As a manuscript editor and freelance manuscript editing coordinator for the JAMA Network specialty journals, I am constantly having to edit out cause-and-effect language from observational studies that are not randomized clinical trials. According to the AMA Manual of Style, the word effect, as a verb, means to bring about a change; as a noun, it means result.
A randomized clinical trial is one of the few types of studies that are designed to assess the efficacy of a treatment or intervention (and thus allowed to use cause-and-effect language) because the participants are treated in controlled, standardized, and highly monitored settings.
Whenever I come across a study in which the authors are trying to determine, for example, whether the use of a certain type of drug will reduce the risk of some complication following a certain type of surgery, I need to verify whether the study is a randomized clinical trial or a report of a controlled laboratory experiment. If it isn’t, and is a report of an observational study (such as a cohort, cross-sectional, case-control, or case series study, or a meta-analysis), then all cause-and-effect language must be replaced. But by what?
Generally, association may be a useful replacement for effect. The AMA Manual of Style defines association as a “statistically significant relationship between 2 variables in which one does not necessarily cause the other. When 2 variables are measured simultaneously, association rather than causation generally is all that can be assessed.” So instead of saying the “effect of this on that,” rephrase as the “association of this with that” or the “association between this and that.”
Sometimes, however, the authors don’t agree and want me to change it back, in which case I calmly let the authors know that it is AMA style to allow cause-and-effect language only for randomized clinical trials and controlled laboratory experiments and that, perhaps in the “Discussion” section of their manuscript, they can try to make arguments to support that the association might be causal. However, to quote from one of our scientific editors, “the expression and ultimate interpretation of the findings can’t be causal.”
The use of cause-and-effect language is quite common in everyday speech, and so it is easy for most people to assume that if one event comes before another, then the first is the cause of the second. In the JAMA Network journals, findings that rely on this type of logic had to have been rigorously tested in a randomized clinical trial.—Paul Ruich
Thank you for this post! Our journal changes “impact” to “effect.” Whenever “effect” is mentioned, we look at it closely for causal language. Your explanation is so helpful!
This post oversimplifies causation. Some randomized controlled trials with positive findings do not support causation, for example, if confounding was not adequately controlled. (Confounding can happen even with a randomized trial). Some observational studies strongly indicate causation. For example smoking causes lung cancer. There are several factors that strengthen the argument for causation, and randomized trial evidence is only one of them. The others are
1. Biologic plausibility. The assertion of cause and effect is consistent with our knowledge of the mechanisms or pathophysiology of disease. For example, cigarette smoke and asbestos contain carcinogens.
2. Consistency across studies. When several studies conducted in different settings with different patients and different study designs reach the same conclusion, the argument for causation is strengthened. For example, many studies confirmed the association between smoking and lung cancer.
3. Strength of association is large. A strong association between a purported cause and effect is better evidence of a causal relationship than a weak association. For example, the relative risk for the association between smoking and lung cancer is 16. The relative risk for the association between smoking and renal cancer is 1.1.
4. Dose-response relationship exists. Varying amounts of the purported cause are related to varying amounts of the effect. For example, increasing number of cigarettes per day is associated with increasing incidence rates of lung cancer.
5. Temporal relationship consistent with causation. Causes should precede effects. For example, critics of the association between estrogen and endometrial cancer argued that estrogen treatment might not cause endometrial cancer. Instead, endometrial cancer causes bleeding which is treated with estrogen. (This is not really true.)
6. Reversible association. A factor is more likely to be a cause if its removal results in a decreased incidence of disease. For example, people who give up smoking decrease their likelihood of getting lung cancer.
7. Specificity is high between cause and effect. One cause, one effect. For example, polio is caused by poliovirus and nothing else. Poliovirus causes polio and nothing else. (But absence of specificity is not a big strike against causation. For example smoking causes bronchitis as well as lung cancer. Lung cancer is caused by asbestos as well as cigarettes.)
8. Analogy with a similar causal association exists. Examples exist where there are causes that are similar to the one in question and the association is well-established. (Not a strong factor supporting causation). In the early days of HIV, it was helpful to know that a slow virus could cause chronic degenerative central nervous system disease, similar to the chronic degeneration of the immune system by HIV.