Nadir PSA predicts survival after radiation and androgen deprivation for unfavorable risk patients

If a treatment isn't working, we want to know as quickly as possible so we can try a salvage therapy while it can still make a difference. We want a measure of effectiveness, called a surrogate endpoint, that will predict survival, and we usually turn to PSA as our best early indicator. But there are different ways of defining mortality, and different measurements utilizing PSA. In an analysis of a randomized clinical trial (available here), the researchers sought to answer these questions for unfavorable risk patients who were treated with external beam radiation (EBRT) + androgen deprivation (ADT).

The purpose of the randomized clinical trial (NCT00116220) was to determine whether adding 6 months of androgen suppression improved freedom from biochemical failure over radiation therapy alone. This was a secondary analysis of the data. The details of the study were as follows:

• All patients were "unfavorable risk," defined as PSA between 10 and 40 ng/ml or Gleason score ≥7 or extracapsular extension or seminal vesicle invasion.
• Men were screened for minimal of no comorbidities.
• Median age was 72.
• They all received 70.2 Gy of 3D-CRT at 6 hospitals in the Boston area between 1995 and 2001
• Half (78 men) got 6 months of ADT with the radiation; half (79 men) had radiation without ADT
• PSA was evaluated every 3 months for 2 years, every 6 months for 3 years, and then annually.
• When PSA climbed above 10 ng/ml they received salvage ADT.

In selecting the kind of "mortality" they wanted to use as the gold standard endpoint, the researchers selected "age-adjusted all-cause mortality (ACM)" rather than "prostate cancer-specific mortality." This was a reasonable choice for several reasons:

• It is often difficult to discern whether  prostate cancer was the final cause of death. Men may die of kidney or liver failure or other final causes that are consequences of their prostate cancer. 
• Prostate cancer has been found to be associated with other causes of death. 
• The men in this study were only included if they had no or minimal comorbidities that might contribute to their death. 
• Men were eventually diagnosed with metastatic castration-resistant prostate cancer, and most had received chemotherapy.
• The mortality was age-adjusted for actuarial death rates. 
• Follow-up was long enough (16.5 years, median) so that even slow-killing prostate cancer would be a significant cause of mortality.

They examined 4 PSA-related metrics as potential surrogate endpoints:

1. PSA failure, defined as nadir + 2.0 ng/ml
2. PSA nadir > 0.5 ng/ml
3. PSA doubling time < 9 months
4. Time to PSA failure < 30 months

They had several criteria for inclusion. Basically, they wanted to find metrics that predicted mortality, and that continued to make a difference in survival time after the effect of the adjuvant ADT no longer extended survival. All but "PSA failure" met their criteria. Of the remaining 3 metrics, PSA nadir > 0.5 ng/ml had the largest effect in explaining survival.

The following table shows the percent of 8-year all-cause mortality for each surrogate endpoint, when it was met and when it wasn't, and the percent of the treatment effect (adjuvant ADT) explained by each metric.

	Percent mortality at 8 years		Percent of treatment effect explained by metric
Endpoint achieved:	YES	NO
PSA failure	32.5%	11.8%	NA
PSA nadir > 0.5 ng/ml	47.4%	13.6%	104%
PSA doubling time < 9 mos.	40.7%	14.1%	43%
Time to PSA failure < 30 mos.	39.4%	15.3%	41%

If the PSA nadir was over 0.5 ng/ml, it predicted the biggest difference in mortality. It also explained essentially all of the treatment effect of the added ADT.

Before anyone gets worried that their PSA is over 0.5 ng/ml, we must remember what "nadir" means. Because this analysis was done in hindsight, nadir is the lowest PSA ever achieved after treatment. It was not, in this case, the lowest value achieved so far. It often takes 5 or more years to achieve the nadir after radiation. For those who received adjuvant ADT with their radiation, the nadir will be achieved while they are still on ADT, and the PSA may rise above the nadir as the effect of the ADT wears off.

A nadir of only 0.5 ng/ml among those taking ADT in this clinical trial suggests that the ADT was not working completely. I assume that ADT was begun 2 months before the EBRT, continued during the 2 months of EBRT, and was continued for 2 months after that (6 months total). If the first PSA was taken 3 months after EBRT completion, the effect of the ADT had not worn off yet. Some of the cancer must already have been castration resistant. Patients received a bone scan, but some must have already had metastases that were too small to be detected by it. We see this reflected in how quickly the metric predicted mortality. In as quickly as one year from the start of treatment, mortality was 20% among those who had already reached a nadir, and it was over 0.5 ng/ml vs. 0% in those who hadn't reached it. At year one, the percent who had met the endpoint was negligible for the other endpoints. Clearly, patients with a PSA that never goes down below 0.5 ng/ml after radiation +ADT are at greater risk.

The authors recommend that patients whose PSA never achieves a nadir below 0.5 ng/ml after EBRT plus 6 months of ADT should be recommended for clinical trials of early use of second-line hormonal agents, chemotherapy, and other new therapies. This is a logical implication, but it is not likely to occur very often because standard of care has changed since this clinical trial began.

The DART 01/05 GICOR randomized clinical trial proved that among high risk patients, 28 months of adjuvant ADT was superior to the 6 months of adjuvant ADT that were used in the present study. It is less likely that the nadir will stay above 0.5 ng/ml with the longer course of ADT and with the escalated radiation dose (of about 80 Gy) that is now standard of care. So while a nadir > 0.5 ng/ml in this situation is still an endpoint indicating elevated risk, few patients will be observed to exhibit it.