Letter to the Editor Concerning “Subcutaneous Lumbar Spine Index (SLSI) as a Risk Factor for Surgical Site Infection After Lumbar Fusion Surgery: A Retrospective Matched Case-Control Study” by Shen et al

To the Editor

We recently read the article with great interest by Shen et al entitled “Subcutaneous Lumbar Spine Index (SLSI) as a Risk Factor for Surgical Site Infection After Lumbar Fusion Surgery: A Retrospective Matched Case-Control Study”. The author conducted a retrospective case-control study on patients who underwent transforaminal lumbar interbody fusion (TLIF) and found that subcutaneous fat thickness (SFT) and subcutaneous lumbar spine index (SLSI) had an association with early surgical site infection (SSI) after adjusting possible confounding factors. They showed that SLSI was a novel radiological risk factor and was a better indicator than SFT to predict early SSI risk after lumbar intervertebral fusion. We appreciate the authors’ great work on this field, and we have some concerns about this study.

1. In order to assure that infection group was balanced with control, the authors matched from people without early SSI according to the following criteria: age (±3 y), gender (male or female), diabetes (yes or no), timing of surgery (morning or afternoon). However, we didn’t get the exact method which the authors used to balance the possible confounders, whether by propensity score matching or other methods. What’s more, the outcomes of match between groups weren’t satisfied. Obesity, number of fusion levels, as well as operation time were significant different between groups. Previous studies have confirmed that longer number of fusion levels ¹ and operation time ² may increase the risk of SSI.

2. We noticed that some data in table 2 about risk factors for SSI using multivariate analysis, for example, the odds ratio of fusion levels (n ≥ 3) was 4.14, which wasn’t distributed in the interval range of 95% CI (.07-.84). So were other parameters, like SFT, and SLSI. We doubted that there may be some flaws in those data and I appreciated that the authors could revisit these data.

3. In the statistical analysis part, we didn’t get the exact methods which the authors used to enter the variables by multivariable logistic regression. As confirmed before, it would be better to fulfill that the number of events per variable values is larger than 10. ³ As the case group had a relatively small sample size (N = 33), the regression coefficients maybe biased when too much variables were included in the regression model.

Despite these concerns, the authors made a great contribution to this topic. We once again thank the authors for their contributions and hope that there will be more and more interesting researches in this area.