Letter to the editor: Causal inference and treatment selection in metastatic spinal disease with spinal instability neoplastic score 7–12
Article information
To the Editor,
We read with interest the systematic review by Kildegaard et al. [1] entitled “How to select a treatment method for patients with potentially unstable metastatic vertebrae (spinal instability neoplastic score 7–12): a systematic review.” The clinical question is important, particularly given the limited comparative evidence available. However, we are concerned that some conclusions use causal language that exceeds what can be supported by the underlying observational data.
Survival is not a treatment effect. The review cites the crude median survival reported by Dial et al. [2]—430 days after surgery plus external beam radiotherapy (EBRT; S+E) versus 121 days after EBRT alone—as evidence favoring surgery [1]. However, these groups were not exchangeable at baseline. The mean revised Tokuhashi score was 10.4 in the S+E group versus 8.6 in both the EBRT and cement augmentation plus EBRT (K+E) groups (p<0.001) [2]. In multivariable analysis, revised Tokuhashi score 0–8 was associated with survival (hazard ratio [HR], 3.14), whereas spinal instability neoplastic score (SINS) was not (HR, 1.01; p=0.72), and EBRT did not differ from K+E (HR, 0.91; 95% confidence interval [CI], 0.59–1.39; p=0.43) [2]. Dial et al. [2] also stated that stabilization “is not intended to alter the length of survival.” Thus, the observed survival difference may primarily reflect patient selection and baseline prognosis rather than a survival benefit attributable to surgery.
Adjustment may not fully address this selection. Operative candidacy incorporates host factors associated with prognosis, including albumin, hemoglobin, lymphocyte count, neutrophil-to-lymphocyte ratio, and alkaline phosphatase [3,4]. These factors are not incorporated into the scoring systems used for adjustment, leaving potential residual confounding.
Treatment allocation should not be interpreted as treatment efficacy. The key point that “a SINS ≥10 indicates a need for surgery” appears to derive from studies showing that patients with higher SINS values were more frequently treated surgically [1]. This reflects treatment selection rather than treatment benefit. SINS was developed to identify potential instability and prompt surgical consultation, not to determine treatment allocation [5]. Consistent with this distinction, Lenschow et al. [6] found no significant difference in neurological outcomes between instrumented and non-instrumented patients among 331 patients with SINS 7–12 (p=0.612), while complications were significantly more frequent after instrumentation (15.5% vs. 5.1%, p=0.016).
Health-related quality of life (HRQoL) is similarly vulnerable to selection bias. Poorer-prognosis patients may be preferentially selected for nonsurgical treatment, while death can differentially prevent later HRQoL assessment. This truncation by death is particularly relevant to the 52-week assessment emphasized in the review [1,7] and complicates causal interpretation of HRQoL differences [8].
We do not argue that surgery is ineffective or inappropriate in selected patients. Rather, the available evidence does not establish surgical superiority for patients with SINS 7–12. We therefore suggest that the key points stating that “surgical intervention is superior” and “surgery improved HRQoL” be reframed as associative rather than causal conclusions. Future comparative studies should use designs and analytical methods that adequately address confounding by indication and treatment selection.
Notes
Conflict of Interest
No potential conflict of interest relevant to this article was reported.
Author Contributions
All the authors have equal contribution in conception, drafting and revision of this Letter.
