six degree curve

[kt2009]

I am new to this part of the forum. I am recently recovering from my own surgery. I have watched my own children like a hawk, but surprisingly, my daughter was diagnosed with a 6 degree curve this winter. She is 14, has stopped growing and the doctor said that they wouldn't have done anything for her anyway.

My concern is: I was told as an adolescent that after bracing my curve would not progress. My curve was much larger than hers. But does anyone have any information on adolescents with small curves and their chance of progression? My heart would break if she had to go through what I have been through.

[hdugger]

Hi Kathy,

I don't have the link to the chart handy (there's one that shows the likelihood of progression based on the size of the curve when discovered), but my memory is that a 6 degree curve is extremely unlikely to progress in a teenager who has stopped growing. I'm not even certain it can be diagnosed as scoliosis if it's under 10 degrees - most people have some small degree of curve.

I completely understand your concern, but I think it's very likely that she's safe.

[concerned dad]

Take a deep breath there Kathy, I think your daughter is in pretty good shape.
The only actual data I can share to support that is the attached figure (maybe others can cite other supporting information).
Using the lowest curve amplitude (10 degrees) and figuring your daughter's skeletal age is Stage 3 or above (likely above), the table indicates a 0% liklihood of progressing to surgery (I know that (surgery) wasnt your question, but this is the only data I have to share).
The 95% confidence intervals are very small (0 to 0). Pretty good.
Now yourself on the other hand, after bracing, likley fell into one of the ranges lower on the graph (with much more vague confidence intervals).
In life nothing is guaranteed, but I think there is a vanishingly low chance your daughter will go through what you did.

[kt2009]

Fascinating...it just shows how much need for further research we have to do. Wide variations where they had too small a sampling...

People my age and older are realizing that what we knew years ago is nothing compared to what we know now. After I took my brace off, I thought nothing of it until 1 year ago, when I noticed how disfigured I was. My doctor thirty years ago was Wood Lovell, who was a champion in the field of pediatric orthopedics...and still we've come so far.

Thanks so much for the info. I will rest easy...but keep a close eye!

[Pooka1]

Take a deep breath there Kathy, I think your daughter is in pretty good shape.
The only actual data I can share to support that is the attached figure (maybe others can cite other supporting information).
Using the lowest curve amplitude (10 degrees) and figuring your daughter's skeletal age is Stage 3 or above (likely above), the table indicates a 0% liklihood of progressing to surgery (I know that (surgery) wasnt your question, but this is the only data I have to share).
The 95% confidence intervals are very small (0 to 0). Pretty good.
Now yourself on the other hand, after bracing, likley fell into one of the ranges lower on the graph (with much more vague confidence intervals).
In life nothing is guaranteed, but I think there is a vanishingly low chance your daughter will go through what you did.

There is something funny about that table... there are huge jumps between small probabilities (shaded) in each and every column and then a huge probability (unshaded). That suggests a real physical threshold at each combination of maturity stage and Cobb angle which strains credulity. You would expect a smoother distribution I think with no sharp thresholds, especially at each stage. It almost can't be right and must be an artifact of extremely small sample sizes in the face of what is known to be a wildly variable condition.

If those data are to be believed, someone should be looking specifically for some physical reason each of those combinations should have a huge probability jump at some point. That might be an important clue if real (which I don't think it is). You can fool yourself with large data sets... it's not unusual to fool yourself with small ones. This is very difficult research.

For example

[Pooka1]

For example

WOW! I lost my train of thought there and I can't now figure out where I was going with that!

The neurons... they peter out...

[concerned dad]

Sharon, you make a valid observation. I neglected to include the specific discussion the author made in the text. I did note though that the researchers intent was not so much to assess likelihood of curve progression, but rather to assess the predictive power/utility of skeletal maturity indicators in that determination (edit, I just reread my post, I guess I didnt "note" this. Oh well, I should have).

The risk of progression is determined with use of logistic regression methods. The estimated probability of the final curve being >50 degrees for Lenke type-1 curves (single thoracic curves) and type-3 curves (double major curves with a predominant thoracic curve) is shown in Table III. The average risk of progression is that determined by logistic regression. Where there was no corresponding data point, a logistically created one was added so that the probability could be estimated for all combinations of main curve magnitude and maturity stage. The 95% confidence intervals are shown in parentheses. Wide confidence intervals reflect a small sample size, indicating that the probability may require modification with a larger sample, and therefore those particular cells should be interpreted with caution.

Table III shows the relationship between maturity stage, curve size, and the probability of the curve progressing to >50.The unshaded cells correspond with combinations for which surgery would be a plausible treatment if >50degrees at maturity is accepted as a threshold for surgical treatment. For example, if a patient presents with a 25 degree curve, the prognosis would clearly depend on the maturity stage of the patient. For stages 1 and 2, surgery is quite likely; conversely, for stages 4 to 8, surgery is unlikely. For Stage 3, the probable outcome is not so clear.

As shown in Table III, a 25 degree curve had substantially different chances of reaching 50 degrees, depending on the skeletal maturity stage. Since stages 1 through 5 all typically occur before iliac apophyseal ossification, a 25degree curve in a girl who is in Risser stage 0 and skeletal stage 2 has a 100%(confidence interval, 92% to 100%) chance of reaching >50 degrees despite bracing. On the other hand, the same 25degree curve in a girl who is in Risser stage 0 but skeletal stage 4 has an essentially 0% (confidence interval, 0% to 5%) chance of reaching >50 degrees. A 25 degrees curve in a girl who is in skeletal stage 3 has a 29% chance of reaching >50 degrees, but the confidence interval (3% to 84%) is too wide for accurate estimation. To predict the probability of the curve being >50degrees to within ±10% (with 95% confidence), a power analysis suggests thirty-six or more patients must be followed through to skeletal maturation for each curve type.

There are other caveats the author reports in the paper and we should be cautious to try to read too much into this chart. It is the only recent study that came to my mind to address Kathy's question. There may be better/more relevant information. Sharon, I know you like to see confidence intervals reported. You must be pleased to see them included in this research. The “take away message” from this data is the stuff I bolded above for emphasis. I think this is the information he wanted to get across. Risser 0 is NOT the important thing. The important thing is skeletal maturity stage (as assessed from an xray of the hand).

I am also attaching Table 1 from the same paper.

[Pooka1]

The average risk of progression is that determined by logistic regression. Where there was no corresponding data point, a logistically created one was added so that the probability could be estimated for all combinations of main curve magnitude and maturity stage.

Well I have my standard "what isn't linear on a log-log plot?" comment?

To predict the probability of the curve being >50degrees to within ±10% (with 95% confidence), a power analysis suggests thirty-six or more patients must be followed through to skeletal maturation for each curve type.

That's interesting. Close to but larger than 30 which is some magic number in stats.

Sharon, I know you like to see confidence intervals reported. You must be pleased to see them included in this research. The “take away message” from this data is the stuff I bolded above for emphasis. I think this is the information he wanted to get across. Risser 0 is NOT the important thing. The important thing is skeletal maturity stage (as assessed from an xray of the hand).

Well me and anyone who wants to know how precise the data are like error bars.

I tried to graph those data in Excel with the error bars and also for the +/- 5 degree error on the angle but I couldn't do it. I need by scientific graphing package at work. But I got far enough to guess why they tabled the data rather than graphed it.

Interesting. This stuff is very hard to do.

[concerned dad]

Logistic regression in this case is not a logarithmic analysis but rather a polynomial expression of risk. A pretty good “Lay explanation” is here on Wiki using death from heart disease as an example. (edit, on second thought, I guess it is a type of logarithmic analysis. Darn, this stuff is confusing)

Your observation about the linear (or non linear) aspect is correct and commented on elsewhere in the paper.

Length of Stages and Correlation with Curve Behavior (Validity)
The stages are not linear. The number of sequential six-month visits for each stage that could be identified as having a lower stage on a previous visit and a higher stage on a subsequent visit was counted. Curve-fitting demonstrated the progression of the stages relative to the curve acceleration phase. The overall progression from stage 2 to stage 8 occurred over four years, starting about twelve months before the curve acceleration phase and lasting three years beyond the start of the curve acceleration phase. Stage 4 occurred about twelve months after the curve acceleration phase began, and stage 8 occurred about thirty-six months after the curve acceleration phase began. Once the patient reached stage 4, each stage progressed rapidly, typically within six months of the previous stage, until stage 7. The Pearson correlation coefficient of the maturity measurements with the curve acceleration phase was 0.91 (p < 0.001)

Correlation of the Stages with Scoliosis Curve Behavior
In addition to reliability, a useful maturity staging system for scoliosis must correlate with curve behavior. Similar to the DSA scores from which they are derived, the stages of this system are highly correlated with the timing relative to the curve acceleration phase (Pearson r = 0.91). This correlation is stronger than those associated with either the Risser sign or the Greulich and Pyle atlas1. The correlation is even stronger using an S-shaped curve fit (a fourth-order polynomial [r = 0.94, r2 = 0.88]), reflecting the rapid skeletal maturity at the middle stages (stages 4 through 6) and the slower maturity changes at either end.

I am not aware of the number “30” having any special significance to statistical studies. As I understand it, when they design studies researchers use existing information about expected outcomes they wish to measure and figure out how large of a sample size (power) they need to reach statistical significance. This study was underpowered (he said they would have needed 36 more patients for each curve type) but still has scientific merit, it can be used to design future studies. It was powered sufficiently to demonstrate the superiority of the method of measuring skeletal maturity, just not powered sufficiently to predict curve progression.

Somewhat related, as an investor, I am patiently awaiting the results of an ongoing study for the treatment of pancreatic cancer. The study was powered to detect a 20% increase in longevity based on the use of the drug they’re testing. They calculated they needed 330 patients for the trial and scheduled interim “peeks” at the data after 1/3, and 2/3 of “events” happened. About a year ago, they had the first “peek” and the results are reported here. Check out Table 2 and note the difference between the control (SOC- standard of care) and the drug they’re evaluating (SOC+drug). At the 75% mark, the data looks very promising with 75% of the people using the SOC passing before 11.8 months and 75% of the people using SOC and the “drug” passing before 19.4 months. An 8 month increase in longevity is huge for a drug that treats a deadly form of cancer. However, those numbers in parenthesis tell a different story. They are the 95% confidence intervals, and for this study, so far, at this current power/stage, the ranges overlap. The next “peek” at the data will come any day now. The big question is: will the data show significance at this interim point or will the study need to continue.

I guess I’m getting pretty far off topic. Why is this relevant? Well, before our discussions on this forum I never gave much thought to medical trials, treatments, ethics of medicine, statistics involved with medical issues, genetics, DNA, RNA, RNAi…. I still only understand a small fraction of the issues. But our discussions have indirectly pointed me toward some interesting things. As Pnuttro pointed out just the other day on another thread, administering RNAi therapies into the target organ is a challenge. Just this morning I read of a new study showing success with inducing dystrophen in the heart of a mouse model of muscular dystrophy. Yeah just a mouse, but baby steps. Interesting stuff.

[Pooka1]

Length of Stages and Correlation with Curve Behavior (Validity)
The stages are not linear. The number of sequential six-month visits for each stage that could be identified as having a lower stage on a previous visit and a higher stage on a subsequent visit was counted. Curve-fitting demonstrated the progression of the stages relative to the curve acceleration phase. The overall progression from stage 2 to stage 8 occurred over four years, starting about twelve months before the curve acceleration phase and lasting three years beyond the start of the curve acceleration phase. Stage 4 occurred about twelve months after the curve acceleration phase began, and stage 8 occurred about thirty-six months after the curve acceleration phase began. Once the patient reached stage 4, each stage progressed rapidly, typically within six months of the previous stage, until stage 7. The Pearson correlation coefficient of the maturity measurements with the curve acceleration phase was 0.91 (p < 0.001)

Okay that was pretty interesting I must say.

Correlation of the Stages with Scoliosis Curve Behavior
In addition to reliability, a useful maturity staging system for scoliosis must correlate with curve behavior. Similar to the DSA scores from which they are derived, the stages of this system are highly correlated with the timing relative to the curve acceleration phase (Pearson r = 0.91). This correlation is stronger than those associated with either the Risser sign or the Greulich and Pyle atlas1. The correlation is even stronger using an S-shaped curve fit (a fourth-order polynomial [r = 0.94, r2 = 0.88]), reflecting the rapid skeletal maturity at the middle stages (stages 4 through 6) and the slower maturity changes at either end.

It seems like if they know when the curve is moving rapidly, and braces are known to though to work during this critical phase, then all attention should be paid to this aspect. And braces should mainly if not only be worn during this relatively short period. Yet while I have heard folks claim this is a critical time to brace, I don't think I've seen anything on efficacy of bracing during this rapid acceleration phase versus any other phase. It might make sense to hit this phase but it might also be the case that no brace can possibly halt most curve acceleration during this phase and may only work during the slower acceleration phases.

In re S-curve and 4th-order polynomial, you can use any order polynomial to empirically fit data though it is always best to use the lowest order that captures the data. These are just fitting curves in an empirical way, not linking anything to a physical process in a rigorously deterministic way. But that is okay... if you can predict it accurately and precisely, you don't need a mechanistic model necessarily.