Impact of body mass index on outcomes after lumbar spine surgery

The impact of body mass index (BMI) on outcomes after lumbar spine surgery is currently unknown. Previous studies have reported conflicting evidence for patients with high BMI, while little research has been conducted on outcomes for underweight patients. This study aims to examine the impact of BMI on outcomes after lumbar spine surgery. This prospective cohort study enrolled 5622 patients; of which, 194, 5027, and 401 were in the low (< 18.5 kg/m2), normal (18.5–30), and high (≥ 30) BMI groups, respectively. Pain was assessed via the numerical pain rating scale (NPRS) for the lower back, buttock, leg, and plantar area. Quality of life was assessed via the EuroQol 5 Dimension (EQ-5D) and Oswestry Disability Index (ODI). Inverse probability weighting with propensity scores was used to adjust patient demographics and clinical characteristics between the groups. After adjustment, the 1-year postoperative scores differed significantly between groups in terms of leg pain. The proportion of patients who achieved a 50% decrease in postoperative NPRS score for leg pain was also significantly different. Obese patients reported less improvement in leg pain after lumbar spine surgery. The outcomes of patients with low BMI were not inferior to those of patients with normal BMI.


Statistical analysis.
To investigate the association of BMI with patient-reported outcomes, all patients were classified into one of three groups: low (< 18.5), normal (18.5-30), or high (≥ 30) BMI groups. We used inverse probability weighting to adjust for differences in demographic and clinical characteristics between groups 23 . We chose this method over regression analysis because it allows for a more straightforward comparison between the three groups. First, a multinomial logistic model was used to calculate propensity scores (i.e., the probability that a patient belongs to a particular group) 24 . The model used age, sex, disease type, ASA-PS, DM, HD, RA, smoking habits, primary or revision surgery, and surgical procedure. Next, the weighted groups were created using the inverse probability weighting method with stabilized weights from propensity scores. Background data and clinical characteristics were compared using chi-square tests and one-way analysis of variance (ANOVA) for categorical and continuous variables, respectively. The differences concerning pre-and postoperative NPRS, ODI, and EQ-5D scores were examined by one-way ANOVA. Previous literature suggests that a 50% reduction in pain can be considered substantially important for patients with chronic pain 25 . Thus, a chi-square test was used to compare the rate of achieving this threshold between groups. Bonferroni corrections and Tukey's multiple comparisons tests were used for all comparisons of categorical and continuous variables, respectively. In the Bonferroni corrections, each p value was tripled and the significance threshold was set to be the same as in other analyses.
To examine the nonlinear trend between BMI and NRS in more detail, restricted cubic spline logistic regression analysis was performed. Logistic regression analysis was performed for unweighted patients to calculate the OR of not achieving a 50% reduction in NRS using the same variables as inverse probability weighting with the addition of BMI. Patients with a BMI score of 25.0 were defined as the reference 26 .
Ethical approval. The manuscript submitted does not contain information about medical device(s)/drug(s).

Results
Patient background of unweighted groups. Of all 8575 consecutive patients who underwent lumbar surgery, this study enrolled 5622 patients of whom 194 (3.5%), 5027 (89.4%), and 401 (7.1%) were in the low, normal, and high BMI groups, respectively (Fig. 1). Table 1 shows the patient background of unweighted groups. The proportion of males was lower in the group L (25.8%) than in the groups N (63.6%) and H (65.3%). www.nature.com/scientificreports/ Inverse probability weighting and background of weighted groups. Following the inverse probability weighting using propensity scores, 192, 5027, and 471 patients were in the low, normal, and high BMI groups, respectively ( Table 2). The mean and standard deviation (SD) of BMI was 17.4 (0.8), 24.0 (2.7), and 32.2 (2.6) in the low, normal, and high BMI groups, respectively. Significant differences were observed between groups in terms of mean age, mean height, ASA-PS, mean operative time, mean estimated blood loss, and the rate of unintended dural tear.
Clinical outcomes in the weighted groups. Table 3 shows the comparison of each clinical outcome between the three weighted groups. The NPRS was completed by 69.4% of patients in the study ( Pairwise comparisons between groups demonstrated that postoperative leg pain in the high BMI group was worse than in the low and normal BMI groups (p = 0.001 and p < 0.001, respectively). The proportion of patients who achieved a 50% decrease in postoperative NPRS score differed significantly in terms of leg pain (low BMI, 76.9%; normal BMI, 67.2%; and high BMI, 59.6%; p = 0.001). Pairwise comparison between groups showed that the proportion was lower in the high BMI group than in the normal (p = 0.010) and low BMI group (p = 0.002). The proportion appeared better, but was not significant (p = 0.077), in the low BMI group than in the normal BMI group.
Of the 5690 patients enrolled in the study, 4046(completion rate: 71.1%; 128, 3,564, and 354 in the low, normal, and high BMI groups) and 3,907 (completion rate: 68.7%; 124, 3,434, and 349 in the low, normal, and high BMI groups) completed the EQ-5D and ODI, respectively (Table 3). No significant differences were noted between groups in either pre-or postoperative EQ-5D scores. On the contrary, a significant difference was  [16.6]; p = 0.047). The high BMI group demonstrated a worse ODI score than the normal BMI group in the pairwise comparisons (p = 0.037). However, the change in preoperative to 1-year postoperative score in terms of EQ-5D and ODI were not significantly different between groups (EQ-5D, p = 0.136; ODI, p = 0.358). Patient satisfaction data were collected from 3,940 (69.2%) patients (Table 3). No significant difference in postoperative satisfaction was observed among the three weighted groups (satisfaction rates: low BMI group, 89.1%; normal BMI group, 82.1%; and high BMI group, 82.7%; p = 0.126). Figure 2 shows the results of restricted cubic spline logistic regression analysis for leg pain, wherein the proportion of patients achieving a 50% reduction in postoperative NRS score was found to be significant. This shows that the OR of not achieving a 50% reduction in leg pain increased as BMI increased (Fig. 2).

Discussion
Leg pain in obese patients. In this study, the percentage of patients achieving a 50% reduction in leg pain worsened as BMI increased. Thus, patients with lower BMI had greater odds of achieving a 50% reduction in leg pain than patients with higher BMI. The relationship between BMI and leg pain remains controversial. Djurasovic et al. retrospectively reviewed patients undergoing lumbar fusion (N = 270) and reported that obese patients had slightly higher leg pain scores two years after surgeries than nonobese patients (mean NPRS 5.10 vs. 4.29; p = 0.043) 27 . De la Garza-Ramos et al. also investigated the impact of obesity on patients who underwent one-to three-level posterolateral fusion for degenerative spine disease 28 . They found that a higher proportion of obese patients had radiculopathy after surgery than nonobese patients (44.3% [31/70] 29 . These discrepancies may be partially explained by differences in patient demographics and baseline characteristics or by small sample sizes. The results of the current study may be more reliable due to the large sample size and www.nature.com/scientificreports/ statistical analysis, which adjusted for the potentially confounding differences in demographics and preoperative characteristics between groups. Since many studies define a BMI score ≥ 30 as obesity, the cutoff value was set to 30 30 . What outcomes are exhibited by those with a BMI score of 25-30 (overweight) and ≥ 40 (super-obese) would be interesting to find out, but the stratification used in this study does not allow us to see this. Therefore, we drew a restricted cubic spline and showed that as BMI increases, the OR of not achieving a 50% reduction in leg pain also increases. Hence, we conclude that at least for leg pain post lumbar spine surgery, the lower the BMI, the better the postoperative outcome.
One possible reason that obese patients are less likely to experience a significant reduction in leg pain after spinal surgery could be related to cytokines secreted from excess adipose tissue (i.e., adipokines) [31][32][33][34][35] . In obese patients, increased secretion of proinflammatory cytokines and decreased secretion of anti-inflammatory cytokines from adipose tissues have been observed, which can lead to increased levels of proinflammatory cytokines (tumor necrosis factor-α, interleukin 6, and so on) and systemic inflammation 35 . Inflammation can cause peripheral and central sensitization of the pain-transmitting system, resulting in hyperalgesia and allodynia 35 . Visceral and subcutaneous truncal white adipose tissue, often found in obese patients, is an active endocrine organ that secretes cytokines 33 , which may affect decompressed nerve roots and prolong pain in obese patients.
Oswestry disability index scores in obese patients. Patients in the high BMI group were observed to have worse preoperative ODI scores than patients with normal BMI and higher levels of back pain than patients in the low and normal BMI groups. Although obese people tend to have worse back pain and ODI scores than nonobese people 26,27,36 , a controversy exists regarding the impact of obesity on ODI score improvement after lumbar spine surgery [11][12][13]26,27,37   www.nature.com/scientificreports/ Table 3. Comparison of NRS, EQ-5D, ODI, and postoperative satisfaction in the low, normal, and high BMI groups after adjustment by inverse probability weighting method. BMI, Body mass index; SD, Standard deviation; NPRS, Numerical pain rating scale; EQ-5D, EuroQol 5 Dimension; ODI, Oswestry disability index.    26 . However, postoperative improvement in ODI, EQ-5D, and patient satisfaction was found in this study to not significantly differ between obese and nonobese patients. These results suggest that obese patients should expect similar levels of improvement in QoL as nonobese patients.

Patient-reported outcomes in underweight patients. Patients in the low BMI group were similar
to patients in the normal BMI group in terms of postoperative satisfaction and postoperative improvement of NPRS, ODI, and EQ-5D. Although several studies related to complications for underweight patients seeking lumbar spine surgery were noted, only a few reports are available on their postoperative outcomes relative to patients with normal or high BMI. Knutsson  Limitations. This study has several limitations. First, the sample size of the low and high BMI groups was smaller than the normal BMI group. Second, the completion rate for each clinical outcome was not high, which could have biased the results. Third, detailed surgical variables (e.g., the number of operated disc levels, the type of implant used, or years of surgeon experience) were not adjusted. These variables may differ between groups and could have unknowingly impacted the results of the current study. Fourth, knowing why a given patient has a high or low BMI is difficult. BMI can be affected by genetics, lifestyle, exercise habits, or various diseases (diabetes, heart failure, chronic kidney disease, liver dysfunction, lung disease, cancer, and so on). Patients who are underweight due to their lifestyle are likely to differ in important ways from patients who are underweight secondary to cancer. Furthermore, data related to frailty or sarcopenia were not collected. Fifth, although related to the third limitation, some selection bias may exist regarding which patients undergo surgery. Some obese and underweight patients may be unable to undergo surgery due to their underlying diseases that impact their BMI. Unfortunately, this potential selection bias was not tracked or accounted for before the decision to undergo surgery. Finally, a detailed image evaluation, which could have been useful to further adjust for baseline differences between groups, was not conducted.

Conclusion
In this study, patients in the high BMI group had worse postoperative leg pain and were less likely to experience significant leg pain improvement than patients in the low and normal BMI groups. Patients in the high BMI group also had worse preoperative ODI scores, but demonstrated similar postoperative ODI improvement compared to patients in the low and normal BMI groups. Patients in the low BMI group demonstrated similar postoperative satisfaction and improvements in NPRS, ODI, and EQ-5D compared to patients in the normal BMI group.

Data availability
The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.