admin 很多临床研究都涉及统计学方法的应用。发表SCI论文时，准确、详细地描述文章中使用的统计学方法非常重要。
如描述不妥，可能会导致审稿人直接质疑统计学方法及其结论，这对于文章的发表是致命的。在阅读文章的时候，看到几篇文章的统计学方法的描述得很好，在这里和大家分享一下。
1. Papadopoulos,E.I., et al., L-DOPA decarboxylase mRNA levels provide high diagnostic accuracyand discrimination between clear cell and non-clear cell subtypes in renal cellcarcinoma. Clinical Biochemistry, 2015. (IF= 2.229)
”
这篇文章影响因子不高，但推荐文章的统计学描述，非常有逻辑。文章的试验流程分三大块：（1）检测了肾细胞癌癌组织和邻近的非癌组织中L-DOPA脱羧酶（L-DOPA decarboxylase, DCC）的mRNA水平;（2）评价了其表达水平在诊断肾细胞癌中的精确性;（3）分析了DCC mRNA的表达水平与肾细胞癌临床特征之间的相关性。以下是文中对统计学方法的描述：
A pairwise comparison between the relative expression levels of DDC in cancerous and adjacent non-cancerous tissue specimens was performed by applying the non-parametric Wilcoxon signed-rank test. The diagnostic accuracy of DDC levels in RCC was evaluated using receiver operating characteristic (ROC) curve analysis and logistic regression. The statistical relationship between DDC mRNA levels and clinicopathological features of the RCC patients was estimated as appropriate per variable type. In detail, the statistical significance between DDC levels and continuous variables was determined by Spearman correlation coefficient. Differences observed in DDC expression among the different groups of categorical clinicopathological parameters were analyzed by either the non-parametric Mann–Whitney U test for binary variables or the respective Kruskal–Wallis omnibus test for those consisted of several independent groups. In the latter case,variations of statistical significance were further subjected to post hoc pairwise analysis by applying the Mann–Whitney U test and Bonferroni's correction, while ordinal categorical variables were also analyzed by the Jonckheere–Terpstra test.
文章是按照试验设计的流程来做相应的描述的。针对试验设计（1）的统计学方法描述为第一句。此处比较的肾细胞癌病人的癌组织以及邻近非癌组织中基因的表达水平，属于两配对样本间的连续型变量的差异比较分析。由于数据不符合正态分布，采用了非参检验中的Wilcoxon秩和检验的方法。此处的描述语句也可以借鉴一下，即“A pairwise comparison between 检验变量in 样本A and 样本B was performed by applying 某种统计方法”。
针对试验设计（2）的描述为第二句。本文使用ROC分析和logistic回归分析来评价诊断的精确性。这里可以扩展一下，指出是用ROC分析的曲线下面积（AUC）及其95%置信区间和统计分析的p值来评价DCClevel的诊断价值；用logistic回归的OR值来进一步量化。
接下来的是对试验设计（3）的描述。根据变量的类型使用不同的统计学方法来研究DCCmRNA的表达水平与肾细胞癌临床特征之间的相关性：若临床特征为连续性变量，使用Spearman相关分析；若为二分类变量，则使用Mann–Whitney U检验；若为多分类变量，则使用Kruskal–Wallis多项检验以及事后两两检验；若为有序变量，则使用Jonckheere–Terpstra检验。
文中对使用的统计学方法做了较为详尽的描述，这样会给审稿人留下认真严谨的印象，对于文章的接收无疑是锦上添花。值得借鉴。
2. Zhang,J.X., et al., Prognostic and predictive value of a microRNA signature in stageII colon cancer: a microRNA expression analysis. LancetOncol, 2013. 14(13): p. 1295-306. (IF=24.725)
”
We investigated the prognostic or predictive accuracy of each feature and multi-miRNA-based classifier using time-dependent receiver operating characteristic (ROC) analysis. We used the area under the curve at different cutoff times to measure prognostic or predictive accuracy. We used R software version 3.0.1 and the “survivalROC” package to do the time-dependent ROC curve analysis.
We compared two groups using the t test for continuous variables and chi-square test for categorical variables. For survival analyses, we used the Kaplan-Meier method to analyse the correlation between variables and disease-free survival, and the log-rank test to compare survival curves. We used the Cox regression model to do the multivariable survival analysis, and Cox regression coefficients to generate nomograms.
文章使用R软件的survivalROC软件包（R software version 3.0.1 and the “survivalROC” package）做的时间依赖ROC曲线（time-dependent receiver operating characteristic），分析了miRNA对 stageII结肠癌不同时间（at different cutoff times）复发的预后价值。评价指标是曲线下面积（area under the curve）。
文中还使用了K-M生存分析及其log-rank test研究了各项指标是否会影响无病生存率。并且使用了Cox比例风险回归模型（Cox regression model）做了多因素分析。
3.Han, W.K., et al., Urinary biomarkers in the early detection of acute kidney injury after cardiac surgery. Clin J Am Soc Nephrol, 2009. 4(5):p. 873-82 (IF=5.25)
”
Continuous variables were compared using the t test or the non-parametric Mann–Whitney U test, as appropriate. For analysis of single biomarkers, receiver-operating characteristic (ROC) curves were generated and the areas under the curve (AUCs) calculated and compared using the method of Hanley and McNeil for comparison for a single time point. Two-tailed P values < 0.05 were considered statistically significant. For joint analysis of multiple biomarkers, a fitted multiple logistic regression model was used to give maximum sensitivity and specificity (Table 1). Statistical analyses were performed using SAS Version 9.1 (SAS Institute, Cary,NC) and MedCalc Version 9.3.1 (Med-Calc, Inc., Mariakerke, Belgium).
文章研究的是心脏外科手术后发生急性肾损伤的早期诊断标记物。首先是对差异显著性分析的描述：连续变量间的比较采用t检验或Mann–Whitney U非参检验的方法。这里的asappropriate的潜在含义就是符合正态分布的数据用t检验，不符合的话用Mann–WhitneyU检验。当然，潜在含义写出来是更好的。然后说明，对单个诊断标记物的检验，使用的是ROC分析的曲线下面积（AUC），其中AUC的估计采用的是Hanleyand McNeil非参方法。给出检验阈值是0.05。而对于多个标记物的联合诊断，采用多因素logistic回归模型的分析给出最大灵敏度和特异度的拟合方程。最后指出本文中所用到的软件（SAS以及MedCalc），并给出版本号和软件信息。