Systematic Reviews & Meta Analysis

Question: Systematic reviews Meta-analysis by Kim G Bae J-H. Answer: The following questions concern the systematic review meta-analysis by Kim G Bae J-H. Vitamin D and atopic dermatitis: a systematic review and meta-analysis. Nutrition 2016 [in press]. Please download this paper from the Assessment 2 folder and read it thoroughly. Question1 This systematic review and meta-analysis of 4 small RCTs examined the effects of vitamin D on atopic dermatitis (AD) symptoms and reported a pooled effect of 5.81 reduction [95% CI: -9.03 to -2.59] in the SCORAD index (see Fig 3). Based on this forest plot: Question 1.1 Which RCT was the largest? Identify the study by the author name and year of publication and the total sample size. (1 sentence) Answer: Camargo 2014, total sample size 104. Question 1.2 Which study contributed more heavily to the pooled effect size? Identify the study by the author name and year of publication and explain why. (2-3 sentences) Answer: Amestejani 2012, as its pooled effect weight was 43.5% Question 1.3 Given the level of statistical heterogeneity reported (50%) was the use of a random-effects model appropriate? (1 sentence) Answer: Yes, the use of random-effects model was appropriate in this case. Question 2 Which factors may have explained the relatively high levels of heterogeneity? List at least three based on the characteristics of the RCTs included in this meta-analysis (3 sentences). Answer: In eight studies, vitamin D was supplemented orally, whereas in one study the subjects were categorized into two groups based on their exposure to sun (subtropical or temperate). Seven studies used cholecalcifereol, one used ergocalciferol, whereas one looked at the effects of sun exposure. Patient and practitioner blinding or performance bias possibilities were low in six studies, unclear in two studies, unreported in one study. Question 3 Considering the sources of error, what is the main assumption of a random-effects model and how does it differ from that of a fixed-effect model? (2 sentences) Answer: In random-effects model, the eà ¯Ã‚ ¬Ã¢â€š ¬ects of the studies are assumed to be randomly distributed and the central point of this distribution is the focus of the combined (pooled) eà ¯Ã‚ ¬Ã¢â€š ¬ect estimate. In fixed-effect model, the assumption is that a single common (or à ¯Ã‚ ¬Ã‚ xed) eà ¯Ã‚ ¬Ã¢â€š ¬ect underlies every study in the meta-analysis.[1] Question 4 Based on your critical appraisal of the validity of the RCT by Amestejani et al to what extent do you accept the validity of the effects reported by this meta-analysis? Justify. (3 sentences). Answer: Except blinding of outcome risk bias and other bias, the validity of the effects reported by this meta-analysis is acceptable. Random sequence allocation, allocation concealment, blinding of participants and personnel, incomplete outcome data, and selective reporting all have Low risk of bias. Blinding of outcome assessment has Unclear risk of bias due to insufficient information to permit judgement of Low risk or High risk and the study did not address this outcome. Other bias also has High risk of bias as longitudinal studies with different doses of vitamin D would be better suited to address this issue. Large cohort studies and clinical trials with more sample size also warrant better assessment of this issue. Question 5 Considering the range of the SCORAD index (investigate this), the main outcome of the meta-analysis, how do you deem the clinical relevance of the pooled effect? (3 sentences) Answer: Out of the nine studies, only five studies considered SCORAD index as their main outcome. Since rest four studies used different parameters as the main outcome, considering the range of the SCORAD index as the main outcome of the meta-analysis would not be a justified pooled effect clinically. Question 6 In the discussion section of the meta-analysis by Kim Bae the authors mention that dietary factors may affect AD (2014). Specifically they mentioned that according to a meta-analysis of the six RCTs of infants, the combined risk ratio of the prevalence of AD associated with probiotics was lower compared to placebo (note that prevalence of AD is not an appropriate term for the outcome in the context of RCTs). That aside, the pooled risk ratio (RR) reported in this meta-analysis was 0.86 (0.77 to 0.96). Considering this effect size, calculate the number needed to treat for patients with a baseline AD risk of 15% (show the details of your calculations). References Perera Heneghan. EBM 2008; 13(3):96-9.