newbie443 wrote:But I do see them as relevant for the purpose the study was intended. That study did not included all failures. That study was done to look at differences in tissue expansion over a set number of years. Those cases were selected for that purpose. Those MTBF numbers only have relevance to that study group of selected cases.
That study was not done to determine failure rates of the different types of implants. To be able to claim that, all failures for those years would need to be included. They were not. That is why other published information that studied failures for the purpose of device longevity do not even come close to being as low as in that study.
That is why the last link I posted found no noticeable difference in longevity between the different devices. It used data from publications that studied failure rates.
This is getting more philosphical now than maybe relevant for this thread.
And an interesting discussion not only for the purpose of establishing a possible lower MTBF for the LGX.
This is still a random sample. They took a random sample of guys who did revisions. Of course some of them will be for other reasons than implant failure. But our experience tell us that an absolute majority of revisions are because of implant failure.
And if one type is overrepresented in revisions for other reasons than implant failure (let us say infection), then this is just as relevant anyway. What we want to avoid is a revision. No matter if the reason is implant failure or infection. As a matter of fact, this is a strength in this study. They looked at revisions for whatever reason. Not only for implant failure.
What we care about is what implant gives us the longest Mean Time Between Revisions. Not Mean Time Between Implant Failure.And when they took this random sample of guys who had done a revision, they asked them how long they had their implant.
Voila, we have a reliability study.
Not the main target of the study, but with great validity and reliability nonetheless.
Random sample from the same population that would be used for a reliability study. I.e. people who needed to replace their implant.
Another example:
A researcher want to see if the second knee prosthesis surgery requires more cutting of the bones or if it can just be inserted where the first prosthesis was.
So he finds 1000 randomly selected patients who did their second knee prosthesis surgery.
He thinks one relevant variable in his study is how long time the prosthesis had been in there. He suspects that maybe those who had it there for long time would have a larger chance of requiring more bone sawing to mount the second prosthesis.
As a control question, he also checks what brand the first prosthesis was. Maybe that makes a difference for the bone sawing required.
If he then would find that the average time to second prosthesis by those operated with brand A was only half vs those with brand B, wouldn't that say something about the reliability of prosthesis brands A and B?
Of course it would.
Random sample from the same population that would be used for a reliability study. I.e. people who needed to replace their knee prosthesis.