Understanding Clinical Papers. David Bowers

Figure 6.1 Findings from an ecological study about smoking and domestic fires.

      Source: From Diekman et al. (2008), © 2008, BMJ Publishing Group Ltd.

      But you may by now have spotted a flaw in this type of study: we don't know whether the individual people who died in house fires were smokers. Put another way, it is possible for a study of this design to come up with these findings even if every person who died in a house fire was a non‐smoker. This flaw is sometimes called the ecological fallacy and is a consequence of the use of aggregated data rather than the more usual research method of collecting data for each individual study participant. The other three types of analytic study set out below are more satisfactory approaches to cause‐and‐effect questions because they are able to relate the supposed risk factor directly to the outcome in each study participant.

CROSS‐SECTIONAL, TWO‐GROUP STUDIES

      Some cross‐sectional studies aim to shed light on cause and effect by recording whether people with a disease were more likely than people without the same disease to have experienced exposure to a risk factor. For example, for more than half a century researchers have recognized that patients with schizophrenia, when compared with the general population, are more likely to come from lower socio‐economic classes. What is a lot less clear is whether lower socio‐economic status is a risk factor for schizophrenia or, conversely, whether schizophrenia causes a slide down the socio‐economic scale. The cross‐sectional study design – in which the researcher measures in each subject a supposed risk factor at the same time as recording the presence of a condition – will nearly always have this chicken or egg problem (which comes first?).

In a study of bullying (Figure 6.2), the researchers persuaded 904 co‐educational secondary school pupils aged 12–17 years to declare whether they were bullied or not and to self‐report their feelings – including a scale that measured their level of anxiety. They found that those who reported being bullied also reported more anxiety. Notice that the study design does not preclude either possibility: that bullying exacerbates anxiety or that anxious children are more likely to be targets for bullies.

Figure 6.2 Extract from table of summary statistics from cross‐sectional study of bullying and self‐reported anxiety (values are numbers of schoolchildren unless stated otherwise).

      Source: From Salmon et al. (1998), © 1998, BMJ Publishing Group Ltd.

      It is a limitation of cross‐sectional designs that the direction of any effect cannot be determined because the supposed risk factor and the outcome are identified at the same time. The next two analytic study designs tackle this weakness and are able to identify the direction of any effect.

CASE–CONTROL STUDIES

      A more satisfactory way of investigating cause and effect is to concentrate on a clinical scenario in which the characteristic that you suspect might be a risk factor and the outcome can only have arisen in that order. Consider for a moment smoking and lung cancer: it is plain that contracting lung cancer cannot have led someone to become a long‐standing heavy smoker.

Notice though that it is the pre‐existence of heavy smoking that defines the difference between this example and the cross‐sectional example above. If a researcher wanted to know whether high blood pressure made stroke more likely and chose to measure blood pressure in two groups of patients, who were and were not victims of stroke, then any finding that hypertension was more prevalent in stroke patients might be a consequence of the stroke rather than a contributory cause. If, on the other hand, each set of patients had previously had their blood pressure recorded some years before, then the finding that stroke patients had a past excess of hypertension might very well point to high blood pressure being a risk factor for stroke.

      Case–control study is the label applied to a study such as the one just mentioned, about high blood pressure and stroke. The group of people with the condition are called the cases and they are compared with another group who are free of the disease and are called the controls. The comparison to be drawn is the exposure of each of the two groups to a supposed risk factor: were the cases more often exposed to the risk than were the controls? For further discussion of cases and controls, see Chapters 16 and 17.

In the study in Figure 6.3, mental health researchers undertook a case–control study to help to determine the role of cannabis in the incidence of psychosis. They recruited, from 11 sites across Europe and Brazil, 901 consecutive patients with first‐episode psychosis. They also recruited 1237 controls – people who did not have psychosis – from the same geographical catchment areas, using various sampling strategies involving random selection, from lists of postal addresses and from general practitioner lists. Using a modified version of the Cannabis Experience Questionnaire, they asked all participants about their use of cannabis and other recreational drugs. They found, for example, that there was a much higher risk of psychosis among people who reported daily cannabis use than among those who had never used the drug (after adjusting for various possibly confounding factors): the adjusted odds ratio was 3.2 – see Chapter 28 for an explanation of odds ratios in case–control studies, and Chapter 17 for material about confounding.