The ethics of bracing (and PT) with a Scoliscore <41

[hdugger]

Actually, in the BIOCHEMISTRY game, results CAN be 100%. This is molecules we are discussing.

I mean that, if they didn't somehow control for bracing when developing the test, then the biochemical results they ended up with aren't solid.

[Pooka1]

I mean that, if they didn't somehow control for bracing when developing the test, then the biochemical results they ended up with aren't solid.

Yes if they didn't control for bracing in a retrospective study that I would say that was "sloppy." I am not sure anyone can claim the results are confounded though. They might be but nobody can KNOW that.

[mamandcrm]

I also think that a lot of doctors (and parents) are going to get out of their comfort zone with this test when confronted with children with curves somewhere over 30* with growing left to do, who have a low score. That would just be human nature. So my point is, no matter how predictive the manufacturer says the test is (and I just assumed that none of the study participants had been braced, but perhaps that's not correct), how the scores will be used is still a very open question. It may not get the broad application that is being proposed here in this thread by the OP.

[Pooka1]

It doesn't. Assuming I understand what you're asking or saying. The three cohorts they used to test/validate the AIS-PT were selected in order to model the standard populations. 1. those referred to the doctor by in-school screening. 2. those who are seen in an active spine surgery clinic. 3. Same as #2 but with males instead of females.

In the validation tests, group 1 had 0 severe curves with an AIS-PT <41. Groups 2 and 3 had a total of 5 with a severe curve and an AIS-PT <41.

You are wondering if the scoliscore has a 100% accuracy/prediction rate for small curves, but that's not what the study was about.

I think the claim is that of the cohort of AIS girls with small curves and a score <41, 100% ended up <40* at maturity. That is what I think the abstract said and what I have been saying. Is that incorrect?

I realize the two other groups were <100% (though they were still high).

[hdugger]

The small risk group is children sent in from in-school screening. I'm not sure if their curve size was any different from the other groups, and I think there were both boys and girls in that group.

[Ballet Mom]

I wonder if it's made clear to parents that their low scoring child's curve could increase 20 degrees (given that they came in with a 20 degree curve) and still be considered a complete success.

I suspect this is the issue that will eventually be duked out in court by unhappy parents/patients if they haven't been explained the full ramifications of the decision to be set free from bracing by use of the Scoliscore test.

[skevimc]

I didn't mean that the markers would change with bracing, but rather that a braced child with a 35 degree curve and an unbraced child with a 35 degree curve would end up in same low risk pool when studied retrospectively, but they might have a very unequal risk when studied prospectively with bracing controlled for.

That is, the braced 35 degree child, if unbraced, might have ended up with a 50 degree curve, and thus would not be included with the low-risk group.

Am I missing something here?

The "low risk pool" isn't based on the curve size. It's based on the test score. But I think I understand what you're saying. The only thing they mention about bracing is:

At present, 178 of the 183 patients (97%) with AIS-PT scores >190 have progressed to a severe curve. Indeed, 175 of 178 of those patients (98%) have undergone fusion surgery. Of these 178 patients, 46.7% reported compliance with prescribed brace treatment.

They don't mention how many people with a score <190 were in a brace. What the test does the best is to say that people with a score of >190 have ~97% chance of progressing to a severe curve. The cut off of 190 was picked because the regression curve drops off a cliff and people in the 180's appear to have a risk anywhere from 40%-90%. 170's appear to be anywhere from 20%-45%. And anything below 150 has <20% chance of progressing to severe.

I think the claim is that of the cohort of AIS girls with small curves and a score <41, 100% ended up <40* at maturity. That is what I think the abstract said and what I have been saying. Is that incorrect?

I realize the two other groups were <100% (though they were still high).

The group to which you are referring is the "low risk group". This did not necessarily imply a small curve. This group was based on the population that generally gets referred to a doctor due to in-school screening. They then populated that group by selecting a specific percentage of patients for the mild-moderate-severe curves (85%, 12%, 3% respectively). It's kind of a weird way to pick the groups and I only partially understand why they did it. At any rate, of this group (n=277) there were 176 patients that had a score <41. Of these 176, 0 had a severe curve. Additionally, there were 101 patients with a score >40. 93 had a moderate-mild curve and 8 had a severe curve.

[skevimc]

The small risk group is children sent in from in-school screening. I'm not sure if their curve size was any different from the other groups, and I think there were both boys and girls in that group.

No, there were only females in the "school screening" group. The curve sizes were different in each group, however all three groups had mild-moderate-severe curves represented. They were just in different percentages. This was done by design, i.e. they wanted the school screening group to have fewer severe curves in it and the higher risk group to have more.

[Ballet Mom]

skevimc,

Could you let us know what are the Cobb angle of the curves that are classified as mild, moderate or severe in this study?

Thanks.

[hdugger]

They don't mention how many people with a score <190 were in a brace. What the test does the best is to say that people with a score of >190 have ~97% chance of progressing to a severe curve. The cut off of 190 was picked because the regression curve drops off a cliff and people in the 180's appear to have a risk anywhere from 40%-90%. 170's appear to be anywhere from 20%-45%. And anything below 150 has <20% chance of progressing to severe.

Ah, that's very helpful (and the exact reverse of how I understood it). So, the best use of this test is in predicting the people who need the *most* treatment (those above 190), rather then in recommending less treatment for those with lower scores?

[Ballet Mom]

A crystal ball for the spine
DNA test helps predict likely progression of scoliosis

By M.B. Sutherland, Special to the Tribune

January 5, 2011

http://www.latimes.com/health/ct-x-s...,7162790.story

[skevimc]

skevimc,

Could you let us know what are the Cobb angle of the curves that are classified as mild, moderate or severe in this study?

Thanks.

Mild = 10-25°
Moderate = 26-40°
Severe = "Progression to a severe curve was defined according to the usual clinical criteria (i.e., progression to a >40° curve in an individual still growing, or progression to a >50° curve in an adult)

Ah, that's very helpful (and the exact reverse of how I understood it). So, the best use of this test is in predicting the people who need the *most* treatment (those above 190), rather then in recommending less treatment for those with lower scores?

While there is strength in the "Positive predictive value" (PPV), which is what you are saying. The real finding of the study is more in the NPV "negative predictive value". But putting it all together show that those with a value <41 have an extremely low (>98%) chance of NOT progressing to >40° and those with a score >190 have ~ 97% chance of progressing beyond 40°. Or to put it another way, score <41 has <1-2% chance of progressing beyond 40°; score >190 has >97% of progressing beyond 40°.

They basically avoid using the PPV though, because they don't have the statistical power or confidence to prove it. As well, they say that the NPV is more clinically relevant than the PPV due to the percentage of over-treatment.

[Pooka1]

The group to which you are referring is the "low risk group". This did not necessarily imply a small curve. This group was based on the population that generally gets referred to a doctor due to in-school screening. They then populated that group by selecting a specific percentage of patients for the mild-moderate-severe curves (85%, 12%, 3% respectively). It's kind of a weird way to pick the groups and I only partially understand why they did it. At any rate, of this group (n=277) there were 176 patients that had a score <41. Of these 176, 0 had a severe curve. Additionally, there were 101 patients with a score >40. 93 had a moderate-mild curve and 8 had a severe curve.

This is interesting. I will get the paper. Hopefully the reasoning is clever. I love that sort of thing.

[Pooka1]

While there is strength in the "Positive predictive value" (PPV), which is what you are saying. The real finding of the study is more in the NPV "negative predictive value". But putting it all together show that those with a value <41 have an extremely low (>98%) chance of NOT progressing to >40° and those with a score >190 have ~ 97% chance of progressing beyond 40°. Or to put it another way, score <41 has <1-2% chance of progressing beyond 40°; score >190 has >97% of progressing beyond 40°.

They basically avoid using the PPV though, because they don't have the statistical power or confidence to prove it. As well, they say that the NPV is more clinically relevant than the PPV due to the percentage of over-treatment.

I imagine they had so few kids out at the high end of the graph and so many loaded towards the low end that all they could do is emphasize the negative prediction aspect. Maybe a literal handful. The positive prediction might be just as accurate but they can't know it if they didn't have enough patients in that range.

[Pooka1]

Test developers

I can't answer any of those questions. All I know is that the person who developed the test isn't saying that he's relying solely on its results. I'm guessing there's a reason for that.

Lonner didn't develop the test. Lonner is a surgeon in NYC. The group who developed it was out in Utah. These are the guys or at least these are the authors of the paper...

Ward, Kenneth MD*; Ogilvie, James W. MD*; Singleton, Marc V. MS*; Chettier, Rakesh MS*; Engler, Gordon MD†; Nelson, Lesa M. BS*