The Effect of High Obesity on Outcomes of Treatment for Lumbar Spinal Conditions

Subgroup Analysis of the Spine Patient Outcomes Research Trial

Kevin J. McGuire, MD, MS; Mohammed A. Khaleel, MD, MS; Jeffrey A. Rihn, MD; Jon D. Lurie, MD, MS; Wenyan Zhao, PhD; James N. Weinstein, DO, MS


Spine. 2014;39(23):1975-1980. 

In This Article

Materials and Methods

Study Design

SPORT involves prospective enrollment of patients with lumbar pathology at 13 multidisciplinary spine centers across 11 states. The details of the design are thoroughly explained.[16]


Institutional review board approval was obtained at each center; enrollees were older than 18 years. Patients were enrolled between March 2000 and February 2005. For SpS, all had neurogenic claudication or radicular symptoms for greater than 1 week. Stenosis was demonstrated on axial imaging, and the patient was confirmed a surgical candidate by the physician. Patients with spondylolysis or isthmic spondylolisthesis were excluded. Many underwent pre-enrollment conservative management including nonsteroidal anti-inflamatory drugs, opioids, physical therapy, chiropractic care, and epidural injections. For DS, patients had upright lateral radiographs demonstrating listhesis. The IDH group had radicular symptoms for greater than 6 weeks with a neurological deficit and/or positive nerve-root tension sign. Axial imaging demonstrated a disc herniation at the appropriate level and laterality. Patients with cauda equina syndrome, progressive neurology, disc herniation superior to L2, prior lumbar surgery, scoliosis of more than 15°, or any contraindication to surgical intervention were excluded.


For each pathology, random computer-generated treatment assignments of operative versus nonoperative treatments were made in the cohort. Patients in the observational cohort arm chose their treatment after physician consultation. Nonoperative groups had education, physical therapy with a home exercise program, and nonsteroidal anti-inflammatory drugs, if possible. For SpS and DS, surgery consisted of decompressive laminectomy with or without single-level fusion (iliac crest bone autograft × pedicle screw instrumentation). For IDH, surgery entailed discectomy. Extensive crossover occurred in the randomized cohort: patients assigned to nonoperative treatment went on to undergo surgery and vice versa. Therefore, an as-treated analysis was made with both groups combined. Similar baseline characteristics and outcomes between the 2 groups made this possible.


Patients completed questionnaires at baseline, 6 weeks, 3 months, 1 year, 2 years, 3 years, and 4 years after surgery or enrollment. Primary outcome measures were the components of the SF-36[17] and the Musculoskeletal Outcomes Data Evaluation and Management System version of the ODI.[18] The SF-36 is scored from 1 to 100 points; higher scores indicate less severe symptoms. The ODI is scored from 0 to 100 points; higher scores indicate more severe symptoms. Secondary outcomes were patient satisfaction, self-reported improvement, and work status.[19]

Symptom severity was measured via the sciatica bothersome index and the low back pain bothersomeness index.[20,21] The former is scored from 0 to 24, and the latter from 0 to 6; higher scores indicate more severe symptoms. The stenosis bothersomeness index was also used for the group with SpS and DS. This is scored from zero to 24, with higher scores indicating more severe symptoms.

Statistical Analysis

Patients were divided into 3 groups on the basis of the BMI at enrollment: nonobese (BMI <30 kg/m2), obese (BMI = 30–35 kg/m2), and highly obese (BMI ≥35 kg/m2). Baseline characteristics were compared with analysis of variance tests for the continuous variables and χ2 tests for categorical values. Primary analysis involved a comparison of operative and nonoperative outcomes by change from baseline at each follow-up interval. This was analyzed with a mixed effects longitudinal regression model, including a random individual effect to account for correlation between repeated measurements for a particular patient. In the as-treated analysis, necessitated by the crossover, treatment was a time-varying covariate. In the operative group, time was measured from surgery, and in the nonoperative group, time was measured from enrollment. In crossover to surgery, preoperative data were retained from the time of enrollment for inclusion into nonoperative treatment effect analysis. Adjusting covariates were used in the longitudinal regression models to adjust for potential confounding variables (marital status, smoking status, compensation, herniation location, working status, depression, other comorbidities, self-rated health trend, duration of the most recent episode, or treatment preference). Race, center, age, sex, and baseline outcome measure scores were included in the longitudinal regression models. Secondary and binary outcome analysis involved the use of generalized estimating equations assuming a compound symmetry working correlation structure. The outcomes stratified by the obesity subgroups at each time point were compared with a multiple-df Wald test. During 4 years of follow-up, overall comparisons of area under curve between subgroups were made with a Wald test. Computations were done with SAS procedures PROC MIXED for continuous data and PROC GENMOD for binary and non-normal secondary outcomes (SAS version 9.1, Windows XP Pro; SAS Institute, Cary, NC). Significance was P < 0.05 based on a 2-sided hypothesis test. No adjustment for multiple comparisons was made because this study was not looking for a single result, but instead for a longitudinal assessment over time of different dimensions of outcome, including symptoms, function, and disability.[10]