A survival curve can refer to remaining alive, remaining free of a defined event, or another specified outcome. Those labels are consequential. A historical study of methamphetamine-associated pulmonary arterial hypertension, or PAH, shows why the endpoint definition must accompany the percentage.
What did the cohort actually measure?
Zamanian and colleagues prospectively studied 90 patients with (meth)amphetamine-associated PAH and 97 with idiopathic PAH at Stanford during 2003–2015. The composite outcome included death, transplantation or hospitalization for right-ventricular failure. Estimated five-year event-free survival was 47.2% versus 64.5%, respectively. The adjusted hazard ratio was 2.04 (95% CI, 1.28–3.25).
The associated group had 52 events: 31 deaths, three transplants and 18 right-ventricular-failure hospitalizations. Thus, event-free survival was not an all-cause-survival percentage. The study was observational and single-center, with treatment practices evolving during recruitment. Dose thresholds for pulmonary toxicity were not established, and administration routes were incompletely captured. Separate hospitalization-code analyses were not estimates of population disease prevalence.
Why is “more than half died within five years” an unsupported rewrite?
Subtracting an event-free percentage from 100 gives an estimated proportion with the defined event, subject to the study’s survival-analysis framework. It does not transform a composite endpoint into death alone. A nonfatal component remains nonfatal when the endpoint is summarized.
Nor can the cumulative count of deaths during a study simply be substituted for a fixed-time Kaplan–Meier estimate. The denominator and observation times must match the calculation being proposed. A reviewer should ask which quantity is being quoted before comparing percentages from different reports.
Is a hazard ratio an individual’s chance of dying?
No. A hazard ratio describes a relative time-to-event comparison under a model. It is neither an absolute probability nor automatically a causal effect. Applying it to a particular person requires additional reasoning about the relevant population, outcome and time horizon.
Adjustment can address specified measured differences. It does not convert an observational exposure comparison into random assignment, or establish that every possible confounder was measured accurately. An expert interpretation should explain what the analysis supports without assigning it a stronger design than it had.
Which clinical question does this study address?
A prognosis comparison among people already diagnosed with a condition is different from the risk of developing that condition among everyone exposed to a drug. To move between those questions, one needs a suitable population at risk and evidence about disease occurrence.
For the same reason, a cohort comparison cannot establish that a positive drug test diagnosed PAH or identified the cause of an individual’s symptoms. Exposure classification, disease diagnosis and prognosis are separate evidentiary steps.
How should the historical result be presented?
Keep the endpoint definition beside the number. Identify the comparator as idiopathic PAH, rather than an unaffected general population. Label the evidence as a historical observational cohort and explain why it may or may not fit the question under review. This article interprets that study’s reporting; its historical disease definitions are not presented as current diagnostic guidance.
Scientific source
Zamanian RT, Hedlin H, Greuenwald P, et al. Features and Outcomes of Methamphetamine-associated Pulmonary Arterial Hypertension. American Journal of Respiratory and Critical Care Medicine. 2018;197(6):788–800. Historical prospective clinical cohort with a separate administrative-data analysis. DOI: 10.1164/rccm.201705-0943OC. Primary full text.