, dyspnea, history of chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF), hypertension requiring medication, history of esophageal varices, disseminated cancer, steroid use, bleeding disorders, hemodialysis, chemotherapy within 30 days of surgery, and radiotherapy within 90 days of surgery. Operative factors including systemic inflammatory response syndrome (SIRS), sepsis, or septic shock at time of surgery, operative time, wound class, and American Society of Anesthesiologists (ASA) score were also recorded.

Preoperative BMI was used to group patients into one of five categories: underweight (BMI <18.5), normal weight (18.5–24.9), overweight (25–29.9), obese (30–39.9), or morbidly obese (40 or higher) [3]. Patients without a recorded BMI were excluded from the analysis.

Outcome measures

Perioperative complications within 30 days were categorized as either minor or major based on previously published literature using the NSQIP database [11–16]. Minor complications included wound dehiscence, superficial wound infection, pneumonia, and urinary tract infection. Major complications included deep wound infection, organ space infection, myocardial infarction, pulmonary embolism, deep venous thrombosis, cerebrovascular accident, postoperative neurologic deficit, sepsis, septic shock, coma, and death. A third outcome measure—total complications—was determined by identifying all patients who developed at least one minor and/or major complication.

Data analysis

Rates of minor, major, and total complications for each BMI category were calculated and compared using a Chi-square test.

Using a multivariate logistic regression analysis controlling for age, smoking status, ASA score, and medical comorbidities, odds ratios (ORs) for minor, major, and total complications were calculated for each BMI category. Patients with a BMI in the normal range were used as the reference group. This analysis was then repeated using wound-related complications (wound dehiscence, superficial wound infection, and deep wound infection) as the outcomes of interest.

The complete multivariate model is included in the “Appendix”. Predictive accuracy of the logistic regression models was assessed using the concordance statistic (c-statistic), or the area under the ROC curve. Statistical analysis was performed using Stata 12 (StataCorp. 2011. Stata Statistical Software: Release 12. College Station, TX: StataCorp LP) and SSPS Statistics (IBM Corp. Released 2013. IBM SPSS Statistics for Windows, Version 22.0. Armonk, NY: IBM Corp). Significance was set at p < 0.05.

Results

56,299 patients were identified from the NSQIP dataset using CPT codes for orthopaedic trauma procedures (n = 89). As depicted in Fig. 1, 3080 patients without a recorded preoperative BMI were excluded, leaving 53,219 patients available for analysis. Average age was 67.3 years, and 35.6 % of patients were male. As shown in Table 1, among the patients with a recorded preoperative BMI, 10.1 % were underweight, 37.3 % were of normal weight, 28.4 % were overweight, 19.7 % were obese, and 4.6 % were morbidly obese.

Fig. 1
figure 1

Flowchart showing patient selection from ACS-NSQIP database

Full size image
Table 1 Demographic characteristics of included patients (n = 53,219)
Full size table

Rates of minor, major, and total complications by BMI category are displayed in Table 2. Among the 53,219 patients, 6.5 % had minor complications and 7.3 % had major complications, with an overall rate of 11.9 % for total complications. There were statistically significant differences in rates of minor, major, and total perioperative complications between groups, with the highest rates of complications occurring in underweight patients (8.9 % for minor, 10.8 % for major, and 16.9 % for total complications).

Table 2 Rates of minor, major, and total complications by BMI category
Full size table

Results of the multivariate analysis are displayed in Table 3. Compared with patients of normal weight, underweight patients had significantly greater odds of minor [OR 1.12, 95 % CI (1.0, 1.3), p = 0.04], major [OR 1.20, 95 % CI (1.1–1.3), p = 0.0009], and total complications [OR 1.18, 95 % CI (1.1, 1.3), p = 0.0003]. Morbidly obese patients had significantly greater odds of major [OR 1.22, 95 % CI (1.0, 1.5), p = 0.023] and total complications [OR 1.18, 95 % CI (1.1–1.4), p = 0.023] than did patients of normal weight. There was a trend toward greater odds of minor complications in morbidly obese patients [OR 1.18, 95 % CI (1.0, 1.4), p = 0.077]. Having a BMI in the overweight or obese range did not significantly increase the odds of minor, major, or total complications.

Table 3 Multivariate analysis displaying odds ratios (ORs) of minor, major, and total complications by BMI category
Full size table

When wound-related complications were examined independently, obesity was associated with increased odds of superficial [OR 1.67, 95 % CI (1.3, 2.1), p < 0.0001] and deep wound infection [OR 1.52, 95 % CI (1.1, 2.1), p = 0.018] compared with patients of normal weight, as shown in Table 4. Morbid obesity was associated with increased odds of wound dehiscence [OR 2.29, 95 % CI (1.1, 4.9), p = 0.034] and deep infection [OR 2.51, 95 % CI (1.6, 3.9), p < 0.0001]. Trends toward increased odds of wound dehiscence in overweight [OR 1.72, 95 % CI (1.0, 3.1), p = 0.053] and obese [OR 1.71, 95 % CI (1.0, 3.1), p = 0.07] patients did not reach statistical significance. The complete multivariate model is included in the Appendix.

Table 4 Multivariate analysis displaying odds ratios (ORs) of wound complications by BMI category
Full size table

Discussion

We utilized the ACS-NSQIP database to investigate the relationship between BMI and perioperative complications following orthopaedic trauma surgery. Our multivariate analysis demonstrates that patients with morbid obesity have significantly increased odds of major and total perioperative complications compared with patients of normal weight. These findings corroborate previous reports demonstrating an association between obesity and complications in specific orthopaedic trauma injury patterns, including acetabular fractures [7], pelvic ring injuries [8], and spine trauma [9]. A similar study by Hoffmann et al. also demonstrated a trend between BMI and mortality among orthopaedic polytrauma patients in Germany [17]. However, our study is the first to document an association between increased perioperative complications and morbid obesity among a large orthopaedic trauma population in North America.

In our study, obesity and morbid obesity were also associated with significantly increased odds of wound complications including superficial or deep infection and wound dehiscence, as shown in Table 4. These findings are in keeping with other published studies that document higher rates of perioperative wound complications in obese patients. In a retrospective study of more than 7500 lower extremity vascular bypass procedures, Giles et al. identified obesity as an independent risk factor for surgical site infections [18]. In a retrospective comparative study of patients with operatively treated acetabular fractures, morbidly obese patients had a 46 % wound complication rate compared with only 12 % in patients with BMI <40 [7]. Sems et al. reported a wound complication rate of 11 % in a series of obese patients who underwent surgical treatment of pelvic ring injuries [8]. Patients with a BMI >30 were 6.87 times more likely to develop a complication and 4.68 times more likely to require reoperation than those with a BMI <30.

In addition to identifying morbid obesity as a risk factor for perioperative complications, our results demonstrate that orthopaedic trauma patients with a BMI less than 18.5 have significantly increased odds of minor, major, and total complications compared with patients of normal weight. To our knowledge, this finding has not been previously reported in the orthopaedic trauma literature. In elective total joint arthroplasty, underweight patients are at increased risk for 90-day readmission following THA [19]. In the surgical oncology literature, underweight status has been shown to be a risk factor for increased length-of-stay in patients undergoing thoracotomy for lung cancer [20]. Similarly, in patients with gastric adenocarcinoma, underweight status and low serum albumin were shown to be independent risk factors for mortality following gastrectomy [21]. In our study, underweight orthopaedic trauma patients had significantly increased odds of minor (OR 1.12), major (OR 1.20), and total (OR 1.18) perioperative complications compared with patients of normal weight.

Increased complication rates at the extremes of the BMI spectrum—a phenomenon often referred to as the “obesity paradox”—have been previously published in the general surgical literature. In patients undergoing non-bariatric general surgery, Mullen et al. reported the highest mortality rates in underweight and morbidly obese patients, with lower mortality rates seen in overweight and moderately obese patients [22]. Davenport et al. reported similar results for patients undergoing vascular surgery [23]. Our results demonstrated similar findings: compared with patients of normal weight, overweight patients actually had slightly decreased odds of major (OR 0.88, p = 0.004) and total (OR 0.90, p = 0.002) complications, as shown in Table 3. In a systematic review of the cardiac and non-cardiac surgery literature, Valentijn et al. identified this phenomenon—worse outcomes in patients at the extremes of the BMI spectrum, with a slight protective effect seen in overweight and slightly obese patients—in multiple surgical subspecialties [24]. Hypotheses proposed to account for the obesity paradox include genetic factors as well as the potentially protective effect of lean body mass and moderate amounts of peripheral body fat. To our knowledge, the current study is the first to demonstrate the “obesity paradox” in patients undergoing orthopaedic surgery.

Our study has some limitations. First, the study was conducted in a retrospective manner. However, the fact that the NSQIP database contains prospectively collected data and is quite comprehensive in its scope largely mitigates this limitation. Multi-centre, prospective randomized controlled trials are associated with significant expense and other logistical challenges. Large multi-centre studies such as ours, which use a high-quality, prospectively collected database, provide the opportunity to answer relevant clinical questions while avoiding the expense and inconvenience of prospective trials [11–16]. Second, the NSQIP database does not capture any complications that occur more than 30 days after surgery. While many postoperative complications do not occur within the first month, the fact that we identified significant differences in 30-day complication rates between groups underscores the significance of these findings. In addition, the NSQIP database does not currently record polytraumas, which might also serve as a good predictor for complications. Finally, the nutritional status of patients in our cohort could not be determined since serum albumin is not a variable recorded in the NSQIP database. Further research may lead to an improved understanding of the role of nutrition in outcomes following orthopaedic trauma, especially in underweight patients.

Using the ACS-NSQIP database, we demonstrate that, compared with patients of normal weight, morbidly obese patients have significantly increased odds of wound dehiscence (OR 2.29), deep wound infection (OR 2.51), major complications (OR 1.22), and total complications (OR 1.18) following orthopaedic trauma surgery. Additionally, having a BMI under 18.5 is associated with increased odds of minor, major, and total perioperative complications. Interventions aimed at decreasing complication rates should be targeted at these high-risk patient populations on both ends of the BMI spectrum.