You may use the EPI Info software to calculate the sample size. you ca use it for data analysis and computing the sample size.I also used it, whe I am conducting my Efficacy of RUSF to acute malnourished children

The term "causality study" is a little vague. There are a number of approaches to investigating the causality of wasting. The approach developed by UNICEF in Sudan and described by Mara Nyawo and myself in a recent (#42) issue of Field Exchange is (probably) the simplest and least expensive option. This apporach nests a matched case-control study within a SQUEAC likelihood survey. If you are not doing SQUEAC you could adapt this approach by: (a) Taking histories from carers of cases in CMAM programs and using semi-structured interviews with key-informants (e.g. doctors, nurses, traditional healers, traditional birth attendants, &c.) in order to build the study questionnaire. (b) Using cases of SAM in CMAM programs as your cases (cheap and easy) and taking neighbourhood (i.e. neighbours of cases) controls matched by age and sex. I would avoid using clinical controls as these can introduce odd biases. There may also be a bias in using clinic cases due to the self-selected nature of the sample so you may also want to find cases in the community. The sample size required would be something like 40 cases with three or four matched controls per case. This should be sufficient to detect moderately large effects (e.g. odds ratio of about 3 or higher). A note on sample size calculation is available [url=http://www.brixtonhealth.com/simpleSampleCC.pdf]here[/url]. Data analysis would require the use of techniques that can account for the matched nature of the study. Most statistical packages (e.g. R, SAS, STATA, Epi-Info by way of an add-in module) can do this. If you are unfamiliar with case-control studies and with statistical techniques such as conditional logistic regression then you should probably consult an epidemiologist or biostatistician before proceeding. I hope this is of some use.

Thank you so much for your reply. We are planning to compare well nourished against acutely malnourished children, using both MUAC and WHZ. However, our concern is that we might not be able to find enough well nourished children to make a proper control group, like what you are suggesting.

Let me second Mark's recommendation to consult an expert before embarking on a study which intends to investigate potential causality, especially if using case-control methods. Epidemiologic evidence for causality is often a tricky thing to develop and interpret. And the medical literature is full of examples of well-intentioned case-control studies with methodologic flaws for which the results and conclusions were refuted by subsequent studies. Case-control studies can be subject to many biases which are difficult to avoid or detect. Don't spend your time and money unless you are reasonably certain the study you will do will produce useful results.

Dear anonymous 476 ACF is working with TUFTS university and IRD/WFP on this causal approach and will be proposing guidelines early 2013 after a study in burkina. Based on our field experiences, we are moving away from the strategy of statistically demonstrate associations between certain risk factors and undernutrition (association is not causality; only certain risk factors can be measured; too much focus on statistic not enough on qualitative information; time and qualified resources needed exceed our usual potentials...). The support of an epidemiologist at the onset of the survey in as mentioned important. Nevertheless, to calculate your sample size, you need to precise your study design (linear regression / case control / longitudinal study ?). Mark Myatt will certainly help on this. But I would advise to stratify on age groups; separate your models by immediate/underlying and basic causes; take time to choose your indicators (we have compiled standard indicators for a Nutrition Causal Analysis). As working on this causal approach, I would be interested to know more about your study and would be happy to exchange (nca@actioncontrelafaim.org) good luck Julien Chalimbaud

I think that Julien has some good points. I would not recommend doing anything without a qualitative first stage to give you contextual information and some idea of the risk factors that should be investigated and the best way of asking questions. This would be an essential step in any approach. As an epidemiologist I have a fondness for designs such as the case-control study as a tool for investigating causality. These do need considerable skill to get right. That said, qualitative data collection and analysis also requires considerable skill to get right. A proper causal analysis is a lot more than some people sitting in a room doing a "problem tree" analysis although that might be one of many activities. I would include quantitative work amongst those activities. I am sceptical about immediate / underlying / basic causes. I would concentrate on risk factors that are amenable to intervention and risk markers that can identify risk groups so we can target interventions. As an ageing epidemiologist broken down by age and sex (old joke) I prefer to leave political revolution to those with stronger stomachs. Reducing the incidence or severity of diarrhoea (e.g.) is often more doable and less messy than overturning economic systems. Causal analysis is a lot more complicated than (e.g.) a simple prevalence survey. You should take professional advice (or team up with someone like Julien) before doing anything.