A Nomogram for Predicting Suicide Risk and Prognostic Factors in Bladder Cancer Patients Following Diagnosis: More Information Needed

We read with interest this article by Liu et al. (2024) in a recent issue of the Journal of Affective Disorders, which developed a new clinical prediction model based on the SEER database that can help oncologists or urologists quickly identify patients at risk of suicide after bladder cancer (BLCA) diagnosis and take timely and appropriate preventive measures to avoid unnecessary deaths. We agree that the authors developed a predictive model to identify BLCA patients who are most likely to commit suicide. However, we would like to highlight a few important issues. First, the authors used the Cox proportional hazard regression model to examine suicide risk factors in patients with BLCA. Suicidal ideation is a psychiatric disorder influenced by a variety of factors. These factors include chronic illness, advanced age, family history, financial difficulties, adverse life experiences, unemployment, marital status, medications, depression and anxiety (O'Rourke, Jamil, and Siddiqui 2023). These risk factors can influence the association between BLCA and suicide. However, we found that in this study (Liu et al. 2024), four risk factors were identified by multifactorial analysis [male sex (HR = 11.419; 4.661–27.970), being married (HR = 0.592; 0.381–0.919 compared to single), advanced tumour local situation (HR = 1.758; 1.137–2.718 for T2 compared to Ta, T1) and unknown nodal involvement (HR = 2.671; 1.166–6.122 compared to N0)] and a column-line diagram was constructed based on these factors to construct the final clinical prediction model. However, we recommend selecting the model with the smallest AIC as the final prediction model based on the Akaike information criterion (AIC). This is because clinical prediction modelling research focuses not on whether independent variable (X) is meaningful but on whether the final model predicts accurately. Moreover, AIC can minimise the risk when the sample size is limited, and AIC is one of the measure of statistical model fit that weighs the complexity of the model being estimated against the goodness of fit of that model to the data (Vrieze 2012). Second, the authors sought to identify BLCA patients at the highest risk of developing suicide to help oncologists or urologists quickly identify individuals at risk of suicide and enable them to take appropriate preventive measures promptly. We think a decision curve analysis (DCA) is necessary for this prediction model. DCA is a method for assessing the benefits of a prediction model across a range of patient preferences for accepting the risks of undertreatment and overtreatment, to facilitate decisions about test selection and use (Vickers and Elkin 2006). DCA has some advantages compared with the traditional receiver operating characteristic curve (ROC). The ROC metric mainly focuses on the predictive accuracy of the model, which can measure the sensitivity and specificity of a prediction model under different criteria, but it does not consider the clinical utility of a specific model (Mandrekar 2010). DCA focuses not only on the accuracy of models, but also on how models affect actual clinical decisions and patient benefits (Van Calster et al. 2018), and DCA has been recognised by the Journal of the American Medical Association (Fitzgerald, Saville, and Lewis 2015), the British Medical Journal (Vickers, Van Calster, and Steyerberg 2016) and the Journal of Clinical Oncology (Kerr et al. 2016). We suggest that DCA should be added for this model. Finally, this retrospective study was based on BLCA patient data from the SEER database, which has limitations in collecting information related to patient suicidality, such as patients' psychopathological status and previous suicide attempts, whereas prospective studies are better because they have higher accuracy of data collection with regard to exposures, confounders and end points (Euser et al. 2009). In conclusion, the authors provide an attractive model that we still believe can help clinicians identify BLCA patients at risk of suicide and thus make decisions to avoid unnecessary deaths. The authors should be commended for their efforts. However, we think this prediction model needs further improvement before clinical application. The authors declare no conflicts of interest. The authors declared that they do not have anything to disclose regarding funding or conflict of interest with respect to this manuscript.