Your questions reveal you are new to statistical modeling including logistic regression. That’s all right. Every one has to start somewhere.

To make it simple, I will start explaining how you interpret the results with only SNP and sex as covariates.

You have told us that you have coded women as “1” and men as “0”, and you have probably coded the presence of SNP as “1” and the absence of SNP as “0”.

In you logistic regression you want to predict the odds ratio of disease (coded “1”) dependent on the presence of SNP, without and with control for sex:

Without control for sex your model is simple to include the SNP variable as the only explanatory variable. If the presence of SNP predict the disease you will find an odds ratio higher than 1, with an associated p-value less that 0.05 (or a 95% confidence interval not including 1).

Now you want to control for sex: You do that simply by adding the sex-variable to your model so that both SNP and sex are explanatory variables. BUT before you do that it would be a good idea first to see if you get what you expect when you include sex as the only explanatory variable. If women have higher risk of disease than men, then again you will find an odds ratio higher than 1, with an associated p-value less that 0.05 (or a 95% confidence interval not including 1).

Now you are ready to see what happens when you include both SNP and sex as explanatory variables. Let’s imagine you find an odds ratio for SNP = 7 and an odds ratio for women = 2.

Then the interpretation is this:

• The odds of disease for men without SNP is set to “1”. This is because men without SNP with the coding you have chosen automatically becomes the reference group (sex=0 and SNP=0).
• The odds for disease for women without SNP compared to men without SNP is “2” (that is, the odds ratio found for sex).
• The odds ratio for men with SNP compared with men without SNP is 7 (that is the odds ratio fond for SNP).
• The odds ratio for women with SNP compared with women without SNP is also 7 (that is the odds ratio fond for SNP).
• Finely the odds of disease for women with SNP is 14 (that is the odds ratio for sex multiplied by the odds ratio for SNP (2x7)).

The above interpretation is based on the assumption that there is no interaction between SNP and sex. That is, that you assume that the effect of SNP is multiplicative, and is the same for women and men. This might not be true. To check for interaction you will need to also include an interaction term between SNP and sex in your model. In SPSS there is probably a smart way to do that, but to make it quite clear, you can make such an interaction term yourself: You need to generate a new variable which is simple the product of the SNP and the sex variable. E.g. SNP_sex= SNP x sex. This new variable will take on the value “1” whenever a woman has SNP and “0” for everyone else.

Now you repeat the previous model but now you also include the new variable “SNP_sex” in the model.

If the odds ratio for this interaction term is significant, the interpretation of your data will be another. Let’s say you get the following results:

Odds ratio for SNP:                       8                           p=0.03

Odds ratio for sex:                         1.8                       p=0.02

Odds ratio SNP_sex:                    0.8                       p=0.04

Then the interpretation is this:

The effect of SNP for men is an increase in the odds by a factor of 8, whereas the effect of SNP for a women is only 8 x 0.8=6,4, and the odds of disease for a women with SNP as compared with a man without SNP is 8 x 1.8 x 0.8 = 11,52.

If the odds ratio for the interaction term is in-significant, the interpretation of your data will be that from the previous model without the interaction term included.

In the example above, the odds ration for the interaction term was less than 1 (OR1.

Regards Kim

There is this line of questioning in Research Gate that might be of use to you. The link is below.

https://www.researchgate.net/post/How_can_I_adjust_age_and_gender_in_spss_using_logstic_regression

Hopefully that helps!

Since you mention SNPs, you might find the attached paper interesting. Best Wishes.

 add them as covariates, there is a step in spss that enables this function

Thanks Douglas,David,Jagdish.

To be more cleare i have to add the sexe codded 1(w)and 0(m) and the age (30-40)1,(40-50)2.......as covariates  together ?(in the same bloc) and then compare the new OR (adjusted) with the first(unajusted).Is i'am in the good way.thanks

You just enter the independent variables in the "Covariates" box in SPSS, i.e. in your case "sexe" and "age"

Thanks Ovidiu Tatar...And then I compare the new Genotype OR(afer adding sexe and age) with the  first?If it is high i said for example the adjust modele predict better the presence or the absence of the disease

In STATA: you use logit command and keep your outcome variable, exposure variables and other variables (covariates) you would like to adjust for. It  gives clean output of log odds for each variables in the model.

Be careful about categorical variables since you have to enter them as dummy variables !

Your questions reveal you are new to statistical modeling including logistic regression. That’s all right. Every one has to start somewhere.

To make it simple, I will start explaining how you interpret the results with only SNP and sex as covariates.

You have told us that you have coded women as “1” and men as “0”, and you have probably coded the presence of SNP as “1” and the absence of SNP as “0”.

In you logistic regression you want to predict the odds ratio of disease (coded “1”) dependent on the presence of SNP, without and with control for sex:

Without control for sex your model is simple to include the SNP variable as the only explanatory variable. If the presence of SNP predict the disease you will find an odds ratio higher than 1, with an associated p-value less that 0.05 (or a 95% confidence interval not including 1).

Now you want to control for sex: You do that simply by adding the sex-variable to your model so that both SNP and sex are explanatory variables. BUT before you do that it would be a good idea first to see if you get what you expect when you include sex as the only explanatory variable. If women have higher risk of disease than men, then again you will find an odds ratio higher than 1, with an associated p-value less that 0.05 (or a 95% confidence interval not including 1).

Now you are ready to see what happens when you include both SNP and sex as explanatory variables. Let’s imagine you find an odds ratio for SNP = 7 and an odds ratio for women = 2.

Then the interpretation is this:

• The odds of disease for men without SNP is set to “1”. This is because men without SNP with the coding you have chosen automatically becomes the reference group (sex=0 and SNP=0).
• The odds for disease for women without SNP compared to men without SNP is “2” (that is, the odds ratio found for sex).
• The odds ratio for men with SNP compared with men without SNP is 7 (that is the odds ratio fond for SNP).
• The odds ratio for women with SNP compared with women without SNP is also 7 (that is the odds ratio fond for SNP).
• Finely the odds of disease for women with SNP is 14 (that is the odds ratio for sex multiplied by the odds ratio for SNP (2x7)).

The above interpretation is based on the assumption that there is no interaction between SNP and sex. That is, that you assume that the effect of SNP is multiplicative, and is the same for women and men. This might not be true. To check for interaction you will need to also include an interaction term between SNP and sex in your model. In SPSS there is probably a smart way to do that, but to make it quite clear, you can make such an interaction term yourself: You need to generate a new variable which is simple the product of the SNP and the sex variable. E.g. SNP_sex= SNP x sex. This new variable will take on the value “1” whenever a woman has SNP and “0” for everyone else.

Now you repeat the previous model but now you also include the new variable “SNP_sex” in the model.

If the odds ratio for this interaction term is significant, the interpretation of your data will be another. Let’s say you get the following results:

Odds ratio for SNP:                       8                           p=0.03

Odds ratio for sex:                         1.8                       p=0.02

Odds ratio SNP_sex:                    0.8                       p=0.04

Then the interpretation is this:

The effect of SNP for men is an increase in the odds by a factor of 8, whereas the effect of SNP for a women is only 8 x 0.8=6,4, and the odds of disease for a women with SNP as compared with a man without SNP is 8 x 1.8 x 0.8 = 11,52.

If the odds ratio for the interaction term is in-significant, the interpretation of your data will be that from the previous model without the interaction term included.

In the example above, the odds ration for the interaction term was less than 1 (OR1.

Regards Kim

 Thanks Dr Kim,It is very kind to give me this deep reflexion.It is true i'am new in statistical modeling.

 Thanks Kamala and sophie

 Dear all,I have found OR=2,5 ,P=0,03 without adjustement,and OR=1,98 ,P=0,177 after adjustement with sexe and age,knowing OR=1,8 P= 0,8 for the sexe(1:women,0:men) and OR2,6=  P= 0,01 for the age(coded 0(40-49),1(50-59),2(60-69).[additif model]The interpretation is there is an association between the genotype and the disease independant on the sexe of the patients.Howeever the old patients with AA or AG SNP have more chance to developing  the disease.what do you think?