New Zealand, 1986–1991. The evidence is arriving. So is the opposition.
"A year before I was born, my parents lost their healthy three-month-old son—their third child—to 'crib death.' Without a remarkable study, you or your children might have shared my late brother's fate."
The SIDS case is often described as a triumph of observational epidemiology. But in real time, the decision to act was not obvious — it was contested. At each evidence stage, you will face a credible dissenting voice raising a legitimate scientific objection, and a stakeholder pressure creating a different kind of noise. Your task is to engage with both before making a recommendation.
Each stage has five steps that must be completed in sequence. None can be skipped.
The dissenting voices in this exercise are not strawmen. Each raises a concern that would be considered methodologically legitimate in many contexts. The analytical challenge is not to dismiss them, but to weigh them against the specific features of this evidence body — and to distinguish scientific caution from obstruction.
1. Read the evidence · 2. Rebut the dissenting voice · 3. Classify the stakeholder pressure · 4. Assign the asymmetric cost · 5. Make your recommendation
Each step unlocks the next. Feedback after each stage addresses all five inputs.
The pattern in the records
What Mitchell found
Dr. Ed Mitchell, the most junior member of his department, is assigned to review infant deaths occurring outside the hospital — deaths that hospital physicians never saw because these patients never arrived alive.
Reviewing the records, he finds a striking and repeated pattern: otherwise healthy infants who went to sleep and did not wake up. Dr. Shirley Tonkin, who had spent years visiting bereaved families, noticed that many apparently healthy babies had been placed to sleep on their stomachs. Dr. Susan Beal had made the same observation independently.
At Dr. Tonkin's insistence, sleep position is added to Mitchell's review. The review confirms a strong pattern: many infants who died suddenly had been placed prone.
A consistent clinical pattern across independent observers: prone sleeping and sudden infant death appear to co-occur. But a pattern alone does not establish cause. You do not yet know whether all babies are put to sleep prone — the association could be an artefact of universal practice.
Based on this evidence — and having engaged with the objection and the pressure — what do you recommend?
The case-control study — year one data
What the data showed
A three-year case-control study launches, comparing infants who died suddenly to infants who did not. This design directly addresses Prof. Pemberton's objection: it controls for the possibility that prone sleeping is universal, by measuring rates of prone sleep in cases against a matched group of living controls.
At the end of year one, Mitchell takes the data on sabbatical to London. He is himself skeptical: "As a pediatrician, I couldn't believe something as simple as placing babies prone could increase the risk of sudden death substantially. Basically thought it was rubbish."
He shows the data to a prominent statistician. On the back of an envelope, the statistician calculates that prone sleeping accounts for more than half of all SIDS deaths. It is rare for a single risk factor to account for more than half of deaths from any cause.
A well-designed case-control study is underway. Year-one data suggest prone sleeping may account for the majority of SIDS deaths — an unusually strong association. The study is not yet complete — two more years of data remain.
Based on the evidence now available — and having engaged with Dr. Stanmore and the political crossfire — what do you recommend?
The mechanism question — and the confounding challenge
What remains unknown
Mitchell reflects: "As a pediatrician, I couldn't believe something as simple as placing babies prone could increase the risk of sudden death substantially."
The precise biological mechanism by which prone sleeping causes sudden infant death is not known. It remains unknown to this day. However, biological implausibility weakens considerably once you accept that sleep mechanics could affect airway function and that a prone infant's ability to rouse in response to physiological stress may be compromised.
The intervention — placing infants to sleep on their backs — is simple, low-cost, and has no plausible mechanism of harm.
A rigorous case-control study shows an unusually strong association. The mechanism is not fully understood, but biological implausibility does not hold once sleep mechanics are considered. The proposed intervention carries no plausible risk of harm.
Based on everything now available — and having engaged with the confounding objection and the media pressure — what do you recommend?
What an RCT would have required
The RCT that could not be conducted
An RCT to confirm that prone sleeping causes SIDS would have required randomizing thousands of infants to sleep on their stomachs — deliberately exposing them to a risk that the existing evidence already suggested was substantial.
The trial would have taken years. Once evidence about prone sleeping became public, adherence in the control arm would have been impossible to maintain. Randomization was not merely inconvenient — it was ethically problematic. The evidence could not be blinded. The time frame was not feasible.
The New Zealand government did not wait. It launched the Back to Sleep campaign, educating parents to place infants on their backs. SIDS deaths declined spectacularly — by more than half. To this day, no one knows exactly why prone sleeping causes sudden infant death.
The New Zealand government acted on: a well-designed case-control study; an unusually strong association; no biological implausibility once sleep mechanics were considered; a simple, low-cost intervention with no plausible harm; and an RCT that was ethically and practically impossible to conduct. The dramatic decline in SIDS deaths confirmed the hypothesis. "Waiting for the RCT" was not a cautious position — it was a choice with a cost measured in preventable infant deaths per year of delay.
The three dissenting voices you faced represent a progression: procedural caution (no comparison group), statistical caution (interim data instability), and epistemological caution (confounding without mechanism). Each is a legitimate scientific concern in the abstract. What made them insufficient here was not that they were wrong in principle — it was that the specific features of this evidence body addressed them. The case for action did not require dismissing the objections. It required showing why they did not apply.
You have engaged with three dissenting voices and three stakeholder pressures. Looking back across your decisions: did any of the objections actually change your position? Which pressure was hardest to set aside — and why? What does that reveal about how you apply an evidentiary threshold under opposition?
Your deliberation record
The most important question for evaluating evidence is not "was this an RCT?" but "is this evidence rigorous enough for this decision?" The dissenting voices you faced were not wrong to raise their concerns — they were wrong to treat those concerns as decisive in the face of an unusually strong association, a no-harm intervention, and an RCT that was structurally impossible. The SIDS case is not a story about ignoring caution. It is a story about knowing when caution becomes obstruction.
Program-based evidence and evidence-based programs are both forms of technical rigor — the difference is sequencing, not standard. The Back to Sleep campaign was itself an epidemiological instrument: the spectacular decline in SIDS deaths was confirmatory evidence for the causal hypothesis that no ethical RCT could have produced.
Choose a public health question to audit
Three of these questions involve interventions for which demands for RCT evidence have been used to argue against action. One is a question for which an RCT is exactly the right instrument. Work through each to develop the analytical skill to tell them apart — the verdict will not always be the same.
As you work through each audit, ask: who benefits from demanding RCT evidence for this question? The RCT fallacy is not only an intellectual error — it is a strategy.
RCT Feasibility Audit
Work through each feasibility criterion for an RCT of this question.
RCT Feasibility Verdict
The absence of RCT evidence and the absence of evidence are not the same claim. The evidence that exists for these questions is not insufficient — it is a different kind of evidence that must be evaluated on its own terms.
Treating a pharmaceutical RCT and a population-level policy question by the same evidentiary standard is a category error that systematically advantages interventions that can be packaged as products over interventions that benefit populations.
Lessons and debrief
The threshold question
Not "was this an RCT?" but "is this evidence rigorous enough for this decision?" Rigor means systematic collection, honest interpretation, and calibrated confidence — whatever form the evidence takes.
Program-based evidence
Implement on the basis of strong observational evidence, monitor outcomes systematically, revise as needed. Not a lower standard — the appropriate standard for a different kind of question.
Scientific caution vs. obstruction
Legitimate scientific caution is specific and addressable. Obstruction uses the form of scientific caution to delay action indefinitely. The distinction lies in whether the concern is proportionate to the evidence and the cost of waiting.
The RCT fallacy
Demanding RCT evidence when it cannot be obtained, then treating its absence as evidence of no effect, is not only an intellectual error. It is a tool of obstruction used strategically to delay action.
Asymmetric cost
Evidence decisions are not symmetric. The cost of acting too early and the cost of waiting must both be made explicit. In the SIDS case, "waiting" meant preventable infant deaths per year of delay — a cost that rarely appeared in the scientific debate.
Humility and generalizability
Even excellent evidence is evidence about a specific time, population, and context. Comstock's call for replication of his own findings was not self-doubt — it was the highest form of scientific integrity.
Discussion questions
1. Of the three dissenting voices you faced, which gave you the most pause — and why? Looking back, was your hesitation justified, or was it a form of being swayed by the authority and confidence of the objection rather than its substance?
2. Dr. George Comstock called for replication of his own landmark epidemiological findings: "That was then, this is now. The bacteria might have changed. The nutritional status of infected people might have changed. The inoculum might have changed. We won't know unless we study it." How does this kind of scientific humility differ from the caution raised by the dissenting voices in this exercise?
3. In your RCT audit, what evidence does exist for your chosen question? Evaluate it on its own terms — not by whether it is an RCT, but by whether it is rigorous enough for the decision at hand.
4. The nasal spray influenza vaccine was supported by several well-designed RCTs showing it superior to the injected vaccine in children aged two to eight. Real-world surveillance the following year showed it was no better — and subsequently completely ineffective against the most common circulating influenza strain. What does this reveal about internal versus external validity — and when is real-world surveillance a more reliable guide than a rigorous RCT?