The End of Automatic Chemo: Inside India’s Homegrown Breast Cancer Risk Test
By Arunima Rajan
Indian oncologists know that roughly 70 percent of women with early, hormone-positive breast cancer will not benefit from chemotherapy, yet in the absence of affordable, local tests, close to 90 percent still receive it, taking on enormous toxicity and cost. Dr Manjiri Bakre, CEO and Founder of OncoStem Diagnostics, built a homegrown risk stratification test on Indian tumour samples and protein markers that predicts which patients need chemo and which patients do not. Dr Bakre did a PhD in cell biology at the Indian Institute of Science, Bangalore. She then did research fellowships at Mount Sinai in New York and UC San Diego. Later on, she worked in Singapore for five years, and returned to India in 2007.
In an interview with Arunima Rajan, Dr Bakre talks about building and validating CanAssist Breast, persuading hospitals to share archival samples, winning over oncologists and insurers, and moving from a single test in one city lab to a kit-based platform that could reshape breast cancer decision-making in multiple countries.
In India, around 90% of early-stage, hormone-positive breast cancer patients end up getting chemotherapy, even though only about 30% actually need it. Western tests like Oncotype DX cost upwards of Rs 3-4 lakh. How did you prove to Indian doctors that this was a real problem?
By the time I started, these kinds of tests had already been around for about 6–7 years, so clinicians were well aware of the problem and value of prognostic tests. Doctors face it everyday but didn’t have an effective solution for our context.
When I spoke to many oncologists early on, they were quite candid. They said they were probably using more chemotherapy than necessary but didn’t have a reliable way to reduce it. Tests like Oncotype DX were known to them and had strong visibility globally, but they were very expensive.
The real gap was that these tests were not developed on our patient population and were not accessible or cost-effective in India. When I asked clinicians if they would use a test we developed if it was reliable, affordable and built on Indian patient data, the answer consistently was yes.
You approached 60 hospitals and got 12 to collaborate. How did you convince them to hand over samples?
For a hospital to collaborate, you need multiple levels of approval, including the pathologist, the department of medical records, the medical and surgical oncologists, and finally the medical director. All of them need to align. Among the first were Manipal Hospital in Bangalore and Jaslok Hospital in Mumbai.
What helped was presenting a strong scientific rationale. My PhD and prior experience were valuable here, and when they asked questions, I was able to answer them with clarity and confidence. With most hospitals, I had to visit three to four times before they could decide. We conducted formal presentations to the scientific review boards and ethical review boards across every hospital we partnered with.
Everything was completely anonymised. No patient identifiers were shared. This was retrospective. Patients had already completed treatment, so there was no risk to them and no impact on their care.
CanAssist Breast analyses five proteins in a tumour, combines that with clinical data, and an AI algorithm produces a risk score from 0 to 100. How was the AI trained, and how was the cutoff of 15.5 arrived at?
We worked closely with our statistician Ljubomir Buturovic who is based in the US. He brought deep expertise in AI and used approaches such as support vector machines to maximise prediction accuracy.
Every test corresponds to a certain percentage risk of recurrence. In the competing Western test, the boundary between low and intermediate risk corresponded to about a 10% chance of breast cancer recurrence over ten years. We wanted to set a slightly more conservative benchmark. In our calibration, a score of 15.5 corresponds to about a 9% risk of recurrence over five years. We chose this cutoff because the model was stable at that point and it aligned well with global standards, while being slightly more stringent.
Our score is not the only factor in decision-making. Clinicians also consider the patient’s age, comorbidities, medical and family history, and the specific subtype of breast cancer. Our test provides a strong, evidence-based input, but the final decision always rests with the treating doctor.
Your test costs roughly a quarter of what the alternatives charge abroad. Has insurance coverage substantially changed uptake?
We continue to work with Indian government health insurance schemes. While some of them do reimburse the cost of CanAssist Breast, it is still on an ad hoc basis. It is slowly happening in schemes like CGHS, but it is not yet reimbursed under Ayushman Bharat.
If this test is included under Ayushman Bharat, patients at tertiary care government hospitals such as Kidwai, GCRI, and AIIMS institutions could all benefit. The system itself would benefit, because avoiding unnecessary chemotherapy for nearly 70% of early-stage patients would result in substantial cost savings. At present, more than 90–95% of our patients come from private hospitals.
So far OncoStem has been built around a single test. What’s the next chapter?
At present, the test is offered as a service from our central lab. However, with increasing international reach, we have realised that logistics, especially Customs delays in releasing tissue samples, can become a bottleneck. To address this, we are now working on developing a CanAssist Breast kit. The idea is to package the key reagents required for 20 to 50 tests, which can then be used locally. The test itself runs on standard Roche machines that are already available in most large hospitals and labs, so it can be easily done there. We are currently working on the necessary approvals for the kit, including CDSCO and CE marking.
We are also looking at expanding our portfolio into other cancers, but adding new tests requires substantial investment. There is strong interest from clinicians to collaborate with us in this area.
Got a story that Healthcare Executive should dig into? Shoot it over to arunima.rajan@hosmac.com—no PR fluff, just solid leads.