Dale Lehman writes:
Let me be the primary (or not) to ask you to weblog about this simply launched NEJM examine. Listed below are the study, supplementary appendix, and data sharing statement, and I’ve additionally included the editorial statement. The examine is receiving large media consideration and is the continuation of a long-term trial that was reported on at a ten 12 months median follow-up. The present publication is for a 15 12 months median follow-up.
The general image is according to many different research – prostate most cancers is mostly gradual to develop and kills only a few males. Intervention can have severe unintended effects and there’s little proof that it improves long-term survival, besides (maybe) specifically subgroups. Therapy and analysis has undergone appreciable change up to now decade. The problem is of appreciable curiosity to me – for statistical causes in addition to private (since I’ve a prostate most cancers analysis). Listed below are my considerations in short:
This examine as soon as once more brings up the problem of intention-to-treat vs precise therapy. The teams have been randomized between lively administration (545 males), prostatectomy (533 males), and radiotherapy (545 males). The evaluation was based mostly on these teams, with deaths within the 3 teams of 17, 12, and 16 respectively. Determine 1 within the paper reveals that throughout the first 12 months, 628 males have been really within the lively surveillance group, and 488 in every of the opposite 2 teams: this isn’t stunning since many individuals resist the invasive therapy and potential unintended effects. I’d think about those who selected totally different teams than the random project throughout the first 12 months because the true efficient group sizes. Nevertheless, the paper doesn’t present knowledge on the precise deaths for the those that switched between the random project and precise therapy throughout the first 12 months. So, it isn’t potential to find out the precise dying charges within the 3 teams.
The paper experiences dying charges of three.1%, 2.2%, and a couple of.9% within the 3 teams. If we simply change the denominators to the precise dimension of the three teams within the first 12 months, the three dying charges are 2.7%, 2.5%, and three.3%, making intervention look even worse. If we assume that half of the deaths within the random prostatectomy radiotherapy teams have been amongst those who refused the preliminary therapy and opted for lively surveillance, then the three dying charges could be 4.9%, 1.2%, and 1.6% respectively, making lively surveillance look reasonably dangerous. In fact, I feel allocating half of the deaths in these teams on this method is a reasonably excessive assumption. Given the small numbers of deaths concerned, the deviations from random project to precise therapy might matter.
The authors have the info to conduct each an intention to deal with and precise therapy acquired comparability, however didn’t report this (and didn’t point out that they did such a examine). If that they had reported particulars on the 45 whole deaths, I might try this evaluation myself, however they don’t present that knowledge. Actually, the info sharing assertion (connected) is sort of outstanding – will the info be supplied? “No.” That actually irks me. I don’t see that there’s actually any concern about privateness. Withholding the info serves to bolster the careers of the researchers and the status of the journal, but it surely doesn’t must be that means. If the journal launched the info publicly and it was rigorously documented, each the authors and the journal might obtain widespread recognition for his or her work. As an alternative, they (and far of the institution) select to depend on their evaluation to bolster their reputations. However nowadays the evaluation is the straightforward half, it’s the knowledge curation and high quality that’s onerous. As soon as once more, the incentives and rewards are at odds with what is sensible.
One other query that isn’t analyzed however may very well be if the info was supplied, is whether or not the time of randomization issues. The article (and the editorial) cites the improved monitoring as MRI photographs are more and more used together with biopsies. Given this evolution, the relative efficiency of the three teams is perhaps altering over time – however no evaluation is supplied based mostly on the 12 months upon which an individual entered the examine.
One different factor that you simply’ve blogged about usually. For me, probably the most attention-grabbing determine is Determine S1 that truly exhibits the 45 deaths for the three teams. Taking a look at it, I see an inclination for the deaths to happen earlier with lively surveillance than both surgical procedure or radiation. In fact, the p values counsel that this may simply be random noise. Certainly it is perhaps. However, as we frequently say, absence of proof isn’t proof of absence. The paper seems to overstate the findings, as does all of the media reporting. Statements akin to “Radical therapy resulted in a decrease danger of illness development than lively monitoring however didn’t decrease prostate most cancers mortality” (web page 10 of the article) quantities to a discovering of now impact reasonably than a failure to discover a vital impact. Null speculation significance testing strikes once more.
Yeah, they need to share the goddam knowledge, which was collected utilizing tons of taxpayer {dollars}:
Relating to the intent-to-treat factor: Yeah, this has come up earlier than, and I’m undecided what to do; I just have the impression that our present commonplace approaches right here have severe issues.
My quick reply is that some modeling must be completed. Sure, the ensuing inferences will rely on the mannequin, however that’s simply the best way issues are; it’s the precise state of our information. However that’s simply low cost discuss from me. I don’t have a mannequin on supply right here, I simply suppose that’s the best way to go: assemble a probabilistic mannequin for the joint distribution of the all of the variables (which therapy the affected person chooses, together with the well being end result) conditional on affected person traits, and go from there.
I agree with Lehman that the intent-to-treat evaluation isn’t the primary purpose right here. It’s positive to try this evaluation but it surely’s not good to cease there, and it’s actually not good to cover info that may very well be used to go additional.
As Lehman places it:
Intent-to-treat evaluation is sensible from a public well being viewpoint if it carefully displays the precise medical apply. However from a affected person viewpoint of creating a choice relating to therapy, the precise therapy is extra significant than intent-to-treat. So, when the 2 estimates differ significantly, it appears to me that they need to each be reported – or, no less than, the info must be supplied that might permit each analyses to be completed.
Additionally, the subject is related to me cos impulsively I must go to the toilet on a regular basis. My physician says my PSA is okay so I shouldn’t fear about most cancers, but it surely’s annoying!
I informed this to Lehman, who responded:
Sadly, the examine in query makes PSA testing even much less worthwhile than beforehand thought (I get mine checked repeatedly and that’s my solely present monitoring, however it isn’t wanting like that’s price a lot, or ought to I say there isn’t a statistically vital (p>.05) proof that it means something?
Rattling.
Dale Lehman writes:
Let me be the primary (or not) to ask you to weblog about this simply launched NEJM examine. Listed below are the study, supplementary appendix, and data sharing statement, and I’ve additionally included the editorial statement. The examine is receiving large media consideration and is the continuation of a long-term trial that was reported on at a ten 12 months median follow-up. The present publication is for a 15 12 months median follow-up.
The general image is according to many different research – prostate most cancers is mostly gradual to develop and kills only a few males. Intervention can have severe unintended effects and there’s little proof that it improves long-term survival, besides (maybe) specifically subgroups. Therapy and analysis has undergone appreciable change up to now decade. The problem is of appreciable curiosity to me – for statistical causes in addition to private (since I’ve a prostate most cancers analysis). Listed below are my considerations in short:
This examine as soon as once more brings up the problem of intention-to-treat vs precise therapy. The teams have been randomized between lively administration (545 males), prostatectomy (533 males), and radiotherapy (545 males). The evaluation was based mostly on these teams, with deaths within the 3 teams of 17, 12, and 16 respectively. Determine 1 within the paper reveals that throughout the first 12 months, 628 males have been really within the lively surveillance group, and 488 in every of the opposite 2 teams: this isn’t stunning since many individuals resist the invasive therapy and potential unintended effects. I’d think about those who selected totally different teams than the random project throughout the first 12 months because the true efficient group sizes. Nevertheless, the paper doesn’t present knowledge on the precise deaths for the those that switched between the random project and precise therapy throughout the first 12 months. So, it isn’t potential to find out the precise dying charges within the 3 teams.
The paper experiences dying charges of three.1%, 2.2%, and a couple of.9% within the 3 teams. If we simply change the denominators to the precise dimension of the three teams within the first 12 months, the three dying charges are 2.7%, 2.5%, and three.3%, making intervention look even worse. If we assume that half of the deaths within the random prostatectomy radiotherapy teams have been amongst those who refused the preliminary therapy and opted for lively surveillance, then the three dying charges could be 4.9%, 1.2%, and 1.6% respectively, making lively surveillance look reasonably dangerous. In fact, I feel allocating half of the deaths in these teams on this method is a reasonably excessive assumption. Given the small numbers of deaths concerned, the deviations from random project to precise therapy might matter.
The authors have the info to conduct each an intention to deal with and precise therapy acquired comparability, however didn’t report this (and didn’t point out that they did such a examine). If that they had reported particulars on the 45 whole deaths, I might try this evaluation myself, however they don’t present that knowledge. Actually, the info sharing assertion (connected) is sort of outstanding – will the info be supplied? “No.” That actually irks me. I don’t see that there’s actually any concern about privateness. Withholding the info serves to bolster the careers of the researchers and the status of the journal, but it surely doesn’t must be that means. If the journal launched the info publicly and it was rigorously documented, each the authors and the journal might obtain widespread recognition for his or her work. As an alternative, they (and far of the institution) select to depend on their evaluation to bolster their reputations. However nowadays the evaluation is the straightforward half, it’s the knowledge curation and high quality that’s onerous. As soon as once more, the incentives and rewards are at odds with what is sensible.
One other query that isn’t analyzed however may very well be if the info was supplied, is whether or not the time of randomization issues. The article (and the editorial) cites the improved monitoring as MRI photographs are more and more used together with biopsies. Given this evolution, the relative efficiency of the three teams is perhaps altering over time – however no evaluation is supplied based mostly on the 12 months upon which an individual entered the examine.
One different factor that you simply’ve blogged about usually. For me, probably the most attention-grabbing determine is Determine S1 that truly exhibits the 45 deaths for the three teams. Taking a look at it, I see an inclination for the deaths to happen earlier with lively surveillance than both surgical procedure or radiation. In fact, the p values counsel that this may simply be random noise. Certainly it is perhaps. However, as we frequently say, absence of proof isn’t proof of absence. The paper seems to overstate the findings, as does all of the media reporting. Statements akin to “Radical therapy resulted in a decrease danger of illness development than lively monitoring however didn’t decrease prostate most cancers mortality” (web page 10 of the article) quantities to a discovering of now impact reasonably than a failure to discover a vital impact. Null speculation significance testing strikes once more.
Yeah, they need to share the goddam knowledge, which was collected utilizing tons of taxpayer {dollars}:
Relating to the intent-to-treat factor: Yeah, this has come up earlier than, and I’m undecided what to do; I just have the impression that our present commonplace approaches right here have severe issues.
My quick reply is that some modeling must be completed. Sure, the ensuing inferences will rely on the mannequin, however that’s simply the best way issues are; it’s the precise state of our information. However that’s simply low cost discuss from me. I don’t have a mannequin on supply right here, I simply suppose that’s the best way to go: assemble a probabilistic mannequin for the joint distribution of the all of the variables (which therapy the affected person chooses, together with the well being end result) conditional on affected person traits, and go from there.
I agree with Lehman that the intent-to-treat evaluation isn’t the primary purpose right here. It’s positive to try this evaluation but it surely’s not good to cease there, and it’s actually not good to cover info that may very well be used to go additional.
As Lehman places it:
Intent-to-treat evaluation is sensible from a public well being viewpoint if it carefully displays the precise medical apply. However from a affected person viewpoint of creating a choice relating to therapy, the precise therapy is extra significant than intent-to-treat. So, when the 2 estimates differ significantly, it appears to me that they need to each be reported – or, no less than, the info must be supplied that might permit each analyses to be completed.
Additionally, the subject is related to me cos impulsively I must go to the toilet on a regular basis. My physician says my PSA is okay so I shouldn’t fear about most cancers, but it surely’s annoying!
I informed this to Lehman, who responded:
Sadly, the examine in query makes PSA testing even much less worthwhile than beforehand thought (I get mine checked repeatedly and that’s my solely present monitoring, however it isn’t wanting like that’s price a lot, or ought to I say there isn’t a statistically vital (p>.05) proof that it means something?
Rattling.