We greatly appreciate the interest that Dr Kittelson and Dr Bonagura took in our paper entitled “Efficacy of Spironolactone on Survival in Dogs with Naturally Occurring Mitral Regurgitation Caused by Myxomatous Mitral Valve Disease.” We think that the issues raised are important and we thank both authors for the opportunity to provide more information about the study than what was possible in the original paper.
The study presented in the paper was part of the dossier submitted to the European Medicine Agency (EMA) for marketing approval of Prilactone ®(Spironolactone) within the EU. Because this study was a licensing study, the initial intent was not to publish it. However, a decision was made to make the results available to the public in order for everyone to have access to more information than the drug label and the possibility to review the data and form their own opinion. We hope that the decision to publish this licensing study will encourage drug companies in the future to do the same and that this may encourage an “open environment” between drug companies and the veterinary community.
Licensing studies often differ considerably from postlicensing studies because they must meet the regulatory authorities' guidance and demands regarding design, conduct, statistical analyses, reporting, and quality assurance. The overall guidance, the Good Clinical Practices (GCP),1,2 means that studies must be well organized by experienced staff (R&D departments of pharmaceutical companies or contract research organizations) relying on Standard Operating Procedures to assure the appropriate level of quality and data reliability. Licensing studies must provide many pieces of data and documentation, and this influences study design. Postlicensing studies often differ from this because they are usually designed to answer only one specific question, with no influence of regulatory authority on study design and not always conducted according to GCP standards. We agree with Drs Kittleson and Bonagura that the optimal way to plan and conduct a clinical trial is to engage a group of independent veterinary specialists to produce the trial protocol and then carry out the trial. However, because of the above-mentioned processes of licensing a drug, this is not as simple as compared with a postlicensing trial. EMA strongly recommends involving veterinary specialists when designing the studies to legitimize the medical content. For the present study, several veterinary specialists were consulted and they were of different opinions concerning study design and inclusion criteria. Finally, the company chose the present design of 2 short-term lead-in studies (designed to answer questions other than survival) and a long-term follow-up study to meet EMA's demands of documentation. It is recommended to conduct efficacy field trials under conditions close to those of the future marketed drug, which include both veterinary specialists and nonspecialists. Therefore, both centers with veterinary specialists and nonspecialists were engaged in the present study.
We certainly acknowledge that the present study had flaws, well described by Dr Kittleson and Dr Bonagura. We believe that some of these flaws were addressed in the paper. The most important ones were difficulties associated with the different inclusion criterion for the lead-in studies (furosemide allowed in one but not in the other), and with the lead-in studies into the follow-up study. This design led to a large number of dogs not continuing from the lead-in studies to the 12-month study because of the owners' wishes. Indeed, 56 out of the 94 cases withdrawn for other than cardiac reasons withdrew at the end of the 2 lead-in studies, but the numbers were well balanced between the treatment groups. In Europe, money incentives for owners to participate in clinical trials are forbidden. This may influence owners' motivation to engage themselves in a long-term study, which may have contributed to the high withdrawal rate.
Although comparisons of survival among different clinical trials must be done with caution, we certainly agree that the dogs of the present study did not have as advanced disease as did dogs included in previously published clinical trials, as indicated by number of events and appearance of survival curves. The inclusion criteria of the 2 lead-in studies were presumably one important reason for this. For instance, cardiogenic pulmonary edema was not mandatory at inclusion. In 2002–2003, when the study was designed, there was not a consensus (at least in Europe) on classification of heart failure and on the way to assess the efficacy of a cardiac drug, as indicated by the different inclusion criteria and endpoint definitions used in published studies conducted during that time. Heart failure classifications (even in humans) focus more on clinical symptoms, such as exercise intolerance and dyspnea, than on edema. For example, in the ISACHC classification, pulmonary edema is not mentioned. However, we agree that the presence of radiological signs of pulmonary edema at inclusion may have generated a more homogeneous study population in terms of stage of disease. Another reason could be the interpretation of thoracic radiographs at inclusion, which has previously been shown to be a source of variation within and among observers. We certainly agree that this variation could have been limited by having all of the thoracic radiographs evaluated by a panel of specialists, or one specialist as in the SVEP study.3 However, because of the size of the present study and the absence of digital radiography, this was not considered practical at the time of the present study.
Clinical trials are always at risk for systematic errors that may potentially affect the outcome, and this study is no exception. We can only try to limit the risk that these errors will occur with two important tools – randomization and blinding – and both were applied in the present study (this study was designed as randomized and double-blinded). During the analysis process, we can control for these errors with appropriate statistical analyses. We consider the statistical analysis of the present study one of its strengths. First, the analysis verified that the randomization was effective by showing that the treatment groups were homogeneous at inclusion. Because the study started with 2 lead-in studies, baseline data were compared between treatment groups (treatment effect) and between lead-in studies (study effect); furosemide was allowed at inclusion in 1 study, but not in the other, which caused significant differences between studies (dogs from 1 study presented with different characteristics than those in the other study), but these did not produce any differences between treatment groups, because of the randomization. The results were further analyzed by investigating the influence of covariates such as which of the lead-in studies the dog entered, the duration of cardiac treatment before inclusion, and the prescription of furosemide. Only furosemide prescription was associated with a worse survival probability. However, regardless of whether the dogs were treated with furosemide or not, dogs receiving the spironolactone had a better outcome (survival probability) compared with those not receiving it. Because dogs were evenly distributed between treatment groups with respect to the proportion of dogs on furosemide treatment, and because the proportions of dogs in each heart failure class were similar over the treatment groups, these factors did not appear to introduce any bias to the results. Because no obvious systematic errors were identified, it was concluded that the difference observed between treatment groups was caused by a treatment effect in favor of spironolactone.
We would like to end this response by pointing out that this study was only 1 piece of extensive documentation submitted to the EMA to support market authorization. Some other pieces of this documentation have already been published.4,5 Furthermore, we would like to point out that this and the other published studies of spironolactone should be regarded as the first in a line of future postlicensing studies. In fact, postlicensing studies conducted more in the manner pointed out by Drs Kittleson and Bonagura are currently either ongoing or in the planning phase.