We have seen many published reports and data sets portending to be ‘what you need for a robust oncology forecast’, which seem to include a lot of erroneous bulk in terms of dataset that are peripheral to what is actually needed. These additional data sets just seem to be there to increase the page count and justify the price of said report, rather than being of actual necessary value.
How many patients are there? Let’s start here. Oncology is different to most other diseases in that it is focused on and reports mostly incidence of a cancer, rather than prevalence. It is critical to understand the difference between these two measures since they are fundamentally different and will have a radical impact on the final forecast depending on which one is used and in what way.
Incidence vs. prevalence. Incidence is the number of new cases of a condition in a period of time. In the case of cancer data, incidence tends to be reported as number of new cases per 100,000 per year. If using registry data or data provided by Globocan of CIS aggregated registry data, these will be representative of the DIAGNOSED and reported cases. It is likely that additional cases are at large in a country, but if these patients are undiagnosed, you should question the validity of including them in a forecast model (unless you can run a campaign to increase diagnosis / disease awareness and increase the diagnosed population). Prevalence is the number or proportion of a population that has a condition at any given point in time. For example, 0.7% of the UK population has rheumatoid arthritis. This is a proportion of the UK population with RA. That prevalence estimate INCLUDES any newly diagnosed patients that have been diagnosed at any point in time prior to this measure being taken. Thus it includes the incident population within its definition.
The two measures are of course linked – roughly speaking; prevalence will be equal to incidence multiplied by duration of survival with the disease. Thus if there are 10 new patients diagnosed with “Disease X” each year and they are expected to survive for ~ 15 years with the disease then there should be ~ 150 prevalent cases of “Disease X”. There are exceptions and nuances to this, but largely the rule remains.
Unlike most traditional drug markets which tend to be chronic and therefore use prevalence as the most appropriate patient metric on which to base a forecast valuation, oncology is considered a more ‘acute’ disease. Patients trend to be diagnosed and subsequently receive immediate treatment (by and large, but there are exceptions) which is then associated with an outcome or a progression to another treatment option. In some cancers life expectancy is short (less than a year) meaning that prevalence is not a very helpful or useful metric to use in these markets for forecasting purposes.
The other point of note that differentiates oncology from other diseases is that a patient that has had a diagnosis of a cancer but has been cured and is disease free is still technically included within the definition of ‘prevalence’. This can lead to a mis-estimation in oncology markets of the actively treated / treatable population for a given therapy.
Extreme care should be taken when assessing the eligible population for treatment!
Other terms rife in the data surrounding oncology: of course, incidence and prevalence are not the only measures that are common when considering oncology patient data. If you search some of the registry sites you will come across terms such as 5 year survival, 5, 10 and 15 year prevalence etc. While these are metrics that are routinely reported and people seem to seek out, they have a limited amount of applicability in terms of valuing an oncology product. Please bear in mind that epidemiology and public health have a remit outside generation of patient numbers for commercial forecasts. Many of these measures are generated to compare different diseases, impact on life years lost / quality of life, estimating mortality or impact of different cancers etc. These metrics drive different decisions and comparisons, so take care when asking for these data if you are unsure what they are, how to apply them or what their utility really is to what you are needing to do…
Please also remember that within oncology, patients are staged at the point of diagnosis. Even if their disease progresses, they do not become another incident patient at another stage. For example, a patient is diagnosed with Stage II renal cell carcinoma. They may receive a nephrectomy (surgery to remove the tumour from the primary site) plus some chemotherapy or radiotherapy and enter remission. If the tumour progresses, the patient is still a Stage II patient, but they have now progressed and may be eligible for systemic treatment. These patients are NOT included within a later stage, even through had they been diagnosed in their present state they might meet the criteria for a diagnosis of a higher stage than their initial grade.
So, in essence for a good oncology forecast, the data you will need the most are:
- Good INCIDENCE data for DIAGNOSED patients by stage or relevant defining factor that would predispose a patient to receiving your treatment
- An estimate of progression rates by stage and over time often associated with different treatment options (e.g. Kaplan Meier curves for patients treated with surgery vs. radiation therapy vs. chemotherapy).
You will most likely NOT need:
- 5 year survival data (unless you’re working in a public health setting and comparing survival rates of different populations with different cancers) or 5, 10 or 15 year prevalence data (these will most likely not be treatment active patients unless you are considering a long term cancer such as prostate or breast where treatment durations can be in excess of 5 years).
If your issue is that your likely indication is within a relapse / refractory treatment setting then actually what you need to know is the number of patients starting in 1st line (which can be derived from the incident population in the relevant disease stage plus an estimation of relapse / progression of earlier disease stages) that then fail on first line and become relapsed or do not respond to their first line treatment. If this is the case, then clinical trials results from studies can help to estimate the progression rates in this case. But be aware that not all patients coming off first line treatment will get a second line treatment.