Evidence-based medicine — structured SAQ (15 marks)

(a) The well-built PICO question (2 marks)

One mark for naming each of P, I and C; one mark for O. Full marks require specificity. [1]

• Population — pregnant women at less than 28 weeks without prior preterm birth, no comorbidities (the patient), or if you are framing a broader question — pregnant women without prior spontaneous preterm birth.
• Intervention — the specific probiotic preparation the patient is asking about (name the strain or formulation).
• Comparator — placebo or no probiotic.
• Outcome — preterm birth under 37 weeks (and secondarily, neonatal morbidity, maternal satisfaction, side effects, cost).[1]

(b) Hierarchy of evidence, most appropriate design, and appraisal tool (3 marks)

One mark per element. [2][3][7]

• Hierarchy — top of the hierarchy for therapy: systematic review and meta-analysis of randomised controlled trials; second-best is a single well-conducted randomised trial. Filtered sources (Cochrane, RANZCOG, RCOG, NICE) come before a PubMed search.[7]
• Most appropriate design — randomised controlled trial (RCT) of the probiotic versus placebo, ideally synthesised in a systematic review and meta-analysis.[7]
• Appraisal tool — PRISMA 2020 for reporting of the systematic review; AMSTAR 2 for the methodological quality of the review; RoB 2 for the risk of bias within the individual RCTs; CONSORT 2010 for trial reporting.[2][3][7]

(c) Calculate ARR and NNT (3 marks)

One mark for the calculation, one mark for the answer, one mark for interpreting in context. [7]

• RR = 0.78, baseline 8% → risk in probiotic arm = 0.08 × 0.78 = 0.0624 (6.24%).
• ARR = 0.08 − 0.0624 = 0.0176, that is, 1.76 percentage points.
• NNT = 1 / ARR = 1 / 0.0176 ≈ 56.8, that is, NNT ≈ 57.
• Interpretation: the probiotic prevents one preterm birth per approximately 57 women treated. State RR, ARR and NNT together; do not quote RRR (1 − 0.78 = 22%) without the absolute baseline.[7]

(d) Funnel plot and I² under GRADE (3 marks)

One mark per meaningful interpretation. [4][5][6]

• Funnel plot asymmetry suggests possible publication bias — small negative trials missing from the literature. Under GRADE this is a reason to downgrade for publication bias.[4][5]
• I² of 12% indicates low statistical heterogeneity — variation across trials is mostly chance, not real differences between populations or interventions. This is reassuring; under GRADE it does not trigger a downgrade for inconsistency.[4][6]
• Overall GRADE — starts at high (RCT evidence) and is likely downgraded once for publication bias to moderate certainty. Check also risk of bias in individual trials (RoB 2) and indirectness if your patient differs from trial populations.[4]

(e) Integrate evidence with patient values (4 marks)

One mark per substantive point; show the structure of shared decision-making, not just the conclusion. [7][8]

• Applicability check — confirm whether the trial population matches your patient (risk status, gestational age, formulation, duration, baseline risk). The trial's baseline rate (8%) is closer to a higher-risk population; the absolute benefit in a low-risk woman (5–6% baseline) is correspondingly smaller and the NNT would be higher.[7]
• GRADE recommendation direction — likely a strong recommendation in favour for higher-risk women, weak/conditional recommendation for low-risk women like yours where the absolute benefit is smaller and values/preferences matter.[4]
• Share the decision — present the evidence in plain language: the relative reduction (~22%), the absolute benefit (around 1–2 per 100), the NNT (~50–100 for a low-risk woman), uncertainty around long-term outcomes and any side effects, cost and burden. Ask what matters to her — safety, doing something proactive, avoiding interventions, cost.[7]
• Document and review — record the PICO, the evidence considered, the GRADE certainty, the recommendation direction, the patient's informed preference and a review date. Audit whether outcomes in your service match the trial.[4][7]