Predictive Value of Preoperative Serum Albumin in Patients With Metastatic Spine Diseases: A Statistical Comment

Dear Editor:

We read with great interest the article by Hirase et al. ¹ entitled “Preoperative Serum Albumin Level Predicts Length of Stay and Perioperative Adverse Events Following Vertebral Corpectomy and Posterior Stabilization for Metastatic Spine Disease.” We applaud the authors for determining, for the first time, a valuable cut-off of preoperative serum albumin level in the prediction of perioperative adverse events (AEs) among spinal metastases patients. Specifically, low preoperative nutritional status in terms of serum albumin (<3.25 g/dL) was associated with higher risks of unfavorable outcomes and AEs following vertebral corpectomy and posterior stabilization. Nevertheless, several statistical errors (Table 1 and Table 2) have drawn our attention. Here we would like to make a comment.

To make comparisons regarding multiple baseline characteristics between the low and normal albumin group, “the Chi-Sqre (Chi-Sqre seemed a typo which should be Chi-Square or ${\mathrm{\chi }}^2$ ) or Fisher’s exact test was used to analyze categorical data” as stated in Methods. In Table 1, functional status should have been treated as one categorical variable, ² which included 4 levels—independent, partially dependent, completely dependent, and not recorded (The sum of the individual count exactly equals the sample size). Instead, the authors split it into 4 variables, performed 4 ${\mathrm{\chi }}^2$ or Fisher tests, and reported 4 P values. We were able to replicate these inappropriate tests using SPSS 21.0 statistical software (IBM Corp, Armonk, NY) and believed that they did make the mistake. We performed the correct ${\mathrm{\chi }}^2$ test and suggested the revision here (Supplementary Table). Likewise, we have statistical concerns about several other categorical variables like corpectomy location (Table 1) and discharge disposition (Table 2).

Moreover, the area under the receiver operative characteristic curve (AUC) was utilized to validate the predictive performance of serum albumin (Figure 1). In the footnotes of Table 3, however, another value (“AUC .75 based on a multivariate logistic regression model including all listed variables”) was reported and not mentioned elsewhere at all. We hold that it warranted further clarification in terms of statistical description and results’ interpretation, given that the AUC and cut-off value for the main predictive variable derived from receiver operative characteristic analyses accounting for other covariables were more than complicated. Lastly, the authors included white blood cell and platelet count as potential risk factors or baseline covariables. It was interesting somehow that preoperative hemoglobin or red blood cell count were not even considered. After all, intraoperative blood loss and transfusion in large amount was typically involved in spine surgeries for metastases.^3,4 It would be better that preoperative hemoglobin, as another important index for nutritional status in addition to serum albumin could be incorporated as one covariable here.

Establishing serum albumin as a valid prognostic biomarker in patients with metastatic spine diseases would be of great clinical significance to improve perioperative outcomes. We hope that the authors could check and address these statistical concerns to validate their findings.

Supplemental Material

Supplemental Material - Predictive Value of Preoperative Serum Albumin in Patients With Metastatic Spine Diseases: A Statistical Comment

Supplemental Material for Predictive Value of Preoperative Serum Albumin in Patients With Metastatic Spine Diseases: A Statistical Comment by Jingyi Wei, Zhenxi Li, Xinghao Zhang, and Askar Amar, in Global Spine Journal