Author + information
We appreciate the comments regarding our article (1). Dr. Takkenberg and colleagues have calculated mortality rates of the thromboembolic and bleeding complications and found no differences between high-intensity and low-intensity therapy with vitamin K antagonists. Unfortunately, only approximately 50% of the included studies reported on mortality rates. As a result, the mortality event rate is too low to draw any statistically confident conclusion. Therefore, it remains uncertain whether the results of this small number of studies would be representative for the total mortality rate of all studies analyzed in our meta-analysis.
Dr. O'Kane raised an important point about the target international normalized ratio (INR). We agree that it is more sound to evaluate the achieved INR rather than the target INR. However, as already mentioned in our article, most reports used for our analysis were based on an intention-to-treat INR range. Furthermore, Dr. O'Kane recommends an INR between 2.5 and 3.0 based on a single-center observational study, which unfortunately lacks information on the time spent in the target range and the achieved INR. In addition, a part of the study population in this study received dipyramidole in combination with vitamin K antagonists, which increased the risk of bleeding complications (2).
Drs. Butchart and Gohlke-Bärwolf are confused when they state that meta-analyses can be performed only on randomized controlled trials. It should be clarified that meta-analysis is a statistical method defined as the quantitative analysis of two or more independent studies to integrate the findings. Studies used for meta-analysis can vary from randomized trials, non-randomized trials, or observational studies, and even from more than one of these types of studies (3).
Furthermore, they postulate that the reported thromboembolic rates are influenced by definitions, data collection methods, and patient characteristics. In our study, all events were analyzed according to the guidelines for reporting morbidity and mortality after cardiac valvular operations of Edmunds et al. (4), which minimize the potential for bias. Mean age at valve implantation and gender did not differ between the groups. Other characteristics were not specified and, therefore, it is unknown whether these factors influence the outcome.
We agree that retrospective conversion of prothrombin time ratios to an INR has the potential to introduce errors. However, the conversion of the results of the thrombotest and prothrombin time ratios to an INR was required in only 7 of 35 studies. Furthermore, for all 7 of the studies, the ISI value of the thromboplastin reagent used was known, as indicated in our article.
Surprisingly, Drs. Butchart and Gohlke-Bärwolf state that we did not acknowledge five randomized trials comparing different intensities of anticoagulation. However, three of these five studies were excluded according to our predefined exclusion criteria (5–7) and are listed on the web site (www.cardiosource.com/jacc.html). The fourth study did not fulfill the inclusion criteria, because the study included patientswith tissue valves (8) and the fifth study has been presented only in part in a supplement journal and was therefore not retrieved by the Pubmed search (9). Drs. Butchart and Gohlke-Bärwolf conclude that four of these randomized trials showed a higher incidence of bleeding with higher intensity of anticoagulation. We disagree with this conclusion because a thorough analysis of these studies reveals that only one of the randomized trials (in patients with mechanicalvalves) comparing low-intensity versus high-intensity vitamin K antagonist therapy showed a significant increased risk of bleeding in the high-intensity group (7). It should be emphasized that the patients in this study received dipyramidole and aspirin in addition to vitamin K antagonists.
In conclusion, our analysis, including a total of 23,145 patients followed for more than 100,000 patient-years, shows that the total number of tromboembolic and bleeding eventsis lowest with high-intensity vitamin K antagonist therapy. The absolute incidences of thromboembolism and bleeding are low, and therefore it would be impossible to perform a large prospective trial comparing different intensities of vitamin K antagonist therapy. Therefore, combining the results of several studies through the techniques of meta-analysis can provide stronger evidence for or against a treatment effect than one can derive from any of the individual studies.
- American College of Cardiology Foundation
- Vink R.,
- Kraaijenhagen R.A.,
- Hutten B.A.,
- et al.
- ↵Medical Use of Statistics. In: Bailer III J, Mosteller F, editors. 2nd edition. Boston, MA: NEJM Books, 1992.
- Edmunds L.H.Jr,
- Clark R.E.,
- Cohn L.H.,
- et al.
- Acar J.,
- Iung B.,
- Boissel J.P.,
- et al.
- Huth C.,
- Friedl A.,
- Rost A.