ONQIVA includes an exploratory survival analysis, a cross-sectional prediction benchmark, and a fictional testing-allocation simulation. They use different data and answer different questions.
The public-data cohort includes 515 adults who reported a previous cancer diagnosis, had a measured 25(OH)D result, and were eligible for mortality linkage. There were 35 observed deaths and a maximum of 37 months of follow-up.
Vitamin D is entered as a continuous measurement, scaled per 10 ng/mL. The interactive curve shows a scenario from this fitted model. It does not estimate a person’s future.
h(t | x) = h₀(t) exp(β₁ age + β₂ sex + β₃ 25(OH)D / 10)The estimate describes an association in this sample. It cannot establish cause, treatment benefit, or who should be tested.
The analysis is observational and unweighted. It uses a small number of deaths, short follow-up, and self-reported cancer history. Confounding, selection, and differences in cancer type or treatment may affect the association.
This separate experiment predicts whether measured 25(OH)D was below 20 ng/mL using age, sex, race/ethnicity, examination season, and income-to-poverty ratio. It does not predict cancer outcomes or decide who receives a test.
These results show modest discrimination in this cohort. Cross-validation is internal; independent validation and calibration are still needed.
The clinic exercise uses fictional inputs and fixed scoring logic. It is not fitted to NHANES, the survival model, or a clinical dataset. Its purpose is to make assumptions visible and open to critique.
Age, body mass index, reported routine, and available test capacity are example values, not patient records.
Each factor contributes a visible number of points. The rule is illustrative and has not been clinically validated.
The example compares the rule with random allocation. It cannot establish improved access or health outcomes.
The next work is independent statistical review, clearer uncertainty analysis, and a carefully defined research question. Molecular data and clinical pilot work remain future possibilities, not current product features.
Request a methods review