Eliciting the Impact of Metformin and Statins on Prostate Cancer Outcomes from a Real-life National Database Analysis - Beyond the Abstract
Moreover, Individuals with T2D have a two to six times higher risk of death from CVDs compared to those without diabetes. Consequently, approximately 35 million Americans are on statins (65 and over, usage was about 60% in 2018), with men constituting a significant proportion due to higher CVDs risk. At the same time, Metformin is the most prescribed medication for T2D in the U.S., with over 86 million prescriptions in 2022. So, according to clinical guidelines, Statins and Metformin usage patterns reflect the prevalence of these conditions and the emphasis on preventive care.
The relationship between metabolic disorders and PC remains ambiguous. Some studies suggest that conditions like T2D, obesity, and dyslipidemia may influence PC risk or outcomes, but findings remain controversial. Conversely, statins have been proposed as protective against the development of PC and potentially beneficial during androgen deprivation therapy (ADT) for advanced stages of PC. Because ADT itself increases the risk of developing metabolic syndrome and increases mortality from CVDs.
To clarify these relationships, we conducted an extensive analysis using French nationwide medical-administrative databases from 2006 to 2018. The study included data from 521,052 men diagnosed with PC and 1,827,345 control men without PC. Medication (Statins and Metformin) use was examined, PC's treatment types, and mortality outcomes to better understand the impact of statins and metformin in the context of PC. At the end of the observation period, 37.8% of PC patients had been treated with statins, 4.3% with metformin, and 11.5% with both. In comparison, among non-PC men, 34.6% used statins, 4.6% used metformin, and 11.7% used both. The average age of users was around 76–78 years, with slightly shorter durations of metformin and statin use in non-PC patients.
As anticipated, mortality was generally lower among non-PC men without treatment. The highest mortality rate was seen in PC patients taking both statins and metformin (Hazard Ratio [HR] = 2.29), likely reflecting the combined impact of comorbidities and age. When broken down by age groups over 50, the highest mortality in non-PC men occurred with dual therapy, while in PC patients, it was associated with metformin use alone.
To refine the analysis, the study examined survival outcomes in two subgroups of PC patients diagnosed between the ages of 50 and 70: those treated with radical prostatectomy (RP) and those receiving ADT as first-line therapy. These treatment pathways roughly correspond to different prognostic groups, with RP generally used for potentially curable cancer and ADT for more advanced diseases.
In the RP subgroup (133,616 men), use of statins and/or metformin was associated with higher mortality. This likely reflects patients with more comorbidities. Importantly, no benefit was seen on overall survival or recurrence rates after RP.
In contrast, in the ADT subgroup (123,198 men), a protective effect was observed for statins use. Statins alone were associated with improved survival (HR = 0.92), and the combination of statins and metformin showed an even stronger effect (HR = 0.85). Metformin alone, however, was linked to worse survival outcomes (HR = 1.07), suggesting that T2D may have a detrimental impact on cancer prognosis that is not sufficiently mitigated by metformin.
Additionally, progression to chemotherapy was less frequent in statin users, potentially due to the reduced tolerance for chemotherapy in patients with CVDs, rather than the beneficial effect of statins themselves.
To further explore these associations, a Bayesian Network (BN) model was used to perform causal inference. To construct a supervised causal model examining outcomes in relation to PC aggressiveness (defined by first-line treatment) and comorbidities (indicated by statins or metformin use), we calculated the direct effects of PC treatment and comorbidities using a Bayesian causal model with backdoor adjustment. This causal approach makes it possible to estimate direct effects (comorbidities, estimated by statins and metformin use, to survival and disease aggressiveness, estimated by first line treatment for PC, to survival), taking into account and excluding indirect effects (impact of comorbidities on survival via changes in decision-making for PC management) inherent in real-life data studies.
A new approach to discussing the results has also been introduced in this study. It was based on the use of semantic causal analysis powered by generative AI platforms. Indeed, beyond systematic reviews and meta-analyses, scientific discussions often rely on authors’ selective interpretation of the literature. Incorporating generative AI tools provides a broader, consensus-based perspective and allows assessment of information relevance. In this way, we then interpreted the results with two knowledge resources: first, using our own clinical and research expertise, and second, by comparing interpretations from five different generative AI platforms (OpenAI’s o3-mini, Mistral’s Pixtral Large, Anthropic’s Claude 3.5 Sonnet, and Google’s Gemini 1.5) that analyze existing online knowledge to identify consensus views. Moreover, the AI tool used in our study—Hellixia in BayesiaLab™—uses large language models and machine learning to identify causal relationships and assess the reliability of online information. It provides a confidence score ranging from +1 (full certainty) to 1 (complete contradiction), with 0 indicating uncertainty. For instance, and trivially: “PC is a male disease” is scored +1; “PC is a female disease” is scored –1. The GAI-generated reports on the impact of statins and metformin on PC outcomes remain controversial, with consistency scores ranging from –0.4 to +0.2 for statins and –0.35 to +0.2 for metformin across the various search engines.
Although our study was based on a large national dataset, we are aware that it has limitations common to administrative data: the lack of clinical detail, particularly with regard to detailed PC staging. In conclusion, this large-scale, real-world analysis highlights the complex interaction between PC, metabolic diseases, CVDs, and associated medications. The results suggest: firstly, that T2DM worsens outcomes in men with or without PC despite the use of Metformin alone, which does not improve sufficiently the mitigation of worsening outcomes; secondly, that statins are associated with improved survival only in PC patients receiving ADT, suggesting that careful assessment of CVDS risk prior to ADT use is essential in practice. Clearly, the potential benefit of routinely prescribing statins to men on ADT but without diagnosed CVDs risk warrants further study through randomized controlled trials. These insights underline the importance of personalized treatment strategies that consider both cancer and comorbid conditions. Moreover, the integration of AI-driven semantic analysis offers a promising approach for enhancing medical research through the synthesis of data and existing knowledge.
Written by: Olivier Cussenot, MD, PhD, Adjunct Professor, Medical University Vienna, Department of Urology, Währinger Gürtel, Vienna, Austria
Reference:
Read the Abstract