[Evaluation of the heterogeneity of systematic reviews on nutrition support for burn patients with Meta-regression algorithm].
摘要
OBJECTIVE: To evaluate the heterogeneity of systematic reviews (SRs) on nutrition support for burn patients with Meta-regression algorithm quantitatively. METHODS: SRs on nutrition support for burn patients since 1980 were searched in 7 databases, including PubMed, EMBASE on CD, Science Citation Index, Cochrane Library, Chinese Biomedicine Database, Chinese Medical Current Contents, and Chinese Journal Full-text Database. SRs were identified with inclusion criteria: the research was interventive SR or Meta analysis of clinical data of burn patients aged equal to or over 18 years who had received clinical nutrition intervention, with regimen including special nutrient given including enterally or parenterally; the outcome indicator was mortality. A two-step method was established to evaluate the heterogeneity of SRs. (1) Qualitative analysis: the publication time of SRs and the trial design, patient type, regimen or route of nutrition therapy, energy intake, and outcome reported by the included clinical trial were analyzed. (2) Quantitative analysis: a mortality Meta-regression model was established by trial design, publication time, regimen or route of nutrition therapy, and energy intake to calculate the expected mortality. Observation/expectation (O/E) ratio and its 95% confidence interval were calculated to identify the heterogeneity among the clinical trials of SRs. RESULTS: (1) Six SRs were retrieved which were published within 10 years. Only three SRs reported two of the three important parameters of nutrition intervention: regimen or route of nutrition therapy, energy intake, and clinical outcome. Totally, 11 clinical trials that gave outcome information were included in the SRs, including 10 RCTs and 1 case control trial. (2) Meta-regression analysis showed that significant heterogeneity existed in 4 RCTs (including five study groups), which were different from the other 7 trials. The patient samples of these 4 RCTs were from different population. No heterog