A targeted therapy arrives with everything in its favor. Strong trial data, a clear biological rationale, and a defined patient group that should respond. The forecast looks excellent. Then uptake stalls, and the reason has nothing to do with the drug. The companion diagnostic that identifies eligible patients is not being ordered often enough in the places that treat most of them.
Stelios Tzellos of the UK has seen this pattern hold across the oncology market. As a professional in business insights, analytics, and oncology marketing at AstraZeneca, with earlier roles at GlobalData and IQVIA, he treats diagnostic adoption as one of the most underweighted variables in pharmaceutical forecasting.
The Assumption That Quietly Breaks Forecasts
Many forecasts treat biomarker testing as a solved problem. If a drug requires a specific mutation, the model assumes that patients with that mutation will be found and treated accordingly. The biology supports the connection, so the spreadsheet treats it as given.
The clinic does not work that way. A community oncologist juggling dozens of patients may not order the test for every eligible case. Tissue samples may be insufficient. Turnaround times may push a decision past the point where the result matters. Reimbursement for the test itself may be unclear. Each of these frictions reduces the number of identified patients, and every identified patient is a precondition for a prescription.
Where Adoption Actually Lags
The gap is widest outside academic centres. Major cancer hospitals test routinely and sequence broadly. Community practices, which treat the majority of patients in many markets, adopt more slowly. They face cost pressures, workflow constraints, and less access to the latest testing platforms. A forecast that assumes academic-centre testing rates across the whole market will overstate the eligible population badly.
Tzellos saw this dynamic during his time at IQVIA, working on oncology disease insights with global clients. The drugs that met their forecasts were often the ones whose commercial teams understood that the diagnostic and the therapy had to grow together. Investment in test education and access was not a side project. It was part of the launch.
The Science Behind the Test
Understanding why a diagnostic matters requires understanding what it measures. A biomarker is not a simple yes-or-no switch. Expression levels vary, testing methods differ in sensitivity, and the same mutation can carry different weight depending on the tumour context. Tzellos studied molecular biology at Imperial College London, where his doctoral research examined gene regulation in Epstein-Barr virus. That background gives him a working feel for why two patients who both appear eligible on paper might respond very differently.
This matters for forecasting because it changes how the eligible population should be defined. The number is not everyone with the disease. It is everyone with the disease who gets tested, whose test returns a usable result, and whose result points toward the therapy.
Forecasting With the Diagnostic in View
The correction is straightforward to describe and hard to execute. Build the forecast around realistic testing rates rather than theoretical eligibility. Segment the market by care setting, because academic and community testing behaviour differ. Track diagnostic adoption as a leading indicator of drug uptake, not a trailing one.
At GlobalData, Tzellos built epidemiology models for oncology and haematology indications including Hodgkin’s lymphoma, where eligible and treated populations rarely match. At AstraZeneca, he leads cross-functional projects that connect medical, commercial, and market access perspectives, which is where diagnostic strategy has to live if it is going to work.