Turning conventional wisdom upside-down^†

Abstract

A revolution in traditional oncological concepts has shifted our gaze from high serum testosterone to low testosterone as a potential predictor of high-risk prostate cancer. The results of a study by Salonia et al add to a growing literature indicating that preoperative low serum T increases the likelihood of high-risk prostate cancer.

Circumstantial evidence is a very tricky thing, answered (Sherlock) Holmes thoughtfully. It may seem to point very straight to one thing, but if you shift your own point of view a little, you may find it pointing in an equally uncompromising manner to something entirely different. Sir Arthur Conan Doyle. The Boscombe Valley Mystery.

For seventy years it has been axiomatic that high testosterone (T) is risky for prostate cancer (PCa) and low T is protective. Yet recent work has shown this concept to be little more than a myth,1 based on what the fictional detective Sherlock Holmes might have called circumstantial evidence. In this issue of Cancer, Salonia et al2 explore a totally different idea, namely whether low T may be associated with high-risk PCa.

In the mid-1980s when I did my residency in urology at Harvard Medical School, men routinely underwent castration for metastatic PCa, resulting in rapid pain relief and reduction in serum acid phosphatase and prostate-specific antigen (PSA). There seemed no reason to doubt the teaching that administering T to an individual with PCa was like “pouring fuel on a fire” or “feeding a hungry tumor”.2

In the early 1990s, I was concerned that increasing a human's serum T might stimulate rapid growth of occult, quiescent cancers, and I, therefore, began performing sextant prostate biopsies in T-deficient (hypogonadal) men with normal PSA (<4.0 ng/mL) and digital rectal examination (DRE) to exclude, as best I could, the presence of PCa before offering T therapy. Remarkably, 11 of the first 77 (14%) men had PCa.3 This was the first indication that low T may be a PCa risk factor, as the cancer rate was similar to contemporaneous series with established risk factors such as abnormal DRE or elevated PSA. Clearly, low T was not protective against PCa.

This led to an investigation into the origin of the concept that for the prostate “more T is bad, less T is good.”1 An extensive review found no compelling evidence in the PSA-era that men with higher endogenous T or who had undergone T therapy were at increased risk of PCa.4 Multiple longitudinal studies failed to show any significant association between higher T and PCa risk. Serum PSA did not correlate with serum T. Men who underwent T therapy were found to be at no greater risk of developing PCa than men who received placebo.5 And supraphysiologic doses of T for up to 9 months caused no increase in prostate volume or PSA in healthy research subjects.6

And yet it was an everyday clinical experience that androgen deprivation caused rapid, dramatic declines in PSA and also that cessation of androgen deprivation resulted in increased PSA as serum T rose. Since the landmark work by Huggins and Hodges7 in 1941 to the present day, androgen deprivation has been a mainstay of treatment for advanced PCa.

This presented a paradox. How can PCa be so sensitive to androgen deprivation or its cessation yet appear to be indifferent to variations in serum androgens under other circumstances? The answer was provided in part by re-examination of historical studies. Whereas it was well-known that several studies showed that T administration produced rapid cancer progression in androgen-deprived men with metastatic PCa, those very same studies revealed that T administration in hormonally intact men produced no negative effects.1 Animal studies and PCa cell lines demonstrated a dose-response curve for androgens at low concentrations, but at higher concentrations even logarithmic increases in androgen concentrations elicited no further growth.8 The resolution of the paradox is that there is a limit to the ability of androgens to stimulate PCa growth, and this limit is achieved at low androgen concentrations.8

This androgen-prostate relation has been formalized as The Saturation Model.8 It describes an exquisite sensitivity of PCa to androgens at very low concentrations, followed by indifference of PCa to androgens at higher concentrations. A likely mechanism is provided by the finite ability of the androgen receptor (AR) to bind androgen, with maximal binding (ie, saturation) occurring at approximately 120 ng/dL (4 nmol/L) in human prostate tissue.9 The Saturation Model explains why manipulation of serum T into and out of the castrate range produces large changes in PSA, whereas variations in serum T within the naturally occurring range do not.8

PCa is androgen-dependent in the same way that a house plant is water-dependent; they both require adequate amounts to achieve optimal growth. However, once the requirement has been met, additional amounts have little, if any, effect. A house plant will never grow to the size of a tall tree regardless of how much water is available. For this reason, I favor replacing the old rubric that “T is like food for a hungry tumor” with “T is like water for a thirsty tumor.” Once the “thirst” has been “quenched,” additional androgen serves merely as excess.

The traditional concept that “high T is bad and low T is good,” never made much sense. It has always been contradicted by the inescapable fact that PCa (especially high-grade PCa) becomes increasingly prevalent with age, and age is associated with low T. If PCa, and similarly benign prostatic hyperplasia (BPH), were truly androgen-driven, then PCa and BPH would primarily affect men in their teens and twenties when T is at lifetime highs. On the contrary, if one wished to consider how androgens may be associated with high-risk PCa, surely it would be more logical to investigate low T rather than high T. It is in this context that we must view the work of Salonia et al2 in this issue of Cancer.

Although longitudinal studies have failed to show any relation between low serum androgens and PCa, with blood drawn years before the diagnosis of PCa,10 there is now an impressive accumulation of studies that demonstrate a significant relation between low T and PCa when serum androgens were measured just before prostate biopsies or radical prostatectomy.

In a larger series of biopsies in T-deficient men, PCa was found in 15% of 345 men (mean age, 59 years) with PSA <4.0 ng/mL.11 Although this value is similar to the cancer rate for men in the placebo arm of the Prostate Cancer Prevention Trial (PCPT)12 with PSA <4.0 ng/mL, these men were a decade younger (59 years vs 69 years). Moreover, the cancer risk was doubled for men in the lowest tertile of serum T compared with men with milder T deficiency.11 A low ratio of serum T to PSA predicted cancer in a group of men presenting for biopsy with PSA of 3-10 ng/mL,13 and similar results were obtained in T-deficient men with PSA <4.0 ng/mL.14 These results suggest that for any given PSA value, a lower T concentration confers an increased risk of PCa.

Low T has been associated with high-grade PCa in several studies, but not all (reviewed by Salonia et al). In 1 study, men with a Gleason score of >7 had lower intraprostatic dihydrotestosterone (DHT) than men with Gleason <6,15 raising the possibility that a low-androgen environment predisposes to development of high-grade PCa. This must be considered among the various explanations for the greater number of high-grade tumors noted in the National Cancer Institute's Prostate Cancer Prevention Trial (PCPT) among men who received finasteride, because finasteride lowers intraprostatic DHT to near-castrate levels.

In radical prostatectomy populations, a number of studies (but not all) have reported low T in association with increased stage at presentation,16 higher rate of positive surgical margins,17 increased risk of biochemical failure,18 and worse survival.19

In this issue of Cancer, Salonia et al2 examine the relation of preoperatively obtained serum T to high-risk PCa pathological features in a large surgical series of 673 Italian men. Low serum T was defined as <300 ng/dL, a frequently used but arbitrary value. Unfortunately, no data were available for free T, which, as a more reliable indicator of androgen status, might have offered more robust results. Nonetheless, this article is an important contribution due to its rigorous design, careful pathological assessment, and generous population size.

The primary findings were that low T was significantly associated with seminal vesicle invasion and high-grade disease (Gleason 7 [4 + 3] or higher) but not extracapsular extension. However, on multivariate analysis, low T lost its significant independent predictor status. Severe testosterone deficiency (<100 ng/dL) maintained a significant relation to seminal vesicle invasion, with an odds ratio of 3.11.

On the basis of the loss of statistical significance for low T on multivariate analysis, the authors concluded that preoperative circulating T was not a predictor of high-risk disease. I believe their data argue powerfully for the opposite conclusion. The risk of high-grade disease was increased by >50% in men with low T compared with men with normal T (33% vs 19.8%, P = .0009), seminal vesicle invasion was nearly doubled (21% vs 11%, P = .003), and for men with severe T deficiency, the risk was trebled (59.5% vs 19.8%, P < .0001). These are impressive differences in pathologic features that matter, features known to herald a worse prognosis.

Multivariate analysis is a useful tool that identifies which variables provide predictive value independent of all other variables in the analysis. However, if 2 or more variables comigrate to any appreciable degree, neither may prove to be an independent predictor of results. Yet 1 or both variables may still provide crucial clinical information. In this multivariate analysis, age was also a nonsignificant variable, yet one would be hard-pressed to argue that age is not an important predictor of high-risk disease. The finding that severe T deficiency maintained a statistically significant association confirms the general point that low T at some threshold is an independent predictor of high risk disease. Experts quibble as to what threshold should be used to categorize low T, and it is quite possible that the use of a T threshold lower than the value used in this study would have provided a more robust statistical outcome.

What are we to make of all this? First, there has been a revolution in thought concerning androgens and PCa. Indeed, selected T-deficient men with untreated PCa may even receive T therapy without evidence of cancer progression.20 Second, the primary results of the study by Salonia et al2 fit nicely into a growing body of evidence indicating that the presence of low T is associated with an increased risk of PCa, particularly high-risk PCa. I believe it is time for clinicians to consider routinely obtaining a serum T before PCa treatment, and perhaps before prostate biopsy, as part of their risk assessment.

Finally, after 7 decades of circumstantial evidence pointing us in the wrong direction, perhaps it is time to consider the once unthinkable: a T therapy trial of sufficient size and duration to determine whether normalization of serum T in older men may reduce the risk of PCa, particularly high-risk PCa.

CONFLICT OF INTEREST DISCLOSURES

The author has received honoraria, consulting fees, or research funds from GlaxoSmithKline, Endo, Abbott, Auxilium, Slate, and Bayer.