Evidence-Based Medicine and Critical Appraisal — Written Clinical Reasoning

SAQ 1 — Appraising, Applying and Sharing the Evidence for Anticoagulation in a Complex Patient (20 marks, 30 minutes)

Prompt: A 75-year-old woman (Mrs K) has non-valvular atrial fibrillation, a CHA2DS2-VASc score of 5 (age, hypertension, diabetes, prior TIA), a HAS-BLED of 4 (hypertension, abnormal renal function, prior fall with a fractured wrist), an estimated glomerular filtration rate of 32 mL per minute per 1.73 m² (stage 3b chronic kidney disease), and a recent mechanical fall. The pivotal direct oral anticoagulant trials largely excluded patients with an eGFR below 30 and did not specifically capture patients with recurrent falls. Outline your structured approach, including: (a) the focused PICO question; (b) the appraisal of the applicability of the trial evidence to Mrs K; (c) the computation of the absolute benefit (NNT for stroke) and harm (NNH for major bleeding) for her baseline risk; (d) the GRADE strength and quality of the evidence; (e) the shared decision-making conversation; and (f) the documentation and the follow-up. [1]

Model Answer

(a) The focused PICO question (3 marks): [1]

The question must be focused and answerable. Population: a 75-year-old woman with non-valvular atrial fibrillation (CHA2DS2-VASc 5), stage 3b chronic kidney disease (eGFR 32), and a recent mechanical fall. Intervention: a direct oral anticoagulant (apixaban, given the renal profile and the lower bleeding signal) versus warfarin. Comparator: warfarin, the standard of care, or no anticoagulation if the bleeding risk is judged prohibitive. Outcome: the patient-centred composite of ischaemic stroke prevented (the benefit) against major bleeding and intracranial haemorrhage caused (the harm), over a one-year horizon, and the patient's own priority (remaining independent, avoiding a disabling stroke, avoiding a bleed). A precise PICO makes the search tractable (apixaban versus warfarin in atrial fibrillation with moderate CKD), the eligibility criteria explicit, and the applicability judgement structured. [1]

(b) Appraisal of applicability (4 marks): [1]

The pivotal apixaban trial (ARISTOTLE) demonstrated a reduction in stroke or systemic embolism and a lower rate of major bleeding (including intracranial haemorrhage) compared with warfarin. I would appraise its applicability to Mrs K using the CASP "will the results help my patients" section [1]. Two applicability concerns arise. First, the trial excluded patients with an eGFR below 25 mL per minute; Mrs K at 32 is just inside the eligibility threshold, so the renal subgroup data (apixaban is partially renally cleared, and the dose is reduced at two of three renal criteria) apply, but the marginal renal function increases the risk of accumulation and I would dose at the reduced regimen (2.5 mg twice daily) if any two of age 80 or above, body weight 60 kg or below, or creatinine 133 micromol per litre or above are present. Second, the trial did not specifically capture patients with recurrent mechanical falls; Mrs K's recent fall and her fractured wrist signal a fall risk that increases the absolute risk of traumatic bleeding (including intracranial haemorrhage) but does not negate the stroke-prevention benefit. The trial's relative effect (the RRR is consistent across the renal and elderly subgroups) is likely to apply; the absolute benefit and the absolute harm must be recomputed for her baseline risk. The judgement is that the trial evidence is applicable with these two caveats, and that the relative effect can be applied while the absolute numbers are individualised.

(c) Absolute benefit and harm (4 marks): [1]

The CHA2DS2-VASc of 5 gives an estimated annual stroke risk of approximately 6 to 7 per cent without anticoagulation. The RRR for apixaban versus placebo (and versus warfarin, which itself reduces stroke by about 64 per cent) is approximately 80 per cent versus no therapy. The ARR for apixaban versus no therapy is therefore approximately 6 per cent multiplied by 0.80, equal to 4.8 per cent per year; the NNT is 1 divided by 0.048, approximately 21 — about 21 patients like Mrs K must be treated for a year to prevent one stroke. Against this, the major bleeding risk on apixaban in an elderly patient with CKD and a fall history is elevated; the annual rate of major bleeding is approximately 2 to 3 per cent, and of intracranial haemorrhage approximately 0.3 to 0.5 per cent. The NNH for a major bleed is approximately 33 to 50. The balance favours anticoagulation: a stroke prevented for every 21 treated, against a major bleed for every 33 to 50 treated, and apixaban (with its lower intracranial haemorrhage rate than warfarin) is the preferred agent given the fall risk and the renal function. I would present these numbers to Mrs K in natural frequencies. [1]

(d) GRADE strength and quality (3 marks): [1]

The GRADE rating for apixaban versus warfarin or no therapy for stroke prevention in atrial fibrillation is a strong recommendation on high-quality evidence [2]. The underlying evidence is from large, well-conducted randomised controlled trials (high quality at baseline), with no serious risk of bias, no serious inconsistency, no serious indirectness for Mrs K (with the renal and fall caveats addressed above), and no serious imprecision (the confidence intervals exclude clinically important differences). The recommendation is strong because the desirable effect (prevention of a disabling stroke) clearly outweighs the undesirable effect (the bleeding risk, which is lower with apixaban than warfarin). The strength means I can offer anticoagulation as the default, while still sharing the decision because the fall and the renal function introduce individual considerations.

(e) The shared decision-making conversation (4 marks): [1]

I would use a structured conversation. First, I would establish her understanding of her condition and her priorities (does she most fear a stroke, a bleed, the tablets, or a fall?). Second, I would present the numbers in plain language: without treatment, about 6 or 7 in 100 people like her would have a stroke this year; with apixaban, that falls to about 1 or 2 in 100 — so for every 21 people treated, one stroke is prevented. Against that, about 2 or 3 in 100 would have a major bleed on the treatment, and a smaller number (less than 1 in 200) a bleed in the brain. I would explain that apixaban was chosen over warfarin because it has a lower rate of brain bleeding, which matters given her fall. I would address the fall directly: a fall does not remove the stroke risk, and a presumed safety concern about anticoagulation in fallers is not supported by the evidence (the stroke-prevention benefit persists). I would invite her to weigh the benefit and the harm against her own values, and I would support whichever decision she reaches, documenting the reasoning and offering to review it. [1]

(f) Documentation and follow-up (2 marks): [1]

I would document the PICO question, the trial evidence and its applicability (with the renal and fall caveats), the computed NNT and NNH for her baseline risk, the GRADE strength and quality, the conversation and her values, the decision, and the monitoring plan (renal function every 6 to 12 months, blood count, adherence, bleeding review, and a planned review of the decision if her circumstances change). EBM is a continuous cycle of ask, acquire, appraise, apply, assess [1], and the assessment step is the scheduled review of the outcome.

SAQ 2 — Critical Appraisal of a Meta-Analysis of a New Antiplatelet Agent (10 marks, 15 minutes)

Prompt: A systematic review and meta-analysis of seven randomised controlled trials (4,800 patients total) of a new P2Y12 inhibitor (drug X) versus clopidogrel for secondary prevention after myocardial infarction reports a pooled odds ratio for the composite of cardiovascular death, myocardial infarction or stroke of 0.84 (95 per cent confidence interval 0.74 to 0.96, p equal to 0.01). The I-squared is 48 per cent. A funnel plot shows asymmetry, with smaller studies favouring drug X. Two of the seven trials were funded by the manufacturer of drug X. Outline: (a) the validity assessment; (b) the interpretation of the results; (c) the heterogeneity and its implications; (d) the publication bias and its implications; and (e) your bottom-line conclusion for a patient in your clinic. [1]

Model Answer

(a) Validity assessment (2 marks): [1]

I would apply the CASP systematic review checklist and the PRISMA framework [4]. I would check whether the review asked a focused PICO question (patients with prior myocardial infarction, drug X versus clopidogrel, composite cardiovascular outcome), whether the search was comprehensive (multiple databases, trial registries, the grey literature, no language restriction) and reproducible (two independent reviewers for screening and data extraction, with a published protocol), and whether the quality of the included trials was assessed with the Cochrane risk-of-bias tool. The two manufacturer-funded trials raise a flag (industry-funded trials are more likely to favour the sponsor's drug through design, analysis and selective reporting), so I would scrutinise their risk of bias (sequence generation, allocation concealment, blinding, incomplete outcome data, selective reporting) specifically and would check whether a sensitivity analysis excluding them materially changes the pooled estimate.

(b) Interpretation of the results (2 marks): [1]

The pooled odds ratio of 0.84 means a 16 per cent reduction in the odds of the composite outcome with drug X. The 95 per cent confidence interval (0.74 to 0.96) does not cross 1, so the result is statistically significant at the 5 per cent level. However, the odds ratio overestimates the relative risk for a common outcome (the composite occurred in well over 10 per cent of these patients), so the true relative risk reduction is somewhat smaller than 16 per cent. I would seek the absolute risk reduction and the NNT, which are more clinically meaningful than the odds ratio; if the baseline composite event rate was 12 per cent, the ARR is approximately 1.9 per cent and the NNT approximately 53. [1]

(c) Heterogeneity and its implications (2 marks): [1]

The I-squared of 48 per cent indicates moderate heterogeneity [7] — the trials are not fully consistent in their estimates. I would explore the sources: are there clinical differences (the dose of drug X, the clopidogrel loading, the patient population, the duration of follow-up, the concomitant use of aspirin) or methodological differences (trial quality, funding source, blinding) that explain the variability? I would perform subgroup and sensitivity analyses (for example, by dose or by funding source) and I would expect the authors to have used a random-effects model, which accounts for the between-study variance but does not eliminate the heterogeneity. A random-effects pooled estimate from heterogeneous studies must be interpreted with caution — it represents an average of genuinely different effects, and the estimate for an individual patient depends on which trial they most resemble.

(d) Publication bias and its implications (2 marks): [1]

The funnel-plot asymmetry, with smaller studies favouring drug X, is consistent with publication bias — the small negative trials that would have pulled the estimate toward the null have not been published [4]. This biases the pooled estimate in favour of drug X, and the true effect is likely to be smaller than the reported 0.84. I would check whether the authors searched trial registries for unpublished data and performed a formal asymmetry test (Egger); if the asymmetry persists, I would downgrade my confidence in the pooled estimate.

(e) Bottom-line conclusion (2 marks): [1]

Drug X appears to reduce the composite cardiovascular outcome compared with clopidogrel, but the estimate is modest, inflated by publication bias, derived from heterogeneous trials, and supported by industry-funded studies that warrant scrutiny. Before adopting drug X in my patient, I would seek the absolute benefit (the NNT) and the harm profile (especially bleeding), I would judge whether the largest and highest-quality non-industry-funded trial confirms the benefit, and I would apply a GRADE rating that downgrades the evidence for publication bias and heterogeneity [2]. If the largest high-quality trial confirms a clinically meaningful absolute benefit with an acceptable bleeding NNH, I would offer drug X to suitable patients as a weak-to-moderate recommendation, sharing the decision until the evidence firms up.