PDA

View Full Version : Interesting aspect of the SpineCor



concerned dad
03-12-2009, 10:54 AM
I was at St Justine’s this week and had the chance to speak with Dr Rivard about some of the SpineCor data. On the wall of the clinic was a Table (similar to the one attached) which aggregates the results of SpineCor bracing.

I had two main questions relative to this Table. One, if not both, of them we have discussed here (well, maybe not here in ‘research’, but on the bracing forum).

I asked for clarification on weaning explaining that on some of the powerpoint presentations and technical papers the term is used to suggest the present tense, ie, weaning is ongoing or occurring for a long period of time. He said that the when they say 5 years weaning, they mean 5 years w/o any brace wear at all.

The other question I had related to the population of patients reported in the attached table who reached 5 years post bracing (N=69). They report an average POST BRACE Cobb angle for this population but only an average initial Cobb angle of the whole population (N=298). I wanted to know what the initial Cobb angle was for this smaller population (the last column in the attached table). Had these early patients all had smaller initial curves (than the later patients), the aggregated data suggesting a continued decrease in curve size would be highly questionable.

I guess I was pretty surprised at what he did next. He got on the phone and called another fellow (perhaps a student or assistant involved with the data processing). This other fellow came in and Dr. Rivard explained the question to him. He opened the main data base and extracted the data of initial Cobb angle and Risser of just those patients who are 5 years post bracing (now in the N=90’s). He calculated the average initial Cobb angle for those patients and, you know what, it was 26 degrees, the same as the average initial Cobb angle for the whole population. (For anyone not following along, in my mind at least, this increases the validity/significance of the Spinecor results.) I have to say I was impressed with the openness of the calculation. Had the calc shown a significantly lower avg initial Cobb angle, it would have really brought into question the results. His response wasn’t ‘let me look into it and get back to you” it was more along the lines of ‘lets see’.

Now, I would have loved to sit down and play with them and the data for the rest of the afternoon. Not only was the brace clinic busy but my daughter started to look a bit agitated with me.

But, I still wanted to talk about what these results imply because that is really why we are giving the SpineCor a try. These data (again, attached figure) show that, even after SpineCor bracing has stopped the curve continues to decrease. This is in stark contrast to the Daniellson data (posted on another thread) showing a gradual (or not too gradual – not enough data points there to define the shape of the curve) increase over time post TSLO bracing.

I asked Dr. Rivard about the mechanism of continued improvement after bracing ceases. He said that they believe that because the brace is dynamic, it trains the muscles over time to apply the corrective movement necessary to reduce the curve. After enough time wearing the brace, the muscles are trained to do it w/o the brace and continued correction is often seen (not a direct quote, just the gist I took away from the chat). Now the next question for us was what is “enough time”? Dr. Rivard said that if you quit bracing too early you loose all correction and it is as though all the bracing you endured was a waste of time. He would like to see my daughter braced until Risser 4 (and I lean more towards sticking with it till her digital skeletal age indicates maturity - the whole hand xray thing as my daughter Riser and skeletal age are discordant).

The last thing we talked about that may be of general interest is related to strap tension. I asked if we couldn’t just tighten the straps up a bit to shorten the treatment time. He said that would be a big mistake. There is a sweet spot in the necessary tension (my gist, not a quote). He said that a SpineCor clinic in Europe aggressively tightened the tension and the results were terrible.

mamandcrm
03-12-2009, 02:03 PM
Hi Concerned Dad,

Did you by any chance ask the doctor your earlier question on the other thread about the status of kids braced pre-growth spurt (10 or 11) who are now on the other side of the adolescent spurt (15 or 16)?

Thanks,

concerned dad
03-12-2009, 03:27 PM
Hi Concerned Dad,

Did you by any chance ask the doctor your earlier question on the other thread about the status of kids braced pre-growth spurt (10 or 11) who are now on the other side of the adolescent spurt (15 or 16)?

Thanks,

No I didn’t ask. I wasn’t sure how much time I had and I wanted to keep it focused on our particular situation. Besides that, I sort of forgot. I know the answer would be important to many of the folks here with young kids using SpineCor.

After glimpsing the database in action though I would guess it would be trivial for him to calculate the results. He would need to show data JUST for patients who were braced at Risser 0. Sanders said in a study looking at Maturity Assessment and curve progression that “The period and pattern of curve acceleration began during Risser stage 0 for all patients” (CD Note – "for all patients" in his study). SRS Criteria also allow inclusion of Risser 1 and 2 which seems to have the potential of biasing the results in favor of bracing by including girls already past their peak curve acceleration stage.

Of course, Sanders went on to show that Risser was not the best correlate to curve progression (see attached table) – something I thought we could discuss too if any one is interested. But Risser is what most everyone is using because it is easy to measure from a scolio x ray. However, as an aside, Dr Dolan said (on the “other” scolio forum) that they are taking hand xrays as part of the info they are collecting for the BrAIST study.

PNUTTRO
03-12-2009, 03:46 PM
Here is a graphical depiction of your table.

I don't see any difference among these groups. Admittedly, your 5 year follow up is underrepresented. The spread of the data is so large that none of these groups is significantly different from the others. Confounding this table is the fact that any Cobb measurement can vary up to +/-5 degrees.

Are there fewer people follow up group because they opted out of treatment, or did they not participate in follow up care? or did they opt for surgery? How many from each group opted out for surgery. I think that might be a good measure of efficacy.

Also, my error bars are plus or minus 1 SD. No one falls into the greater than 45 degree surgery group (about 68% of all patients). To cover the entire group, you have to go out 3 SDs. Again, the spread of numbers is so large, its hard to tell who failed bracing.

I have to say, I am not against bracing, per se. I'm just saying that based on your evidence here the facts are not very compelling. We also don't have a wait and see group to compare with. I would guess that they are similar.

p

concerned dad
03-12-2009, 04:53 PM
I think your error bars are showing something other than error.

Since we are dealing with a population mean it is more meaningful to look at the standard error of the mean. (the standard deviation divided by the sq root of the sample size).

You see, it is entirely likely that a population of post braced patients will have a large standard deviation in their measurements. However, as you sample more and more of these patients, you should converge on the true sample mean. At the same time however, the sample standard deviation could remain totally unchanged. If N= 1million, the standard deviation could still be 10, but our guess at the true sample mean would be pretty darn accurate.

So, I guess I would submit my attached graph showing standard error for the error bars is a more meaningful rendition of the error involved. I do admit however that I am still uncertain how exactly we deal with BOTH the error in determining the sample mean and the error involved with determining the angle of an individual measurement (5 degrees).

Of course, a test of significance is what is really required. But using SD as error bars for a sample like this is incorrect (IMO).

Here is a link to an interesting discussion on error bars
Most researchers don't understand error bars (http://scienceblogs.com/cognitivedaily/2008/07/most_researchers_dont_understa_1.php)

jillw
03-12-2009, 06:50 PM
Hi Concerned Dad,

Did you by any chance ask the doctor your earlier question on the other thread about the status of kids braced pre-growth spurt (10 or 11) who are now on the other side of the adolescent spurt (15 or 16)?

Thanks,

Mamandcrm, I did ask a similar question... about those with JIS (who therefore are clearly before their growth spurt) who had gone on to reach maturity. He said they haven't yet compiled results for the JIS folks (only the AIS which are presented in their studies), but that they would be working on putting results together for JIS next. I don't know the time frame (and didn't think to ask until I had left)

Concerned Dad, Interesting, thanks for sharing. I haven't checked yet to see if you posted about your daughter's results, but i'll check for other posts now. I hope all went well.

Pooka1
03-12-2009, 09:16 PM
A few thoughts on the Coillard data table...

1. The most obvious issue is what Pnuttro said about the fate of the 136 folks who are not accounted for at 2 years and the 229 folks who are unaccounted for by five years. For all we know they all went on to surgery.

2. There is something funny about how the original population had about half the Standard deviation as any other group in the table. I can't put my finger on it but that bothers me.

3. I don't think it is incorrect to plot the error bars from the table onto the graph and interpret it in the manner Pnuttro did. CD, why is that wrong?

4. I'd like to know how confident Coillard and Rivard are that the patients are not still wearing the braces at least some of the time.

PNUTTRO
03-13-2009, 08:43 AM
Of course, a test of significance is what is really required. But using SD as error bars for a sample like this is incorrect (IMO).

Your table doesn't say if the errors are standard deviation or standard error, so your interpretation of my graph is based on my assumption of the data--that the table had SDs.

So do your data approach significance?

If so, does it change the outcome for the patients?


Did patients drop out of follow up because they opted for surgery or no treatment?

concerned dad
03-13-2009, 09:00 AM
A few thoughts on the Coillard data table...

1. The most obvious issue is what Pnuttro said about the fate of the 136 folks who are not accounted for at 2 years and the 229 folks who are unaccounted for by five years. For all we know they all went on to surgery.

I think we had a similar discussion already. Isnt it intuitive that some of the folks haven’t hit the two year or five year mark? Also, is it reasonable to expect a 100% follow up at 5 years? (We’re not traveling back to Montreal 5 years after we’re done on my dime to provide a data point for a paper). Although it would be reasonable to ask what percentage of the 5 year patients were available for follow-up (a number I don’t see reported). From the text of the paper:

This prospective interventional study was carried out on a group of 639 patients (92.3% females) having idiopathic scoliosis treated with the SpineCor brace. Some 583 patients respected the inclusion criteria, 234 (40.1%) did not complete the treatment by brace at the time of the analysis and 51 immature patients required surgical fusion while receiving treatment, leading up to 298 patients who had reached skeletal maturity at the end of bracing. Out of this cohort of patients, 162 patients had 2 years and 69 patients had 5 years follow-up post-bracing.

But this is why I asked the question about what was the initial Cobb angle of the 5 year patient group. One is left to assume that it is a random sample of the total initial cohort. When they calculated it for me it indeed was the same average as the total initial cohort. Of course Risser could be different, curve type could be different etc. But lets just say that it increased my confidence in the results to see the Cobb was the same between the two groups.



2. There is something funny about how the original population had about half the Standard deviation as any other group in the table. I can't put my finger on it but that bothers me.

Good point, I pondered that (and still do). I go back to a result from Sanders study where he made the observation that
At this stage, curves also separated into rapid, moderate, and low-acceleration patterns, with specific curve types in the rapid and moderate-acceleration groups.
(The “Stage” he is referring to is a maturity indicator of TW3 Stage F to G – the hand xray derived thing). So, in answer to your question about what happened to increase the standard deviation (SD), the girls went through their growth spurt. Some advanced, some stabilized, some reduced. The advanced and the reduced groups would increase the SD. Back to the Coillard data,

Some 137 patients (46.0%) out of 298 stabilized their Cobb angle ( ±5°) at skeletal maturity at the end of bracing, 122 patients (40.9%) corrected their initial Cobb angle and 39 patients (13.1%) had 6° or more
progression of their initial Cobb angle.



3. I don't think it is incorrect to plot the error bars from the table onto the graph and interpret it in the manner Pnuttro did. CD, why is that wrong?

I explained why I think it is wrong in my post above. I thought you in particular would get a kick out of the blog article I linked to as it goes sort of along with your “Why most published research results are false” paper.

I would ask you to rethink my argument. The number we are trying to determine is the mean of a group. The standard deviation (SD) is indeed related to the confidence interval. However, the more measurements you include in your sample (N) the closer you will be to determining the true mean of the group. (a group with a small SD would require fewer measurements than a group with a high SD to reach the same level of confidence that you indeed have a reasonable estimate of the mean). PNuttro completely excluded consideration of N in his presentation of error bars.

To really determine the significance of the results requires an appropriate statistical test. Coillard did not do that. A possible test methodology might be what Daniellson used. They say:

Statistical Methods. Distributions of variables are given as mean, SD, and range. For comparisons between 2 groups, the Mann-Whitney nonparametric U test was used. For comparisons of proportions between 2 groups, Fisher exact test was used.
All significance tests were 2-tailed and conducted at the 5% significance level. Survival curves were produced according to the Kaplan-Meier method.



4. I'd like to know how confident Coillard and Rivard are that the patients are not still wearing the braces at least some of the time.

Well, I have to say that after talking with my daughter, this is a non issue. This brace isn’t a teddy bear to be cuddled at night. When folks are done, they’re done. Unlike the Boston brace, the Spinecor deteriorates with use. How many 16 year old girls are going to be wearing their tattered brace when they don’t need to?

concerned dad
03-13-2009, 09:30 AM
Your table doesn't say if the errors are standard deviation or standard error, so your interpretation of my graph is based on my assumption of the data--that the table had SDs.

Yes, the Coillard table presented SDs


So do your data approach significance?
Good question. I suspect it does but I dont know how to do the appropriate statistical tests. (BTW, It's not MY data.)


If so, does it change the outcome for the patients?

I guess it all depends on what you mean by "outcome".
Did it reduce the need for surgery? I dont think we can say yes or no on that.
Did it reduce final curve amplitude? Well, IF the results are significant, then we could say yes.


Also, regarding the 5 degree error in measuring Cobb angle. It seems to me that as long as this error is constant between the two groups you are comparing it should not make a difference. (if the sample size is large enough). Also, I'd like to know more about this 5 degree number. Is it intrAobserver or interobserver? I have remeasured my daughters xrays and come out with the same number (well, within 1 degree plus minus). I suppose double curves would be more difficult to measure. But, I dont even know what I'm doing and I get the same number. (Perhaps I'm doing it wrong each time. )



Did patients drop out of follow up because they opted for surgery or no treatment

another good question. From their paper...

Eight mature patients out of 298 (2.7%) require surgery after weaning of the brace.

This is at the 2 year point. They didnt present data for the 5 year point.

Pooka1
03-13-2009, 07:03 PM
I'm going to have to buttonhole a friend of mine to explain to me how to properly represent the average about the mean of observations that have a precision of +/-5.

I don't work with large numbers of averaged observations too much but when I have, I ignore the measurement precision if/when it is very small in relation to the error about the mean. In doing this, I might be part of the problem rather than the solution to the high falaciousness(!) of the literature. :eek:

Nevertheless, I don't know if that is technically correct but it's what I have done. But I would not take that approach with these Coillard et al. data because it seems like the measurement precision is not very small in comparison to the error about the mean.

The other issue that I learned the hard way is somewhat rarefied... you will get a different answer depending if you:

1. average the start points and then average the end points and compare those two averages and say the average changed x%

vice

2. calculating the change fo each pair of start and end points individually and then averaging those changes.

I know for a fact that at least working with percentages, the former is incorrect and the latter is correct. I assume it applies to non-percentages also but I don't know that without crunching some test data. I'll try that in a minute.

The last issue is that there is something NQR about simultaneously saying:

1. all patients go thru at least one 6 month weaning period,

and

2. Showing data from "2 years" wherein supposedly the patients didn't wear a brace at all.

So if it's 2 years after any brace wear then it is at least 2.5 years after the treatment stopped. The table should be much clearer it seems.

I also want to see the stats of how many patients required one, two, three, etc. weaning periods.

I'd also like to hear what Coillard says about whether patients continue to wear the brace if they fail at least one weaning. I suspect some do.

Pooka1
03-13-2009, 07:09 PM
I think we had a similar discussion already. Isnt it intuitive that some of the folks haven’t hit the two year or five year mark?

No. I don't think it is clear that ONLY 298 completed the treatment to date based on the table. Some must have dropped out for various reasons.

Essentially, working ONLY with the 298 who completed treatment and reached skeletal maturity is data selection. This group could have a much lower average Cobb angle to start with than the folks who dropped out. I'd like to see the distribution of angles in the folks who dropped out (for any reason) vice those in the 298.

Pooka1
03-13-2009, 07:22 PM
The other issue that I learned the hard way is somewhat rarefied... you will get a different answer depending if you:

1. average the start points and then average the end points and compare those two averages and say the average changed x%

vice

2. calculating the change fo each pair of start and end points individually and then averaging those changes.

I know for a fact that at least working with percentages, the former is incorrect and the latter is correct. I assume it applies to non-percentages also but I don't know that without crunching some test data. I'll try that in a minute.


Well it seems that doing it both ways returns the same number in my test data. Hmmm.

PNUTTRO
03-16-2009, 10:35 AM
If you really want to look at this critically, I have more questions.


Some 583 patients respected the inclusion criteria, 234 (40.1%) did not complete the treatment by brace at the time of the analysis and 51 immature patients required surgical fusion while receiving treatment, leading up to 298 patients who had reached skeletal maturity at the end of bracing. Out of this cohort of patients, 162 patients had 2 years and 69 patients had 5 years follow-up post-bracing.

What was the inclusion criteria?

Because you have a large SD. I will assume that most of the patients probably had a curve in the 12-24 degree range to start and a few outliers in the greater than 40 degree range. Are these the ones that opted out for surgery? Did any of them have improvement?

Anmore interesting number to me is the curve degrees of those that didn't complete the treatment--2 years later or 5 years later. Those might be a pretty good control population--wait and see group.



PNuttro completely excluded consideration of N in his presentation of error bars.

Actually, the SD is a calculation based on the number of samples in the group. The standard error is based on the average of the distribution. It doesn't account for outliers.


Finally, what is the take home message here? In your opinion.


p

concerned dad
03-16-2009, 12:43 PM
Finally, what is the take home message here? In your opinion.



The Take home message to me is that it is really difficult to discuss a technical paper unless we're able to post the entire paper.

All your questions and points are good.

PNUTTRO
03-16-2009, 06:36 PM
The Take home message to me is that it is really difficult to discuss a technical paper unless we're able to post the entire paper.

All your questions and points are good.

LOL
Let me know if there is any more I can do.
gotta reference? I can look for myself.

concerned dad
03-28-2009, 02:01 AM
Just to sort of wrap this thread up, I no longer think the topic an "interesting aspect". It is just as likely BS
:(

PNUTTRO
03-28-2009, 08:14 AM
why the change of heart? or would you rather not disclose?

p

concerned dad
03-28-2009, 01:02 PM
why the change of heart? or would you rather not disclose?

p

well, I'm not ready to write the "Why I decided to STOP bracing my daughter with the SpineCor" thread, but will someday. I did edit the first post of my SpineCor thread in the bracing forum.

But basically it boils down to my loss of faith in the researchers and their data. The revelation was prompted by conversations with you and Sharon regarding margin of error on reading Cobb angles.

In short, I believe my daughter's reported in brace correction was significantly (greater than 10 degree difference) overstated.

This is just my opinion and I am not a medical professional.

It was very disheartening to come to this conclusion. I could be wrong but we all must make up our own minds. I set the bar high on the Spinecor because of the lack of literature from folks not involved in the design of it. For me, they didnt make it over the bar.

We are now seriously considering the option of NOT bracing at all. As irritating as I may be to Pam, she has opened my eyes to the notion that the decision to brace should not be entered into lightly. The other alternative we are considering is a night brace.

PNUTTRO
03-28-2009, 08:03 PM
Sorry you aren't getting the results that you had hoped for. Its hard to separate what is good for a population vs what is good for me. Even if the results are great for a group, you never know where you will fit into the distribution. If it isn't working for you, then its just time to move on I guess. Do the best you can do.

p

concerned dad
03-29-2009, 08:07 AM
Its hard to separate what is good for a population vs what is good for me.



Unfortunately it goes deeper than that. I could come to grips with my daughter being a statistical outlier. It's the "overstatement" part of my post above that troubles me. I mean, there we were having a discussion about error bars and statistical significance and a light goes on in my head saying, and I'll put this nicely, what about researcher bias?

There is an interesting editorial style paper in this months Spine Journal. There is no abstract and I dont have the paper in front of me now. But it discussed the current state of the spine related orthopedic field and its' history of techniques and procedures that turned out to be just money making opportunities with no medical gains. As I recall, they primarily discussed questionable surgical techniques and apparatus. Very interesting. I dont know what made me think of this just now.
:D

Haleysmom
03-31-2009, 02:26 PM
Concerned Dad,
Always love reading your posts.
I am interested in your reasons that you believe you can wean before risser 4-5 or before the recommended 18 months standard treatment. If I have followed correctly, your daughter is a 0, but older (14-15)??
My daughter is NOT a typical patient...at risser 4 she was progressing rapidly even at 2 years post-menarch. So much for peak velocity growth spurt...that was done 2 years ago! According to the guidlines we were not a good candidate for Spine-cor, but made our decision for various reasons to brace. We are now approaching 1 year and doing well, and going soon for our first Out-of-brace xray. We have started discussing when we can wean, but I am leary even at a risser 4.

concerned dad
03-31-2009, 03:11 PM
Thanks for the kind word about my posts.
Weaning isnt the word for it, we abandoned the Spinecor.
Re-read my post 19 above for the extent of the details I feel comfortable sharing right now.

I dont wanrt to discourage anyone who is getting good results. But I would suggest they at least consider an independent review of their x-rays.

CAmomof2
03-31-2009, 03:50 PM
... I would suggest they at least consider an independent review of their x-rays.

Good sugguestion! That's why we get our x-rays on disc and take them to our daughter's Ped. Ortho. He is more than happy to have them! Hopefully if something's not quite right - he'll spot it......

The more sets of eyes on my child's spine the better!!!! No matter who the DR's are! It makes me wonder why they are so eager to give us a copy of the x-rays at Ste Justine's if they are giving inaccurate info.. ?? Just a thought. :confused:

concerned dad
03-31-2009, 04:16 PM
You are correct. There was no problem or hesitation at all with getting the x-rays.

edit