Thursday, September 15, 2016

The first randomized clinical trial comparing active surveillance, surgery and external beam radiation tells us little :-(

This was supposed to be HUGE! The first clinical trial ever where patients were randomly assigned to active surveillance (AS), radical prostatectomy (RP) or external beam radiotherapy (EBRT). The results were published in The New England Journal of Medicine (see this link). They started signing up men in the UK in 1999 and continued recruitment for 10 years. By 2009, they screened over 82,000 men for prostate cancer and found 1,643 men with newly diagnosed localized prostate cancer who were willing to be randomized to initial treatment with AS, RP or EBRT, about a third in each. They then followed them for a median of ten years to see how well they did with each therapy. Imagine the effort involved! Sounds good so far -- what could go wrong?

The bottom line was that all 3 therapies did about the same in preventing death. AS was found to cause higher rates of disease progression and metastases. We will explore why below.

There were several problems that arose.

1. They planned to detect mortality differences, but couldn't.

They thought there would be more deaths in the ten years of follow-up, but almost all the men defied those expectations. That's partly because of all the great new life-prolonging drugs that became available in the 21st century; drugs like docetaxel, Xtandi, Zytiga, and Xofigo. Also, in a clinical trial, patients are very closely diagnosed, treated, and monitored. They get far better care than the average patient in community practice. There were only 17 prostate-cancer related deaths

Men also survived longer because of progress in treating other diseases. But most of all, men lived longer because they frequently visited doctors as part of the study, during which they were  closely monitored for other illnesses. There were only 152 deaths from all other causes, only 9% of the total sample size. Men were 50 to 69 years of age  (62 years median) at the start of the study and were tracked for 10 years. On average, based on US actuarial tables, about 18% should have died from all causes. So the mortality rate was half of what was expected. On the average, men in the UK live two years longer than men in the US - not enough to account for the difference.

No worries. Instead of looking for mortality differences, the researchers had a secondary objective to look for differences in disease progression and rates of metastases. Those are excellent surrogate endpoints. But...

2. The intended treatment wasn't always what patients wound up doing

Although men were randomized to one of the 3 therapies, a lot of the men apparently changed their minds, as was their right. The authors of the study analyzed everything based on the intended treatment at the time they were randomized. This is how they said they would analyze the data, and they stuck with the plan. The switching that occurred was as follows:
  • Of the 545 men randomly assigned to AS,  482 (88%) stayed with it at least for 9 months. The rest decided to have surgery, radiation, no therapy, or dropped out.
  • Of the 553 men randomly assigned to RP, 391 (71%) did have surgery within the first 9 months following randomization. Most of the remainder switched to AS, the rest to radiation or other treatment, and a few chose no treatment or dropped out.
  • Of the 545 men randomly assigned to EBRT, 405 (74%) did have EBRT within the first 9 months following randomization. Most of the remainder switched to AS, the rest to surgery, other treatment, no treatment or dropped out.
  • In all, 22% of the men did not have the therapy they were originally randomized to, yet they are including in the analysis as if they did. It is unknown how this may have skewed the findings.
3. Their AS protocol was nothing like contemporary protocols.

     a. Inclusion criteria were much less restrictive

In contemporary AS protocols, almost all men are in the "low risk" category. "Low Risk" means they are stage T1c or T2a, their Gleason score is 6, and their PSA is less than 10. Some of the more restrictive AS programs, like Johns Hopkins, also include the "Epstein criteria." That means there were no more than 2 positive cores, no more than 50% cancer in any positive cores, and the PSA density must be less than 0.15 ng/ml/g. For the first time this year, NCCN included AS as an option for men with Gleason score 3+4 if no more than half the cores were positive, but only if they were otherwise low risk.

In the ProtecT trial, the only inclusion criterion was that the men had to have localized prostate cancer. See this link for their protocol. This means that they allowed men who were higher stage (T2b and T2c), higher grade (Gleason score ≥ 7), and higher PSA (PSA could be as high as 10-20 ng/ml). In fact, they previously reported that, among the AS cohort:
  • 10% had an initial PSA between 10 and 20 ng/ml
  • 22% had an initial Gleason score≥ 7 (2% were GS 8-10)
  • 25% had a clinical stage of T2 - they do not break that into subcategories, presumably most were T2a
So, many of those higher risk men would have been screened out of a contemporary AS program. The authors did not analyze this higher risk subgroup to tell us how many of the 33 cases of metastases or 112 cases of clinical progression were among them, but they do report (Table 2) that of the 8 prostate-cancer deaths in the AS group, 5 were among men with Gleason score ≥ 7 at diagnosis (vs. 2 each for RP and EBRT). The remaining 3 deaths among those diagnosed as Gleason 6 was similar to the number for RP (3) and EBRT (2). It seems that all extra deaths were attributable to higher Gleason scores in their AS program.

     b. Monitoring of men on AS was below contemporary standards.

In contemporary AS protocols, there is always a confirmatory follow-up biopsy within a year of the first screening biopsy. The repeat biopsy schedule varies from that point on, and may be every year, as it was originally at Johns Hopkins. Some AS protocols utilize mpMRI to search for suspicious areas and only biopsy as suspicion arises, others implement a biopsy schedule that may vary depending on the findings of the last biopsy. Some do TRUS biopsies, some do mpMRI-targeted biopsies, some combine the two, and some do follow-up transperineal template-mapping biopsies. But all good AS programs include follow-up biopsies.

In the ProtecT trial, patients were screened for a high PSA (> 20 ng/ml), emergent symptoms, or a 12-month PSA increase ≥ 50%. So those who had a form of prostate cancer with a low PSA output (such as some of those with predominant Gleason pattern 5) would never be discovered until symptomatic metastases occurred. I don’t know what percent ever got a second biopsy.

We recently saw what happened in Göteborg when there was no pre-determined biopsy schedule: 54 out of 474 men (11%) failed on AS. They used a similar monitoring system as the ProtecT trial: quarterly, and then semi-annual PSA tests, and re-biopsy at the discretion of the doctor.

I sometimes talk to patients who get periodic PSA tests and claim they are on active surveillance. They are putting themselves in danger. Time and again, PSA kinetics have been rejected as a sole indicator of progression for very good reasons, mainly (1) PSA is affected by many non-cancer causes, and (2) some of the most virulent prostate cancer cells put out very little PSA. There is no substitute for confirmatory and follow-up biopsies.

Let's put perspective just how egregious a difference it is when active surveillance does not include follow-up biopsies. Current estimates are that one in three TRUS-guided biopsies (12 through the rectum) will miss a higher grade of cancer. So, if one biopsy failed to detect a higher grade cancer with odds of 33%, then the odds of missing it on two biopsies is (.33) squared, etc. As the following table shows, the odds of missing the higher grade cancer with annual biopsies for ten years is about 1 in a hundred-thousand.

Odds of missing higher grade in ALL the biopsies

Now, at Johns Hopkins, for example, it was their active surveillance policy to have annual biopsies, and they used the Epstein criteria discussed above. After 15 years of follow-up, the metastasis-free survival rate was 99.4%. Laurence Klotz at Sunnybrook in Toronto has the longest running trial of active surveillance in North America. They allowed some patients as high as favorable intermediate risk, and while there was always a confirmatory biopsy in the first year, their biopsy schedule was not as rigorous as Johns Hopkins. After 20 years, of follow-up, they report metastasis-free survival of 97.2%. In the ProtecT trial, there were 33 men out of the 545 men in the AS cohort - 6.1% had already been diagnosed with metastases after only 10 years of follow-up. The outcomes of the AS cohort are very out-of-line compared to active surveillance programs that have more rigorous selection criteria and monitoring protocols.

Selection criteria
Biopsy schedule
Active Surveillance Program
Metastasis-free survival
Epstein protocol
Johns Hopkins
15 years
Less strict:
favorable risk only
Confirmatory and periodic thereafter
20 years
Any localized regardless of PSA or grade



10 years


4. Their EBRT protocol was below today's standards.

In the years prior to 1999 when they were planning this study, there were very different radiation therapies in place than have now become standard of care. This is a problem with all long-term clinical trials involving radiation technology. By the time we get the results, they are irrelevant because the technology and understanding has progressed so much. For an expanded discussion of this issue, see this link.

They used an older technology (3D-CRT) to deliver only 74 Gy in 37 treatments while adding 3-6 months of hormone therapy before and during treatment. Now, with IGRT/IMRT technology, the patients would safely receive about 80 Gy. Low and favorable risk patients probably do not benefit from adjuvant ADT -- it adds sexual side effects without adding to cancer control in most of them. Some have questioned whether the increase is justified for low or intermediate risk patients (see this link), but, as we saw, 10 years is not long enough to judge that, and there is no consequence to the higher dose in terms of side effects. It is entirely possible that the low dose they gave patients only delayed progression but did not cure the cancer.  If that is true, we may see the EBRT outcomes deteriorate when they present their planned 15-year follow-up.

ProtecT was a vast and expensive undertaking. It will probably never be repeated, and there isn't likely to ever be a US equivalent. Sadly, we can't learn very much from their current analysis of this major study, although it may yield more fruit with some subsequent analyses.

No comments:

Post a Comment