Interpreting Case Control Studies
…and the rocky road to recommendations…!
It is important to realise that simply finding an association, however ‘big’ or statistically significant, is not the same as finding evidence of causation. That is to say; a case-control study may find a strong association between smoking and lung cancer, however that association in and of itself is not proof that smoking causes lung cancer.
Case-control studies provide useful information about factors that are associated with increased or decreased risk of SIDS. When findings of increased or decreased risk are identified consistently among studies they can allow researchers and public health advisers to make recommendations about how risk can be reduced. An example of this was given in the previous section; we do not know exactly how or why sleeping alone increases risk of SIDS compared to room-sharing. Nevertheless a large number of studies conducted on different populations in different locations over a long period of time have consistently identified a protective effect from room-sharing. Researchers are in agreement about the veracity of the association, and the likelihood of a causal link between parental presence and reduced SIDS risk. Despite the precise mechanism of this link being unknown, this has led to the public health recommendation to always sleep babies in the same room as a parent or carer.
Because of the variation between studies, and the lack of information about exactly how individual or multiple factors affect SIDS, it is inadvisable to simplistically use ORs derived from single or small-sample studies to provide point estimates of the magnitude of individual risks (e.g. “x y or z ‘doubles’ the risk”), or how they compare to other risk factors (“x is three times more risky than y”). For some risk factors this creates a frustrating problem: many studies have included bed-sharing as a potential risk factor, and most have found some increased risk compared to room-sharing. This consistent finding led to broad discouragement of bed-sharing on safety grounds, however this is not the full story. For pragmatic reasons bed-sharing is often combined with sofa-sharing (to increase sample sizes). Very few studies have taken alcohol, drug use and smoking into account in analysis of bed-sharing as a risk factor – again because in the early days of research these factors were not known to be relevant, and more recently because to do so again reduces effective sample sizes to too-small levels. Recent research has suggested that bed-sharing in the absence of smoking, alcohol, drug use and sofas is no more risky than room-sharing; however due to the decrease in SIDS prevalence in recent years it is unlikely that any case-control study will be done which manages to obtain samples large enough to allow for these variables to be unpicked with a greater level of statistical confidence. The same goes for other factors which may affect bed-sharing risk such as breastfeeding. A recent meta-analysis aimed to combine results from multiple studies to provide a more robust estimate of the bed-sharing risk (Venneman 2012); however this simply pooled a large number of studies which themselves omitted relevant variables, so was unable to inform us further about the factors which might increase the risk of bed-sharing in some groups.