A new article by Australian researchers in the Journal of the National Cancer Institute has suggested that the more widespread use of androgen deprivation therapy (ADT) among unscreened patients (controls) in the ERSPC and the Göteborg studies may have substantially affected patient mortality data in these trials.
The implication of their hypothesis is that the published data may, consequently, have been biased to suggest superior mortality data among the screened patients (which would then seriously undermine perceptions that these trials have demonstrated a significant survival benefit favoring screening for prostate cancer).
According to Haines and Mikos, as discussed in some detail in an article on the Medscape Oncology web site:
- In the ERSPC trial,
- 8.8 percent of men in the screening group received ADT as a primary therapy.
- 19.6 percent of men in the control group received ADT as a primary therapy.
- 8.7 percent of men in the screening group received radiation + ADT as primary therapy.
- 19.2 percent of men in the control group received radiation + ADT as primary therapy.
- In the Göteborg trial,
- 7.0 percent of men in the screening group received ADT as primary therapy.
- 22.6 percent of men in the control group received ADT as primary therapy.
And we do know that the use of ADT as a primary form of treatment for men with locally advanced prostate cancer is associated with some degree of risk for serious cardiovascular side effects (up to and including death).
Now an hypothesis is neither a fact nor a reasonable conclusion. The hypothesis put forward by Haines and Miklos is just that — an hypothesis and only an hypothesis. At this point in time, we don’t even know if the raw trial data could be accurately analyzed by independent investigators to determine whether this hypothesis is viable or hogwash, so getting hot under the collars about this hypothesis is not going to be very constructive.
As you will see from the Medscape article, however, the predictable “expert” is already calling for a detailed re-analysis of the ERSPC data. It would, after all, support his views on the lack of benefit of prostate cancer screening if Haines and Miklos were proven to be correct.
What we can say, however, is that the difference in treatment patterns between the men in the screened and control arms of these studies is significant. Is it possible that this could have affected the mortality rates in the ERSPC and Göteborg trials? Yes. It is. Do we know that that occurred? No, at present we absolutely do not.
Filed under: Diagnosis, Management, Risk, Treatment | Tagged: bias, ERSPC, results, screening, Treatment |
THE "NEW" PROSTATE CANCER INFOLINK 
CART LIKELY BEFORE THE HORSE — PRIMARY ADT AS A SIGN OF MORE ADVANCED CANCER AT DIAGNOSIS
I was one of those guys back in the late 1990s who was influenced enough by the theme that PSA testing was of questionable value (the debate is not new) that I deferred getting the test. My first-ever test was at age 56, and my PSA was 113.6, with other characteristics of an advanced case. I went straight to ADT, and after being rejected for surgery due to the advanced case characteristics, I considered radiation, but was doing so well by then on ADT that I stayed on it, making it my primary therapy until this spring.
I believe this is what is happening to the men in the control arm of the ERSPC — far more of them were not screened (except for non-compliers) and are therefore being diagnosed with more advanced disease that has greater potential for lethality. In short, because significantly more of them were not screened, their cancer was discovered later in its course, likely often not localized. ADT has long had more appeal for cancers that are not localized, hence the association with such patients.
Therefore, my hypothesis is that ADT is serving as a proxy in the ERSPC for men whose cancers were discovered later, and that lateness is driving the higher mortality rates among the control group. To my way of thinking, the higher use of ADT among the men in the screening group diagnosed with prostate cancer is further evidence of the effectiveness of screening.
Jim:
I don’t blame you for proposing what you want to believe to be true … but like the authors of this paper, you are formulating an hypothesis. You need to appreciate that there is no actual evidence to suggest (in any way, shape, or form) that the patients in the ERSPC trial who were not screened were any more likely to have advanced disease than the men who were screened at the start of the study. The investigators made quite sure of that because otherwise it would have been a pointless trial from Day 1.
Unless there’s more to this study than is provided by these brief descriptions, I’m not sure that these results really show that there was anything wrong with the methodology of the European study.
You do screening more in one group than another. Therefore, in the prostate cancer screening group, you catch prostate cancer earlier. You are therefore more likely to do surgery or radiation. In the non-screening group, you tend to catch the cancer later, after it has spread, and so you do more hormonal treatments only.
This has nothing to do with whether initially the two groups were at different stages of prostate cancer. Due to random assignment, it is highly likely that this was the case. But the screening itself would be likely to lead to a difference of prostate cancer stage at the time the prostate cancer is detected.
It is also the case that, in the screening group of the European study, a higher percentage only received watchful waiting treatment. Is that a “flaw” in the study? No, we would expect that in any study that leads to any differences in when prostate cancer is detected, it would lead to at least some differences in how prostate cancer is treated.
For there to be a flaw, the authors of this new study would have to show that, conditional on stage at which the cancer was detected — and one would have to be sure that this was very precisely measured — that the screening and control groups received different treatments for some reason.
And by the way, all of this was reported in appendix Table 9B of the European study when it was published. It seems to be ungated at NEJM. Furthermore, both in the main text of the study, and in Table 2 of the appendix, it is clear that for men diagnosed with prostate cancer in each group, a higher percentage of the men in the control group had a more advanced stage and also higher Gleason scores.
The main study is also now ungated, so you can read the text which mentions this and refers to Table 2.
So I really don’t see what is the “news” in this new study. It hardly shows “bias” in a study to show a result that would be expected to occur if the study was done perfectly.
Dear Tim:
Re: “a study to show a result that would be expected to occur if the study was done perfectly.”
This would appear to suggest a presumption on your part that a perfectly conducted screening trial would necessarily result in a prostate cancer-specific or (better still) an overall survival benefit.
While I grant you that such a perfectly conducted trial hasn’t been implemented to date, and that the necessary follow-up period is probably 30 rather than 15 years, I would also note that — as a corollary — we don’t yet have any reliable demonstration of the effectiveness of screening in extending prostate cancer-specific or overall mortality! You are making an unjustifiable assumption based on wishful thinking. There’s nothing wrong with wishful thinking, but as an experienced statistician I expect to hold you to a pretty high standard when it comes to presumptions like the one above!
As to why this paper is news … It’s news because it is yet another controversial opinion on a controversial topic — regardless of whether the authors are right or wrong (which may or may not be demonstrable).
What Tim said! Amen! (That’s what I was trying to get to in my initial post.) Yes, agreed and understood that the degree of cancer in each group should have been equal at the outset of this screening trial, but it was the diagnosis point where we should expect differences if screening is effective, and that is what happened per Tim’s look, which I just observed for myself. It occurred to me later today that the data on stage of cancer at diagnosis had no doubt been tracked, and thank you Tim for providing navigation aids to the source.
Table 2B, re Gleason score for the total ERSPC cohort with a cut-off of December 31, 2008, shows that 8.1% of patients in the screening group were diagnosed with a Gleason greater than 7, while 14.0% of the control group had a Gleason greater than 7, demonstrating greater aggressiveness in cancers at diagnosis in the control group.
Similarly, Table 2A, with the same cut-off, shows totals for diagnoses of stage T3 and T4 — no longer localized — of 9.0% for the screening group but 17.4% for the control group … nearly double! That again demonstrates a far greater proportion of uncontained prostate cancer in the control group at diagnosis.
One endpoint for success in any screening trial could be the avoidance of ADT and other advanced therapies such as chemotherapy and Provenge. Mortality is not, should not be the only measure of success from the patient’s viewpoint.
I do not believe that the authors have made a strong case for bias in the study.
(1) The incidence of metastatic disease and T3+T4 disease was much higher in the control group so the use of either ADT alone or ADT + XRT would be expected to be higher in the control group. Of course, we do not know whether the duration of ADT was used appropriately in both groups and if not, that could account for some of the difference. However, a meta-analysis of randomized studies of ADT + XRT has failed to demonstrate the same excess mortality that has been seen in uncontrolled, retrospective analyses that suggested ADT caused more deaths. For that reason, I believe that it is hard to make a strong case that more ADT was the reason for more deaths in the control group.
(2) At the same time that advanced disease in the control group led to more use of ADT, watchful waiting was used much more often in the screened group (18.6% vs 10.1%). At least from the Scandinavian trial of watchful waiting vs. surgery, we know that mortality was higher by about 6% in the watchful waiting group, so this would allow the pro-screeners to argue that the true benefit of screening in this trial might have been even higher if watchful waiting was used less often in the screened group.
For these reasons, I find it hard to accept that treatment bias would explain the small difference in favor of screening.