Age, Education and Treatment Choices Shape Prostate Cancer Survival in Landmark Brazilian Study of 10,556 Men

Prostate cancer remains one of the most commonly diagnosed malignancies in the world, and a new large-scale analysis from Brazil is shedding fresh light on which patients are most at risk of dying from the disease. Researchers followed more than ten thousand men treated within the oncology care network of the southeastern state of Espírito Santo and found that five-year prostate cancer-specific survival reached 87.7 percent, a figure that outperforms several previous Brazilian estimates but still lags far behind the survival rates reported in wealthy countries such as the United States, where five-year survival can approach 98 percent. The study, published in the open-access journal Cancer Reports, draws on a retrospective cohort of 10,556 men diagnosed between 2000 and 2016, making it one of the most comprehensive investigations of prostate cancer outcomes ever conducted in a Brazilian state.

The scale of the prostate cancer problem in Brazil is considerable. According to estimates from the country’s National Cancer Institute, prostate cancer is the most frequently diagnosed malignancy among Brazilian men when non-melanoma skin cancers are excluded, with nearly 78,000 new cases expected annually in the coming years. Globally, roughly 1.4 million new cases and more than 375,000 deaths were estimated in 2020, ranking prostate cancer as the second most frequently diagnosed cancer and the fifth leading cause of cancer death among men. Yet while high-income countries have accumulated extensive survival data, low- and middle-income countries such as Brazil have historically lacked robust, registry-based studies linking patient characteristics to long-term outcomes, a gap the new research set out to close.

To build the cohort, the team tapped the Hospital Cancer Registry of Espírito Santo, an institutional system recognized by the International Agency for Research on Cancer, and linked its records with the national Mortality Information System. Record linkage relied on key identifying variables including patient name, mother’s name, date of birth, sex and municipality of residence, with phonetic Soundex encoding applied to name fields to improve matching accuracy. Only records showing exact agreement across the selected identifiers were retained, and every matched pair was manually reviewed using complementary information such as residential address and dates of diagnosis and death. Patients were followed from diagnosis until death from any cause or administrative censoring at the end of 2021, producing a follow-up window that spans more than two decades of clinical care.

The demographic and clinical profile of the cohort reveals much about the population the Brazilian public health system serves. The average age at diagnosis was 69.2 years, and 70.39 percent of patients were 65 or older. Nearly two-thirds of the men identified as non-White, 65.92 percent were married, and almost two-thirds had no formal education or only basic schooling. The overwhelming majority of tumors, 98.52 percent, were adenocarcinomas, and 91.65 percent of patients had no distant metastases at diagnosis. First treatments were distributed across surgery in 23.71 percent of cases, hormone therapy in 21.97 percent, radiotherapy in 19.47 percent and combined radiotherapy and hormone therapy in 15.15 percent. By the end of follow-up, 18.34 percent of patients had died from prostate cancer and 21.14 percent had died from other causes, a pattern that underscores the importance of accounting for competing causes of death in elderly cancer populations.

Methodologically, the study is notable for the care with which it separates two distinct statistical questions. Because many of these older patients died of causes unrelated to their cancer, the researchers used cause-specific Cox proportional hazards models to estimate the instantaneous hazard of prostate cancer death among men who remained at risk, rather than the competing-risk Fine–Gray framework the same team had previously applied to this cohort. The authors emphasize that the two approaches answer complementary questions: Fine–Gray models estimate the cumulative probability of cancer-specific death in the presence of competing mortality, while cause-specific Cox models are better suited to etiological inference and the identification of prognostic determinants. Kaplan–Meier curves, which treated other-cause deaths as censored observations, were used only descriptively, since they can overestimate cumulative cancer-specific mortality when competing events are frequent.

The univariate survival analyses told a striking story. Five-year prostate cancer-specific survival fell from 92.3 percent in men diagnosed before age 65 to 85.7 percent in those diagnosed at 65 or older. Survival dropped steeply with advancing tumor stage, from 96.5 percent for stage I disease to just 49.9 percent for stage IV, and the presence of distant metastases cut five-year survival to 42.5 percent, less than half the 91.7 percent seen in patients without metastatic spread. Patients who underwent surgery or radiotherapy as initial treatment showed five-year survival above 95 percent, while those receiving hormone therapy had markedly lower survival at 77.4 percent. Education also mattered: men with no formal education had five-year survival of 83.4 percent, compared with 90.7 percent among those with middle or high education.

When the researchers adjusted for multiple factors simultaneously in a multivariable Cox model restricted to 6,481 patients with complete data, several independent predictors of prostate cancer-specific mortality emerged. Each additional decade of age at diagnosis raised the hazard of prostate cancer death by 14 percent. Education showed a clear protective gradient, with basic education associated with an 18 percent lower risk and middle or high education a 36 percent lower risk compared with no formal schooling. Patients referred from private or insurance-based services rather than the public Unified Health System had a modestly higher risk of prostate cancer-specific death, a finding the authors note warrants careful interpretation. Surgery was associated with a substantially lower hazard of death, while hormone therapy was associated with nearly double the risk.

The authors are careful to caution that the treatment associations should not be read as causal effects. In real-world oncology, patients selected for surgery or radiotherapy typically have localized disease and preserved functional status, whereas hormone therapy is generally reserved for more advanced disease. This confounding by indication is compounded in the study by a substantial proportion of missing data: TNM staging was absent for 48.92 percent of records and disease status after first treatment was missing for 51.71 percent, which prevented these key severity variables from entering the adjusted models. A sensitivity analysis comparing the complete-case sample with excluded patients, using standardized mean differences, found negligible imbalance for core variables such as age, race, marital status and metastasis status, suggesting the complete-case analysis was unlikely to have introduced major selection bias, though residual confounding cannot be excluded.

The findings carry clear public health implications for Brazil and other middle-income countries. Lower educational attainment is associated with delayed diagnosis and treatment, reduced chances of cure and higher mortality, likely reflecting gaps in health information, access and understanding of symptoms. The study’s authors argue that strengthening early diagnosis, reducing social inequalities in access to care and improving the quality and completeness of clinical data in cancer registries are essential steps toward better outcomes. They also highlight the value of applying complementary statistical frameworks to the same population, noting that competing-risk models support population-level risk prediction while cause-specific models illuminate etiological relationships. Together, the two analyses of the Espírito Santo cohort offer a more complete picture of prostate cancer prognosis in Brazil than either method alone, and provide a template for how registry-linked data can drive evidence-based cancer policy in resource-constrained settings.

Subject of Research: Prognostic factors for prostate cancer-specific survival and mortality in a large retrospective Brazilian cohort

Article Title: Survival and Mortality Predictors in Prostate Cancer: A Retrospective Cohort of 10 556 Brazilian Patients

Article References: Survival and Mortality Predictors in Prostate Cancer: A Retrospective Cohort of 10 556 Brazilian Patients. (n.d.). https://doi.org/10.1002/cnr2.70667

Image Credits: AI Generated

DOI: 10.1002/cnr2.70667

Keywords: prostate cancer, survival analysis, Brazil, cancer registry, Cox regression, competing risks, health inequalities, hormone therapy, education level, mortality predictors, Espírito Santo, oncology