This quality assurance project, conducted within real-life clinical practice, found that the long-term effectiveness and safety of anticoagulant treatment during PSM is comparable to conventional treatment administered by GPs. The same patients were followed from the start of conventional treatment at the GP and during PSM, and no statistically significant difference was observed in the annual risk of hospitalizations due to severe clinical complications, such as severe bleeding or thrombosis. Nonetheless, a noticeable trend towards a reduced annual risk of hospitalization due to severe clinical complications was apparent during PSM compared to conventional GP treatment. This trend, however, did not attain statistical significance, likely due to the limited number of patients included in the study. Nevertheless, patients undergoing PSM demonstrated statistically significant higher TTR and achieved greater stability in anticoagulation, as evidenced by reduced variability in INR and fewer occurrences of extreme INR values, compared to conventional treatment.

Previous reported annual risk of serious bleeding/major hemorrhages and thromboembolic events during PSM in a real-world setting from other European countries are comparable with our results. Specifically, the risk of serious bleeding/major hemorrhages during PSM in real-world settings ranges from 0.5% to 2.3% [6, 7, 9, 11,12,13, 24] whereas our study reported a risk of 1.3%. Similarly, the annual risk for thromboembolic events during PSM varies between 0.50% and 1.6% in previous studies [6, 7, 9, 11,12,13, 24], compared to 0.67% in our study (Table 2). However, it's noteworthy that only one of these previous studies directly compared the outcomes of PSM with conventional treatment [13]. In that study, no significant differences were observed in the rates of serious bleeding/major hemorrhages and thromboembolic events between conventional treatment and PSM, even after 1 or 5 years of follow-up [13]. Similarly, a meta-analysis from RCTs concluded that there was no reduction in the risk of major hemorrhages when comparing self-testing or PSM with conventional treatment [2,3,4], consistent with the findings of our study. However, it's important to note that the meta-analysis also included patient self-monitoring alongside PSM, and the participants included in RCTs represent a highly selected group. As a result, direct comparisons with our study may be limited. Nevertheless, it's noteworthy that the proportion of spontaneous bleedings observed in our study was lower during the PSM period (26%) compared to the conventional treatment period (43%), potentially indicating a clinical benefit associated with more stable INR values.

The meta-analysis revealed that the risk ratio (RR) or hazard ratio (HR) for thromboembolic events was 0.58 and 0.51, respectively, during PSM compared to conventional treatment [2,3,4], which is slightly lower than our study's findings (RR = 0.65). Reported annual mortality during PSM in clinical practice have been documented at various intervals: 0.33% (after one year of follow-up), 1.08% (after 5 years of follow-up) [13], and 2.4% (after 3.3 years of follow-up) [6], compared to 0.69% in our study. Discrepancies in reported risks for complications and mortality may be attributed to variations in patient selection criteria, sample size, indications for VKA therapy, comorbidities, patient demographics (such as age), duration of observation, quality of training, and the quality of management of VKA therapy during the conventional treatment period.

As clinical events such as severe bleeding and thrombosis are relatively uncommon and primarily associated with TTR [25], the lack of a significant reduction in the yearly risk for thrombosis during PSM compared to conventional treatment in our study is likely attributable to the relatively small sample size [25].

Comparison of TTR between conventional treatment and PSM in clinical practice has not been previously reported. However, the observed increase in TTR from conventional treatment to PSM aligns with findings from previous meta-analyses conducted in other countries [1, 3, 26]. Moreover, the TTR achieved during PSM in our study (78.6%) is consistent with values reported in previous studies conducted in real-world clinical settings, ranging from 71.5% [6] to 78.5% [11]. This is slightly higher than the TTR results reported in meta-analyses (ranging from 67.1% [25] and 74.8%) [3] and comparable to findings from a prior study conducted in Norway (TTR 78.1%) [19]. However, direct comparisons are challenging due to variations in study methodologies, as elucidated above. Nevertheless, previous research has demonstrated a correlation between TTR and clinical outcomes, as summarized by Samsa and Matchar [24]. Therefore, the significant increase in TTR from conventional treatment to PSM observed in our study suggests an improvement in the quality of anticoagulant therapy. Given that there were no differences in baseline characteristics between the “Patients that provided INR-results”-group (n = 285) and the entire cohort (“All patients”) (n = 400), it is reasonable to infer that the INR group represents the same population as the entire cohort.

The risk of thromboembolism and bleeding increase significantly at INR levels below 2.0 and above 5.0, respectively [26, 27]. Moreover, variability in INR measurements correlates with the incidence of thromboembolic events, bleeding events, and mortality [28]. Therefore, the reduction in INR values ≤ 2.0 and the decrease in INR variation observed during PSM compared with conventional treatment in our study further validate the safety and quality of PSM in our cohort. Although the percentages of extreme INR values are higher than those reported in previous studies [19, 20, 29], this is likely attributable to the real-world context, as the figures are calculated from a higher number of INR measurements and longer intervals between INR testing in remote areas.

In this study, less than 5% of the patients selected for training by GPs in PSM at Nordland Hospital in Bodø were unable to complete the training program due to various reasons, including difficulties in adhering to weekly INR control and parallel INR measurements, cognitive impairment, or suspicion of abuse. Only a small subset of patients discontinued PSM due to a significant deviation exceeding 20% in repeated parallel INR measurements. Notably, two of these patients were diagnosed with triple-positive antiphospholipid syndrome (APS), highlighting the importance of identifying APS patients, as interactions between antiphospholipid antibodies and thromboplastin used in point-of-care (POC) INR testing may influence results [30]. In summary, the selection criteria for identifying suitable patients eligible for PSM in this study have proven to be successful. Nevertheless, some support from GPs also during PSM remains crucial for long-term safety, particularly for identifying cognitive dysfunction or the need for assistance in managing interactions with other medications or advanced surgical interventions.

This study represents the first longitudinal cohort study conducted in clinical practice settings in Norway, wherein the same patient cohort was followed from the initiation of warfarin therapy under conventional follow-up by a GP (median 2.45 years, 1958 patient-years) to long-term PSM) (4.99 years, 2242 patient-years). Comparable longitudinal studies conducted in clinical practice across other European countries have reported median follow-up durations ranging from 1 [11] to 4.3 years [7], and includes between 296 [11] and 2068 patients [31].

A unique aspect of this study is the acquisition of data directly from patient journals, ensuring that all events from the initiation of warfarin therapy were captured from original documents. The patient cohort represents approximately 20% of individuals trained in PSM across Norway, exhibiting diverse indications for warfarin therapy and spanning an age range from 15 to 85 years. A potential limitation is the uniform training of patients at a single hospital; however, it is important to note that the training program adheres to standardized protocols and is implemented across ten different hospitals nationwide. Consequently, the findings of this study are representative for patients undergoing PSM training by Noklus.

One limitation of this study is that patients were not treatment-naive to warfarin at the time of the intervention, which may introduce a learning effect that could impact the results. Consequently, the comparability between the two periods (conventional treatment and PSM) may be compromised. Ideally, a RCT design would have been employed to ensure that the observed outcomes were solely attributed to the training in PSM. However, RCTs involve highly selective patient, limiting the generalizability of the findings [32], and necessitating validation through real-world studies. Nevertheless, the results of previous meta-analyses of RCTs align with our findings [2,3,4]. Another limitation is that 13 patients received warfarin treatment for less than 3 months prior to the initiation of PSM training, potentially increasing their risk of adverse events during the early stages of warfarin therapy. However, as no severe clinical complications were observed in these patients during this period, this did not affect our results. Additionally, information regarding the pharmacogenetic genotypes of CYP2C9 and VKORC1 related to warfarin dosage and efficacy was not included, as these analyses are not routinely performed in Norway. Morever, the lack of adjustment for the increased risk of mortality associated with advancing age. As this study primarily serves as a quality assurance study, its primary objective is to ascertain the safety equivalence of PSM compared to conventional treatment. Furthermore, it is important to note the decline in the number of patients receiving warfarin treatment, from approximately 2% in 2012 to 0.6% in 2020 following the introduction of DOACs [33]. Consequently, GPs may possess reduced proficiency in managing these patients, potentially resulting in inferior quality conventional treatment compared to the outcomes observed in this study.

Previous estimates and economic models suggest that PSM becomes more cost-effective compared to conventional treatment after approximately ten years [5, 25]. Moreover, for our cohort, geographical conditions such as long travel distances to the GP office and limited accessibility may restrict the frequency of INR measurements during conventional treatment. Consequently, periods characterized by INR deviations from the therapeutic range may go undetected with conventional treatment, thereby increasing the risk of bleeding or thrombosis.