Real-Life Peak and Trough Dabigatran Plasma Measurements Over Time in Hospitalized Geriatric Patients with Atrial Fibrillation

Abstract

Background

Few geriatric patients were included in studies on direct oral anticoagulants and data on dabigatran concentration and safety are needed in this population. Our objectives were to evaluate peak and trough dabigatran plasma concentrations over time in a geriatric population and to identify factors associated with dabigatran plasma concentrations and to assess the relationship with bleeding events.

Methods

Peak and trough dabigatran plasma concentration were performed 4,8,15,30,45 days after inception of dabigatran treatment in 68 consecutive patients≥75 years old hospitalized in a geriatric hospital with atrial fibrillation. Bleeding events were monitored for 1 year.

Results

Mean age was 85.8(5.1) years old and 76.5% were women. Overall, 541 dabigatran plasma measurements (270 peak, 271 trough) were performed. Mean dabigatran concentrations of the 5 sequential measurements ranged 106-146ng/mL for peak and 66-84ng/mL for trough. Renal failure was associated with high peak and trough dabigatran concentration. Inter- and intra-individual coefficients of variation were 59.5% and 44.7% for peak and 74.5% and 44.6% for trough. Participants in the lower two tertiles of dabigatran concentration at day 8 (D8) remained below the 90^th percentile (243.9ng/ml) on the next measurements. Bleeding events were associated with high trough dabigatran concentrations. Trough dabigatran concentration at D8>243.9ng/mL significantly predicted bleeding.

Conclusion

In this geriatric population, renal function and low albumin were associated with dabigatran concentrations. Despite large variability, participants in the lower two tertiles of dabigatran concentration at D8 remained below the 90^th percentile on the following measurements. D8 dabigatran trough concentration≥243.9ng/mL identified patients at risk of bleeding.

Introduction

The prevalence of atrial fibrillation (AF) increases with aging up to 15% in people over 80 years old (1). AF is associated with an increased risk of stroke particularly in the elderly population, with an annual risk of stroke over 8% per year after 75 years (2). Thus anticoagulation is essential for thromboembolic prevention in this particularly high risk population (3).

Vitamin K antagonists (VKA) significantly reduce the risk of stroke in the AF elderly patients (4) but they have significant limitations due to multiple drug and food interactions. Furthermore, VKAs are associated with an increased risk of bleeding and are the leading cause of iatrogenic side effects with numerous hospitalizations in the elderly related to bleeding episodes (5). The bleeding risk of VKA is age-dependent and contributes to a significant underuse of anticoagulation in old patient with AF (6, 7) Therefore 30 to 50% of eligible patients are not treated with VKAs due to fear of bleeding (8).

Non-vitamin K oral anticoagulants (NOAC) including direct thrombin inhibitor (dabigatran) and factor Xa inhibitors (rivaroxaban, apixaban and edoxaban) have been proposed as an alternative to VKA for the prevention of stroke and systemic embolism. NOAC have a more favorable risk-benefit profile than VKA; they have a significant lower risk of ischemic stroke, intracranial hemorrhage, and mortality and a similar risk of major bleeding than VKA despite their higher risk of gastro-intestinal bleedings (9). In analyses of patients over 75 years old who were included in clinical trials, NOAC are associated with equal or greater efficacy than VKA and do not cause more bleeding than VKA (10, 11). Concerns remain over the applicability of data for the NOAC to geriatric patients with multiple comorbidities, renal failure, and polypharmacy or compliance issues. Indeed, geriatric patients were mostly excluded from the randomized trials.

In the RELY trial, a significant treatment-by-age interaction has been reported for extra-cranial bleeding with dabigatran (12). Age influences pharmacokinetics properties of dabigatran and could be responsible of drug level variability (13, 14, 15, 16). In addition, elderly patients often receive multiple drugs including P-glycoprotein inhibitors that may increase dabigatran level. Moreover elderly patients often have multiple comorbidities such as renal insufficiency that increases the risk of accumulation (17, 18) since dabigatran, after conversion from the pro-drug dabigatran etexilate, is eliminated up to 80% through the kidneys. Pharmacokinetics studies have shown a higher concentration of dabigatran in patients with renal impairment (19, 20). Therefore there is a risk of overdosing in patients over 80 years. Hence data in real life are needed in geriatric population especially for dabigatran, since data from RE-LY have shown a relationship between plasma concentration of dabigatran and unfavorable outcomes with age as the strongest covariate (21).

The primary objective of this study was to evaluate in “real-life the peak and trough dabigatran plasma concentrations over time in a very old geriatric population, to identify the determinants of dabigatran level. The secondary objective of this study was to assess the relationship between dabigatran concentration and bleeding or embolic events during a 12-month follow-up.

Materials and methods

The study was conducted between January 2012 and January 2014 among patients hospitalized in the geriatric department of Broca University Hospital, (Paris, France). Sixty-eight consecutive patients ≥ 75 years old with non valvular atrial fibrillation not previously treated with oral anticoagulation and for whom treatment with dabigatran was started were proposed to participate in the study. Patients were hospitalized in acute or rehabilitation care unit and were hemodynamically stable. Participants who accepted blood draws 4, 8 15 days after dabigatran inception were included in the study; additional blood draws at 30, 45 days were also proposed. Oral and written informed consents were obtained from each participant or their legal representative. The study protocol was reviewed by local Ethics committee (Comité de Protection des Personnes de Paris Ile de France VI) and the study was conducted in accordance with the declaration of Helsinki.

Data collection

Demographic characteristics, health and medical history, medication intake, and cardiovascular risk factors were obtained through medical files at baseline. Hypertension was defined as blood pressure ≥ 140/90 mmHg and/or use of antihypertensive medication. Height and weight were measured in lightly dressed patients and body mass index (BMI) was calculated. Global cognitive function was assessed with the Mini Mental State Examination (MMSE) (22). Cognitive disorder was defined as an MMSE < 24. Before receiving dabigatran, all participants had a blood draw for complete blood cell count, measurement of blood creatinine, serum albumin and liver function tests. Creatinine level was also assessed at day (D) 4, 8, 15, (and eventually at D30 and D45) after dabigatran inception. Estimated Glomerular filtration rate (eGFR) was calculated using MDRD formula (eGFR-MDRD) (23) and Cockcroft-Gault formula (eGFR-CG) (24). Co-medication was noted with drugs that can interact with dabigatran, e.g., aspirin, nonsteroidal anti-inflammatory drugs, P-glycoprotein inhibitors (amiodarone and verapamil), serotonin reuptake inhibitors, proton-pump inhibitors. Co-morbidity was evaluated with Charlson comorbidity index (25).

Thromboembolic risk score CHA₂DS₂-VASC (26) and bleeding risk score HAS-BLED (27) were calculated at baseline.

Measurements of dabigatran in plasma coagulation assay

Treatment decisions and dosing were managed according to clinical practice, and based on current guidelines. All participants received dabigatran 110 mg twice a day. Blood samples were obtained at 4, 8, 15, 30 and 45 days for measurement of peak and trough concentrations of dabigatran etexilate. Trough samples were collected in the morning 10 to 14 hours after dabigatran intake the day before and before the morning dose of dabigatran and peak samples were collected in the morning 2 to 3 hours after the morning dose of dabigatran. Blood samples were drawn by direct venipuncture and collected into a standard vacuum tubes (3,5mL) containing sodium citrate (3.2%, Greiner bio-One GmbH, Austria) and centrifuged within 20 minutes at 2500 G for 20 minutes at room temperature. Plasmas were frozen at – 80°C in 3 aliquots of 1mL.

Hemoclot thrombin inhibitor (hyphen Biomed, Neuville-sur-Oise, France) test that is CE approved was used to assess the dabigatran anticoagulation activity according to manufacturer's recommendations (28). All samples were analyzed within three hours after thawing of frozen samples according to manufacturer's instructions.

Clinical survey

During follow-up period, all clinical events were recorded. Any thrombotic complications, or bleeding events, or any adverse events were gathered by phone 6 and 12 months after the dabigatran inception through the patients' general practitioners. Only hemorrhage occurring while the participant was using dabigatran or within 7 days of dabigatran withdrawal were considered related to dabigatran. Major bleeding was defined as (i) a reduction in the hemoglobin level of at least 20 g/L or (ii) transfusion of at least 2 units of red blood cells concentrates or (iii) symptomatic bleeding in a critical site (29). Non-major clinically relevant bleeding was defined as overt bleeding not meeting criteria for major bleeding but (i) requiring medical intervention or (ii) unscheduled contact (visit or telephone) with a physician or (iii) pain or (iv) impairment of daily activities (30).

Analytical sample

Sixty-eight participants who had at least 3 peak and 3 trough dabigatran measurements were included in the study. Dabigatran level was also measured at D30 in 43 participants and at D45 for 36 participants.

Statistical analysis

Because of a better steadiness of concentrations after D8 compared with D4 that had still lower dabigatran concentrations, we used D8 results as the steady state concentrations.

Peak and trough concentrations were analyzed in term of arithmetic mean, standard deviation (SD) and coefficient of variation (CV) calculated as SD / Mean × 100, geometric mean, geometric CV, median value, 10^th and 90^th percentiles and lowest and highest measures at D4, D8 D15, D30 and D45. Median peak and trough concentrations at days D4, D8, D15, D30 and D45 were plotted.

Intra- and inter-individual dabigatran concentration variability was graphically assessed using tertiles of trough dabigatran concentration at D8 (tertile thresholds 72.1 ng/mL and 165.4 ng/mL for peak and 44.9 ng/mL and 111.3 ng/mL for trough). Inter-individual trough and peak dabigatran variability was assessed with the geometric mean and CV at each measurement. Lastly inter- and intra-individual dabigatran variability was also evaluated using mixed-effects models for repeated-measures for trough and peak concentrations.

Mean peak and trough concentrations at D8 were calculated for the different values of the categorical characteristics and according to the median value for the continuous variables and comparison were made using linear regressions with log-transformed measures of peak and trough dabigatran concentrations as dependent variables (because of their non-normal distributions) and the different variables of characteristics as independent variables. For the continuous variables of characteristics, we took the variables in a continuous manner.

The relationship between log-transformed peak and trough dabigatran concentrations and characteristics of the participants was then analyzed using linear mixed-effects regression model with random intercept and slope function. In these models, independent continuous variables were standardized in order to compute the estimate for a 1-SD increase of these variables. In addition to age and sex, we included variables that were univariately associated (p ≤ 0.15) with peak and trough concentrations at D8.

Characteristics of the sample were then analyzed according to the occurrence of bleeding event and compared with Wilcoxon and Fisher's exact test for continuous and categorical variables respectively.

In the EU summary of product characteristics of dabigatran, concentration of trough dabigatran above the 90^th percentile (200ng/mL) is considered to increase the risk of bleeding (31). We therefore further analyzed the characteristics of the sample according to the 90^th percentile of trough dabigatran at D8 and drawn a Kaplan-Meier curve with calculation of log-rank statistic with the 90^th percentile of trough dabigatran at D8. We draw another Kaplan-Meier curve with calculation of log-rank statistic with the threshold concentration of 200ng/mL defined by the EU summary of product characteristics. For these analyses, participants who discontinued dabigatran for a non-hemorrhagic cause were right-censured at the time of withdrawal. Sensitivity and specificity of the 90^th percentile of trough dabigatran at D8 to predict hemorrhagic events was calculated.

Statistical analysis was performed with the R Core Team (2014). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/. In all analyses, the 2 sided α-level of 0.05 was used for significance testing.

Results

Overall, 541 dabigatran plasma measurements were available for analysis, 270 for peak and 271 for trough. Median follow-up was 11.3 months. Over the 12-month follow-up period, 12 participants died, 13 participants experienced a hemorrhagic episode and 21 participants discontinued the dabigatran treatment. All participants received dabigatran 110 mg twice a day.

General characteristics of the whole population are shown table 1. Mean age of the studied cohort was 85.8 (SD=5.1) years old, 76.5% (n=52) were women and mean MMSE was 18.2 (8.5). The main comorbidities were hypertension, malnutrition, cognitive disorders, heart failure, diabetes mellitus, renal function (eGFR-CG and eGFR_MDRD), anemia and history of stroke. Mean Charlson comorbidity index was 7.00 (2.14), mean eGFR-CG was 55.2 (18.4) mL/min and mean eGFR-MDRD was 80.6 (25.4) mL/min/1.73m².

Table 1.

General characteristics of the sample

Characteristics, M (SD)	Whole sample
	N=68
Age (years)	85.8 (5.1)
Women, % (N)	76.5 (52)
Body mass index (kg/m2)	25.7 (5.6)
Hypertension, % (N)	76.5 (52)
Falls, % (N)	11.8 (8)
Diabetes, % (N)	23.5 (16)
Cognitive disorders, % (N)	73.3 (44)
Heart failure, % (N)	41.2 (28)
Stroke, % (N)	36.8 (25)
Cancer, % (N)	8.82 (6)
Malnutrition (albumin < 35 g/L)	67.2 (45)
Mini mental state examination /30	18.2 (8.5)
Systolic blood pressure (mm Hg)	128.1 (17.7)
Diastolic blood pressure (mm Hg)	68.5 (10.9)
Anemia, % (N)	52.9 (36)
Serum Albumin (g/L)	32.0 (4.6)
Hemoglobin (g/dl)	12.0 (1.4)
Serum creatinine at baseline (μmol/L)	74.3 (23.6)
eGFR-MDRDa at base line (mL/min/1.73m2)
≥ 50	91.0 (61)
< 50	9.0 (6)
eGFR-CGb at base line (mL/min)	55.2 (18.4)
≥ 50	55.2 (37)
< 50	44.8 (30)
Co-medication, % (N)
Aspirin	2.94 (2)
Nonsteroidal anti-inflammatory drug	4.41 (3)
Proton-pump inhibitor	47.1 (32)
Serotonin reuptake inhibitor	52.9 (36)
Amiodarone	13.2 (9)
Verapamil	1.47 (1)
Charlson comorbidity index	7.00 (2.14)
CHA2DS2-VASc score	4.84 (1.39)
HAS-BLED score	1.76 (0.71)

M (SD), mean (standard deviation); % (N), percentage (number);

^aestimated with MDRD formula;

^bestimated with Cockcroft-Gault formula.

The overall means of peak and trough dabigatran plasma concentrations were 167.9 (129.8) ng/mL and 107.8 (95.3) ng/mL respectively. Peak and trough dabigatran measurement characteristics of the 5 sequential measurements are summarized table 2. Arithmetic means of dabigatran concentration ranged from 153 to 197 ng/mL for the five peak measurements and from 97 to 121 ng/mL for the five trough measurements. Dabigatran concentration geometric means were as expected lower than arithmetic means and ranged from 106 to 146 ng/mL for the five peak determinations and from 66 to 84 ng/mL for the five trough determinations. The highest dabigatran concentrations were 720 ng/mL and 594 ng/mL for peak and trough measurements respectively.

Table 2.

Peak and trough characteristics of plasma dabigatran concentrations

				Dabigatran plasma concentration
		At peak (ng/mL)					At trough (ng/mL)
	D4	D8	D15	D30	D45	D4	D8	D15	D30	D45
N	66	68	57	43	36	66	68	57	43	37
Mean	153.4	177.2	158.7	197.3	156.1	97.4	110.7	103	120.5	113.5
SD	130.1	136.4	98.0	149.0	136.2	94.3	104.9	79.3	105.5	91.9
CV%	84.9	77.0	61.8	75.5	87.2	96.8	94.7	76.9	87.5	81.0
Geometric mean	111.4	130.4	125.8	146.1	106.0	65.5	74.9	76.7	84.0	78.3
Geometric CV%	98.4	101.1	89.0	103.5	127.8	122.3	117.9	94.5	109.3	123.8
Median	115.5	127.3	141.1	150.0	131.1	59.5	68.4	71.1	77.6	89.7
P10	44.3	46.0	47.1	51.3	30.4	27.6	29.5	29.0	28.4	28.4
P90	320.4	372.9	309.1	425.6	376.9	234.7	243.9	209.9	292.8	241.5
Minimum	17.3	12.4	8.77	8.77	9.97	1.60	3.10	12.9	11.5	8.11
Maximum	720.0	641.6	442.4	648.1	605.8	419.3	593.9	375.3	422.6	347.4

SD, standard deviation; CV%, coefficient of variation; P10, 10^th percentile; P90, 90^th percentile.

Figure 1 showed that participants in the highest tertile of trough dabigatran concentrations at D8 tended to have higher trough dabigatran plasma concentrations in all the subsequent measurements. Similarly, participants in the lowest tertile at D8 tended to have lower trough dabigatran plasma concentrations in all the subsequent measurements. Participants in the lower two tertiles of dabigatran concentration at D8 remained below the 90th percentile on the following measurements.

Figure 1.

A, Box plot and whiskers peak and trough dabigatran concentrations at D4, 8, 15, 30 and 45. B, Box plot and whiskers trough dabigatran concentrations at D4, 8, 15, 30 and 45 according to tertiles of trough dabigatran concentrations at D8 (tertile thresholds 44.9 ng/mL and 111.3 ng/mL)

Inter-individual variability, expressed as the geometric CV of the measurements at day 4, 8, 15, 30, 45, ranged from 89.0% to 127.8% for peak and from 94.5% to 123.8% for trough dabigatran plasma concentrations.

Using a mixed model, the inter-individual coefficients of variation were estimated to be 59.5% and 74.5% for peak and trough dabigatran concentration respectively and were significantly larger than the intra-individual coefficient of variation that were estimated to be 44.7% and 44.6% for peak and trough dabigatran concentration respectively (p<0.0001 for peak and trough concentration).

Table 3 summarized the associations between peak and trough dabigatran concentrations and the different characteristics of the sample. Renal function (eGFR-CG and eGFR-MDRD) was associated with peak (p=0.16 and p=0.01) and trough (p=0.05 and p=0.003) dabigatran concentrations. Further participants with chronic heart failure had significantly higher peak dabigatran concentrations (p=0.01) and subjects with a history of stroke (p=0.04) had significantly higher trough dabigatran concentrations. There were also marginal relationships between peak dabigatran plasma concentrations and sex (p=0.10) and diabetes (p=0.14) and between trough dabigatran plasma concentrations and hypertension (p=0.06) and low albumin level (p=0.14). Co-medication did not significantly modify peak or trough dabigatran concentrations although participants taking amiodarone had a lower concentration of trough dabigatran concentration (p=0.08).

Table 3.

Population characteristic according to peak and trough dabigatran concentrations

Characteristics	Dabigatran concentration (Anti 2a)
	At peak		At trough
	M (SD)	p*	M (SD)	p*
Age
≤ 87 years old	188.6 (136.6)	0.48	117.2 (113.8)	0.86
> 87 years old	163.6 (137.2)		102.9 (94.3)
Sex
Men	149.1 (163.1)	0.10	102.9 (145.4)	0.30
Women	185.9 (127.7)		113.1 (90.6)
Body mass index
≤ 25.5 kg/m2	172.7 (138.6)	0.43	103.7 (81.4)	0.42
> 25.5 kg/m2	180.3 (138.0)		119.4 (126.3)
Hypertension
No	145.7 (110.7)	0.18	89.1 (89.9)	0.06
Yes	186.9 (143.0)		117.3 (109.0)
Diabetes
No	167.3 (131.6)	0.14	103.3 (88.9)	0.28
Yes	209.4 (151.1)		134.7 (146.3)
Cognitive disorders
No	186.1 (121.0)	0.79	114.6 (78.7)	0.56
Yes	178.6 (145.1)		114.5 (116.7)
Chronic heart failure
No	145.2 (124.8)	0.01	92.8 (101.5)	0.16
Yes	222.9 (141.4)		136.2 (106.1)
History of stroke
No	181.2 (118.5)	0.37	118.0 (90.1)	0.04
Yes	170.3 (165.3)		98.1 (127.4)
Serum albumin
≤ 32 g/L	205.8 (167.9)	0.57	139.8 (133.0)	0.14
> 32 g/L	146.7 (89.6)		79.8 (53.5)
Serum creatinine at baseline
≤ 70 μmol/L	148.9 (122.3)	0.05	80.6 (68.5)	0.005
> 70 μmol/L	207.6 (148.1)		142.2 (128.5)
eGFR-MDRDa at base line
≥ 50 mL/min/1.73m2	167.7 (123.8)	0.01	96.1 (80.5)	0.003
< 50 mL/min/1.73m2	271.1 (231.8)		231.3 (223.3)
eGFR-CGb at baseline
≥ 50 mL/min	158.6 (122.0)	0.16	88.5 (75.2)	0.05
< 50 mL/min	199.6 (153.4)		136.4 (130.4)
Co-medication
Proton-pump inhibitor
No	202.1 (154.4)	0.19	124.5 (122.9)	0.40
Yes	149.2 (108.6)		95.1 (78.9)
Serotonin reuptake inhibitor
No	181.5 (126.3)	0.32	106.4 (80.5)	0.60
Yes	173.3 (146.5)		114.5 (123.6)
Amiodarone
No	183.4 (136.9)	0.21	116.6 (108.1)	0.08
Yes	136.6 (133.8)		71.9 (73.1)
Any hemorrhage within 1 year
No	148.3 (98.1)	0.01	89.5 (75.0)	0.01
Yes	299.5 (201.4)		200.4 (159.1)
Charlson comorbidity index
≤ 7	157.4 (113.8)	0.07	101.7 (88.9)	0.59
> 7	224.7 (173.8)		132.3 (136.2)
CHA2DS2-VASC
≤ 5	164.7 (117.8)	0.06	104.2 (89.6)	0.41
> 5	211.8 (177.9)		128.7 (140.4)
HAS-BLED score
≤ 2	178.2 (130.0)	0.35	107.5 (91.3)	0.47
< 2	170.8 (182.5)		131.7 (176.3)

^*p values computed with log-transformed peak and trough dabigatran plasma concentration using T-test or Wilcoxon rank-sum test; M (SD), mean (standard deviation);

^aestimated with MDRD formula;

^bestimated with Cockcroft-Gault formula.

In a mixed model adjusted for age, sex and hypertension, low eGFR-MDRD and low albumin level were associated with increased trough dabigatran concentration over time (estimate -0.16 (95%CI -0.49; -0.21), p < 0.0001 and -0.18 (95%CI -0.36; -0.003), p < 0.05 for eGFR-MDRD and albumin respectively). Only low eGFR was associated with increased peak dabigatran concentration (estimate -0.20 (95%CI -0.32; -0.07), p=0.002) over time in a mixed model adjusted for age, sex, diabetes and chronic heart failure (figure 2).

Figure 2.

Plot of the multivariate mixed models of determinants of peak and trough dabigatran concentration

During the one year follow-up, 13 participants experienced a hemorrhagic event, 3 major bleedings (2 lower and 1 upper gastrointestinal bleedings) and 10 non-major clinically relevant bleedings (3 lower gastrointestinal bleedings, 2 hematuria episodes, 2 epistaxis, 1 metrorrhagic episode, 1 conjunctive hemorrhage and 1 hemorrhage in an unspecified site). Participants who experienced a hemorrhage had higher peak and trough dabigatran plasma concentrations at all measurements. They also had a lower level of serum albumin at baseline (see table 3 and 4).

Table 4.

Study population characteristics according to the occurrence of hemorrhagic event

Characteristics, M (SD)	Hemorrhage within 1 year
	No	Yes	p*
	N=55	N=13
Age (years)	85.6 (5.3)	86.8 (4.2)	0.65
Women, % (N)	76.4 (42)	76.9 (10)	0.99
Body mass index (kg/m2)	26.1 (5.8)	24.1 (4.6)	0.30
Hypertension, % (N)	78.2 (43)	69.2 (9)	0.49
Diabetes, % (N)	21.8 (12)	30.8 (4)	0.49
Cognitive disorders, % (N)	72.3 (34)	76.9 (10)	0.99
Heart failure, % (N)	38.2 (21)	53.8 (7)	0.36
History of stroke, % (N)	36.4 (20)	38.5 (5)	0.99
Anemia, % (N)	52.7 (29)	53.8 (7)	0.99
Mini mental state examination	18.9 (8.4)	15.5 (8.9)	0.23
Systolic blood pressure (mm Hg)	127.2 (17.8)	132.2 (17.2)	0.33
Diastolic blood pressure (mm Hg)	68.0 (11.0)	70.6 (10.9)	0.66
Hemoglobin (g/dl)	12.1 (1.4)	11.7 (1.6)	0.42
Serum Albumin (g/L)	32.6 (4.5)	29.2 (4.2)	0.02
Serum creatinine at baseline (μmol/L)	74.9 (20.9)	71.8 (33.5)	0.13
eGFR-MDRDa at base line, % (N)
≥ 50 mL/min/1.73m2	90.7 (49)	92.3 (12)	0.99
< 50 mL/min/1.73m2	9.3 (5)	7.7 (1)
eGFR-CGb at base line, % (N)
≥ 50 mL/min	53.7 (29)	61.5 (8)	0.84
< 50 mL/min	46.3 (25)	38.5 (5)
Co-medication, % (N)
Proton-pump inhibitor	45.5 (25)	53.8 (7)	0.81
Serotonin reuptake inhibitor	49.1 (27)	69.2 (9)	0.32
Amiodarone	14.5 (8)	7.69 (1)	0.84
Charlson comorbidity index	6.85 (2.16)	7.62 (2.06)	0.25
CHA2DS2-VASc	4.78 (1.45)	5.08 (1.12)	0.48
HAS-BLED ≥ 2	60.0 (33)	69.2 (9)	0.77
Trough concentration (ng/mL)
D4	86.7 (87.5)	141.0 (111.4)	0.03
D8	89.5 (75.0)	200.4 (159.1)	0.01
D15	93.6 (76.8)	142.4 (80.7)	0.04
D30	88.7 (69.3)	213.0 (138.5)	0.005
D45	100.3 (72.2)	154.4 (133.9)	0.62
Peak concentration (ng/mL)
D4	135.8 (121.0)	225.1 (146.3)	0.04
D8	148.3 (98.1)	299.5 (201.4)	0.01
D15	145.4 (91.3)	214.2 (110.1)	0.07
D30	146.5 (95.7)	344.9 (180.6)	0.0008
D45	134.0 (107.1)	222.6 (192.9)	0.32

^*Wilcoxon rank sum test or Fisher exact test; M (SD), mean (standard deviation); % (N), percentage (number);

^aeGFR, glomerular filtration rate estimated with MDRD formula;

^bglomerular filtration rate estimated with Cockcroft-Gault formula.

During the one year follow-up, 12 participants died. Among them 1 patient died within 7 days of a major bleeding, and 21 (30%) patients discontinued dabigatran because of hemorrhagic events (n=8), renal insufficiency (n=2), patients preference (n=3) and unspecified reason (n=8). Lastly one patient had an embolic event (ischemic stroke).

Seven participants had a trough dabigatran concentration at D8 above the 90^th percentile of trough dabigatran concentration at D8 (243.9 ng/mL). They had a lower renal function and a lower albumin level (creatinine: 98.4 (35.9) vs.71.5 (20.4), p=0.004; eGFR-CG: 42.7 (13.2) vs. 56.7 (18.4), p=0.03; eGFR-MDRD: 59.3 (17.1) vs. 83.1 (25.1), p=0.007; and albumin: 28.7 (2.0) vs. 32.4 (4.7), p=0.02). Kaplan Meier curves (see figure 3) show the occurrence of hemorrhage according to the 90^th percentiles of trough dabigatran concentration at D8 (243.9 ng/mL) (log-rank tests p<.0001). When the threshold for trough dabigatran concentration at D8 was set at 200ng/mL (as proposed by the EU summary of product characteristics for increased risk of bleeding) trough dabigatran concentration above 200 ng/mL was still significantly associated with hemorrhagic events (log-rank tests p=0.0003). For these 2 analyses, the 13 participants who discontinued dabigatran for a non-hemorrhagic cause were right-censured at the time of withdrawal.

Figure 3.

Kaplan-Meier curve of hemorrhagic events according to D8 trough dabigatran concentration ≥ 243.9 ng/mL (90^th percentile of D8 dabigatran plasma concentration)

The sensitivity and specificity of the 90^th percentiles of trough dabigatran concentration at D8 (243.9 ng/mL) for predicting bleeding events was determined via receiver operating characteristic curves. The specificity was 98% (95%CI 0.95-1.00) and the sensitivity was 54% (95%CI 0.23-0.77).

Discussion

In this study prospectively conducted among 68 very old participants (mean age 85.5 (5.1) years old) with AF hospitalized in geriatric care setting, peak and trough dabigatran plasma concentrations, measured 4, 8, 15, 30 and 45 days after dabigatran inception, were relatively stable among individuals, but there was a large inter-individual variability.

Low renal function was a key determinant of dabigatran plasma concentrations. Bleeding episode occurrence was more frequently observed among participants above the 90^th percentile of trough dabigatran plasma concentrations measured at D8.

To date there are very few studies specifically focusing on geriatric population using NOAC although prevalence of atrial fibrillation and its thromboembolic complications dramatically increase with age. To the best of our knowledge, this is the first “real world longitudinal study conducted in geriatric population treated with dabigatran with repeated measurements of peak and trough dabigatran concentrations with assessment of bleeding events.

In our study, the geometric mean dabigatran peak and trough concentrations (130.4 and 74.9 ng/mL respectively) in an geriatric population (mean age 85.8 (5.1) years old) taking 110mg twice a day were slightly higher than in RE-LY study composed of a population of mean 71.4 (8.6) years old in which the geometric mean dabigatran peak and trough concentrations were 126 and 64.7 ng/mL.

The main determinant of dabigatran plasma concentration was renal function. High creatinine and low eGFR were associated with higher peak and trough dabigatran plasma concentrations. These results are consistent with previous studies (21, 32, 33). In RELY study, concentrations of dabigatran were increased 1.2 fold and 1.8 fold for patients with eGFR between 50 to 80 and 30 to 50 mL/min respectively. However in RELY study, the relationship between dabigatran concentrations and eGFR was not anymore significant in a model adjusted for age (21). In our study of very old patients, this relationship with eGFR remained statistically significant even after adjustment for age, suggesting that renal function is still the main determinant factor of dabigatran plasma concentrations among the very old.

In our population, dabigatran plasma concentration had a stronger association with eGFR calculated with MDRD formula than eGFR calculated with Cockcroft-Gault suggesting that eGFR-MDRD was a better predictor of dabigatran plasma levels (34).

Low albumin level was also associated with higher trough dabigatran plasma concentration values. Although dabigatran has a low binding to proteins around 30 to 35 %, hypoalbuminemia may have some effect on dabigatran pharmacology and could lead to an increase of dabigatran plasma concentration. This could explain the significant relationship observed between low plasma albumin levels and hemorrhagic events. Co-medication did not significantly modify peak or trough dabigatran concentration.

Our data showed that the trough concentrations remained stable after day 8 until day 45. Although the half-life of dabigatran varies between 13 and 18 hours in subjects with eGFR > 30 mL/min, in our population of very old participants, the steady state was only reached after 8 days. Half-time in our older participants seemed longer that previously reported possibly related to the high number of comorbidities and low renal function.

We observed a large inter-individual variability of peak and trough levels of dabigatran concentration and a lower intra-individual variability. Meanwhile, our inter-individual and intra-individual variability was higher than previously reported (21, 32, 33). This is, in all likelihood, related to our patient population heterogeneity characterized by older people with multiple comorbidities. Nonetheless, even though the intra-individual variability was relatively large for trough dabigatran plasma concentration, participants in the two first tertiles of dabigatran concentration at day 8 remained below the 90^th percentiles (243.9 ng/mL) on the consecutive next measurements. Furthermore dabigatran plasma concentration at the steady state (at day 8) was associated with bleeding events. Interestingly, renal function was not associated with bleeding event even though renal function was associated with dabigatran concentrations. Our findings are in agreement with the EU summary of product characteristics which stated that exceeding the 90^th percentile of trough dabigatran levels (i.e., 200 ng/mL) is associated with an increased risk of bleeding (31). In RE-LY study the 90^th percentile of trough dabigatran concentration is 200 ng/mL. In our geriatric population however the 90^th percentile of trough dabigatran concentration was slightly higher at 243.9 ng/mL. Interestingly HAS-BLED score was neither associated with dabigatran concentrations nor bleeding events in our population.

The 90^th percentile of trough dabigatran concentration threshold had a very high specificity (98%) and a moderate sensitivity (58%) for predicting hemorrhagic events. Considering the high inter-individual variability and quite moderated intra-individual variability of plasma dabigatran concentrations (21, 35), this result calls for more studies to investigate the ability of plasma dabigatran measurement at D8 to identify subjects at high risk of bleeding occurrence. These results also suggests that in these participants with a high trough dabigatran level at D8, more frequent monitoring of the renal function should be implemented and potential P-glycoprotein inhibitor drugs should be avoided whenever possible and eventually dabigatran dosage reduction should be discussed (36). Lastly a switch to another NOAC or even to VKA might also be considered. Further studies are needed to assess the benefit/risk of a reduced dose of dabigatran (75 mg twice a day) in such a very old AF population with an elevated Dabigatran plasma concentration (36). There is also a need of more studies in geriatric population at higher risk of bleeding to evaluate the clinical utility of measuring dabigatran level for dose adjustment to prevent bleeding event.

In this study, dabigatran concentrations were not directly measured by mass spectrometry. Plasma dabigatran concentrations were derived from the plasma diluted thrombin time measured with the Hemoclot® thrombin inhibitor (37). However this coagulation method has been validated and demonstrates a close linear relationship and a strong correlation with physical determination dabigatran concentrations in both in vitro and ex vivo studies (38). This method is therefore suitable for monitoring purposes of therapeutic and supratherapeutic dabigatran plasma concentrations (38, 39).

The strength of our study lies in the data describing such a very old geriatric population, characterized by multiple comorbidities with high Charlson comorbidity index, and which profile is never included in classical randomized controlled studies. It thus provides “real-life data in a population with high-risk of biological over-dosage and/or clinical bleeding. There was a large number (n=541) of blood measurements with measurements up to 45 days after dabigatran inception and a follow-up of bleeding events for 1 year. All participants had at least had 3 measurements of trough and 3 measurements of peak dabigatran plasma concentrations. Because all measurements were performed in the same laboratory, measurements methods were perfectly standardized with no inter-center variability.

The study has also some limitations. The follow-up period for blood measurements was only 45 days while anticoagulant treatment of AF lasts for years. Only 66% and 50% of patients had a dabigatran concentration measurement at D30 and D45 respectively. Bleeding follow up was done over the phone and some minor events may have been overlooked. There were only 13 clinically relevant bleeding events with only 3 major bleeding events, so there was limited power in detecting predictors of bleeding. Moreover the results concerning the cut-off value to identify patient at higher risk of bleeding need to be confirmed in larger populations. Not all drugs interacting with dabigatran were monitored but only drugs frequently prescribed in elderly population like aspirin, nonsteroidal anti-inflammatory drug, proton-pump inhibitor, serotonin reuptake inhibitor, amiodarone and verapamil. Lastly, there was no plasma dabigatran concentration measurement at the time of the bleeding events.

Conclusions

This is the first NOAC study in a “real life cohort of very old geriatric population with high comorbidity level that collected more than 500 measurements. Renal function was the key determinant of peak and trough dabigatran plasma concentrations. Low albumin was also associated with trough dabigatran concentrations. Inter-individual dabigatran concentration variability was larger than intra-individual variability. Despite this large variability, participants in the two lower tertiles of dabigatran trough concentration at D8 remained below the 90^th percentiles (243.9 ng/mL) on the following measurements. Lastly, our results suggest that in this old geriatric population, a single measurement of dabigatran trough concentration performed 8 days after dabigatran initiation may identify patients at significantly higher risk of bleeding.

Conflict of Interest

IE received consultant/advisory/lecture fees from Bayer Healthcare, Boehringer-Ingelheim, Bristol-Myers Squibb, Daiichi-Sankyo, Pfizer, Sanofi-Aventis, Stago, Werfen. GG received consultant/advisory/lecture fees from Boehringer- Ingelheim. OH received consultant/advisory/lecture fees from Daiichi-Sankyo, Bayer, Boehringer-Ingelheim, BMS, Pfizer, Novartis, Servier, Astra-Zeneca. Sanofi. JSV, EC, LC, CB, FL have no conflict of interests.

Ethical standard

The study was conducted in accordance with the declaration of Helsinki.