The aim of the study was to evaluate the effect of INR monitoring on BP control in primary care patients with hypertension. The results suggest that the hypothesized effect exists and that it is clinically relevant. After adjustment for confounders, both systolic and diastolic BP were significantly lower in the VKA group by about 9 mm Hg and 2 mm Hg, respectively.

These findings lead to the question, ‘Which factor associated with INR monitoring could cause the effect?’ One explanation is that INR monitoring requires regular consultations with the GP, where INR values are available right within the same consultation due to a point-of-care device. Indeed, comparison showed that patients under VKA treatment had significantly more consultations per year with their GP than controls. However, in the multivariate analysis, we controlled for the number of consultations and the effect sustained. Similarly, we controlled for the number of chronic conditions, because one could argue that patients under VKA treatment are more like to suffer from multiple chronic conditions which require regular monitoring and increase GPs’ awareness for BP control. A very common chronic disease that could unfold such a confounding effect is diabetes. In fact, the prevalence of patients with diabetes was approximately 15 % in our study population and therefore in the expected range considering age and morbidity [14]. Moreover, previous studies have shown that the majority of patients with diabetes in Swiss primary care undergo regular BP measurement [15]. Apparently, GPs are aware of the necessity of BP control in patients with diabetes. So, we analyzed patients with comorbid diabetes separately. Again, a lower BP under VKA treatment was observed.

Another explanation is that VKA agents have an inherent BP lowering effect. For instance, animal experiments revealed that specific coumarin agents or compounds have a dose-dependent, BP lowering effect [16, 17]. However, Phenprocoumon is not among these. Moreover, if Phenprocoumon had a physical effect on BP, normotensive patients under VKA treatment would regularly suffer from hypotension. To conclude, we consider this explanation as rather implausible.

Another explanation for our findings is that INR monitoring promotes patients’ adherence to medical therapy and subsequently enhances the intake of antihypertensive medication more than other monitoring measures. If so, it would be interesting to investigate whether the effect could also be observed in patients who routinely receive INR measurement by different healthcare providers or who perform self-monitoring at home.

Our study has several limitations. Due to the design of our study, we can only speculate about causality between VKA treatment and better blood pressure control. We also did not measure compliance to VKA treatment directly because our dataset did not contain INR-values or time in therapeutic range. However, also after adjusting for several possible confounding factors, we still observed a strong association between VKA and lower blood pressure. However, there remains the possibility of residual confounding due to missing data on e.g. BMI and limitations due to undercoding for reasons for anticoagulation.

A major strength of our study is that it is based on the substantial FIRE database and consequently a large sample size. As in February 2013, the database contained records from 56,756 patients and led to the inclusion of 4,412 patients in this study. Moreover, our inclusion criteria were rather strict, defining hypertension as antihypertensive medication intake for at least six months. As newly induced treatments can cause great variations in blood pressure control, we ensured that patients at the beginning of antihypertensive treatment were excluded. Patients who had received VKA treatment for less than three months were also excluded, in order to avoid the inclusion of patients under short-term anticoagulation therapy after thrombosis. Therefore, data can be considered as representative.

A significant association between VKA treatment and BP control could be shown. Although, our study method is limited to a cross-sectional design and not qualified to prove causality, our findings are of special relevance for daily practice in primary care. In the light of an increasingly older and multimorbid population, new models of care are required that are tailored to the needs of chronically ill patients. Indeed, health services research has yielded several promising concepts, e.g. the Chronic Care Model or the Patient-Centered Medical Home Model [18, 19]. However, the implementation of these concepts into primary care routine of European countries is challenging and still in the early stage of its development [20]. Meanwhile, regularly recurring processes of care such as INR monitoring in GP practices fill in the gap and constitute an important opportunity for GPs to structure care provision, evolve a stable physician-patient-relationship and provide comprehensive medical care. Our findings show the importance of this «gap filling» and highlight that VKA should not be simply replaced by NOACs without a thorough implementation of new care models.

Further research is needed, in order to clarify which aspect of INR monitoring is causative for the observed BP lowering effect and how it could be systematically exploited, particularly in multimorbid patients.