Analysing Epidemiological Data — Pattern Recognition and Risk Factor Quantification

Q1.1 — Relative risk (Heart Protection Study)

RR = risk in simvastatin group ÷ risk in placebo group = 19.8% ÷ 25.2% = 0.786 ≈ 0.79 [1 mark for correct calculation]. This means the simvastatin group had approximately 79% of the risk of the placebo group — about 21% lower relative risk [1 mark for correct interpretation].

Q1.2 — ARR, NNT and interpretation

ARR = 25.2% − 19.8% = 5.4 percentage points (0.054) [1 mark]. NNT = 1 ÷ 0.054 ≈ 19 [1 mark]. Plain language: approximately 19 high-risk patients with existing cardiovascular disease must take simvastatin for 5 years to prevent one additional major vascular event (heart attack, stroke, or revascularisation) compared to placebo. For a high-risk group, NNT = 19 is considered clinically meaningful [1 mark for interpretation].

Q1.3 — RRR and the misleading headline

RRR = ARR ÷ risk in control = 5.4% ÷ 25.2% ≈ 21.4% ≈ 21% [Note: the trial's reported headline figure of ~24% reflects a more refined statistical model; accept 21–24%]. [1 mark for correct method]. Why it could mislead a low-risk patient: the 24% (or ~21%) RRR is accurate but is calculated relative to the HPS patients' baseline event rate of 25.2%, which is very high (existing CVD). If the same drug is given to a low-risk person with an event rate of only ~5% over 5 years, a 21% RRR would translate to an ARR of only ~1% and an NNT of ~100 — far less impressive in absolute terms. A low-risk patient hearing "24% risk reduction" may overestimate their personal benefit [2 marks for complete explanation].

Q2.1 — 24-month survival values

At 24 months: combination immunotherapy ≈ 0.70 (70%); ipilimumab monotherapy ≈ 0.47 (47%) [1 mark for reading both values, ±0.05 acceptable]. Absolute difference = 70% − 47% = approximately 23 percentage points — roughly 23% more patients in the combination group were alive at 24 months [1 mark].

Q2.2 — Diverging curves interpretation

The curves diverge from ~month 6 and continue to separate through 60 months, meaning the survival advantage of combination immunotherapy grows over time rather than diminishing [1 mark]. This is consistent with the mechanism of immunotherapy: nivolumab and ipilimumab block immune checkpoints (PD-1 and CTLA-4), releasing the patient's own T-cells to attack melanoma cells and establish immunological memory. This durable immune response provides continuing benefit over time, producing a progressively wider gap compared to ipilimumab alone, which has a more limited and shorter-lived immune activation [1 mark].

Q2.3 — Censoring tick marks

The vertical tick mark represents a censored patient — a participant who left the study before experiencing the primary outcome (death), either because the study ended, they were lost to follow-up, or they withdrew [1 mark]. Multiple censoring marks indicate that not all participants reached the end of follow-up — a common feature of clinical trials. This is important because survival estimates become less reliable later in follow-up when fewer participants remain "at risk", so caution is needed when interpreting the far-right portion of the curve [1 mark].

Q2.4 — Supported and unsupported conclusions

Supported: At 60 months, a higher proportion of patients who received combination immunotherapy remained alive compared to those who received ipilimumab alone (approximately 52% vs 24%). Not supported: That combination immunotherapy cures advanced melanoma — 48% of combination patients also died within 60 months, and we do not know what happens beyond 60 months. [½ mark each — 1 mark total].

Q3.1 — ARR and NNT for NSAID trial

ARR = 58% − 40% = 18 percentage points (0.18) [1 mark]. NNT = 1 ÷ 0.18 ≈ 5.6, so approximately 6 patients [1 mark]. For every 6 patients treated with the new NSAID instead of standard NSAID, one additional patient achieves ≥30% pain reduction at 10 weeks.

Q3.2 — Methodological limitations (2 marks each)

Limitation 1 — Single-blind design (2 marks): Clinicians who know which patients received the new NSAID may unconsciously rate or document those patients more favourably, or encourage them to report greater improvement during assessments (assessment bias). For a subjective outcome like self-reported pain on the WOMAC scale, assessor knowledge of treatment allocation substantially undermines confidence in the result. Double-blinding (where both patients and all assessors are unaware of allocation) would be required to rule this out [2 marks for identifying the limitation AND explaining how it affects the conclusion].

Limitation 2 — Industry funding / single site / small sample (any one fully explained): (a) Industry funding: the manufacturer of the new NSAID funded the trial, which creates a conflict of interest — industry-funded trials are statistically more likely to report favourable results than independently funded trials, possibly through selective outcome reporting or endpoint choice. This does not invalidate the result but reduces confidence without independent replication [2 marks]. OR (b) Small sample from one clinic: 90 participants from a single Sydney rheumatology clinic is a small, geographically narrow sample. Patients at this clinic may differ systematically in age, comorbidities, or prior treatment history from the broader Australian osteoarthritis population, limiting generalisability [2 marks]. OR (c) Short follow-up: 10 weeks may not be sufficient to capture clinically relevant outcomes for a chronic condition like osteoarthritis — late-onset adverse effects, the sustainability of pain relief, and disease modification over time are all unmeasurable from this trial [2 marks].

Q3.3 — Why not immediately switch all patients

p = 0.048 is barely above the conventional 0.05 threshold — with a small sample (90 patients), there is a higher probability this represents a chance finding or is sensitive to minor analytical choices [1 mark]. Additionally, the trial excluded patients over 75, with severe kidney impairment, or on anticoagulants — precisely the patients most likely to need osteoarthritis pain relief in real practice. The drug's safety and efficacy in these excluded populations is completely unknown, meaning the result cannot be generalised to a large proportion of the real clinical population [1 mark].

Q4.1 — Impact of three design changes

(1) Larger sample (1,200) across 8 hospitals: Increases statistical power to detect a real but modest effect (reduces false-negative risk), and improves generalisability — multiple sites with different patient demographics and clinical practices provide a more representative sample of the broader Australian osteoarthritis population [1 mark]. (2) Double-blinding: Eliminates assessment bias by ensuring clinicians cannot unconsciously influence how they assess or record outcomes based on knowing treatment allocation. For a subjective outcome like pain, this is essential for a reliable result [1 mark]. (3) 24-month follow-up: Captures sustained efficacy (whether pain relief persists), allows detection of late-onset adverse effects, and is more clinically relevant for a chronic condition like osteoarthritis, where long-term management is the goal [1 mark].

Q4.2 — Limitation that cannot be removed

The industry funding conflict of interest cannot be removed by these design changes — the manufacturer still has a financial interest in a positive result [1 mark]. Even a perfectly designed trial can be selectively reported (outcome-switching, publication of only the favourable findings, suppression of adverse event data). The only resolution is independent replication by researchers with no financial ties to the manufacturer, and ideally inclusion in an independent systematic review [1 mark for explaining how this persists despite design improvements].