Sir, we read Du's study [1] with great interest. In this study, they aimed to investigate the predictive capacity of lymphocyte subpopulations, sarcopenia and myosteatosis for clinical outcomes in patients who underwent gastric cancer surgery. Additionally, the prognostic significance of CD3+/CD4+ cells in conjunction with myosteatosis was explored. Based on their statistical analysis, they found that CD3+/CD4+ cells, CD4+/CD8+ ratio and CD19+ cells are indicative of clinical prognosis in gastric cancer surgery patients. Sarcopenia and myosteatosis are also associated with the patient's prognosis. Despite promising results, in this letter, we raise some statistical concerns about the results of this study.

Firstly, we noted that there were 38 variables in Table 2 or Table 3 for univariate and multivariate analysis for progression-free survival or overall survival. Thus, according to the basic statistical rule demands that 1 covariate per 10 outcomes for the prediction analysis [2-5], analysing 38 covariates demands at least 380 death gastric cancer patients in Du's study. In contrast, there were only a total of 190 gastric cancer patients, let alone fewer death patients (approximately <100 death patients) in this study. Thus, these logistic regression analysis models were too overfitted, the statistical result may not be accurate, and it needs another larger cohort to validate their risk factors's results.

Secondly, the author could reduce the number of covariates by making the comparison between the dead and survival groups, screening the reduced covariates and using these reduced variables to do the logistic regression analysis. It can get more reliable results.

Thirdly, did these gastric cancer patients receive chemotherapy in this study, while this chemotherapy could severely influence the results of haematological parameters? As staff worked in the clinical laboratory, we know that the haematological parameters were severely influenced by chemotherapy drugs. The same patient may have different results on the same day. Thus, the author should discuss the chemotherapy drug's effect on these covariates, such as the influence on the lymphocyte subsets and other haematological parameters. Without considering this influence, these gastric cancers may have different baselines, resulting in inaccurate risk factors results.

Finally, we show great respect for Du's outstanding study despite these comments.

先生，我们饶有兴趣地阅读了杜的研究[ 1 ]。在这项研究中，他们旨在研究淋巴细胞亚群、肌肉减少症和肌脂肪变性对接受胃癌手术的患者临床结果的预测能力。此外，还探讨了 CD3+/CD4+ 细胞与肌脂肪变性相关的预后意义。基于他们的统计分析，他们发现CD3+/CD4+细胞、CD4+/CD8+比率和CD19+细胞可以指示胃癌手术患者的临床预后。肌肉减少症和肌脂肪变性也与患者的预后相关。尽管结果令人鼓舞，但在这封信中，我们对这项研究的结果提出了一些统计方面的担忧。

首先，我们注意到表 2 或表 3 中有 38 个变量用于无进展生存期或总生存期的单变量和多变量分析。因此，根据基本统计规则要求每10个结果需要1个协变量进行预测分析[ 2-5 ]，杜教授的研究中分析38个协变量至少需要380名死亡胃癌患者。相比之下，本研究中总共只有190名胃癌患者，更不用说死亡患者了（约<100死亡患者）。因此，这些逻辑回归分析模型过于拟合，统计结果可能不准确，需要另一个更大的队列来验证其危险因素的结果。

其次，作者可以通过死亡组和存活组的比较，筛选减少的协变量，并利用这些减少的变量进行逻辑回归分析，从而减少协变量的数量。可以获得更可靠的结果。

第三，本研究中这些胃癌患者是否接受了化疗，而化疗会严重影响血液学参数的结果？由于工作人员在临床实验室工作，我们知道血液学参数受到化疗药物的严重影响。同一个病人在同一天可能会有不同的结果。因此，作者应该讨论化疗药物对这些协变量的影响，例如对淋巴细胞亚群和其他血液学参数的影响。如果不考虑这种影响，这些胃癌可能具有不同的基线，导致危险因素结果不准确。

最后，尽管有这些评论，我们还是对杜的杰出研究表示极大的敬意。