Help with stats interpretation!! Odds Ratio in dose-response model

• Sep 11th 2012, 01:29 AM
jennaturally
Help with stats interpretation!! Odds Ratio in dose-response model
Hi all
just wondering if someone can please help me understand this a bit better. I am reading an article and I am soooo confused about the odds ratios. I'm not very good at all this stats stuff. Any help is much appreciated!

The paper is looking at serum TSH (blood test) and the relationship between risk of thyroid cancer - they found that higher TSH is associated with a higher risk of thyroid cancer.

In the non linear dose response meta-analysis, they said all models tested showed a positive response between TSH and OR for thyroid cancer, all non linear models (showing interaction of the relationship by differing TSH levels) explained the data better than conventional linear models. Spline models (allowing a change in magnitude of thyroid cancer-serum TSH relationship at a given TSH concentration) were the best fitting; with a change occuring at serum TSH = 1mU/L.

**Now this is the bit I don't understand:**
The OR for TSH below 1 mU/L was 1.72 (1.42, 2.07) per milliunit per litre, but above this TSH level became 1.16 (1.12, 1.21) per milliunit per litre of higher serum TSH.

Are they saying that as your TSH is lower than 1, you are 1.72 more times likely to get thyroid cancer per miliunit less than 1?
And as your TSH is greater than 1, for each miliunit greater, you are 1.16 times likely to get thyroid cancer?

But with the conclusion that the greater the TSH, the greater cancer risk, I would have thought that the OR should be greater for a TSH > 1 vs TSH < 1?
Can someone please explain this to me? If it helps I can email the article but I think I should have put enough info in here

thanks
Jen
• Sep 11th 2012, 01:52 AM
jennaturally
Re: Help with stats interpretation!! Odds Ratio in dose-response model
Is it maybe talking about rate of change of TSH risk is greater at smaller TSH?

They said that 'putting it into perspective, the model predicts that the odds of thyroid cancer is 3x greater in a patient with a TSH of 4 mU/L compared with TSH of 0. Within the normal range, the model predicts a doubling of odds of thyroid cancer between a TSH of 0.65-4 mU/L; from a normal range to an elevated TSH there is a doubling of odds of thyroid cancer between a serum TSH of 2.2 and 7 mU/L.