Evidence-Based Recommendations for Spine Surgery

Alexander R. Vaccaro, MD, PhD, MBA; Charles G. Fisher, MD; Srinivas K. Prasad, MD; Alpesh A. Patel, MD, FACS; John Chi, MD, MPH; Kishore Mulpuri, MBBS, MHSc; Kenneth C. Thomas, MD, MHSc; Peter G. Whang, MD, FACS


Spine. 2020;45(21):E1441-E1448. 

In This Article

Methodological Review

One strength of this investigation is that authors explicitly discuss certain aspects of the methodology including the inclusion/exclusion criteria, the treatments specific to each cohort, and the sample size calculation which is clearly elucidated and easily reproduced. However, there are numerous methodological issues which considerably limit the relevance and applicability of the study findings. While the randomization scheme may have been effective since the mean age of the subjects in each group is similar, no other characteristics are provided so the initial status of the patient populations is otherwise ambiguous and there does not appear to have been any adjustment for the baseline values of the various outcomes (e.g., RSA, LSA, VAS). It is also important to point out that the pain management protocols are not noted so it is unclear whether they were consistent between the cohorts in terms of the standardized opioid equivalents of the medications administered to patients; without this information, the correlation of the difference in pain scores cannot be generalized. The authors also do not enumerate any complications related to treatment in either group.

Another major issue is that other than identifying the type of software (SPSS 2.0), there is essentially no discussion of the statistical methods that was used for this analysis. No defined treatment effect is stipulated nor is there any mention of the clinical significance of this effect. The authors do not indicate the extent of cross-over between the groups, nor do they share how many subjects dropped out of the study. No confidence intervals for the group differences are calculated and exact P values are not documented in many instances which reduces the precision of the estimates and minimizes the robustness of the results. Finally, no adjustments were made to the P values for multiple comparisons (i.e., multiple time points and multiple outcomes).

Furthermore, the credibility of the investigation is undoubtedly affected by several glaring discrepancies within the manuscript. Perhaps most concerning is that the mean time to return to work time was higher for the operative group than the nonoperative cohort (3.2 mo vs. 1.05 mo, respectively) yet the authors repeatedly state the opposite which is a gross misstatement of the facts and may potentially represent the inherent biases of the authors. There are also other errors evident in the description of figures (e.g., Figure 2 appears to be a comparison of RSA but is labeled as VAS pain scale, Figure 3 appears to be for RSA but is labeled as LSA) and many of the axes are not even labeled; moreover, some of the tables are incorrectly referenced with the body of the manuscript (e.g., Table 1 does not display age and sex as stated in the text). Without question, this lack of oversight serves to further detract from any significance of these findings.