When should the patient and doctor consider salvage radiation therapy?

Patients and their doctors often have to make a critical decision soon after surgery – at what point after surgery, if at all, should salvage radiation therapy be started? Immediate treatment is often too early, and waiting can be too late. Two new papers give us much-needed help in finding the “Goldilocks moment.”

Let’s begin with a shared understanding of the definitions of some commonly used terms and abbreviations:

  • Aggressive pathology means that the post-op pathology report indicates that cancer was found in one or more of the following places:
    • Outside of the prostate capsule (pathological stage T3a), or
    • In the seminal vesicles (pathological stage T3b), or
    • Locally, but at a distance from the prostate (pathological stage T4), or
    • At the surgical margin, where the surgeon has cut through the cancer — a “positive surgical margin” (PSM).
  • Adjuvant radiation therapy (aRT) means radiation given after prostatectomy when there is aggressive pathology, but before the PSA becomes detectable.
  • Salvage radiation therapy (sRT) means radiation given after prostatectomy but also only after a biochemical recurrence (BCR).
  • Early salvage radiation therapy (early sRT) means radiation given after the point that aRT would be given, but before sRT would be given.
  • Wait-and-see is the strategy of waiting until after a BCR to decide what to do. The wait-and-see decision may be sRT at any time after BCR, or the patient may decide to forgo radiation, use hormone therapy, or do nothing until there is evidence of clinical progression.
  • Biochemical recurrence (BCR) post-surgery is now defined as a confirmed PSA ≥ 0.2 ng/ml. This was chosen in 2007 because it was the most frequently used threshold in published studies. When those studies began, a PSA of 0.1 ng/ml was as low as could reliably be measured. Anything below that was undetectable at the time, and 0.2 ng/ml was arbitrarily deemed a biochemical recurrence.
  • An ultrasensitive PSA (uPSA) test is any PSA test that can reliably detect PSAs below 0.1 ng/ml. While the definition of biochemical recurrence has not been changed, detectable levels of PSA are now as low as 0.001 ng/ml on some commercially available ultrasensitive tests.

The above definitions can be illustrated as potential decision points along a line showing uPSA values after prostatectomy and aggressive pathology:


Why start sooner than sRT/wait-and-see, but later than aRT?

Three major randomized clinical trials have shown that there is an oncological advantage to aRT over a wait-and-see strategy in patients with aggressive pathology after prostatectomy. This is hardly surprising, especially because “wait-and-see” includes patients who never even received salvage radiation, or may have only received palliative hormone therapy. There has never yet been a randomized clinical trial comparing aRT to sRT.

Based on those studies, both the American Urological Association (AUA) and the American Society of Radiation Oncologists (ASTRO) endorse aRT in their guidelines. However, in spite of those guidelines, only 43 percent of men who get radiation after prostatectomy do so within the first 6 months of their surgery (see Sheets et al.). Why aren’t more patients choosing aRT?

Patients (and many doctors too) worry about over-treatment, and the adverse effects of radiation on recently cut tissues. Immediate aRT may represent over-treatment for many men for whom small, detectable amounts of PSA are leaked into the serum from benign tissue left behind by surgery, or men in whom any tiny amounts of malignant tissue left behind may be indolent or susceptible to scavenging by the immune system. Kang et al. found that, among men with capsular perforation, PSMs, or seminal vesicle invasion after surgery, only 17 percent actually went on to have a true biochemical recurrence. The other advantage to waiting is that it may allow for better recovery of continence and erectile function after surgery in at least some patients.

All three of those above-mentioned clinical trials accrued participants before uPSA tests became routinely available. Although the use of uPSA testing is controversial, its widespread use has led many patients and clinicians to wonder whether waiting for some low PSA value – “early salvage radiation” – might be equivalent in outcomes to aRT.

There are randomized clinical trials underway in Canada and the UK, in Australia and New Zealand, and in France to determine whether early sRT might be equivalent to aRT in terms of survival. The combined results of those trials may have sufficient power to answer the question. Those findings will be definitive, but in the meantime two groups of researchers have retrospectively analyzed their patient outcomes for clues.

Ultrasensitive PSA reliably predicts eventual biochemical recurrence (a UCLA study)

Researchers at the University of California, Los Angeles (UCLA) looked at available evidence that the uPSA test might afford radiation oncologists the opportunity to treat patients late enough that they are assured to be on a path to clinical recurrence, yet early enough that waiting for treatment does no oncological harm. Kang et al. conducted a retrospective analysis of data from 247 patients treated at UCLA between 1991 and 2013 who were found on post-op pathology to have aggressive disease characteristics — stage pT3-4 disease and/or positive surgical margins — and who received uPSA tests. That cohort had the following characteristics:

  • Positive margins in 79 percent of patients
  • Patients were excluded if
    • They had already received radiation and/or hormone therapy, or
    • They were found to be node-positive at the time of surgery
  • Pathological stage T3/T4 in 55 percent of patients
  • Gleason score ≥ 7 in 81 percent of patients
  • Initial, pre-surgical PSA ≥ 10 in 29 percent of patients
  • Time to first post-op PSA, 3 months
  • Median number of PSAs post-surgery and before subsequent treatment, 4
  • Median follow-up, 44 months

Kang et al. found that a uPSA ≥ 0.03 ng/ml was the optimal threshold value for predicting biochemical recurrence (BCR). Other findings included:

  • uPSA ≥ 0.03 ng/ml was the most important and reliable predictor of BCR. It predicted all relapses (no false negatives: no one was under-treated), and hardly ever predicted relapses incorrectly. Only 2 percent would be over-treated by waiting for this cut-off.
  • It was especially prognostic if found on the first uPSA test after surgery.
  • Even if the first uPSA test was undetectable, any subsequent test where uPSA ≥ 0.03 ng/ml predicted BCR.
  • Other lesser predictors of recurrence were pathologic Gleason grade, pathologic T stage, initial PSA before surgery, and surgical margin status.
  • At 5 years of follow-up, 46 percent of patients had a BCR using the “standard” PSA ≥ 0.2 definition, 76 percent using the PSA ≥ 0.03 definition.
  • Treating when an ultrasensitive PSA level reached 0.03 ng/ml gave a median lead time advantage of 18 months over waiting until PSA reached 0.2 ng/ml.
  • It was necessary to monitor PSA for at least 5 years post-op, and to test at least every 6 months.

What is not known at this time is whether there is a survival disadvantage from waiting until uPSA reaches 0.03 ng/ml if it is not at that level immediately after surgery. It is also not known at this time whether there is a survival advantage from treating as soon as the uPSA level reaches 0.03 ng/ml rather than waiting until PSA reaches 0.2 ng/ml. That is, does the lead time advantage actually translate to a survival advantage?

If the findings of this study by Kang et al. are confirmed by randomized clinical trials, there is certainly a strong argument that all patients with aggressive post-op characteristics should be monitored routinely using ultrasensitive PSA tests, and offered treatment with salvage radiation when their PSA level reaches 0.03 ng/ml. It is also arguable that the definition of biochemical recurrence after prostatectomy should then be changed to 0.03 ng/ml, which would be more practical.

Ultrasensitive PSA can reliably predict eventual biochemical recurrence at 2 months after surgery (a Czech study)

A Czech study (Vesely et al.) looked at a group of 116 patients who had PSMs after surgery. Unlike the UCLA study, staging was not a selection criterion. The two studies’ goals were somewhat different. While the UCLA study didn’t start uPSA testing until 3 months after prostatectomy, in this study uPSA testing was begun at 2 weeks post-surgery. Most urologists wait for 3 months because surgery sheds a lot of PSA into the serum, and it takes a while for that excess to clear out. The goal in this study was to find out just how early in time after prostatectomy they could detect a uPSA prognostic for BCR, whereas the UCLA study sought to find out how late in uPSA progression they could detect a PSA prognostic for progression. The Czech cohort had the following characteristics:

  • Only patients with PSMs were included
  • Patients who received aRT or hormone therapy were excluded
  • Pathological stage T3/T4 in 54 percent of patients
  • Gleason score ≥ 7 in 51 percent of patients
  • Initial, pre-surgical PSA ≥ 9.2 in 50 percent of patients
  • Time to first post-op PSA, 14 days

All patients’ PSA levels were measured on days 14, 30, 60, 90, and 180 post-surgery, and at 3-month intervals thereafter; the median follow-up was 31 months.

Vesely et al. found that the uPSA on day 30 had predictive accuracy of 74 percent for recurrence, and reached a maximum of 84 percent by day 60, when the uPSA was 0.04 ng/ml (increases in accuracy afterwards were not statistically significant). The following table summarizes their findings:


  • Applying the uPSA cut-off at day 60 as the indicator for sRT would result in the decrease of over-treatment from 53 to 4 percent. Of the 33 percent who would be under-treated, 86 percent would eventually be discovered at subsequent follow-up at 3 months, and 98 percent by 4 years.
  • uPSA at day 14 was not prognostic for recurrence.
  • The following were not predictors of recurrence in men with PSMs: pathologic Gleason grade, pathologic T stage, initial PSA before surgery.
  • Neither the location nor the extent of PSMs had any significant impact on the frequency of BCR.
  • At 5 years of follow-up, 47 percent of patients had a BCR using the “standard” PSA ≥ 0.2 definition.

The main conclusion of this study is that uPSA tracking can begin earlier. Even as early as 30 days post-op, uPSA has good accuracy for predicting BCR in men with PSMs, and at 60 days, the accuracy is even better. If duplicated in larger studies, this implies that uPSA testing ought to begin 1 or 2 months earlier than it usually does at present.

The predictive accuracy of this study is somewhat lower than the UCLA study, perhaps in part because the sample size was half as big. The results in terms of uPSA prognostic threshold values, however, are surprisingly similar. Here, the threshold was 0.04, 0.01, and 0.02 ng/ml at 2, 3, and 6 months, respectively. In the UCLA study, the threshold was 0.03 ng/ml at any time from 3 months onward. Because the uPSA ≥ 0.03 at 90 days and onwards was 100 percent predictive of BCR in the UCLA study, and led to almost no under-treatment, it may well obviate the need for earlier uPSA testing advocated in the Vesely et al. study.

As in the UCLA study, however, it is not yet known whether early sRT translates to a survival advantage over waiting for BCR.

For the first time, these studies give the patient and doctor new insight into the timing and use of uPSA to predict BCR. If confirmed, setting a uPSA threshold at about 0.03 ng/ml would reduce over-treatment compared to aRT, and would reduce under-treatment compared to sRT. We await the completion of three randomized clinical trials before we have more reliable data.

Editorial comment: This commentary was written by Allen Edel. We also thank Dr. Chris King of UCLA for making the full text of the article by Kang et al. available to us, and for making a number of clarifying suggestions regarding the content of this commentary.

11 Responses

  1. [Ed: The following comment was left in relation to an earlier version of this commentary.]

    Thanks, Allen, for bringing this important preliminary finding to wider attention —- that post-RP PSA of 0.03 appears to be nearly determinative of eventual biochemical recurrence (BCR). In light of the finding, it seems to me that it behooves us to learn the following:

    (A) For a successful radical prostectomy (RP), what is the typical PSA decay curve?
    (B) What are the commonest deviations from the typical curve, and how does a particular deviation pattern correlate with outcome?

    My experience is probably typical of poor data-gathering and analysis. My first post-prostatectomy PSA at 25 days was considered “much too early” because the “half-life is about 3 days.” When I pointed out that a 25-day delay should therefore be expected to show a diminution by a factor of more than 300, they looked at me as though I were Martian.

    Presumably, after a successful prostatectomy, there is a large immediate rise for (say) P days, to a peak, PSA(P), followed by an exponential decay — a half-life of D days. If this simple model is typical, then the PSA value n days after the peak would be:

    — PSA(n) = PSA(P) / 2^(n/D)

    It would be easy enough to measure ultrasensitive PSA levels on 100 patients successfully treated with RP (yes, “success” would be determined retrospectively) to gather 400 data points:

    — Point 1: Pre-surgery
    — Point 2: 2 to 6 days post-surgery
    — Point 3: 7 to 21 days post-surgery
    — Point 4: 42 days post-surgery

    This would be enough to establish the validity or otherwise of such a model, to determine whether the “3-day half-life” is real or folkloric, and to get a rough guess at how soon the peak occurs following RP.

    If the model is promising, then the first follow-up study could establish error bars and refinements to the “successful RP” model, and subsequent studies could incorporate corrections for factors found to be correlated — perhaps age, prostate volume, positive margins, and/or Gleason sum. Thereafter, RP patients could know during their convalescence whether their postoperative course is typical or atypical:

    If typical, such men might be given the all-clear after only 1 year, rather than 10 years or longer of quarterly/yearly PSA-test anxiety. If atypical, the pattern of atypicality could be a valuable guide to whether and how to follow up with monitoring or treatment.

  2. As Allen Edel is already aware, he and the sitemaster have subtle differences in interpretation of the hypothesis suggested by Kang et al. and supported to some degree by Vesely et al.

    The sitemaster is of the (entirely personal) opinion that while uPSA testing may well help to identify early signs of biochemical recurrence, there is another problem: not every biochemical recurrence translates into a clinical recurrence, for all sorts of reasons. Kang et al. report that, in theri study cohort, “At 5 years of follow-up, 46 percent of patients had a BCR using the “standard” PSA ≥ 0.2 definition, 76 percent using the PSA ≥ 0.03 definition.” In other words, not all patients who had a PCA ≥ 0.03 ng/ml necessarily went on to have a PSA ≥ 0.2 ng/ml within 5 years. And even a PSA ≥ 0.2 ng/ml is not a definitive indicator of clinical recurrence. Thus, while uPSA testing may well be helpful in limiting the inappropriate use of adjuvant radiation therapy (which would definitely be a good thing) immediately post-surgery, there will still be hard choices to be made by patients and their doctors if the uPSA level is rising slowly over time. Those choices may depend on all sorts of factors that are important in individual cases.

    In addition, as emphasized in the above commentary, one of the “great unknowns” at this point in time is whether there is real benefit to implementing salvage radiation therapy early (i.e., at, say, 0.05 ng/ml) as opposed to at the “traditional” level of 0.2 ng/ml. We still have a good deal to learn before we will known how to optimize the execution of second-line radiation therapy for individual patients and minimize the risk for over-treatment.

  3. Paul:

    I think you may have posted that before I added the findings of the Czech study. Table 1 of the Vesely et al. study gives some info that you may be able to use in your calculations. For the men who went on to BCR, their PSA fell from 9.2 to 0.3 in 14 days, which I think would be a half-life of about 3 days. Similarly, PSA of the BCR-free men started at 7.6 and fell to 0.2 in 14 days — also a half-life of roughly 3 days. Then the two groups started to diverge. The BCR-free men continued on the same half-life rate over the next 16 days, but the BCR men did not. Vesely et al. found that the absolute level of uPSA at the 30-day mark, and especially at the 60-day mark, was a better predictor than the decline in PSA.

    There’s a lot we still don’t know about uPSA kinetics, and I hope a researcher will take you up on your suggestion in a much larger study.

  4. Sitemaster,

    Among the men whose PSA reached 0.03 ng/ml, 98% progressed further to conventionally-defined BCR of 0.2 ng/ml. So if the conventionally-defined level is a good predictor of clinical BCR, then setting the threshold to the lower figure would certainly be a good idea. I agree that clinical progression is different from biochemical progression. Unfortunately, with adjuvant or salvage radiation, we have limited tools for detecting clinical progression while the PSA is still low. Even at a PSA of 0.2 ng/ml, it is difficult to find the site of clinical recurrence using current technology. Since we don’t want the cancer to metastasize outside of the local region and become incurable, which may happen if we delay, we feel the urge to treat when there is biochemical recurrence.

    The real question, which you and I both asked above, is whether treating at biochemical recurrence (either at the conventional level, or at the .03 ng/ml level) translates to a survival advantage. Of the 3 randomized clinical trials of aRT, only one was able to show a survival advantage to aRT over wait-and-see, and that study had some serious flaws. Because PC is slow to progress, randomized clinical trials have to run for many years, perhaps 15-20 years, before any statistically valid conclusions about median survival can be reached. The sample size has to be very large to allow for drop-outs and deaths from other causes. The findings of the 3 studies investigating early sRT will have to be combined to have sufficient power to have statistically significant results.

  5. With the surgical removal of the prostate gland — the organ that produces PSA — there should be no PSA level higher than possibly 0.03 ng/ml, and if one’s PSA level is higher the concern then is that tissue has been left behind that is still able to produce PSA. Certainly if the PSA level post-RP is anything ≥ 0.1 ng/ml there should be reason for concern accompanied by close monitoring to see the progression of PSA elevation while at the same time determining if/when sRT should be administered or, alternatively, if/when ADT should be prescribed. In any case, one or the other should be implemented with elevating PSA.

  6. Good articles. Interesting concept, to redefine BCR from 0.2 down to 0.03 ng/ml. On the flip side, there are also advocates of using 0.4 ng/ml to define BCR, an order of magnitude higher.

    ASTRO/AUA, in their discussion of Guideline Statement 5 (which recommends defining BCR at a PSA of 0.2 ng/ml), cites Grade C data supporting their recommendation. Freedland (2014), who represents ASCO (Clinical Oncologists), while generally endorsing the guidelines, disagrees. He states “Defining an exact cut point for PSA recurrence is challenging. We acknowledge this difficulty, and although we agree that 0.2 ng/ml is a reasonable cutoff and is widely used in research publications and in clinical practice, the benefits of using this cut point versus others are unclear. As such, we believe that the evidence to support this recommendation was clinical practice rather than clinical evidence.”

    ASTRO/AUA Guideline Statement 8 makes it a clinical principal (i.e., a belief that is not supported by research) that sRT is more effective when given at lower levels of PSA, “ideally before PSA rises to 1.0 ng/ml”. This statement in the discussion is difficult to reconcile with their BCR definition of 0.2 ng/ml, which is also the level at which they would initiate sRT. They go on to admit “… there is no evidence to suggest a threshold above which RT is ineffective.”

    In the end, the important issue to the patient is not the level at which BCR is defined by the experts, but when to subject himself to further treatment. The potential side effects of RT are very real, and the patient wants to optimize quality of life over his remaining lifespan. Postponing or avoiding RT is certainly a goal to many of us, especially if the benefits of early treatment are unproven. This is a decision where we are likely to receive conflicting messages from our urologist, radiation oncologist, and clinical oncologist. Despite the latest research, there are too many other variables to consider in making this decision other than some semi-arbitrary definition of BCR.

  7. Dear Archie:

    Do you feel that the data presented help to confirm that immediate, adjuvant RT can be avoided in all but the most aggressive cases? Who do you think should still get such immediate aRT?

  8. Yes, the data presented here help confirm my belief, which was partly based on Kang’s earlier published research (April 2014), that ASTRO/AUA recommendations on aRT lead to potential over-treatment in up to 80% of the cases. However, with the retrospective, relatively small study nothing is proven. I did not have access to the full paper so I am relying on the data presented here.

    Regarding your second question, I can only present my personal experience. My pathology came back with G(3 + 4 = 7) with a focal positive margin. My 30-day PSA was 0.008 ng/ml. I opted for close surveillance with ultrasensitive PSA. After 2 years my PSA started slowly rising. Now I am 5 years post-surgery, and my PSA is now close to 0.1%. Do I need sRT? Perhaps. Do the new guidelines make me regret my decision to forgo aRT? In the absence of proof that I have compromised my survival, or any clinical evidence, no.

    If I had extensive multiple positive margins, and other adverse features, or a very high 30-day PSA, I probably would have opted for aRT out of desperation. But with a slowly rising PSA it seems rational to wait for a change in kinetics rather than pulling the trigger at some predetermined level that is not even sufficient to accurately calculate PSA doubling time.

  9. Thanks for the addition info Archie.

  10. Thank you for presenting this research, and for your insightful commentary.

  11. Some readers may be interested in Arthur’s response to a question about this issue on the “Ask Arthur” page.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: