How We Completed an Ambitious Study Without a Grant

In 2024, I and my research group (affectionately known as ThinkTank Cafezinho) at Federal University of Sao Paulo in Brazil set out to answer a deceptively simple question: What is the burden of oncologic emergencies in Brazil?

It seemed like the kind of question that should have been answered years ago. Oncologic emergencies—conditions like febrile neutropenia, spinal cord compression, hypercalcemia of malignancy, and tumor lysis syndrome—are among the most time-sensitive, life-threatening events in cancer care. In high-income countries, there are robust epidemiological data on their incidence, trends, and outcomes. But in low- and middle-income countries, the page is almost blank. We knew these emergencies were common. We knew they killed people. What we didn't know was the scale: how many hospitalizations and deaths they caused each year, whether the burden was concentrated in certain regions, or whether two decades of advances in cancer care had actually narrowed the gap—or widened it.

That gap was our starting point.

We designed a rather ambitious study: a nationwide analysis of all oncologic emergencies recorded in Brazilian public databases from 2008 to 2024. An ecological design, using data from DATASUS, mortality registries, and hospital information systems. The goal was to map the historical trends, identify regional disparities, and quantify the burden across Brazil's five macro-regions.

On paper, it was feasible. The data existed. The methods were established. We had a team of bright, motivated students. The only thing we did not have was money.

Best Laid Plans of an Epidemiology Data Study

The study was straightforward in concept. We would extract data from Brazil's Unified Health System databases, covering all hospitalizations and deaths attributed to oncologic emergencies over a 17-year period. We would stratify by region, age, sex, and type of emergency. We would use interrupted time series analysis to detect changes in trends over time.

The hypothesis was that the burden of oncologic emergencies would be higher in the North and Northeast regions—the poorest parts of the country—and that, despite advances in cancer care, the disparities would have widened over time. Were this to be true, our data could provide a solid foundation for public policies, offering a renewed focus on the unique challenges of oncologic emergencies.

We had a clear research question, a solid design, and a team eager to work. All we needed was a budget.

When Grants Only Provide a Reality Check: Learning to Run on Empty

We applied to at least seven grants—national, regional, disease-specific—confident that one would land. The first rejection came from CNPq, Brazil's National Council for Scientific and Technological Development. The proposal was deemed "relevant but not a priority." A second rejection followed from a nonprofit foundation. The message was consistent: oncologic emergencies were not on the agenda.

Then, unexpectedly, we received a small grant—a Scientific Initiation scholarship—from the São Paulo Research Foundation (FAPESP). It was enough to support one student's stipend for one year. Enough to keep the project alive. But not enough to cover the software licenses, the processing power, or the articles locked behind paywalls.

The group was restless, as was I. Could I protect them from the brutal pragmatism of a system that measures success in publications, not in perseverance?

We had no seed money for this study. No dedicated workspace. No administrative support. The students were working on borrowed time and borrowed machines.

The first decision was to rewrite the analysis plan. The ideal analysis—the one we had proposed in the grant applications—required more than we had. We scaled back. We used simpler models. We prioritized what was feasible over what was elegant.

For the next six months, we operated on goodwill. We adopted R for all statistical analyses. It was free, open-source, and had a steep learning curve. But the students learned. They taught each other. We created a shared repository with scripts and troubleshooting notes—a digital survival manual for the project.

We ran analyses on borrowed computers. We accessed paywalled papers through institutional proxies that often failed.

The servers at our university were notoriously unstable. I remember staring at my screen at 2 a.m., watching a regression model run for the fourth time after the server crashed at hour 23. The student responsible for that analysis, a first-year undergraduate, had been trying to complete the same routine for three weeks. Each crash meant starting over.

As a workaround, we negotiated access to a server at a collaborating institution, a teaching hospital in São Paulo. It wasn't fast, but it didn't crash—and for the first time, the student could let a model run overnight without checking on it.

When the student stipend was delayed for three months—a common occurrence in Brazilian public universities—we pooled our own resources to cover basic costs. We bought coffee, not software. We bought time, not hardware.

The most important decision was to keep the team informed. Transparency became our currency. I showed them the rejection letters. I showed them the budget spreadsheet, the empty columns. I told them, "This may go nowhere. But if it goes somewhere, it will be ours."

The entire team stayed.

The Data Rewards of Persistence

The data finally cohered. What we found was sobering. Over the 17-year period, the absolute number of oncologic emergencies in Brazil increased substantially, driven by an aging population and improved cancer diagnosis. But the rates—the burden adjusted for population size—told a different story. They remained stubbornly high in the North and Northeast, where access to oncology care is limited and health infrastructure is fragile.

Our analysis revealed an enormous disparity. The North and Northeast recorded the sharpest increases in hospitalizations and deaths, despite the Southeast and South having the highest overall totals. We published these findings in Cancer Epidemiology, with a companion piece in Current Opinion in Oncology.^{1, 2} The data were clear: Oncologic emergencies in Brazil are a neglected public health crisis, and the gap between regions is not an accident of geography. It is a structural failure.

What the Budget Couldn't Buy

This study taught me something that no grant application or mentorship workshop could convey: Science in resource-limited settings is not about doing more with less. It is about deciding what to protect when everything is in peril.

We could not afford the ideal data infrastructure. We could not afford the software. We could not afford the time. But we could afford transparency, shared ownership, and a commitment to the question itself.

The experience reshaped how I lead. I no longer wait for funding to pursue important questions. I start with the question and build around it. I prioritize feasibility without abandoning ambition. I assume that the infrastructure will fail and plan for it.

Brazil's science funding system is fragmented, underfunded, and unpredictable. But it is not empty. FAPESP, CNPq, and CAPES sustain a remarkable volume of research despite chronic constraints. The problem is not the absence of money. It is the absence of consistent, long-term investment in the questions that matter for the majority of the population.

To young PIs in similar contexts: Your moment of scarcity will come. It will not be dramatic. It will be a paywall, a server crash, a delayed stipend. In that moment, remember that the question matters more than the budget. The team matters more than the hardware. And the answer—when it finally arrives—will be entirely yours.

We mapped the terrain of oncologic emergencies. We gave Brazilian clinicians and policymakers a baseline. And we proved that a question, a team, and an open-source statistical tool can produce high-impact science, even without a grant.

References