Biomarkers and oncology forecasting: How to hit a moving target
Peter Mansell investigates why biomarkers and other molecular diagnostic tools may be the most undervalued area in oncology
By on Aug 3, 2010Peter Mansell investigates why biomarkers and other molecular diagnostic tools may be the most undervalued area in oncology
Much of the current boom in the oncology market can be attributed to the emergence of more targeted products that use biomarkers and molecular diagnostics to hone efficacy and limit side effects in better-defined patient populations.
At best, this is a win-win situation for companies, physicians, patients, and healthcare systemsthe right treatment for the right patient at the right dose at the right time.
It is especially important in a climate in which cost-effectiveness and predictability are increasingly the arbiters of drug choice and uptake, even in a category that historically has been spared this level of scrutiny.
Another consequence of the shift towards personalized medicine in oncology is to add layers of complexity to a market in which forecasting is already muddied by issues such as multiple therapy regimens, health technology assessment, pressure for more flexible pricing and reimbursement models, wavering political status, and diminishing elbow room around key indications.
More and more, oncology forecasting will need to take biomarkers into account, which means multiplying the assumptions and scenarios that feed into assessments of individual product potential.
The significance of biomarkers
The significance of biomarkers extends far beyond their ability to segment markets by tailoring therapy to the responsive patients.
They are also intrinsic to streamlining research and development so that new treatments can be brought to market more quickly, more efficiently, and at a lower cost.
Despite the availability of high-profile targeted anticancers such as Gleevec (imatinib) or Herceptin (trastuzumab), drug development in oncology remains painfully slow, note Eric Groves, Jason Hill, and Christopher Ung of Quintiles in a recent white paper.
Only 8% of new molecular entities in oncology make it to market and late-stage failures are common, with approximately six out of ten drugs in development falling by the wayside in Phase III trials.
Stripping out some of that risk calls for tools that can elucidate a drugs mechanism of action and steer selection of the most appropriate patients to enrich the clinical study population, mitigate risks to non-responders, and facilitate earlier assessment of efficacy.
As the Quintiles executives point out, over the past two decades molecular biomarkers have become established components of clinical research in a way few could have foreseen.
At the same time, this rapid evolution is reshaping clinical practice and the oncology marketplace.
Now a particular tumor type is no longer classified by its histological properties alone, Groves, Hill, and Ung comment.
Biomarkers are increasingly used to segment tumor types by molecular profile.
So, for example, breast cancer is now recognized as containing luminal A, luminal B, basal type, erbB2, and triple negative sub-populations, each of which have a different management strategy, and vary in patient prognoses and therapy response characteristics.
The end result is that marketed oncology products live in an environment where the markets size (target population) is continually being affected by evolving patient selection biomarkers, and these effects on the potential use and sales of a drug can be large, the white paper authors add.
The most undervalued area in oncology
Karol Sikora, medical director of CancerPartnersUK, sees biomarkers and other molecular diagnostics as the most undervalued area in a category heading for meltdown under the competing pressures of cost containment, ageing populations, earlier intervention, an explosion of new and expensive therapies, more aggressive patient expectations and advocacy, and hard questions about the real value of incremental survival benefits. (For more on forecasting challenges, see Forecasting for complex diseases.)
These tools can radically alter the cost-effectiveness ratio for cancer therapies and, in doing so, address stakeholders concerns across the board, Sikora believes.
By 2020, Sikora foresees at worst risk prediction in small patient subsets and at best population risk banding for cancer.
Micro-bands will reclassify cancers by targeting specific disease pathways.
There could even be routine molecular phenotyping before prescribing decisions are taken.
That means overhauling pharmaceutical industry strategy in the category, Sikora told eyeforpharmas recent Oncology Market Access Europe 2010 Summit in London.
In the 1990s, it was all about being first in class, achieving market leadership with one or two indications, and spinning it out through lifestyle management.
The next 10 years will require a more subtle approach, with companion diagnostics segmenting the market and driving new line extensions, Sikora predicted.
As things stand, the pool of oncology products with associated biomarkers is still relatively small in terms of patient numbers, notes Dr Robert Ramsey, vice president and chief scientific officer of Kantar Health.
Where biomarkers have become part of standard care, though, such as in breast or colorectal cancer, the impact is substantial.
And in a comparatively minor indication like chronic myelogenous leukemia, the effect of targeting therapy at the BCR/ABL biomarker (Gleevec) has been to extend lifespans from around six months to six or seven years, Ramsey points out.
Moreover, most of the emphasis in oncology drug development is now moving away from what Ramsey calls the blunderbuss approach towards more tailored therapies.
Targeted biomarkers include mutations of the B-RAF gene (melanoma and colorectal cancer), of EML4-ALK (non-small cell lung cancer), FLT-3 (acute myeloid leukaemia) and BRAC 1/2 (variety of tumours). (For more on the related area of biosimilars, see Forecasting the future of biosimilars.)
Biomarker development
To date, though, biomarker development has generally been hit and miss, Ramsey acknowledges.
We dont know enough about the biology of these drugs and what theyre affecting early on in clinical development to know where we should look for biomarkers.
If a drug is targeted at a genetic mutation, it should be detectable, he adds.
But with biomarkers such as HER2 (Herceptin) or oestrogen receptors (tamoxifen, aromatase inhibitors), Were not aiming just at a mutation, were looking at some other aberration in terms of metabolism.
As Ramsey explains, a biomarker such as HER2 provides a measurement along a continuum, rather than a simple yes/no answeras with K-RAS (Erbitux) or EGFR (Iressa) mutations.
Consequently, there needs to be cut-off points indicating a positive result, and so far these have been set arbitrarily.
Even with a yes/no biomarker, where the objective is to determine the presence or absence of mutations, there are complications.
For example, Ramsey notes, tumors are by nature heterogeneous.
So if a biomarker test yields a positive mutation, that raises the question of what proportion of cells in the tumor a positive measure actually represents.
And while current biomarkers that may be predictive of drug response have derived from the tumor cells themselves, there may be other cells present (e.g., endothelial, pericytes, fibroblasts) that support tumour growth.
While drugs such as Avastin have targeted these cells, to date no predictive markers have been found that relate to those drugs, Ramsey observes.
Beware of silver bullets, he warns.
You may have a mutation you can identify, but there may be other mutations present that might nullify it.
For example, Erbitux was approved for colorectal cancer without any K-RAS biomarker designation.
It subsequently emerged that both Erbitux and Vectibix were effective only in the wild-type K-RAS population (around 60% of all CRC patients).
But responses to Erbitux may also be compromised in the 10% of wild-type KRAS patients who carry the B-RAF mutation.
Defining the target population
For all that, biomarker take-up is encouraging.
According to survey conducted last year by Kantar Health in the US, polled community oncologists were using HER2 over-expression to guide treatment selection in 91% of cases, while frequency of use was 90% for hemoglobin or hematocrit, 88% for CD-20 positivity, 69% for 5Q deletion, 76% for K-RAS mutation status and 47% for EFGR positivity.
Payers are taking notice too.
In the US, most of the people who pay the bills in the case of colorectal cancer have said that if you want to use Erbitux and you want to get it paid for, you show that its a wild-type K-RAS patient, Ramsey says.
The starting point for forecasters is to arrive at a proper definition of the target population.
Here the challenge is to make sure every nuance of potential biomarker impact is factored in, including a full understanding of the competitive landscape.
As Ramsey points out, even if youre not segmented yourself, the competition may well be.
At eyeforpharmas recent 4th annual Pharma Forecasting Excellence Europe meeting in Zurich, Ramsey and Kantar Health consultant Michael DeStefano presented some case studies to demonstrate how targeting with biomarkers calls for a tiered approach to oncology forecasts.
One of these studies concerned a Product Y for metastatic colorectal cancer that, like Avastin, was aimed at the vascular system within the tumor and was not geared to any particular first-line colorectal sub-population.
The forecast took into account the existence of products already targeting the wild-type K-RAS sub-population; the pending entry during the forecast period of a drug specifically aimed at the first-line B-RAF mutation population; and the intention to use Product Y as part of a chemotherapy regimen.
What this called for was three forecasts addressing, respectively, the K-RAS mutation population; patients in the wild-type K-RAS population who did not also have the B-RAF mutation; and the B-RAF mutation sub-population of the wild-type K-RAS segment.
Within these parameters, the following assumptions could be made about the market potential of product Y:
- There would be initial penetration into the first-line K-RAS mutation population, commensurate with the assumed safety and efficacy of Product Y compared with Avastin.
- Due to the presence of Erbitux in the first-line wild-type K-RAS population, there would be less penetration here than in first-line K-RAS mutation population.
- When the drug targeting the B-RAF mutation population entered the market, it would steal share from Product Y in the K-RAS/B-RAF subpopulation.
- A percentage of the first-line wild-type K-RAS patients who failed on Erbitux would receive Product Y in an extended-use setting.
- Each of these populations would have differing prognoses, so durations of therapy would need to be adjusted accordingly.
Implications for the oncology marketplace
The introduction of targeted therapies has also scuppered traditional duration of therapy assumptions about oncology products, Ramsey and De Stefano pointed out.
With cytotoxic drugs, duration of therapy comprised a set number of treatment cycles over a set time period.
But targeted agents are often used until progression of disease, and for some of these duration of therapy may be the key to assessment of commercial potential.
There are other ways in which targeted therapies in oncology are changing forecasting assumptions, such as the shift in mode of administration from intravenous to oral, which raises issues of drug concordance and persistence with therapy.
As Ramsey observes, oral therapies in the US fall under part D of the Medicare system, where the patient may have to make a contribution to the cost.
If all of that makes the forecasters job a lot more difficult, the implications for the oncology marketplace as a whole are more upbeat. (For more on oncology forecasting, see The challenge of oncology forecasting.)
As Ramsey and DeStefano indicated, these include reduced drug development costs, faster time to market, access to new indications and market segments, efficacy in refractory patient populations, and better support for premium pricing.
Ultimately, complexity, changing parameters, and incremental advances are part of the deal.
Its a game of inches, Ramsey comments. The gains are hard fought and relatively small, but they all add up.