clogit on mulptle matched pairs

[Nick Dalesio]

Hi,

I have a database of 2000+ twins_siblings, where I want to evaluate 2 potential binary outcomes of sinus disease (sindx). Patient relationships are listed as familyid(same id for family members) and zygosity (mono- & dizygotic twins or sibling). I want to analyze whether there is a higher odds of developing sinus disease if a specific haplotype in a gene is present. There are currently 3 haplotypes (bitfhomo, bithetero, bitnofxn), each their own binary variable. I want to match patients by their zygosity and familyid, with the clogit to give me the OR for each zygosity pairing. Currently Stata is giving me only 1 readout. I have the following so far:

*To match the variables
.ccmatch famid zygosity, id(jhsid) cc(sindx)

*Conditional logistic regression for each gene haplotype
.clogit sindx bitfhomo, group(match) or
.clogit sindx bithetero, group(match) or
.clogit sindx bitnonfxn, group(match) or

For each read-out, I get only 1 OR, however, and want an OR for each of the zygosity listings.
Can you help?
Thanks,
Nick

[Nick Dalesio]

I'm sorry...above it should say "1" potential binary outcome of sinus disease and not "2."
Nick

[Clyde Schechter]

I'm not entirely sure I understand what you want. Perhaps you are looking to model an interaction between zygosity and the haplotypes? If so, you need a zygosity variable--you might code it 0 for sibling, 1 for dizygotic twins, and 2 for monozygotic twins. Then you could run

clogit sindx i.bitfhomo##i.zygosity, group(match) or
lincom 1.bitfhomo, or
lincom 1.bitfhomo#1.zygosity + 1.bitfhomo, or
lincom 1.bitfhomo#2.zygosity + 2.bitfhomo, or

and analogous commands for the other haplotypes.

[Nick Dalesio]

I have the model as described above, however, in addition to matching on zygosity, I have matched on their family ID as well. For example, a MZ twin pair will both have the same family ID, zygosity = 2. The outcome is sinus disease (sindx) and the "exposure" is the genetic SNP (bitfhomo).

This is the Table that I would want to view evaluating the OR to develop sindx:

Monozygotic twins

Sinus Disease No Sinus disease
bitfhomo Yes
bitfhomo No

Dizygotic twins

Sinus Disease No Sinus disease
bitfhomo Yes
bitfhomo No

Does this make more sense?

Thanks!
Nick

[Clyde Schechter]

The code I showed in #3 will get you the odds ratios asked for in #1 (as well as a third odds ratio for non-twin siblings).

But now you seem to want certain tables. I think the simplest way to do that is

by zygosity, sort: tab bitfhomo sindx

But bear in mind that the tables you are asking for do not correspond to the matched-pair analysis, so don't expect the odds ratios from those tables to match those of the analysis. For tables that correspond to the matched-pair analysis, see -help mcc-.

By the way, I also don't understand how monozygotic twin pairs fit into this study. Being monozygotic, unless there are post-conception mutations in the genes you are looking at, the pair will always have the same haplotypes, right? Therefore, within those pairs, there is no possibility that the haplotype is associated with sindx (nor anything else!)

[Nick Dalesio]

Hi. I apologize...the table was just used to show the what data I was trying to acquire. It is true that MZ twins will have the same SNPs. The gene has 3 haplotypes, 1 homozygote functional, 1 heterozygous, and 1 homozygous for non-functional. If I am matching twins who have the same entire genome, but different outcomes, I will know this allele has nothing to do with the outcome. This is the point of the project. Thanks for the help.