ANSWER: B

Rationale:

This question tests precise understanding of the mathematical relationship between RRR, ARR, and NNT, and the clinically important distinction between relative and absolute measures of treatment effect. Absolute risk reduction (ARR) = event rate in placebo group − event rate in treatment group = 4% − 2.6% = 1.4% = 0.014. NNT = 1 / ARR = 1 / 0.014 71. Relative risk reduction (RRR) = ARR / baseline event rate = 0.014 / 0.04 = 0.35 = 35%. Note that NNT can also be calculated as 1 / (RRR × baseline event rate) = 1 / (0.35 × 0.04) = 1 / 0.014 = 71 — confirming the calculation in both Option B and Option E, though Option E's interpretation (RRR meaningful only above 20% baseline) is pharmacologically incorrect and not an established threshold. The critical insight is that RRR and ARR convey very different clinical impressions of the same data: the 35% RRR sounds impressive and may drive prescribing enthusiasm; the NNT of 71 — meaning 70 patients receive five years of treatment, its costs, and its adverse effects for no individual benefit, while 1 patient benefits — provides a more honest reckoning with absolute clinical impact that must inform prescribing, guideline development, and patient counseling. Contexts where this distinction is clinically decisive: high-risk populations (e.g., post-MI patients with 20% five-year event rate) would have ARR = 0.35 × 0.20 = 7%, NNT = 14 — substantially more favorable; in a low-risk population (4% baseline) the same RRR yields NNT = 71. The same drug with the same RRR has profoundly different absolute benefit depending on population baseline risk — illustrating why risk stratification before prescribing is essential for rational therapy. Option A incorrectly uses 1/RRR to calculate NNT — NNT = 1/ARR, not 1/RRR. Option C uses the treatment group event rate as ARR, which is mathematically incorrect. Option D uses the placebo event rate alone as ARR and incorrectly claims NNT and RRR are equivalent measures. Option E arrives at the correct numerical NNT but appends a pharmacologically unjustified threshold claim about RRR meaningful only above 20% baseline.

ANSWER: B

Rationale:

This case presents multiple simultaneous prescribing issues that require systematic application of the rational prescribing framework and Beers Criteria in a complex elderly patient. Ranking by clinical urgency: (1) Ibuprofen — the most immediately dangerous prescribing error. NSAIDs in a patient with eGFR 28 create multiple overlapping harms: direct nephrotoxicity worsening CKD toward dialysis; COX-1 inhibition impairing renal prostaglandin-mediated afferent arteriolar vasodilation, further reducing GFR; pharmacodynamic interaction with apixaban (COX-1 inhibition impairs mucosal cytoprotection, creating ulcer risk and GI bleeding that is then inadequately hemostated by Factor Xa inhibition); cardiovascular risk (sodium retention, blood pressure elevation impairing AF management); and Beers Criteria listing (NSAIDs in elderly with eGFR <30 are potentially inappropriate). Acetaminophen at standard doses is the safe substitute. (2) Metformin — at eGFR 28 (approaching the <30 threshold), metformin should be reduced to 500 mg twice daily maximum or discontinued; if eGFR falls below 30, metformin is contraindicated due to accumulation and lactic acidosis risk in the context of renal hypoperfusion events. (3) Apixaban dose assessment — the dose reduction criteria (any 2 of: age ≥80, weight ≤60 kg, serum creatinine ≥133 µmol/L) must be formally assessed; this patient is 78 (just below the age criterion) — the other two criteria must be checked; at eGFR 28, apixaban exposure is increased but dose reduction is based on the specific criteria, not eGFR alone. (4) Bisoprolol — renally eliminated but at eGFR 28 requires dose monitoring, not immediate cessation; it is appropriate for AF rate control and HF if present. (5) Amlodipine — appropriate for blood pressure management; peripheral edema is a known class effect but not a contraindication in CKD elderly patients. Option A incorrectly prioritizes bisoprolol and inappropriately suggests verapamil (non-dihydropyridine CCB combined with bisoprolol risks AV block). Option C is incorrect — amlodipine is not contraindicated in elderly CKD patients; thiazide diuretics have reduced efficacy at eGFR <30. Option D is incorrect — multiple clinically significant and potentially life-threatening prescribing issues exist. Option E is incorrect — the combination of bisoprolol (cardioselective beta-blocker) with amlodipine (dihydropyridine CCB) is not absolutely contraindicated and does not cause complete heart block; the interaction concern applies to non-dihydropyridine CCBs (verapamil, diltiazem) combined with beta-blockers.

ANSWER: D

Rationale:

AED selection in pregnancy requires integration of seizure control imperatives (uncontrolled generalized tonic-clonic seizures carry significant maternal and fetal risk from hypoxia, acidosis, trauma, and premature labor) with individualized fetal teratogenicity risk assessment based on pregnancy registry data — exactly the framework the PLLR is designed to support through its narrative Pregnancy section. Valproate carries the highest teratogenic risk profile of any commonly used AED: major congenital malformations (neural tube defects including spina bifida, cardiac defects, hypospadias, limb defects — overall malformation rate approximately 6–9% at therapeutic doses vs approximately 2–3% background rate); dose-dependent cognitive impairment (IQ reduction of 7–10 points compared to unexposed children — the most clinically devastating long-term outcome); and significantly increased risk of autism spectrum disorder. The PLLR Pregnancy section for valproate explicitly states that it should not be used to treat epilepsy in pregnant women or women of childbearing potential unless other medications have failed — a strong evidence-based prescribing restriction. Lamotrigine has the most reassuring pregnancy registry data among established AEDs: major malformation rate approximately 2–3% (similar to background rate) with no consistent pattern of organ-specific teratogenicity at doses below 300 mg/day; however, lamotrigine plasma concentrations fall significantly during pregnancy (increased renal clearance and UDP-glucuronosyltransferase induction by pregnancy hormones) requiring dose monitoring and adjustment. Levetiracetam has similarly favorable registry data with major malformation rates approaching background, no established pattern teratogenicity, and stable pharmacokinetics in pregnancy (though renal clearance increases requiring monitoring). High-dose folic acid (4–5 mg/day, rather than standard 400 mcg) is recommended for all women with epilepsy planning pregnancy or in early pregnancy to reduce neural tube defect risk — particularly important for AEDs with folate-antagonizing mechanisms. Option A is completely incorrect — valproate has the highest, not safest, teratogenic risk profile. Option B is incorrect — AEDs differ substantially in fetal risk and the PLLR explicitly differentiates them; drug selection must incorporate fetal risk assessment. Option C is incorrect — phenytoin carries well-documented teratogenic risk (fetal hydantoin syndrome: craniofacial abnormalities, digital hypoplasia, growth restriction, cognitive impairment) that remains clinically relevant.

ANSWER: B

Rationale:

This question tests critical appraisal of systematic review evidence — specifically the interpretation of heterogeneity and its implications for clinical applicability of pooled results. I² (I-squared) is a statistic that quantifies the proportion of variability in a meta-analysis that is due to between-study heterogeneity rather than sampling error (chance). An I² of 74% indicates high heterogeneity — meaning that 74% of the observed variation in effect estimates across trials is attributable to true differences between studies, not sampling variation. This is clinically important because high heterogeneity signals that the studies may not be measuring the same treatment effect in the same population — they may differ in: patient populations (age, disease severity, comorbidities), drug doses and regimens, co-interventions, duration of follow-up, and outcome definitions. When I² is high, the pooled effect estimate is a mathematical average of genuinely different effects — applying this average to an individual patient whose characteristics align with one subset of studies may be misleading. The statistically significant result (OR 0.82, p = 0.03) does not resolve the heterogeneity problem — a result can be statistically significant and still be heterogeneous enough to be unreliable as a clinical guide for an unselected population. The appropriate clinical response: examine subgroup analyses to identify which patient types, doses, or contexts show benefit; assess the clinical plausibility of the heterogeneity sources; apply results only to patients whose characteristics match the subgroup demonstrating benefit; await further trials if the evidence base is insufficient. Option A incorrectly interprets I² as a measure of statistical power — I² measures between-study heterogeneity; adequate power is a separate concept. Option C misinterprets the 95% CI as a harm rate — the CI expresses uncertainty around the point estimate of the odds ratio; an OR <1 indicates benefit (reduced mortality), and both bounds of the CI (0.68–0.98) are below 1.0, both indicating benefit. Option D incorrectly states that p < 0.01 is a required threshold for meta-analyses — the conventional threshold is p < 0.05 for statistical significance; p < 0.01 is not a standard meta-analytic requirement. Option E incorrectly defines I² as a measure of publication bias — publication bias is assessed separately using funnel plot asymmetry and Egger's test; I² measures between-study heterogeneity, not publication bias.

ANSWER: D

Rationale:

Clinical decision support systems represent one of the most important infrastructural advances in pharmacovigilance and rational prescribing — but their effectiveness depends critically on implementation quality, specifically the balance between sensitivity and specificity of alerts. Alert fatigue is a well-documented and serious patient safety phenomenon: when CDS systems generate large numbers of low-specificity alerts (flagging every theoretical interaction regardless of clinical significance, severity, or patient context), physicians rapidly learn to override alerts reflexively without reviewing them. Studies consistently show override rates of 49–96% for drug interaction alerts in electronic prescribing systems — meaning the majority of alerts are dismissed without clinical evaluation. This reflexive override behavior means that when a truly clinically significant alert fires (a dangerous interaction, a critical renal dose adjustment requirement, a high-risk pharmacogenomic contraindication), it is treated with the same reflexive dismissal as the dozens of low-significance alerts preceding it — nullifying the potential safety benefit of the CDS system entirely. Effective CDS design addresses this through: tiered alert severity (with different escalation pathways for critical vs advisory alerts); patient-specific context integration (suppressing alerts when the interaction is already acknowledged or clinically irrelevant for this patient's specific situation); evidence-based alert thresholds (alerting only when interaction severity and clinical evidence exceed a meaningful threshold); and monitoring of override rates as a patient safety quality metric. Override rates persistently above institutional targets should trigger CDS refinement rather than acceptance as normal clinical behavior. Pharmacogenomic CDS represents a particularly high-value application when integrated with patient genotype data — alerting prescribers to HLA-relevant drug prescriptions, CYP-metabolizer status relevant drug dosing, and TPMT/NUDT15-relevant thiopurine dosing at the point of prescribing. Option A is incorrect — alert override rates are a meaningful patient safety metric; reflexive overrides without review have been associated with preventable adverse drug events. Option B is incorrect — high override rates due to low-specificity alert design represent a CDS failure, not appropriate physician behavior. Option C is incorrect — mandatory non-overridable alerts for all interactions would paralyze prescribing workflows and are not recommended; CDS should support, not replace, physician judgment. Option E is incorrect — drug interaction and renal dose adjustment alerts are important CDS categories with demonstrated safety benefit when implemented with adequate specificity; limiting CDS to pharmacogenomics alone would miss the majority of preventable drug-related harm.