Question on how to calculate odds ratio
villandra
interpret.
Gene rs ID Genotype 1-1 Genotype 1-2 Genotype 2-2 OR P
NGT T2D NGT T2D NGT T2D (95% CI)
CDKN2A/2B rs10811661 2,302 (65) 2,230 (71.3) 1,107 (31.2) 820 (26.2)
134 (3.8) 76 (2.4) 0.74 (0.66-0.82) 3.5×10-8
It says below each table that the odds ratios are based on a
multiplicative model, except for two SNP's that they think follows
laws of Mendelian dominance, meaning that the odds ratio must somehow
be based on all six cells of the table. It also says that they
controlled somehow for age, sex, and body mass index (so that our
recalculated odds ratios won't exactly match theirs).
Genotype 11 is TT (two of the dominant allele in the population),
Genotype 1-2 is CT (one copy of the dominant allele and one of the
minor allele), and Genotype 1-3 is CC (two copies of the minor allele
in the population.
This table is inherently confusing, because in this case, it is the
allele that is most common in the population that is the risk allele
for T2D - type 2 diabetes. The reversal makes the table so hard to
make sense of that I had trouble transcribing it accurately, and it
results in a negative odds ratio for a genetic mutation that is
positively associated with type 2 diabetes.
First of all, I can't tell if the odds ratio given for the effect of
rs1081161 on odds of developing T2D applies to each allele of
rs1081161 (one copy of T), or to having one or more risk alleles (T),
or to having two risk alleles (TT). My own actual genotype for this
SNP is CT (according to gene testing at 23andMe). Does the odds
ratio of .74 apply to each instance of T, or only to the TT condition?
Second, I want to recalculate the odds ratio so that my brain can wrap
itself around it. This gene gives me 0.75 odds, relative to not
having two copies of the mutation for diabetes, of NOT developing
diabetes? It doesn't make any sense in terms of my odds of developing
diabetes, and what is more, if the odds ratio were calculated right
way around, I suspect that the effect would probably look smaller.
Plus I've got to figure this into my other 16 genes for diabetes where
I have atleast one positive allele, and in most cases, I have the odds
ratio right way around.
I can't find a simple, easy to understand, formula or explanation
anywhere of how to calculate a single odds ratio for a three by two
table. Only for two by two tables, or else for odds ratios for each
cell, which isn't what I want - though actually, it is. LOL.
However I would not be able to check my results against the table that
was provided.
Feel free to answer me at tigge...@yahoo.com - Google Groups'
method of accessing newsnet is ridiculous!
Rich Ulrich
<vill...@gmail.com> wrote:
>Here is a three cell by two cell table that I need to know how to
>interpret.
>
>Gene rs ID Genotype 1-1 Genotype 1-2 Genotype 2-2 OR P
> NGT T2D NGT T2D NGT T2D (95% CI)
>CDKN2A/2B rs10811661 2,302 (65) 2,230 (71.3) 1,107 (31.2) 820 (26.2)
>134 (3.8) 76 (2.4) 0.74 (0.66-0.82) 3.5�10-8
>
>It says below each table that the odds ratios are based on a
>multiplicative model, except for two SNP's that they think follows
>laws of Mendelian dominance, meaning that the odds ratio must somehow
>be based on all six cells of the table. It also says that they
>controlled somehow for age, sex, and body mass index (so that our
>recalculated odds ratios won't exactly match theirs).
Well, that presentation of a 2x3 table is a mess, even though
I don't know what you are talking about.
Here is a rearrangement, best read with a fixed font.
Gene CDKN2A/2B
Genotype NGT T2D Odds, NGT/T2D
A. 1-1 2302(65.0) 2230(71.3) 1.0322
B. 1-2 1107(31.2) 820(26.2) 1.350
C. 2-2 134( 3.8) 76( 2.4) 1.763
- Now it is easy to see that the numbers in parentheses
represent percentages, adding to 100 in each column.
OR = .74 (.66, .82) p= 3.5 E08
I've added a column of "odds" for the ratio of the two
columns. I don't know where the summary OR comes from,
but it is not tremendously far from what I get for the
ratio of Row A/ Row B or Row B/Row C.
OR_AB= .766
OR_BC= .767
It is true that a 2x3 table of unordered categories
needs more than one OR for its summary.
However, if I was estimating the effect of "adding one
'2' to the Genotype", I might be trying to combine
0.766 and 0.767. I don't remember enough about genetics
to know whether this is your hypothesis, or not; or
I hope you feel in enough need to read the groups, anyway.
--
Rich Ulrich
Ray Koopman
> On Mon, 24 May 2010 08:26:36 -0700 (PDT), villandra
>
>> Here is a three cell by two cell table that I need to know how to
>> interpret.
>>
>> Gene rs ID Genotype 1-1 Genotype 1-2 Genotype 2-2 OR P
>> NGT T2D NGT T2D NGT T2D (95% CI)
>> CDKN2A/2B rs10811661 2,302 (65) 2,230 (71.3) 1,107 (31.2) 820 (26.2)
>> 134 (3.8) 76 (2.4) 0.74 (0.66-0.82) 3.5×10-8
>>
>> It says below each table that the odds ratios are based on a
>> multiplicative model, except for two SNP's that they think follows
>> laws of Mendelian dominance, meaning that the odds ratio must somehow
>> be based on all six cells of the table. It also says that they
>> controlled somehow for age, sex, and body mass index (so that our
>> recalculated odds ratios won't exactly match theirs).
>
> Well, that presentation of a 2x3 table is a mess, even though
> I don't know what you are talking about.
> Here is a rearrangement, best read with a fixed font.
Good job! I couldn't make sense off it.
>
> Gene CDKN2A/2B
>
> Genotype NGT T2D Odds, NGT/T2D
> A. 1-1 2302(65.0) 2230(71.3) 1.0322
> B. 1-2 1107(31.2) 820(26.2) 1.350
> C. 2-2 134( 3.8) 76( 2.4) 1.763
> - Now it is easy to see that the numbers in parentheses
> represent percentages, adding to 100 in each column.
>
> OR = .74 (.66, .82) p= 3.5 E08
The "OR" of .74 is not an odds ratio. It is the geometric mean
of the three Odds when they are taken the other way, T2D/NOT.
>> Feel free to answer me at tiggernu...@yahoo.com - Google Groups'