We read Abu-Arja’s study1 with great interest. They aimed to identify risk factors associated with severe hearing loss (HL) and to evaluate the generalizability of a published HL calculator among medulloblastoma patients treated with passive scattering proton therapy (PSPT) and cisplatin, while the HL is associated with worse neurocognitive outcomes among medulloblastoma patients. In their adjusted Cox models, they found that only higher the mean cochlear dose (Dmc) was associated with grade 3-4 HL. However, in this letter, we raise the concern that a basic statistical rule may ignored in this study, which may change the result of this study. We noted that there were 5 covariates in the multivariable Cox regression for prognostic factors associated with SIOP-Boston grade 3-4 hearing in table 3. Thus, according to the basic statistical rule that 1 variable per 10 outcome events for the logistic regression analysis,2–5 analyzed 5 covariates demand at least 50 outcome patients. However, there were only 27 Grade 3-4 HL patients in this study. In other words, this overfitted multivariate analysis could not obtain reliable statistical results. Moreover, more samples can reach statistical significance, or another validation study could validate the conclusion of this study.
Secondly, covariates were usually selected by the comparison of the outcome and non-outcome groups, then the significant variables were used for the final Cox regression analysis. Indeed, the comparison between the grade 0-2 group and the 3-4 group was made in table 2, but 3 significant covariates in table 2 were not analyzed in table 3 for the final Cox regression analysis. Thus, these 3 significant covariates in table 2 including the disease risk-stratification, the mean CSI dose, and the mean posterior fossa dose may also be associated with SIOP-Boston grade 3-4 HL in this study, selecting covariates at observer bias will result in the inaccurate prognostic results in this study.
Thirdly, we are curious why the covariate (the cisplatin dose) had not been analyzed in table 2 between the grade 0-2 group and the 3-4 group, but it was included in table 3 for the final Cox regression analysis. It seems to be select these covariates at random, not obeying the statistical flow chart.
Finally, it is our great honor to comment on Abu-Arja et al.’s outstanding work despite these comments.
Acknowledgment
We reviewed this manuscript and agreed to submit this manuscript.
Contributor Information
Wenjue Tang, Department of Pediatric Hematology & Oncology, Children’s Hospital, Maternal and Child Health Hospital of Guangxi Zhuang Autonomous Region, Nanning, China.
Huihong Dou, Department of Pediatric Hematology & Oncology, Children’s Hospital, Maternal and Child Health Hospital of Guangxi Zhuang Autonomous Region, Nanning, China.
Conflict of interest
None declared.
Funding
None declared.
References
- 1. Abu-Arja MH, Brown AL, Su JM, et al. The cochlear dose and the age at radiotherapy predict severe hearing loss after passive scattering proton therapy and cisplatin in children with medulloblastoma. Neuro-Oncology. 2024:noae114. doi: 10.1093/neuonc/noae114. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Pavlou M, Ambler G, Seaman SR, et al. How to develop a more accurate risk prediction model when there are few events. BMJ (Clin Res ed.). 2015;351:h3868. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Riley RD, Ensor J, Snell KIE, et al. Calculating the sample size required for developing a clinical prediction model. BMJ (Clin Res ed.). 2020;368:m441. [DOI] [PubMed] [Google Scholar]
- 4. Riley RD, Snell KIE, Ensor J, et al. Minimum sample size for developing a multivariable prediction model: Part I—continuous outcomes. Stat Med. 2019;38(7):1262–1275. [DOI] [PubMed] [Google Scholar]
- 5. Pate A, Riley RD, Collins GS, et al. Minimum sample size for developing a multivariable prediction model using multinomial logistic regression. Stat Methods Med Res. 2023;32(3):555–571. [DOI] [PMC free article] [PubMed] [Google Scholar]