Llorca Javier, Martínez-Sanz Fernando, Prieto-Salceda Dolores, Fariñas-Alvarez Concepción, Chinchón M Verónica, Quinones Dolores, Delgado-Rodríguez Miguel
Division of Preventive Medicine and Public Health, University of Cantabria School of Medicine, Santander, Spain.
J Glaucoma. 2005 Jun;14(3):190-5. doi: 10.1097/01.ijg.0000159124.57112.69.
To study the quality of controlled clinical trials on glaucoma.
Two hundred and twenty-six clinical trials published between 1980 and 1999 were selected from seven international ophthalmological journals. Their quality was assessed by four researchers with epidemiological skills using a structured questionnaire.
Sample size was pre-estimated in 34 (15.0%) papers, which were of greater size (P = 0.05). Randomization was performed in 98.2% of the trials, although the procedure of randomization was scarcely reported. Masking was reported in 56.6% of the papers, and was more frequent in medical treatments (P < 0.001). The basal characteristics of the groups were compared in 139 papers (61.5%). Patient losses during the follow-up period were fully described in only 27 trials. Intention-to-treat analysis was used in 17 (7.7%) papers. Most trials reported P values, but a measure of effect (mean, proportion, or relative risk) appeared in only 16 trials (7.7%). Trials performed in the US more frequently compared baseline characteristics of the groups (P = 0.03), described the patient flow (P = 0.04), and used adequate statistical procedures (P = 0.03). Those trials that included a statistician or an epidemiologist among the authors were more commonly blinded (P = 0.06) and they always avoided the analyses of subgroups (P = 0.006). Several methodological issues have improved throughout the studied period.
Several methodological characteristics should be improved when reporting a clinical trial on glaucoma. Using a checklist like that suggested by the CONSORT can help to achieve this.
研究青光眼对照临床试验的质量。
从七种国际眼科杂志中选取1980年至1999年间发表的226项临床试验。由四名具备流行病学技能的研究人员使用结构化问卷对其质量进行评估。
34篇（15.0%）论文预先估计了样本量，这些论文的样本量更大（P = 0.05）。98.2%的试验进行了随机分组，尽管随机分组过程很少被报告。56.6%的论文报告了设盲情况，在药物治疗中更常见（P < 0.001）。139篇论文（61.5%）比较了各组的基线特征。仅27项试验充分描述了随访期间的患者流失情况。17篇（7.7%）论文使用了意向性分析。大多数试验报告了P值，但仅16项试验（7.7%）出现了效应量（均值、比例或相对风险）。在美国进行的试验更频繁地比较各组的基线特征（P = 0.03），描述患者流程（P = 0.04），并使用适当的统计方法（P = 0.03）。作者中包括统计学家或流行病学家的试验更常采用设盲（P = 0.06），并且总是避免亚组分析（P = 0.006）。在整个研究期间，几个方法学问题有所改善。
报告青光眼临床试验时，应改进几个方法学特征。使用CONSORT建议的清单有助于实现这一点。
