Introduction

Warfarin is the most commonly prescribed oral anticoagulant in resource-limited settings, where it is used for the treatment and prophylaxis of venous thromboembolism, and for the prevention of embolic strokes in patients with atrial fibrillation (AF) and valvular heart disease. Patients receiving warfarin require ongoing monitoring of their International Normalised Ratio (INR) which needs to be kept in a defined therapeutic range determined by the indication for therapy. Under-dosing (leading to a sub-therapeutic INR) and overdosing (leading to a supra-therapeutic INR) places individuals at risk of thrombosis and bleeding, respectively. INR control in patients in sub-Saharan Africa is often poor [1–3] which may result in potentially preventable morbidity [4] and mortality [5]. The burden of HIV and tuberculosis (TB) is high in sub-Saharan Africa [6]. Patients on warfarin who are HIV positive and/or have TB require multiple concomitant medicines, which may interact with warfarin and complicate warfarin dosing and dose adjustment [7].

In high income countries, warfarin dosing is frequently guided by validated dosing algorithms, some of which include genotyping for polymorphisms important for warfarin metabolism [8]. Little work has been done on validated algorithms to guide dosing in sub-Saharan Africa. Direct oral anticoagulants (DOACs) are widely accessible in high-income settings, but higher costs limit their use in resource-limited settings. Warfarin is therefore likely to continue to be used extensively, and strategies to improve anticoagulation control on warfarin are important.

Our aim was to evaluate the quality of anticoagulation services in Uganda and South Africa and to describe the clinical and demographic characteristics of patients receiving warfarin.

Materials and methods

Results

Anticoagulation services included in this audit are described in Table 1. The number of patients receiving warfarin ranged from 30 to 800 patients per month. Specialists or medical doctors were present in all clinics except for the secondary level service in South Africa, which was staffed by nurses. All facilities had an onsite pharmacy dispensing warfarin. At the Ugandan clinics, warfarin stock outs occurred frequently.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of the five outpatient anticoagulant services in Uganda and South Africa.

https://doi.org/10.1371/journal.pone.0227458.t001

INR turnaround time varied (see Table 1). At health facilities with onsite INR testing, results were available on the same day. However, for the tertiary hospital in Uganda and the primary health centre in South Africa, which did not have onsite INR testing services, INR results were returned the next day, requiring two consecutive visits for the patient. Only one site (Uganda Heart Institute) had point of care INR testing available.

At the Ugandan clinics, patients paid approximately £4 for an INR test; occasionally this fee was subsidised where point of care testing was available. Patients at both Ugandan clinics had to purchase warfarin themselves when there were drug stock-outs at the clinic; at a cost of between £3 and £6 for a month’s supply of warfarin depending on warfarin dose. At the South African clinics, patients did not pay for INR testing or for warfarin.

Three of the included clinics used warfarin dose adjustment protocols. The dose adjustment protocol used at the Uganda Heart institute was based on the Modified Henry Ford warfarin maintenance dosing algorithm [11]. The dosing adjustment schedules used in the South African clinics were not referenced. To our knowledge these protocols had not been validated for the clinic population.

We reviewed the clinic records of 229 patients: 68 from Uganda Heart Institute, 32 from Mulago National Referral Hospital haematology clinic, 48 from Groote Schuur Hospital, 47 from Tygerberg Hospital, and 34 from Gugulethu Community Health Centre. The three most common indications for warfarin treatment were venous thrombo-embolism (49%), atrial fibrillation (32%) and valvular heart disease (including mechanical heart valves) (13%). Characteristics of included patients are summarised in Table 2. We found that 126 patients (55%) had at least one concomitant medication which could potentially interact with warfarin [10]. There were 12 patients (5.2%) on efavirenz and one patient (0.43%) on lopinavir-ritonavir.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Characteristics of patients attending five outpatient anticoagulation services in Uganda and South Africa.

https://doi.org/10.1371/journal.pone.0227458.t002

TTR was suboptimal at all included sites (see Fig 1). Median TTR overall was 41% (interquartile range 14% to 69%).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Time in therapeutic INR range (TTR) of patients attending five anticoagulation clinics in Uganda and South Africa (box represents median and interquartile range).

https://doi.org/10.1371/journal.pone.0227458.g001

Discussion

This evaluation, which included outpatient clinics managing patients on warfarin at primary, secondary and tertiary level, found that TTR was suboptimal at all included sites. This occurred despite regular INR monitoring.

Unfortunately, our findings complement other recent sub-Saharan studies in South Africa, Namibia and Botswana, all reporting poor INR control, with mean/median TTR ranging from 29% to 47% [1,12,13]. It is known that a higher TTR is associated with increased efficacy of warfarin and reduction in bleeding complications [14]. In patients with non-valvular AF receiving warfarin, there is a three-fold risk of ischaemic events with under-anticoagulation and a five-fold risk of bleeding events with over-anticoagulation, compared to being within the therapeutic range [14]. Maximising TTR is therefore crucial for attainment of good clinical outcomes, as shown by a study in the UK, where a 10% increase in time out of range was associated with a 29% increased risk of mortality, a 10% increased risk of ischemic stroke, and 12% increased risk of other thromboembolic events [15].

A number of factors may affect INR control. We found that INR testing was generally available at all sites; however, patients at the Ugandan clinics had to pay up to £4 for a laboratory INR test. In addition, frequent pharmacy stock outs of warfarin at these clinics means that patients also have to incur costs of privately purchasing warfarin. These costs may be prohibitive, particularly in the Ugandan population where the average household monthly income is £62 - £144 [16]. However, TTR was poor at South African anticoagulation services as well, despite patients not having to pay for INR testing or warfarin.

Laboratory turnaround time at two clinics was at least one day requiring a second visit for dose adjustment, also contributing to increased transport costs and missed work days. Thus, strengthening laboratory facilities and pharmaceutical supply chain and provision of near-patient INR testing may contribute to improved anticoagulation control. However, TTR was poor even at facilities with onsite INR testing and no problems of drug stock outs, suggesting that there are other factors contributing to poor TTR.

Our population had a high burden of comorbidities such as heart failure and diabetes, which may be associated with lower TTR [17], because of polypharmacy as well as physiological changes due to comorbidity [18]. In addition, high HIV prevalence may further complicate warfarin dosing due to drug-drug interactions with non-nucleoside reverse transcriptase inhibitors and protease inhibitors [19,20]. Although our sample included few TB patients, it is important to consider that rifampicin, which is used in first-line TB treatment, induces warfarin metabolism, which further complicates warfarin dosing [21].

Pharmacogenomic differences due to polymorphisms in VKORC1 and CYP2C9 genes and additional polymorphisms in other genes e.g. CALU rs339097 may affect dosing requirements in patients of African descent [22]. Patients managed using genotype guided dosing algorithms in high-income settings have been shown to have higher mean TTR and fewer episodes of excessive anticoagulation compared to usual dosing [8]. Further studies need to be conducted in our population to develop and evaluate clinical and genetic dosing algorithms relevant for low-income settings.

DOACs require less frequent monitoring and are non-inferior or even superior to warfarin in preventing stroke in non-valvular AF [23,24]. However, their access is severely limited in resource-limited settings due to cost limitations; necessitating urgent improvements in warfarin management.

This study has limitations. We only included urban sites, and findings may not be generalizable to rural settings in South Africa and Uganda. However, in Uganda, we included the clinics known to manage the majority of patients receiving warfarin anticoagulation. It is possible that there are additional challenges facing anticoagulation services in more remote and poorly resourced settings in both South Africa and Uganda, which we have not captured. We audited each included facility at a single timepoint.

In conclusion, in the South African and Ugandan clinics that we evaluated, TTR was suboptimal. Strategies to improve INR control in sub-Saharan patients taking warfarin are urgently needed as the alternative DOACs are not freely available. The findings from our work will inform further research to develop and validate a warfarin dosing algorithm, one such strategy which may improve TTR, for use in this setting.

Supporting information

Acknowledgments

Thank you to the staff of Mulago National Referral hospital, Uganda Heart Institute, Groote Schuur Hospital, Tygerberg Hospital, and Gugulethu Community Health Centre for their help and support during the data collection for this work.