High quality data describing the association between exposure to supplemental testosterone and the risk for prostate cancer has not been available, making this question difficult to answer. We also lacked data describing whether testosterone supplementation increased the risk for aggressive prostate cancers that lead to fatal disease. Fortunately, Dr. Stacy Loeb and her colleagues sought to answer these questions in this week’s Journal of Clinical Oncology article entitled, “ Testosterone Replacement Therapy and Risk of Favorable and Aggressive Prostate Cancer.”
The investigators used data from national heath registries from Sweden and created the Prostate Cancer Database Sweden (PCBaSe) to evaluate the relationship between testosterone exposure and prostate cancer. They hypothesized that if greater levels of testosterone increased the risk of prostate cancer, they would see an association between testosterone supplementation and prostate cancer in the data set, with men exposed to a longer duration of testosterone supplementation being at greater risk for prostate cancer overall, and a greater risk for aggressive prostate cancer specifically.
A great strength of this study was the nested case control design used to test their hypotheses. A nested case control design is challenging, and typically fraught with risk of biases that could invalidate a study’s findings. This is because it is performed by identifying a group of “case” individuals afflicted with the illness of interest (in this instance, prostate cancer) within a pre-defined dataset, and then identifying a group of unaffected “control” individuals who are like the cases in all respects other than having the illness of interest. When investigators choose which patients to include in a study, selection bias, or the chance that certain individuals would have a higher likelihood of being chosen for unmeasured reasons that could affect the risk of developing the disease outcome, can confound the results and make them invalid. In addition to the selection bias of choosing the cases, US datasets tend to have biases introduced by the way that datasets are defined before cases and controls are even identified. For example, US datasets are routinely defined by rules that may affect disease outcomes, including geographic location or insurance provider, inserting selection bias into the entire dataset.
In designing this study, Loeb and colleagues were able to draw upon the strength and rigor of the PCBaSe. This Swedish dataset is a unique one made possible by the universal healthcare system that covers every individual, and includes data on 98% of prostate cancer patients treated in Sweden. This dataset includes registry data that from individual nationwide registries that include all health care costs, including prescription drug data, and outcomes, including prostate cancer outcomes. This makes it all-inclusive, and dramatically reduces the risk of selection bias in the way the dataset was defined and the way the cases were selected. Moreover, when using a dataset comprised of data from an entire country, the number of cases can be many thousands of people, even when the exposure of interest (testosterone supplementation, in this case) is rare, making the power of studies performed in the dataset sufficient to detect even small effects.
In their study, the investigators included all men with a diagnosis of prostate cancer in Sweden between 2009 and 2012 as cases, with a total of 38,570 men diagnosed with prostate cancer during the time period. A matching algorithm matched cases with controls based on birth year and county, identifying five controls for each case, and a total of 192,838 men without prostate cancer. When compared with each other, the cases were more often married, and had higher socioeconomic status, fewer comorbid illnesses, and more negative prostate biopsies than the controls. In all, approximately 1% of cases (men with prostate cancer) and controls (men without prostate cancer) had exposure to testosterone supplementation (unadjusted OR 1.03; 95% CI 0.91-1.17).
To assess the association between exposure to testosterone and the risk of developing prostate cancer, the investigators performed a multivariable analysis including comorbidity index, marital status, and education level. All of these factors have been associated with prostate cancer and health outcomes in prior studies, and were important confounding factors to include in the analysis. In this adjusted model, testosterone exposure was not associated with the odds of developing prostate cancer overall (adjusted OR 1.03; 95% CI 0.90-1.17). Testosterone exposure was associated with increased odds of favorable-risk or intermediate risk prostate cancer (favorable defined as clinical T1-T2, PSA <10 ng/mL, Gleason ≤ 6, N0, M0, and intermediate defined as T1-T2, Gleason 7, PSA 10-20 ng/mL, N0, M0) (adjusted OR 1.35; 95% CI 1.16-1.56), and decreased odds of high-risk or aggressive prostate cancer (defined as local high risk T1-T2, Gleason 8-10, 20>PSA<50 ng/mL, N0, M0; locally advanced T3, PSA < 50 ng/mL, N0, M0; regionally metastatic T4, 50<PSA<100 ng/mL, N1, M0; and metastatic on bone imaging or PSA >100ng/mL) (adjusted OR 0.5; 95% CI 0.37-0.67). The authors also found that there was not an association between risk of prostate cancer and route of testosterone administration.
Short of a randomized controlled trial of testosterone exposure vs no exposure that monitored for the development of prostate cancer, this is the strongest evidence available on this issue. The study includes over 200,000 individuals, and because of this, is able to assess an exposure only present in 1% of the cohort. The complete registry data available in Sweden is not something that can be replicated in the US with our complex set of payers and comparatively higher rates of uninsured patients, many of whom are minority men at higher risk of prostate cancer. This data puts testosterone supplementation into perspective: in men considered at average risk of developing prostate cancer, there appears to be no increase in risk associated with testosterone supplementation. It is important to note, however, that this study does not specifically address questions of whether men with a strong family history of prostate cancer or African American men are at an elevated risk of developing prostate cancer if exposed to testosterone replacement.
All in all, this is a high quality study that is uniquely capable of assessing a rare exposure and its relationship to an outcome that is typically decades in the making. It provides good rationale for men of average risk of prostate cancer to consider using testosterone replacement, though I still recommend that men do so under the guidance of physicians who screen for prostate cancer. Data like this never obviates the need for personalized counseling for the men we see, but it goes a long way to giving us the information we need to answer at least one of the burning questions we face on testosterone each day.
Written by: Alicia Morgans, MD, MPH
Watch the Video Interview with Stacy Loeb
Read the abstract: Testosterone Replacement Therapy and Risk of Favorable and Aggressive Prostate Cancer.