Keywords
INTRODUCTION
The way study results are presented can influence the perception of the value of the study and, as a consequence, clinical decisions. The interpretation of numerical data can vary depending on the way biomedical research results are presented. Random distribution clinical trials generally use relative risk reduction (RRR) to describe the effect of a particular treatment, a measurement that is independent of baseline patient risk and, therefore, is easy to extrapolate. Nevertheless, it is also clear that this measurement augments the effect of an intervention, particularly when there is a low incidence of episodes. For example, an RRR of 25% can be mean absolute episode rates of 30%, 15%, 5%, or 2%. The RRR clearly quantifies the effects of an intervention in a proportional manner, but does not quantify the magnitude of the effect on an absolute scale. An alternative method for presenting clinical data is absolute risk reduction (ARR), which establishes the underlying susceptibility of the patients and provides more complete information than the RRR. Nevertheless, contrary to the RRR, the ARR is difficult to document and incorporate in clinical practice since, if not converted, it tends to be expressed in decimal fractions. The reciprocal corresponding measurement, number of people who need to be treated to prevent an event (NNT), has the same advantages but is more useful than the ARR because it indicates to clinicians, in a definitive way, what the necessary inversion is to avoid an episode and also what the probability is that an effect will be clinically significant.1 As opposed to the RRR, neither the ARR nor the NNT can extrapolate which patients have specific baseline risks. Different studies in the fields of primary care and internal medicine2-5 have shown that the opinions of clinicians about pharmacological treatment are influenced by the predominant use of the RRR.
Our study was performed with the help of Spanish cardiologists who were randomly assigned 3 questionnaires with different measurements to assess the effect of an intervention (RRR, ARR or NNT) and to evaluate its influence on their perceptions of the degree of efficacy of the pharmacological treatment in the prevention of heart disease, as well as the probability of initiating said treatment in simulated patient cases. As a secondary objective, we evaluated the theoretical total and LDL cholesterol that they believed were recommended for patients with myocardial infarction, as well as the percentage of patients they thought could attain these levels.
METHOD
All cardiologists who were members of the Sociedad Española de Cardiología (Spanish Society of Cardiology) were included in the study. Pediatric cardiologists and cardiovascular surgeons were excluded since they do not normally participate in the preventative treatment of heart disease. The data utilized in the 3 questionnaires dealing with treatment effects was taken from the WOSCOPS study,6 Physicians´ Health Study,8 a meta-analysis of beta-blockers after acute myocardial infarction,9 and SAVE,10 although the questionnaires did not reveal from which studies the data had been obtained. We created 3 types of questionnaires (Appendix). Questionnaire A included the effects of treatment as reported by RRR; questionnaire B as reported by ARR, and questionnaire C as reported by NNT. The confidence intervals for the values considered were obtained from the original study or calculated by means of standard methods for categorical variables.11 A summary of the results from each of the 5 studies was given (2 for primary prevention and 3 for secondary prevention), and we evaluated the perception of the pharmacological treatment´s efficacy by a Likert scale with values from 0 (treatment that had no effect) to 10 points (treatment with the most beneficial effect) on which the respondent marked with an X the value they considered most appropriate. We also included simulated clinical cases based on each study, from which the cardiologists had to indicate the probability of using a specific medication on a Likert scale with values from 0 (very low probability of using the medication) to 10 points (very high probability of using the medication).
In addition, the 3 questionnaires all contained a question regarding the recommended levels of total and LDL cholesterol for patients with a history of acute myocardial infarction and the percentage of patients they though could achieve said cholesterol levels.
The cardiologists were chosen randomly from a table of random numbers to receive 1 of the 3 questionnaires. We did 3 mailings at 14-week intervals between May, 1998, and January, 1999, to 1408 clinical cardiologists.
Statistical analysis was performed with the SPSS-PC12 information program. The characteristics of the 3 random groups of cardiologists were compared by the χ² test and variance analysis. We used the Kruskal-Wallis test to compare the scores obtained according to the method used to present the results. We defined a statistically significant level as P≤.05 bilaterally.
RESULTS
Six hundred and fifty-eight questionnaires (46%) were returned; in these questionnaires, 99 replied that the questionnaires were not appropriate for their practice, leaving 559 questionnaires (40%) for analysis.
Characteristics of the respondents
Table 1 shows the characteristics of the 3 groups that were surveyed. There were no differences between groups with respect to age, sex, type of hospital, amount of time taken to answer the questionnaire, or percentage of time devoted to clinical practice.
Comparison of the scores from the 3 groups
Tables 2 and 3 show the comparison of the mean scores, on a scale from 0 to 10 points, of the 3 groups with respect to the efficacy level perceived and the probability of beginning treatment as a function of RRR, ARR, and NNT. Calculation of the mean efficacy level was assessed from the answers to the questionnaire as a function of the presentation method for the study results in terms of RRR, ARR, and NNT. We observed an absolute difference of 0.41 (RRR and AAR) and 1.36 (RRR and NNT), for primary prevention with statins; an absolute difference of 1.78 (RRR and AAR) and 2.59 (RRR and NNT) for primary prevention with aspirin; an absolute difference of 0.70 (RRR and AAR) and 0.62 (RRR and NNT) for secondary prevention with angiotensive converting enzyme inhibitors (ACEI); we observed practically no difference in the results for secondary prevention with statins or secondary prevention with beta-blockers. There was no difference between the groups with respect to the probability of initiating pharmacological treatment except in the case of aspirin used for primary prevention of an acute myocardial infarction--scores were higher when the results were presented in terms of RRR.
On the topic of the recommended total and LDL cholesterol levels for patients with a history of myocardial infarction, 90% of the cardiologists responded with a total cholesterol level of 200 mg/dL or lower, and 99% answered 130 mg/dL or lower for LDL cholesterol. On the topic of the percentage of patients the physicians believed could achieve said levels, 34% of the cardiologists thought that 80% to 100% of their patients could reach these levels, 55% thought 40% to 60% of their patients could achieve the levels, and 23% thought that only 20% or fewer of their patients could achieve these levels.
DISCUSSION
In our study, we observed how the method of presenting the results of a study can result in different perceptions of the efficacy of the intervention studied, a finding that concurs with the findings of previous studies.1-4 Nevertheless, since we used various clinical cases, our findings are more reliable than previous studies from a methodological point of view. We confirmed that using RRR can augment the perception of the efficacy of an intervention, resulting in overestimation of the results.
As opposed to other studies, we did not find differences in the probability of initiating treatment except in the case of aspirin for the primary prevention of myocardial infarction. We believe that the decision to initiate treatment extends beyond the perception of efficacy, and can be influenced by other factors. A possible explanation may be that our study population was made up of clinical cardiologists, and therefore was a more homogenous population than that studied in previous investigations, in which general practioners or internists were selected. It is also of note that the clinical studies used in our study to present the clinical cases are very well known, and have been widely disseminated in recent years among the cardiology community by means of articles in professional journals, medical meetings and symposia, and (as far as the medications in question are concerned) by the pharmaceutical companies. This fact may be the reason we rarely encountered a relationship between the way the study results were presented and the probability of initiating pharmacological treatment.
Another interesting finding of our study is that, overall, there is a slight tendency to use aspirin for primary prevention, and beta-blockers after myocardial infarction, in simulated clinical cases. The first finding may be due to the fear of adverse effects from the use of aspirin, especially with respect to hemorrhagic problems. The Physician´s Health study certainly did not convince the medical community of the efficacy of this intervention, particularly given the British study results13 did not confirm their findings. As a result, it is understandable that there is a low incidence of the use of aspirin in the case of an asymptomatic male patient without risk factors. The second finding is consistent with the results from other studies that have shown that beta-blockers are clearly underused.14,15 This finding may partially reflect the relatively conservative attitude in Spain with respect to the use of beta-blockers for this condition, which has been evident since publication of the results of the EUROASPIRE study, which in 1995 and 1996 examined secondary prevention of heart disease in 9 European countries, including Spain.16 In that study, only 35% of Spanish heart patients used beta-blockers compared with 54% of patients in a combined population of 9 countries. Nevertheless, these results have improved in accordance with the results from the recently published EUROASPIRE II study,17 in which 47% of Spanish heart patients and 63% of the total combined population used beta-blockers. The fact that 50% of clinical cardiologists think that only 50% of their patients could achieve cholesterol values recommended for secondary prevention, including when said levels are reasonably well known, reflects the opinion of the cardiologists with regard to the difficulty in implementing medical guidelines in clinical practice. Among the reasons for this point of view are a number of factors such as professional experience, noncompliance with treatment by patients, and the cost of treatment. Other authors18 consider the lack of communication among specialists and primary care physicians to be an additional barrier. In a study that compared the differences in treatment continuity rates with hypolipemic medications among patients included in clinical trials and patients not included in clinical trials in 2 primary care centers,19 substantially lower rates were found in actual medical practice.
Study Limitations
Our results are restricted to the clinical setttings presented in the study; the perception of the efficacy of an intervention and the likeliness to treat may have varied considerably if different clinical trials or different settings based on the same clinical studies had been used. For this reason, we cannot extrapolate our results to other studies with different clinical trials or different simulated cases, and we cannot easily compare our results with those from other studies.
Forty percent of cardiologists responded to the questionnaires, and although this is not an unusual percentage for a study conducted by mail, it could have produced a skew in the results. However, we believe such a skew is improbable since this was not a study of the cardiologists´ knowledge but instead of their clinical perceptions; the profiles of the cardiologists who did not respond to the questionnaires should not differ in principal from those who did respond. Nevertheless, if the probability of less-informed cardiologists not responding to the questionnaires was greater than that of more well-informed cardiologists, our results could be skewed in the sense of undervaluing the differences observed. It is also probable that the cardiologists who responded were more critical with respect to the efficacy of the medications than those who did not respond, and that their point of view with regard to a particular treatment would be more to the point than those who did not respond. Unfortunately, we did not have information about the cardiologists who did not respond with regard to their age, workplace, type of practice, and other characteristics, preventing comparison with those who did respond.
In conclusion, our study shows that there are differences in the perception of efficacy according to the method of measuring effect, although the decision to initiate pharmacological treatment for those in the simulated cases was independent of the method of presentation of the results from the clinical trials in the questionnaires. The use of ARR and NNT methods, rather than RRR methods, is advisable to describe the expected benefits of an intervention in individual patients for therapeutic recommendations that appear in scientific publications. Nevertheless, these variables should be only one element considered when a treatment decision is to be made; the physician´s own experience and clinical judgment, and the quality of life and the patient's preferences should also be taken into account.
ACKNOWLEDGEMENT
We are grateful to all the cardiologists who took the time to fill out the questionnaires.
Financied by the Fondo de Investigación Sanitaria, Ministerio de Salud (97/1226).
Correspondence: Dr. C. Brotons.
Servicio de Cardiología. Unidad de Epidemiología.
Hospital Universitari Vall d'Hebron.
Pg. Vall d'Hebron, 119-129. 08035 Barcelona. España.
E-mail: cbrotons@eapsardenya.net