Does the prospective cohort study “Migraine and risk of cardiovascular disease in women”, demonstrate good quality science?
What is good science? To properly evaluate the quality of a scientific study, it is important to highlight the guiding principles for proper scientific inquiry. The book “3 Guiding Principles for Scientific Inquiry”, which includes one chapter highlighting important aspects of scientific inquiry, states that science involves “seeking conceptual (theoretical) understanding, posing empirically testable and refutable hypotheses, designing studies that test and can rule out competing counter-hypotheses, using observational methods linked to theory that enable other scientists to verify their accuracy, recognizing the importance of independent replication and generalization.” (cite) It is important to acknowledge that not all scientific studies will be able to meet all of the aforementioned criteria as there may be many inhibiting factors such as cost, ethics, and time. However, the criteria should be something that scientists should strive to meet when conducting a scientific study. The study being evaluated in this critique is, “Migraine and the risk of cardiovascular disease in women: prospective cohort study”, conducted by Tobias Kurth and his co-authors to evaluate the association between migraine and incident cardiovascular disease and cardiovascular mortality in women (Cite). The study involved collecting data from 115 541 female Nurses’ Health Study II participants over a 20-year period to compare incidence for cardiovascular disease events and mortality in those with migraine vs no migraine (cite). The results from the study showed that there was a consistent link between migraine and cardiovascular disease events (including mortality) and that migraine was associated with a significantly increased risk for these illnesses. Overall, the study exemplifies good quality science because it aligns with many principles of scientific inquiry , acknowledges the limitations and accounts for confounding variables, and does not claim conclusions unsupported by the data.
The paper starts by providing insight, mainly the rationale and objectives, of the study in the introduction section. Migraine is a prevalent headache disorder, and an association between migraine and certain vascular diseases (ischemic stroke) has been recognized in other studies. Although scientists speculate a potential mechanism for this observed association (e.g. endovascular dysfunction), the overall understanding of underlying processes is poor (cite). The objective of the study is to describe if people who have migraine will have a different incidence of developing cardiovascular ailments in the future compared to people who do not have migraine based on the data compiled in the Nurses’ Health Study II. Does all the information lead coherently to the purpose of the study? Yes, because there are supporting details which provide sufficient information to explain the relevance of this study. The introduction poses the issue, migraine and cardiovascular disease, and how it’s high incidence is a big public health concern especially when combined with cardiovascular disease. Although there is a demand for studies to determine a mechanism for this association, there is also a demand to track young people with/without migraine over a long period of time for incidence of cardiovascular disease. This is how this study separates itself from other pieces of literature. The data compiled from this helps provide not only a more convincing association but a better understanding of the long-term implications of migraine on cardiovascular health.
Did the study design align with most of the guiding principles of scientific inquiry? Yes, as it has a representative study population and strong data collection method which attempts to limit potential confounding variables. The study population was 115 541 nurses participating in the Nurses’ Health Study II, the largest investigation into risk factors for chronic illnesses in women, between the years 1989 and 2011. The individuals involved were all fairly similar in characteristics; they are all women, all nurses, and between ages 25-42 which is a potential strength as homogenous population could limit potential confounding variables. Studying females was also useful because migraine are more prevalent in women than men, and the age group selected happens to overlap with the demographic of women who have the greatest number of migraines. Therefore, the authors collected data on a large sample size of a high risk population which add relevance and significance of this study. The study population was divided into two, migraine vs no migraine and the study compares the incidence of cardiovascular disease (CVD) between these two groups. What does the study measure for and is it relevant? The study collects data on whether or not an individual has migraine and also incidence of major cardiovascular disease and mortality events (stroke, myocardial infarction, etc). The data collected are definitely relevant because they are exactly the events of concern to public health, and changes in incidence of these events are indicative of migraine as a risk factor. Data was collected via an initial self administered questionnaire and follow-up questionnaires every two years. The follow-up rate for questionnaires was greater than 90%, adding to the power of results collected–another strength of this study. Disease events were physician diagnosed and/or met strict criteria defined by international organizations. Regarding cardiovascular mortality events, data was collected using clearly defined and systematic processes 98% of the time. These methods included collecting data from autopsy reports, medical records and the national death index. These are also strengths as it provides a more complete set of data collected. Inherent in the nature of a cohort study is that people often times begin at a different stage in their lifetime and there are also many differences in lifestyle choices and these things could alter the measurement of the independent variable (migraine) and affect the dependent variable (incidence of cardiovascular disease). Understanding the limitation of cohort studies, the study collects information (e.g. age, elevated cholesterol, diabetes, BMI, etc.) and will used the information to adjust for confounding variables in statistical analysis. Although data was self-reported, concerns are mitigated with the method of data collection. The reasons being 1) large sample size- increased power of the information gathered 2) The study population was all nurses (healthcare professionals) and thus provided a higher likelihood of accurate and reliable self-reporting, 3) self reported information was either assessed or verified by a physician for confirmation and 4) there was consensus between self reporting and national/international criteria for the classification of diseases (ex: ICD-9, International Headache Society criteria). Overall, the study design measures the appropriate variables of interest, makes the best effort in ruling out potential confounding variables, and accurately measures the association between migraine and CVD.
Effect size: In this study, effect size would give a measure of the magnitude, or strength, of the association between migraine and CVD. It is not instantly clear how we can quantify the effect size given the data presented in this study. Typically, effect size can be quantified using a statistical measure such as Cohen’s d which is calculated by taking the difference between two means and dividing that value by the standard deviation.
In this study, to get an idea of the effect size of migraine on the development of cardiovascular disease events, we can compare the hazard ratios derived from the Cox proportional hazards model. Although we are not well versed in statistics, we can describe the Cox proportional hazards model as modelling the number of new cases of disease per population at-risk per unit time. A hazard ratio, which is the parameter of interest derived from a Cox model, describes the relative probability of a given event in two groups over time. In this study, we are looking to compare the hazard for cardiovascular disease in the migraine group, compared to the no-migraine group. A ratio of 1 would mean that the relative risk of the two groups would be the same. The study reports multivariable adjusted hazard ratios for cardiovascular disease outcomes according to migraine status; some of the important hazard ratios included 1.37, 1.39, 1.50, 1.62 and 1.73 for cardiovascular mortality, myocardial infarction, major CVD event, stroke and angina/coronary revascularization, respectively.
In general, this means that more women experiencing migraine are having cardiovascular events proportionally to the comparison group (no migraine). But is this a small or large effect size? Previous studies have defined some parameters for interpreting the magnitude of the effect size. Tannock et al (2007) defined a small size effect as HR≥1.3, a medium size effect as HR≥1.5 and a large size effect as HR≥2.0. Therefore, the effect size for hazard ratio between migraine and various cardiovascular diseases are small to medium in size. Precision of the effect size varies as a function of the sample size and since the sample size of this cohort study is so large, we can conclude that the observed effect size is precise.
Conclusion:
The main conclusion that the authors make for this study is that there is indeed a consistent link between migraine and increased incidence of cardiovascular disease events in women, suggesting that history of migraines could serve as a useful marker for increased risk of many cardiovascular disease events. Their argument finds strength in study characteristics such as sample size, length of study and the characteristics of the sample population. Having a large sample allows for less variance and higher power, giving meaning to the observed association and the ability to classify the observed trends as real. The long length of the study period allows for the observation of diseases that may typically manifest more often with increasing age (such as myocardial infarction). We believe that studying nurses not only improves the reliability of self reporting by providing a better ability to distinguish a headache from migraine, but also controls for variables such as the access to health care. Furthermore, the high degree of agreement between self reporting and various criteria (such as physicians, or the criteria laid out by World Health Organization, National Survey of Stroke, and the International Classification of Diseases) assists to negate uncertainty about whether or not self reporting was a valid method of data collection. By studying women, this study makes up for a large gap in the field of cardiovascular disease research, since studies of this sort have been restricted, historically, by studying males only. Lastly, we believe that since the findings of this study align with those of other similar studies, there is a true association between migraine and cardiovascular disease.
There is still much need for further research if we wish to overcome the study’s limitations and extrapolate these findings to a wider demographic. For example, the study doesn’t provide information for migraine with aura, the group in which, for other studies, the association between migraine and CVD was most strongly observed. This limitation calls into question whether the association described in this study is confined only to those suffering from migraine with aura. Genetic factors, including the relative risks for different ethnic groups are not mentioned in this study, despite observations previously made in the field of cardiovascular research (Mensah et al., 2005). The study population was described as “mostly white”, and so the same association between cardiovascular disease and migraine may not be observed to the same extent in a more diverse population. Future studies should recognize these limitations and work to overcome them in order to further solidify migraine as a new risk marker for cardiovascular disease.
For now, all we can do is speculate an association between migraine and cardiovascular disease events. The health risk that migraine proposes, though small at the level of individual risk, is important to study at the level of the general population because of migraine’s high prevalence in society. If the association between migraine and cardiovascular disease is true, then the study highlights the need for more research into the causes and mechanism of migraine, since CVD is a leading cause of mortality in our population. Further research incorporating the suggestions mentioned above, as well as the potential development of animal models to model this association, may help us to elucidate the mechanism of migraine and be able to apply the association to a more diverse population.