I have 20 patients and I'm in a process of selecting the minimum number of age, sex-matched controls in order produce a statistically significant comparison at the end of the study. My intentions include to investigate particular disease causing mutations in a group of 20 patients and compare it with controls. I need to select the minimum number of age, sex matched controls, so I will be able to compare whatever our findings with the controls. Any simple theory which tells us the number of minimum controls and the suitable age, sex of each control?
If I dont get you wrong, the answer is: 20. You have 20 patients. To match controls you have to have 20 controls with thame age/sex combinations.
Btw, do you know if age and sex has an impact (or is it reasonable to assume such an impact) on the response? If not, matching will likely lower the precision, because you will sacrifice 19 degrees of freedom for nothing...
Thank you Dr Jochen, Below is my patients' information
f 4 y
f 6 y
m 7 y
m 8 y
M 8 Y
F 11 y
m 13 y
m 15 y
m 15 y
f 15 y
m 18 y
m 22 y
f 24 y
f 26 y
f 28 y
m 28 y
m 30 y
m 32 y
f 35 y
m 35 y
f 38 y
f 41 y
f 47 y
m 56 y
f 65 y
These patients are diagnosed with a primary immunodeficiency and we're trying to investigate possible mutations in a gene that is believed to be associated with this PIDs. Therefore there is a genetic predisposition. interestingly some the mutations in this particular gene have also found in controls. Primary diagnosis is based on age specific Ig levels plus there is a considerable diagnosis delay. We would also want to get cell ratios, so I guess age and sex will be important. Any suggestions based on the list above?
Any possibility that we can consider age groups (0-10, 11-20 etc) and few controls to represent that group? Because I need the minimum number due to lack of resources.
The power of any case control study is principally dependent on the number of controls in relation to cases. The best ratio of control:cases, a compromise between power and feasibility, is generally taken as 3:1, depending on the level of precision you need, and the number of variables you may find you want to analyse simultaneously in a multivariable model. The more variables, the higher the control:case ratio you need.
For the subjects who are children, the easiest way to find controls is to approach their school and seek the next pupil above and below the index case on the school register (taking patient sex into account) whose parents consent for inclusion. If the patients are from a geographically scattered population this has the advantage of giving you matching by locality too. It's more difficult with adults, but you can use workplaces for those with jobs. For those not in work you sometimes just have to advertise, either on the internet or in a local newspaper or a shop window.
If you are so resource-limited that you can't even get 20 controls, then unfortunately the work is likely to be of such low scientific quality that you are unlikely to be ethically justified in doing it, and you probably would not get it accepted for publication even if you get some data.
I've used the GMATCH macro available on the mayo clinic website: http://www.mayo.edu/research/departments-divisions/department-health-sciences-research/division-biomedical-statistics-informatics/software/locally-written-sas-macros
You will have to play around with the parameters, and how close you'd like to have your controls to your matches, but it works relatively simply, and it's free!
good luck.
you should have estimated the sample size based on comparison of two groups using any software for sample size calculations. Then at the time of selection of control include the subject of same age and sex. For example if you have included patients in cases all male with 18-35 age then for controls include only male with age 18-35.. This type of sample is called age and sex matched
I think use any Optimization technique for best result and control on global basis..
As per issues of selecting controls for cases, I suggest you read this Reference; "Grimes Schultz. "Compared to what? Finding controls for case-control studies." 2005. Lancet 365 (9468):1429-33. PMID 15836892
Hey,
In case you or others are interested in age and gender matching case and control groups you might find the Matlab-programme below interesting. It randomly matches your groups. It let's you know how many of your subjects in the case group are matched/not matched yet and their 'priority' (i.e. how often can you match each individual if you match both groups randomly):
https://au.mathworks.com/matlabcentral/fileexchange/66984-age-gender-match-year-gap-group1-group2-
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics