Abstract
Dear Editors:
In the article by Christensen et al. 1 entitled "Home Management of Oral Anticoagulation via Telemedicine Versus Conventional Hospital-Based Treatment,” patients consuming the anticoagulant drug warfarin or phenprocoumon were studied. The reactions of patients to the drugs, according to what diseases they had such as heart valve disease, etc., were different.
In this study, patients were not classified by types of diseases for which they needed to consume the drugs. The physical mechanisms of the patients were considered equal, whereas the function of the coagulation system in patients with a history of having liver problems or specific physiology is completely different. 2 On the other hand, there were no considerations as to the mechanism and type of randomization. 3 Also, according to the type of intervention in the study, blinding could not be applied, which could affect the results. 4
According to the demographic differences among the three groups of participatants who were studied and the nonparametric tests that were used in this study, the validity of the conclusions would be low. 5,6
In this study, which had be done by the clinical trial method, the patients were randomized to one of three groups: (1) once weekly measurement and report online, (2) twice weekly measurement and report online, and (3) continued conventional treatment with international normalized ratio (INR) measurement in the lab every 4 weeks and dose adjustment by letter.
Group 1 had the best responses to the treatment. There is an important matter that was not recognized here. If the time of coagulant blood INR measurement is shorter, the control of anticoagulant drugs would be better. This means that if the conditions of measurement and evaluations were equal, like the same interval for the evaluation of Groups 1 and 3, the comparison of the groups' results would be better.
References
Response to Fayaz‐Bakhsh and Karyani
Henry Christensen, MScIT, and Ivan Brandslund, M.D., DMSc
Dear Editors:
Fayaz‐Bakhsh and Karyani comment correctly on the design of our study 1 in the two first paragraphs.
This design was chosen because the purpose of the study was to ensure that a combination of home measurement of international normalized ratio (INR) and dosing, using telemedicine, would achieve a higher time in therapeutic range (TTR) than conventional hospital‐based treatment. This could also be expected based on previous publications. However, our study is not able to answer the question as to what proportion a weekly measurement at home and telemedicine each contribute to the increase in TTR.
In the region of Southern Denmark, the hospital cost of INR measurement and management has been calculated as $300 (USD) per year and home measurement and telemedicine at $650 (USD) per year, of which the INR machine and strips contribute to $350 (USD) and the information technology setup to only $50 (USD) per year per patient. As the cost of telemedicine is relatively low, the effect of a system combining self‐testing and management by telemedicine achieving a high TTR has been judged to be a very good investment. 2
No controlled trials have at this time shown whether self‐management is as good as self‐testing and professional dosing using telemedicine, but we are planning to do this with the 272 patients now using the system.
We agree with Fayaz‐Bakhsh and Karyani's third paragraph that ideally the demographics should be totally equal, which we show in our article's Table 1 that it is not. Ideally the single diagnostic groups should be large enough to enable a trial group by group. This would, however, call for a multicenter trial, which is not possible with the very limited use of telemedicine in Denmark or indeed the world at the time of the investigation.
We also agree with Fayaz‐Bakhsh and Karyani that the nonparametric tests used overestimate the statistic significance. Parametric tests, however, cannot be used for two reasons: For the individual patient, TTR values in days vary depending on the total time in treatment. As this varied by up to 30 days this cannot be used without expressing the TTR as a percentage. To detect a difference of 7% in TTR with a standard deviation of 10%, a power of 80%, and a significance limit of 5% requires 32 patients. Furthermore, the percentage values of TTR are not Gaussian in distribution, excluding the use of a t test, and call for a Wilcoxon or Mann–Whitney test.
However, if we use a Wilcoxon rank sum test on an individual patient's values we obtain p values of 0.0504 between Groups A and C and of 0.025 between Groups B and C.
Finally, a parametric test can only be used on a ratio scale variable. Percentage is not a ratio scale, ending at 100. Nevertheless, if we do it using Welch's t test for the comparison of Groups A versus C, a p value of 0.036 is found, and for Groups B versus C a p value of 0.029 is found.
The results obtained in daily routine support the benefits of self‐testing combined with telemedicine. Now 272 patients are monitored on the system, and the 12‐month average TTR is 75.1%. This TTR includes values for the 272 patients through the startup of treatment and during the intended reduction of INR because of surgical interventions, and hence not just values during maintenance therapy.
We think that Fayaz‐Bakhsh and Karyani's last paragraph states that if the conventionally treated group was taken to a hospital and measured as frequently as the self‐testing groups, the TTR would be more comparable. We agree that the effect on TTR by self‐testing is primarily caused by the frequency of measurement. Cost of transportation of patients prohibits this solution in our country, but may be a possibility in other countries.
The introduction of the thrombin inhibitor dabigatran, which in Denmark has a cost of $1,800 (USD) per year per patient, necessitates that anticoagulation clinics using warfarin should obtain a TTR above 66% to be competitive in clinical and economic terms. 3 And, the higher the TTR is, the more cost‐effective warfarin will be. 4 We believe telemedicine and weekly measurement also prevent bleeding as a serious side effect. No randomized controlled trial has yet been done to compare dabigatran and warfarin with once weekly measurement and dosing.