Interpreting a meta-analysis for a shared decision — OSCE

Station brief (8–10 minutes)

Help the parents decide whether to start the preventive therapy by interpreting the meta-analysis honestly and weighing the guideline recommendation for their child. Address the online 'halves the risk' figure, the review's method and its limitations, the size and precision of the effect, and its applicability to a four-year-old. Do not invent jurisdiction-specific thresholds or specific survival statistics. [10]

Tasks for the candidate

1. Judge the review's method before its result, naming PRISMA, AMSTAR-2, heterogeneity, and the funnel plot. [1] [2]
2. Reframe the 'halves the risk' figure with the absolute baseline, the absolute risk reduction, the number needed to treat, and the confidence interval. [10] [4]
3. Weigh the applicability of adult-trial evidence and a strong guideline recommendation to a four-year-old, and rate the certainty with GRADE. [8] [7]
4. Integrate the appraisal with the family's values through shared decision-making, and arrange follow-up. [8] [10]

Expected performance

Must hit. PRISMA flow and AMSTAR-2 or ROBIS named before the pooled effect; I-squared of 65 percent identified as substantial heterogeneity and the asymmetric funnel plot identified as likely publication bias; absolute baseline and absolute risk reduction given alongside the relative figure; confidence interval read for precision and null-crossing; applicability of adult evidence to a four-year-old weighed and certainty rated down for indirectness with GRADE; decision integrated with values; written summary and follow-up arranged. [1] [2] [10]

Merit. Acknowledges the online figure without dismissing the parent; places the single review in the totality of the evidence; names the indirectness explicitly and downgrades certainty; appraises the guideline with AGREE II and notes the industry funding as a reason for caution; uses teach-back and a decision aid for the preference-sensitive choice. [7] [8]

Fail. Quotes the 50 percent relative figure alone; reads the result before the method; ignores a null-crossing confidence interval; applies adult evidence to a child without weighing applicability; treats a strong recommendation built on low-certainty adult evidence as mandatory; gives no certainty rating, no shared decision, and no follow-up. [4] [10]

Sample candidate structure

"Thank you both for coming in. First, let me be clear about the question: in a preschool child like yours, does this daily preventive therapy, compared with not taking it, reduce the chance of a serious complication. Before I trust the number you read, I need to check whether the study behind it was fair. The review was funded by the company that makes the drug, its studies disagreed with one another more than I'd like, and a check of its funnel plot suggests small negative studies may not have been published — so the headline figure is probably larger than the truth. The 'halves the risk' you saw is a relative figure, and on its own it looks bigger than it is. Let me show you the absolute numbers: out of 100 children like yours, about this many would have the complication without the medicine, and this many fewer with it. The studies were done in adults, and the guideline's strong recommendation rests on them — so I would treat it as a starting point, not a verdict, and rate the certainty down for how far it is from your four-year-old. That makes this a choice where your values and your son's circumstances should guide us, and I'd like to walk you through it together." [10] [8]