Pediatric Oncology and the Caribbean Data Gap

Note from the President

September is Childhood Cancer Awareness Month and Suicide Prevention Awareness Month. Last September, when we last marked both, I wrote that the communities suffering most are the ones with the least voice in the rooms where treatments are designed. Twelve months on, that observation has not aged. The acceleration in AI tools for drug discovery and clinical research is real, the gains for patients in wealthy countries are real, and the gap between those gains and what reaches a child treated in a public ward in Kingston or Port of Spain has, if anything, widened.

Dr. Yolande Pierre-Louis returns to our pages this month with a careful piece on what is happening in AI-assisted pediatric oncology, what is not, and what our region has to do to be in the rooms where it matters. Her piece is harder reading than most, and that is on purpose.

A reminder, our first regional AI Summit opens in Bridgetown on October 9. Two thousand four hundred attendees are registered. If you have not yet, register at the usual address. The full programme is now published.

If you are in crisis or worried about someone who is, the lifelines I listed last September are still good. They are also at the foot of this letter.

Adrian Dunkley Founder and President, Caribbean AI Association

---

Feature

Pediatric Oncology and the Caribbean Data Gap

By Dr. Yolande Pierre-Louis

There are roughly four hundred children diagnosed with cancer each year in CARICOM member states. That is a small number relative to almost any other public health problem we manage. It is also, in another sense, the most important small number our pediatric services handle, because the gap between five-year survival rates in our region and those in high-income countries on the same cancers remains, in 2025, between fifteen and thirty percentage points depending on the cancer type. A child with acute lymphoblastic leukemia diagnosed in Bridgetown is significantly less likely to be alive five years later than the same child diagnosed in Toronto. The biology of the disease is the same. The arithmetic that produces the outcome is not.

A great deal of the global AI in healthcare conversation, in the past two years, has been about the application of machine learning to drug discovery, to genomic risk stratification, to radiomics, and to treatment regimen optimization. The headline results from pediatric oncology research consortia in the United States, the United Kingdom, and continental Europe have been genuinely significant. AlphaFold's third version, released in 2024, has begun to change the structural biology workflow of small molecule discovery. AI-augmented review of pediatric leukemia bone marrow samples has improved morphological classification accuracy in studies at multiple centres. Risk stratification models that incorporate genomic and transcriptomic features alongside conventional clinical variables are now part of treatment selection at several reference centres internationally.

The Caribbean's representation in any of this is approximately zero. I want to explain why, and what we need to do about it, because this is not a problem that resolves itself.

The pediatric cancer datasets that drive contemporary AI research are assembled by long-running consortia, the Children's Oncology Group in North America, the International BFM Study Group in Europe, the relevant Asian consortia, and others. These consortia have spent decades building the infrastructure to enroll children consistently, to standardize their data collection, to coordinate their tissue banking, and to perform the long-term follow-up that supports both clinical trials and the AI models trained on the resulting data. The Caribbean has, historically, not been a meaningful contributor to these consortia. Several individual Caribbean centres have enrolled patients in selected trials, but the systematic regional participation that would put our children's data into the training sets has not happened.

The consequences are concrete. The AI models being trained today on the data from the major consortia are tuned on populations whose ancestry, environmental exposure, treatment histories, and supportive care contexts differ from ours in ways that affect both biological response and clinical outcome. When those models, eventually, become standard of care components, they will perform less reliably on our children than on the children whose data trained them. We have seen this pattern before in adult oncology with risk scores that under-predict mortality in Black patients. There is no reason to expect the pediatric story will be different.

A second consequence is the access timeline. The AI assisted drug discovery pipelines that are now beginning to deliver candidate compounds for rare pediatric cancers will, on the current trajectory, reach Caribbean patients only after they have been deployed in the markets of the countries whose research dollars funded the discovery. The lag has historically been five to ten years for novel cancer therapies. There is no reason to expect AI-discovered therapies will be different unless we change the structural position of our region in the discovery pipeline.

A third consequence is intellectual. The pediatric oncologists in our region, several of whom are colleagues of mine and friends of long standing, are increasingly being asked questions by parents that they would like to be able to answer from a position inside the research, rather than from outside it. The dignity of being a contributor rather than a recipient matters. It matters to our clinicians. It matters to the institutional culture of our teaching hospitals. It matters to whether our best young oncologists choose to build their careers in our region or to leave.

So what do we do.

I want to be specific, because vagueness on this question is not honest.

Step one. Build the registry.

The single most leveraged investment our region can make in pediatric oncology research is a regional childhood cancer registry that meets international data quality standards. This is not a new idea. There have been efforts in the past, and they have not, to be honest, been sustained at the level that produces the kind of data the international consortia would consider contributing partners. The Pan American Health Organization has been a willing convener. The technical work, the data harmonization, the legal framework for cross-border data sharing within CARICOM, the funding for the registry coordinators in each member country, has not been completed.

The cost of a proper regional registry, on conservative estimates we have prepared for the CAIRA Working Group on AI in Health, is approximately one point seven million US dollars in initial setup and four hundred thousand US dollars per year in operating costs across the region. That is not a large sum on the scale of regional health spending. It is, on present arrangements, larger than any single CARICOM health ministry will fund on its own. It is, however, well within the scale of what the Caribbean Development Bank, the Inter-American Development Bank, or a coordinated bilateral support package could deliver, particularly when the case is made on the strength of the AI era opportunities rather than the historical case for cancer registries.

Step two. Join the consortia.

The Children's Oncology Group has, for several years, accepted protocol-specific participation from centres outside North America. The International BFM Study Group is similarly open. The barrier to entry is not, in the main, scientific. It is administrative and infrastructural. The Caribbean centres that could realistically participate, the University Hospital of the West Indies in Kingston and Mona, the Eric Williams Medical Sciences Complex in Trinidad, the Queen Elizabeth Hospital in Barbados, and the Princess Margaret Hospital in Nassau, each have or could develop the clinical capacity. What is needed is a coordinated regional submission, the legal architecture for data sharing, and the IRB infrastructure that international consortia look for before partnering. CAIRA, with the Working Group, will publish a feasibility study on this in December.

Step three. Build the bioinformatics capacity.

Even where Caribbean children participate in international trials, the analytic work, the genomic sequencing, the transcriptomic profiling, and the AI model training is done elsewhere. Our region has, today, a thin bench of clinical bioinformatics expertise. The University of the West Indies has begun a graduate programme in bioinformatics. It will need expansion. The pipeline from Caribbean undergraduate computer science and biology programmes into clinical research bioinformatics is small and inconsistent. CAIRA's Learning Working Group has begun work on a Caribbean clinical bioinformatics curriculum, drawing on the existing materials from the European Molecular Biology Laboratory, Cold Spring Harbor, and the Broad Institute. The first cohort will start in February 2026 if funding holds.

Step four. Insist on local validation of AI clinical tools.

When AI tools for pediatric oncology, for radiomics, for clinical decision support, for risk stratification, begin to be marketed to our region, the regulatory expectation should be that they have been validated on populations that include children whose ancestry, geography, and clinical context resembles ours. This is the harder line of the Charter discussed by Andre last month. The harder line is the right one for pediatric oncology specifically. The cost of deploying a poorly validated tool in a small population with concentrated outcomes is, in this domain, the cost of children. CAIRA's position to the regional ministries of health will be unambiguous on this.

Step five. Connect to the diaspora.

There are, by my own informal count of Caribbean clinicians and researchers in the relevant fields who I know personally, at least forty Caribbean-born pediatric oncologists, hematologists, and clinical research scientists currently working at major North American and European centres. Several of them have, in conversations I have had, expressed willingness to contribute to a regional effort if there is a credible recipient organization. The CAIRA Fellows programme that Aisha described in June is one vehicle. We will be running a focused pediatric oncology cohort in the first half of next year. If you are a Caribbean clinician or researcher in this field and have not yet been contacted, write to health@caribbeanaiassociation.com. We will find you.

A closing word, to parents.

If you are reading this because you have a child currently in treatment, or because you have a memory of one who was, I want to say two things directly.

The first is that the gap I have described is not a fact about the quality of the care your child has received from Caribbean clinicians. The clinicians in our regional pediatric oncology services, given the resources they work with, perform at standards that compare favourably to any peer health system. The work they do, on the wards, with the supportive care they have available, is good work. The gap I am describing is a structural gap in the global research enterprise. Our clinicians did not cause it and cannot, alone, close it. Closing it is a regional project, and the work will take time.

The second is that the existence of this gap does not mean your child's care has been somehow lesser in moral importance. Every child treated in our region, the ones who have survived and the ones who have not, deserves the dignity of being remembered, and the families that have walked this road deserve the dignity of being heard. The work we are now beginning, on the registry, on the consortia, on the bioinformatics capacity, on the validation standards, is work in the names of those children. We will do it carefully, and we will do it for them.

A quiet September to those of you who are carrying this month for personal reasons. We are with you.

---

Dr. Yolande Pierre-Louis is a consultant radiologist at the University Hospital of the West Indies and chairs the CAIRA Working Group on AI in Health.

Crisis lifelines as listed in September 2024 remain current. For an updated list by country, see crisis@caribbeanaiassociation.com.

Originally published in The Monthly Intelligence Report, September 2025.