Author + information
- Jan A. Staessen, MD, PhD∗ ( )(, )
- Yu Jin, MD, PhD,
- Lutgarde Thijs, MSc,
- Alexandre Persu, PhD,
- Michel Azizi, MD, PhD and
- Sverre Kjeldsen, MD
- ↵∗Studies Coordinating Centre, Research Unit Hypertension and Cardiovascular Epidemiology, KU Leuven Department of Cardiovascular Sciences, University of Leuven, Campus Sint Rafaël, Kapucijnenvoer 35, Block D, Box 7001, BE-3000 Leuven, Belgium
Pokushalov et al. (1) reported the first-in-man randomized clinical trial (NCT01117025) comparing the effects of pulmonary vein isolation with (intervention) or without (control) renal denervation in patients with refractory atrial fibrillation and treatment-resistant hypertension. We were struck by several inconsistencies in this trial.
First, of particular concern are the changes in design, primary endpoint, sample size, and inclusion and exclusion criteria from the successive protocols posted at http://www.clinicaltrials.gov (Online Appendix) to the published report (1). The original protocol identified office systolic blood pressure at 2 years—not recurrence of atrial fibrillation at 1 year—as the primary endpoint that should have informed the sample size calculations. The number of patients actually randomized was 27 (1), 1 more than required according to last version of the protocol at the registry Website, and 82% less than first planned for (N = 150). The discrepancy is especially worrisome because the p value for the between-group difference in the incidence of atrial arrhythmia was only 0.033 (1), so that adding a single control patient (14 instead of 13) might have made the statistical significance. The results of atrial ablation in the control group were dismal (2). Previous use of amiodarone was an exclusion criterion in the published paper (1), but not in any version of the design in the trial registry. The original exclusion criterion of secondary atrial hypertension was changed to secondary hypertension without mention of the diagnostic procedures. At variance with the CONSORT statement (3), the number of patients screened and invited remains unreported (1).
Second, the Pokushalov et al. (1) trial was registered as investigator driven without industry sponsor. Researchers were originally located in Novosibirsk and Athens. An employee from industry coauthored the published report (1). The group, who first reported on the blanking period (4) and also received industry support (1), was also added at the publication stage. Pokushalov et al. excluded the first 3 months after pulmonary vein isolation from the analysis of recurrent atrial fibrillation (1). This so-called blanking period (4) should not have been longer than 2 to 4 weeks because recurrence of atrial fibrillation significantly decreases after 1 month and thereafter remains stable (5). Moreover, all patients of Pokushalov et al. (1) were treated with propafenone or flecainide for 6 weeks after the procedure.
Third, the published methods (1) are not detailed enough to allow an independent replication of the study. The catheter used to stimulate the renal nerves is not mentioned. An exhaustive Internet search for the Stimulator B-53 led us to http://www.biotok.ru. This company is currently located in Tomsk, not in St. Petersburg. We could not access their Website because of an HTTP 500 Internal Server Error. Furthermore, the Pokushalov et al. paper does not provide any information on the technique or reproducibility of the echocardiographic measurements (1).
Fourth, there are possible differences between described and applied statistical methods. The authors stated that continuous variables were presented as mean ± SD and analyzed by a Student t test (1). Systolic/diastolic blood pressure averaged 178 ± 8 mm Hg/96 ± 4 mm Hg in 14 control patients and 181 ± 7 mm Hg/97 ± 6 mm Hg in 13 patients of the intervention group. For the blood pressure level, SDs around 8 mm Hg systolic and 5 mm Hg diastolic seem unrealistic unless patients were selected within narrow blood pressure limits. Reportedly, in the interventional group, the decrease in blood pressure at 12 months averaged −25 ± 5 mm Hg systolic and −10 ± 2 mm Hg diastolic (1), whereas the change read from Figure 5 (1) was −25 ± 3 mm Hg systolic. Recalculation of the p values, using an unpaired Student t test and the blood pressure changes digitized from Figure 5, showed that the measure of spread for both blood pressure and blood pressure decrease was probably not SD, but more likely SE. For comparison, in the SYMPLICITY HTN-2 study (6), office systolic/diastolic blood pressure averaged 178 ± 16 mm Hg/98 ± 17 mm Hg in the control group (n = 52) and 178 ± 18 mm Hg/97 ± 16 mm Hg in the renal denervation group (n = 52), and the blood pressure changes at 6 months (± SD) were 1 ± 21 mm Hg/0 ± 10 mm Hg in the control group and −32 ± 23 mm Hg/12 ± 11 mm Hg in the denervation group. SD is not dependent on sample size, whereas SE is.
Fifth, Pokushalov et al. (1) did not report key test statistics and rounded measures of central tendency and spread to a single meaningful digit. In a parallel-group trial, p values for the within-group changes in an outcome variable provide some information, but the main test statistics should rest on baseline-adjusted between-group differences in the endpoint. Pokushalov et al. did not show these key statistical parameters (1). Furthermore, the echocardiographic results were not consistently reported throughout paper. Table 1 shows baseline values only for left atrial diameter and left ventricular ejection fraction (1). Next, Table 2 gives changes from baseline in thickness of the interventricular septum, posterior wall, left ventricular internal diameter, and left ventricular mass index, for which baseline values were not given (1). Moreover, echocardiographic measurements were expressed in unusual units: centimeters instead of millimeters for the interventricular septum, the posterior wall, and the left ventricular internal diameter, and g/m, not g/m2 for left ventricular mass index. This issue is not trivial, as the changes in left ventricular walls and internal diameter were small with average within-group changes ranging from 0.1 mm to 1 mm. Rounding measures of central tendency and spread to a single meaningful decimal and reporting left ventricular echocardiographic measurements in centimeters rather than millimeters makes it difficult to reproduce p values from the published data.
Sixth, 1 other potential inaccuracy pertains to the number of drugs taken. According to Table 1 (1), the number of antiarrhythmic drugs averaged 3.6 (range 2 to 5) in control patients and 3.8 (range 2 to 5) in the intervention group. The number of antihypertensive drugs was the same, albeit with different ranges as reported in the Results: 3.6 (range 3 to 5) and 3.8 (range 3 to 5), respectively. The use of 5 antiarrhythmic drugs in a single patient is incompatible with current guidelines (7).
According to Ioannidis' seminal essay (8), several corollaries for a possible false result might apply to the Pokushalov et al. study (1). The likelihood that research findings are true decreases with a smaller sample size (corollary 1); greater flexibility in designs, definitions, outcomes, and analytical methods (corollary 4); greater financial or other interests and preconceived ideas in a scientific field (corollary 5); and hotter research topics (corollary 6). In conclusion, the methodological drawbacks in the report of Pokushalov et al. (1) cast doubt on its validity, which clearly does not comply with CONSORT standards (3). Our observations highlight the necessity to have the design, methods, statistics, and conclusions of the Pokushalov et al. report revised (1) and to continue remaining vigilant about the peer-review process.
Please note: Dr. Kjeldsen received honoraria for lecturing from AstraZeneca; Bayer; Medtronic; Merck, Sharp & Dohme; Takeda and consulting honoraria from Bayer, Medtronic, Takeda, and Serodus; and grant support from AstraZeneca. Dr. Azizi was an investigator in the Symplicity HTN-2 and the Reduce-HTN trials and received consulting honoraria from Boston Scientific, Vessix, and Cordis. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- American College of Cardiology Foundation
- Pokushalov E.,
- Romanov A.,
- Corbucci G.,
- et al.
- Verdino R.J.
- Camm A.J.,
- Lip G.Y.H.,
- De Caterina R.,
- et al.