Measures of vitamin K antagonist control reported in atrial fibrillation and venous thromboembolism studies: a systematic review.

OBJECTIVE: To aid trialists, systematic reviewers and others, we evaluated the degree of standardisation of control measure reporting that has occurred in atrial fibrillation (AF) and venous thromboembolism (VTE) studies since 2000; and attempted to determine whether the prior recommendation of reporting ≥2 measures per study has been employed.
DESIGN: Systematic review. SEARCH STRATEGY: We searched bibliographic databases (2000 to June 2013) to identify AF and VTE studies evaluating dose-adjusted vitamin K antagonists (VKAs) and reporting ≥1 control measure. The types of measures reported, proportion of studies reporting ≥2 measures and mean (±SD) number of measures per study were determined for all studies and compared between subgroups. DATA EXTRACTION: Through the use of a standardised data extraction tool, we independently extracted all data, with disagreements resolved by a separate investigator.
RESULTS: 148 studies were included, 57% of which reported ≥2 control measures (mean/study=2.13±1.36). The proportion of time spent in the target international normalised ratio range (TTR) was most commonly reported (79%), and was frequently accompanied by time above/below range (52%). AF studies more frequently reported ≥2 control measures compared with VTE studies (63% vs 37%; p=0.004), and reported a greater number of measures per study (mean=2.36 vs 1.53; p<0.001). Observational studies were more likely to provide ≥2 measures compared with randomised trials (76% vs 33%; p<0.001) and report a greater number of measures (mean=2.58 vs 1.63; p<0.001). More recent studies (2004-2013) reported ≥2 measures more often than older (2000-2003) studies (59% vs 35%; p=0.05) and reported more measures per study (mean=2.23 vs 1.48; p=0.02).
CONCLUSIONS: While TTR was often utilised, studies reported ≥2 measures of VKA control only about half of the time and lacked consistency in the types of measures reported. A trend towards studies reporting greater numbers of VKA control measures over time was observed over our review time horizon, particularly, with AF and observational studies. Published by the BMJ Publishing Group Limited. For permission to use (where not already granted under a licence) please go to http://group.bmj.com/group/rights-licensing/permissions.

This large systematic review (N=148 studies) adds to the existing literature by providing updated results and new data regarding the frequency and consistency of vitamin K antagonists (VKA) control measure reporting.

While the previous review by Fitzmaurice et al included studies of all VKA indications; ours evaluated atrial fibrillation (AF) and VTE studies only.

Unlike previous reviews, our systematic review examined VKA control measure reporting over time and differences in reporting between AF and VTE studies and randomised trials and observational studies. In addition, we explored the way in which VKA control measures are concomitantly reported in studies.

Introduction

Adjusted-dose vitamin K antagonists (VKAs) are frequently used, and are the standard-of-care anticoagulants that most new oral anticoagulants for the prevention of thrombotic events in patients with atrial fibrillation (AF) and following venous thromboembolism (VTE) are compared with.1–10 VKAs have substantial evidence from clinical trials supporting their efficacy, and their use is endorsed by multiple national guidelines11 12; however, they are often underused due to difficulty in maintaining the international normalised ratio (INR) in the narrow therapeutic range (often 2.0–3.0).13 14

Fitzmaurice et al15 performed a systematic review of studies published between 1995 and 1999 in order to evaluate the manner in which VKA control was reported and to provide recommendations for reporting of VKA control measures (parameters used to summarise the level of anticoagulation). Their review found that a wide range of measures had been used in the literature, but with little consistency between studies. Since studies also suggest different VKA control measures (eg, percentage of time spent in range, proportion of tests in range, point prevalence) used in the same population can result in different conclusions regarding the quality of VKA control,16–18 researchers recommended ≥2 VKA control measures be reported per study.

In order to aid researchers (eg, clinical trialists and systematic reviewers) and other end users, we performed a systematic review to assess the degree of standardisation in VKA control measure reporting that has occurred in AF and VTE studies since the publication of the paper by Fitzmaurice et al; and to determine whether their recommendation of reporting ≥2 control measures has been widely employed.

Methods

A systematic review of MEDLINE, CENTRAL and EMBASE (from 2000 to June 2013) was conducted to identify published studies (English full-text randomised controlled trials, prospective cohort studies or retrospective analyses) including at least one dose-adjusted VKA treatment arm and reporting a minimum of one VKA control measure in adult patients being treated for AF or VTE as their primary reason for anticoagulation. Our search strategy for MEDLINE (PubMed) is provided in online supplementary appendix 1. Studies were excluded if they included <50 patients or planned to treat patients for <3 months. Manual backwards citation tracking of references from identified studies and review articles was also performed to identify additional relevant studies. All citations were screened by two independent investigators (ESM and J-SS) with discrepancies resolved by a third investigator (CIC).

Through the use of a standardised data extraction tool, we independently extracted all data (ESM, J-SS and JH), with disagreements resolved by a separate investigator (CIC). Collected study-level data included: study identifier and year of publication; indication(s) for VKA therapy; sample size; study design (prospective, retrospective or randomised study); duration of VKA treatment; mean age of participants; and the type(s) of VKA used. The types of VKA control measures reported were also extracted from each study. These included (but were not limited to): percentage of time in range (target international normalised ratio (TTR), calculated using Rosendaal's linear interpolation method19), below and/or above range, TTR in an extended range (ie, 1.8–3.2) and extreme ranges (ie, <1.5 and/or >5.0); proportion of INR measurements in range (PINRR), below and/or above extreme range; mean/median INR; mean/median VKA dose; frequency of INR monitoring (number of INR measures per patient over the course of the study); INR variability; INR monitoring interval (number of days between each INR measure); point prevalence (eg, the proportion of patients in range and/or out of range, proportion of patients in range >50% of time or proportion of patients with ≥50% of INR measures <3.0); number of VKA dosage changes; INR measure after a previously subtherapeutic or supratherapeutic INR; proportion of patients with ≥1 INR measure below range after reaching an adequate INR; number of days until the next INR measure after an extreme measure; proportion of days with treatment stability (two consecutive INR measures in range); days to reach a therapeutic INR; mean time until stable and minimum and maximum INR values per patient.

The types of measures reported were summarised and displayed using tables and figures, and the proportion of studies reporting ≥2 measures along with the mean number of measures per study (±SD) were reported for all identified studies. We also compared these same end points between select study subgroups (primary indication for anticoagulation (AF vs VTE); study design (randomised trial vs observational study); and year of publication (2000–2003 vs 2004–2013)). The year categorisations were chosen based on the year of publication of the review by Fitzmaurice et al.15 Finally, in order to assess the concomitant use of VKA control measures within studies, a diagram depicting per study measure linkages was created.

Between-group comparisons were made using χ2 tests (or Fisher's exact tests, where appropriate) for categorical data and unpaired t tests for continuous data. A p value of <0.05 was considered statistically significant in all situations. Statistical analysis was performed using SPSS V.17.0 (SPSS Inc, Chicago, Illinois, USA).

Results

Of the 5301 citations initially identified, 1119 full-text articles were reviewed for inclusion. A total of 148 studies met all inclusion and exclusion criteria and were included in the analysis (figure 1, table 1).1–9 18 20–157 Of note, 112 VKA studies were excluded from our systematic review because they did not report a VKA control measure although study participants were receiving a VKA for AF or VTE as their primary reason for treatment for greater than 3 months.

Figure 1

Results of the literature search. AF, atrial fibrillation; CCTR, Cochrane controlled trials register; RCT, randomised controlled trial; VKA, vitamin K antagonist; VTE, venous thromboembolism.

Table 1

Studies reporting at least one VKA quality control measure

Study	Disease state	Study design	VKA-treated N	VKA studied	Target INR	TTR, %	PINRR, %	Mean/median INR	Mean/median dose	Monitoring frequency	INR variability	INR testing interval, days*	PPIR	Other*
Abdelhafiz and Wheeldon20	AF	PD	402	W	2–3	•		•		•	•
Abdelhafiz and Wheeldon21	AF	PD	402	W	2–3	•		•		•	•
Agnelli et al 22	DVT	RCT	134	W,A	2–3	•
Agnelli et al 23	PE	RCT	165	W,A	2–3	•
Agnelli et al 24	DVT	RCT	126	W,A,P	2–3	•
Agnelli et al 7	VTE	RCT	2704		2–3	•
Albers et al 25	AF	RCT	1962	W	2–3	•	•						•
Amiwero et al 26	VTE	RCT	126	W	2–3	•		•					•
Anderson27	AF	RD	87	W	2–3		•			•
Ansell et al 18	AF	RD	1511	W, A, F	2–3	•	•			•	•	•		•
Aujesky et al 28	PE	RCT	339	W,A,P,F	2–3	•
Bona et al 29	VTE	PD	98	W	2–3	•
Boulanger et al 30	AF	RD	6431	W	2–3	•				•		•	•
Büller et al 31	PE	RCT	2184	–	2–3	•
Büller et al 32	DVT	RCT	137	W,A,P,F	2–3		•			•
Büller et al 33	PE	RCT	1595	W	2–3	•
Burton et al 34	AF	RD	259	W	2–3	•		•		•
Cafolla et al 35	AF/VTE	PD	871	W,A, other	2–3	•
Cafolla et al 36	AF	PD	112	W	2–3/1.5–2.5	•			•		•
Campbell et al 37	VTE	RCT	749	W	2–3.5								•
Cheung et al 38	AF	RD	555	W	1.5–3	•
Chitsike et al 39	VTE	PD	349	W	2–3	•							•
Chung et al 40	AF	RCT	75	W	2–3	•
Coleman et al 41	AF	PD	65	W	2–3	•								•
Connolly et al 42	AF	RCT	3371	–	2–3	•
Connolly et al 43	AF	RCT	3371	–	2–3	•
Connollyet al 6	AF	RCT	6022	W	2–3	•
Copland et al 44	AF	RD	328	W	1.8–3.3				•	•
Currie et al 45	AF	RD	1513	W	2–3	•	•		•	•	•			•
Daskalopoulos et al 46	DVT	RCT	52	A	2–3				•
Dimberg et al 47	AF	RD	791	W	2–3	•						•
Douketis et al 48	VTE	RCT	1021	–	2–3				•
Easton et al 49	AF	RCT	1643	W	2–3	•
The Einstein Investigators49a	DVT	RCT	1718	W,A	2–3	•
The Einstein Investigators49b	PE	RCT	2413	W,A	2–3	•
Ellis et al 50	AF	RCT	66	T	2–3	•		•	•				•
Evans et al 51	AF	PD	288	W	2–3	•	•
Evans et al 52	AF	PD	214	W	2–3	•	•
Ezekowitz et al 53	AF	RCT	70	W	2–3	•
Ezekowitz et al 54	AF	RCT	6022	W	2–3	•
Fiessinger et al 55	VTE	RCT	1249	W	2–3	•
Ford et al 56	AF	RCT	3665	W	2–3	•	•
Gadisseur et al 57	DVT	PD	266	A,P	2.5–3.5	•	•
Gallagher et al 58	AF	RD	18 113	W	2–3	•				•
Gallagher et al 59	VTE	RD	10 381	W,A,P	2–3	•				•
Garcia et al 60	AF	RCT	9081	W	2–3	•
Go et al 61	AF	RD	6320	W	2–3	•				•
Gomberg-Maitland et al 62	AF	RCT	3624	W	2–3	•
Granger et al 8	AF	RCT	9081	W	2–3	•
Hankey et al 63	AF	RCT	7133	W	2–3	•
Heidinger et al 64	AF/DVT	PD	1375	–	2–3				•
Ho et al 65	AF	RD	476	W	2–3	•	•						•
Hokusai-VTE Investigators65a	VTE	RCT	4122	W	2–3	•
Holmes et al 66	AF	RCT	244	W	2–3	•
Hori et al 67	AF	RCT	108	W	2–3/2–2.6	•
Hori et al 68	AF	RCT	639	W	2–3/1.6–2.6				•
Hutten et al 69	VTE	RCT	1039	–	2–3	•
Hylek et al 70	AF	PD	472	W	2–3	•				•
Hylek et al 71	AF	RCT	3665	W	2–3	•							•
Jacobs et al 72	AF	RD	90	W	2–3	•						•
Jones et al 73	AF	RD	2223	W	2–3	•	•		•	•		•
Kalra et al 74	AF	PD	167	W	2–3	•						•
Kearon et al 75	VTE	RCT	738	W	2–3/1.5–1.9	•	•					•
Kearon et al 76	VTE	RCT	81	W	2–3	•	•
Kearon et al 77	VTE	RCT	703	W	2–3	•
Kim et al 78	AF	RD	129	W	2–3	•		•	•	•		•
Kim et al 79	AF/VTE	PD	646	W	2–3	•
Kulo et al 80	AF	PD/RD	117	W,A	2–3	•	•	•	•	•
Kulo et al 81	AF	PD/RD	117	W,A	2–3	•			•	•		•	•
Kurtoglu et al 82	DVT	PD	246	W	2–3	•								•
Lee et al 83	AF	PD/RD	200	W	2–3	•	•						•
Lip et al 84	AF	RCT	318	–	2–3								•
Lopez-Beret et al 85	DVT	RCT	77	A	2–3				•
Malik and Taylor 86	AF/VTE	RD	328	W	2–3	•
Mant et al 87	AF	RCT	488	W	2–3	•	•
Matchar et al 88	AF	RCT	363	W	2–3	•						•
Matchar 89	AF	RCT	363	W	2–3	•						•
McBride et al 90	AF	PD	324	W	2–3	•
McCormick et al 91	AF	RD	174	W	2–3	•						•
Melamed et al 92	AF	RD	906	W	2–3	•				•			•
Menzin et al 93	AF	RD	600	W	2–3	•				•
Morgan et al 94	AF	RD	2235	W	2–3	•				•
Naganuma et al 95	AF	RD	845	W	1.5–2.5	•
Nakatani et al 96	AF	R	95	W	2–3/1.6–2.6	•			•		•	•
Neree et al 97	AF	RD	395	W,P,A	2–3	•			•	•			•
Nichol et al 98	AF	RD	1107	W	2–3	•						•
Nieuwlaat et al 99	AF	RCT	266	–	2–3	•	•					•
Njaastad et al 100	AF/VTE	RD	936	W	2–3	•
Nozawa et al 101	AF	PD	156	W	1.6–1.9		•			•	•			•
Obata et al 102	AF	RD	110	W	1.6–2.6	•	•	•		•	•	•
Ogawa et al 103	AF	RCT	74	W	2–3/2–2.6								•
Okumura et al 104	AF	PD	501	W	2–3/1.6–2.6	•		•				•
Olsson 105	AF	RCT	1703	W	2–3	•	•
Olsson et al 106	AF	RCT	83	W	2–3	•
Ombandza-Moussa et al 107	VTE	RD	81	–	–		•
Ono and Fujita108	AF	PD	63	W	1.5–2.5	•	•					•
Palareti et al 109	VTE	PD	733	W, A	2–3	•						•
Palareti et al 110	VTE	PD	297	W, A	2–3	•
Patel et al 1	AF	RCT	7133	W	2–3	•
Pengo et al 111	AF	PD	433	W,A	2–3	•	•	•				•
Pengo et al 112	AF	RCT	267	W	2–3/1.5–2.0	•	•					•
Perez-de-Llano113	PE	RCT	50	A	2–3				•
Perez-Gomez114	AF	RCT	479	A	2–3	•	•	•		•		•
Perez-Gomez115	AF	RCT	91	–	2–3	•	•	•		•		•
Perez-Gomez116	AF	RCT	496	–	2–3	•	•	•
PERSIST Investigators117	DVT	RCT	132	W	2–3				•	•
Petersen et al 118	AF	RCT	67	W	2–3								•
Poli et al 119	VTE	PD	182	W	2–3	•		•
Poli et al 120	AF	PD	290	–	2–3	•
Poli et al 121	AF	PD	783	W	2–3	•		•
Poli et al 122	AF	PD	780	W	2–3	•		•
Poli et al 123	AF	PD	578	–	2–3	•	•
Poller et al 124	AF/VTE	RCT	9148	W, A, P	2–3	•
Prandoni (Galilei) et al 125	VTE	RCT	720	W	2–3								•
Prandoni et al 126	DVT	RCT	180	–	2–3								•
Ridker et al 127	VTE	RCT	255	W	1.5–2		•	•
Rombouts et al 128	AF	RCT	104	P	2–3.5	•
Sadanaga et al 129	AF	PD	269	W	1.5–3		•						•
Samsa et al 130	AF	RD	660	W	2–3	•			•	•		•
Sarawate et al 131	AF	PD	470	W	2–3	•
Schulman et al 4	VTE	RCT	1265	W	2–3	•				•
Schulman et al 5	VTE	RCT	1426	W	2–3	•
Sconce et al 132	AF	RCT	70	W	2–3	•		•			•
Shalev et al 133	AF	RD	4408	W	2–3	•			•	•		•
Shen et al 134	AF	RD	18 867	W	2–3	•				•
Shen et al 135	AF	RD	8992	W	2–3	•				•
Sullivan et al 136	AF	RCT	4060	W	2–3	•
Suzuki et al 137	AF	PD	667	W	1.6–2.6		•
Tincani et al 138	AF	PD	90	W,A	2–3	•		•
van Bladel et al 139	PE	PD	86	A	2–3.5	•
van Dongen et al 140	DVT	PD	244	–	2–3	•							•
van Geest-Daalderop et al 141	AF	RD	284	A	2–3.5	•
van Gogh Investigators141a	VTE	RCT	2572	W,A	2–3	•								•
Van Spall et al 143	AF	RCT	6022	W	2–3	•				•
Veeger et al 144	VTE	RD	2304	A	2–3.5	•	•			•		•	•
Vene et al 145	AF	PD	113	W	2–3		•
Voller et al 146	AF	RCT	202	–	2–3	•			•	•	•
Walker et al 147	AF	RD	84	W	2–3	•								•
Weimar et al 148	AF	PD	252	W	2–3							•		•
Weitz et al 149	AF	RCT	250	W	2–3	•								•
White et al 150	AF	RCT	3587	W	2–3	•				•		•	•
Wieloch et al 151	AF/VTE	RD	15 264	W	2–3	•			•	•
Willey et al 152	VTE	RD	225	W	2–3	•	•			•
Wyse et al 153	AF	RCT	4060	W	2–3		•
Yamaguchi154	AF	RCT	115	W	2.2–3.5/ 1.5–2.1		•
Yamashita et al 155	AF	RCT	129	W	2–3/1.6–2.6	•
Yasaka et al 156	AF	PD	88	W	–		•
Yousef et al 157	AF	RD	739	W	2–3		•		•	•		•

*‘Other’ includes: number of dosage changes; INR measure after a previously subtherapeutic or supratherapeutic INR; proportion of patients with ≥1INR measure below range after reaching an adequate INR; number of days until the next INR measure after an extreme measure; proportion of days with treatment stability (two consecutive INR measures in range); days to reach a therapeutic INR; mean time until stable (6 months within target INR range); minimum and maximum INR values per patient.

– Indicates data not reported.

A, acenocoumarol; AF, atrial fibrillation; DVT, deep vein thrombosis; F, fluindione; INR, international normalised ratio; N, sample size; PD, prospective design; P, phenprocoumon; PINRR, proportion of INR measures in range; PPIR, proportion of patients in range; RCT, randomised controlled trial; RD, retrospective design; T, tecarfarin; TTR, time in therapeutic range; VKA, vitamin K antagonist; VTE, venous thromboembolism; W, warfarin.

Overall, 57% of studies reported ≥2 VKA control measures (mean/study=2.13±1.36; table 2). TTR was the most common measure reported (79%), and in a little more than half of these studies, was accompanied by the proportion of time above and/or below range. Other common metrics (used in ≥20% of studies) included mean/median INR, frequency of INR monitoring, INR testing interval and the proportion of patients in/out of range. Subgroup analysis found AF studies were 1.7-fold more likely than VTE studies (table 3), observational studies were more than twice as likely as randomised trials (p≤0.05 for all comparisons; table 4) and recently published studies were 70% more likely than older studies (2000–2003; table 5) to report ≥2 control measures. Moreover, the AF, observational and later time period study subgroups were also more likely to report a greater absolute number of measures per study (p<0.02 for all comparisons). When studies that included a new oral anticoagulant (n=30; all published after 2003) were analysed exclusively, only eight (26.7%) reported ≥2 VKA control measures (mean/study=1.37±0.72). At the same time, however, TTR was reported in all but five studies. Finally, AF and observational studies were more likely to report less common metrics, such as extended range time in the therapeutic range, INR testing interval and frequency of INR monitoring (p<0.05 for all comparisons).

Table 2

Types and frequency of VKA control measures reported in identified studies

Variable	(N=148)n (%)
Number of measures reported (mean±SD)	2.13±1.36
1	63 (42.6)
2	44 (29.7)
3	16 (10.8)
4	13 (8.8)
5	8 (5.4)
6	3 (2.0)
7	1 (0.7)
≥2	85 (57.4)
Percentage of time in range (INR=2–3)	117 (79.1)
Below range (<2)	77 (52.0)
Above range (>3)	77 (52.0)
In extended range (1.8–3.2)	15 (10.1)
In extreme range (<1.5, >5)	19 (12.8)
Proportion of INR tests in range (INR=2–3)	24 (16.2)
Below range (<2)	22 (14.9)
Above range (>3)	20 (13.5)
In extreme range (<1.5, >4)	11 (7.4)
Mean/median INR	38 (25.7)
Mean/median VKA dose	17 (11.5)
Frequency of INR monitoring	38 (25.7)
INR variability	8 (5.4)
INR testing interval	32 (21.6)
Proportion of patients in/out of range*	29 (19.6)
Other†	13 (8.8)

*For example, point prevalence, proportion of patients in range >50% of time or proportion of patients with ≥50% of INR measures <3.0.

†Other measures include: number of dosage changes; INR measure after a previously subtherapeutic or supratherapeutic INR; proportion of patients with ≥1INR measure below range after reaching an adequate INR; number of days until the next INR measure after an extreme measure; proportion of days with treatment stability (two consecutive INR measures in range); days to reach a therapeutic INR; mean time until stable (6 months within target INR range); minimum and maximum INR values per patient.

INR, international normalised ratio; VKA, vitamin K antagonist.

Table 3

Differences in VKA control measures reported between AF and VTE studies

Variable	AF (N=106)n (%)	VTE (N=49)n (%)	p Value*
Number of measures reported (mean±SD)	2.36±1.44	1.53±0.92	<0.001
1	39 (36.8)	31 (63.3)
2	30 (28.3)	13 (26.5)
3	13 (12.3)	4 (8.2)
4	13 (12.3)	0
5	8 (7.5)	0
6	2 (1.9)	1 (2.0)
7	1 (0.9)	0
≥2	67 (63.2)	18 (36.7)	0.004
Percentage of time in range (INR=2–3)	87 (82.1)	36 (73.5)	0.31
Below range (<2)	53 (50.0)	25 (51.0)	0.96
Above range (>3)	54 (51.0)	24 (49.0)	0.96
In extended range (1.8–3.2)	15 (14.2)	0	0.01
In extreme range (<1.5, >5)	13 (12.3)	6 (12.2)	0.80
Proportion of INR tests in range (INR=2–3)	18 (17.0)	7 (14.3)	0.85
Below range (<2)	16 (15.1)	7 (14.3)	0.91
Above range (>3)	14 (13.2)	7 (14.3)	0.94
In extreme range (<1.5, >4)	11 (10.4)	0	0.05
Mean/median INR	30 (28.3)	8 (16.3)	0.16
Mean/median VKA dose	15 (14.2)	3 (6.1)	0.24
Frequency of INR monitoring	32 (30.2)	7 (14.3)	0.06
INR variability	8 (7.5)	0	0.11
INR testing interval	29 (27.4)	4 (8.2)	0.01
Proportion of patients in/out of range†	20 (18.9)	9 (18.4)	0.88
Other‡	11 (10.4)	2 (4.1)	0.32

*p Value for the comparison of AF vs VTE.

†For example, point prevalence, proportion of patients in range >50% of time or proportion of patients with ≥50% of INR measures <3.0.

‡Other measures include: number of dosage changes; INR measure after a previously subtherapeutic or supratherapeutic INR; proportion of patients with ≥1INR measure below range after reaching an adequate INR; number of days until the next INR measure after an extreme measure; proportion of days with treatment stability (two consecutive INR measures in range); days to reach a therapeutic INR; mean time until stable (6 months within target INR range); minimum and maximum INR values per patient.

AF, atrial fibrillation; INR, international normalised ratio; VKA, vitamin K antagonist; VTE, venous thromboembolism.

Table 4

Differences in VKA control measures reported between randomised trials and observational studies

Variable	Randomisedcontrolled trials(N=72)n (%)	Observational studies(N=76)n (%)	p Value*
Number of measures reported (mean±SD)	1.63±1.08	2.58±1.46	<0.001
1	48 (66.7)	18 (23.7)
2	12 (16.7)	29 (38.2)
3	6 (8.3)	10 (13.2)
4	3 (4.2)	10 (13.2)
5	3 (4.2)	5 (6.6)
6	0	3 (3.9)
7	0	1 (1.3)
≥2	24 (33.3)	58 (76.3)	<0.001
Percentage of time in range (INR=2–3)	56 (77.8)	61 (80.3)	0.87
Proportion of INR tests in range (INR=2–3)	10 (13.9)	14 (18.4)	0.60
Mean/median INR	14 (19.4)	24 (31.6)	0.09
Mean/median VKA dose	6 (8.3)	11 (14.5)	0.36
Frequency of INR monitoring	8 (11.1)	30 (39.5)	<0.001
INR variability	2 (2.8)	6 (7.9)	0.31
INR testing interval	6 (8.3)	24 (31.6)	<0.001
Proportion of patients in/out of range†	13 (18.1)	16 (21.1)	0.80
Other‡	2 (2.8)	10 (13.2)	0.04

*p Value for the comparison of randomised controlled trials vs observational studies.

†For example, point prevalence, proportion of patients in range >50% of time or proportion of patients with ≥50% of INR measures <3.0.

‡Other measures include: number of dosage changes; INR measure after a previously subtherapeutic or supratherapeutic INR; proportion of patients with ≥1INR measure below range after reaching an adequate INR; number of days until the next INR measure after an extreme measure; proportion of days with treatment stability (two consecutive INR measures in range); days to reach a therapeutic INR; mean time until stable (6 months within target INR range); minimum and maximum INR values per patient.

INR, international normalised ratio; VKA, vitamin K antagonist.

Table 5

Change in VKA control measures reported in studies published between 2000–2003 and 2004–2013

Variable	Studies published in 2000–2003 (N=23)n (%)	Studies published in 2004–2013 (N=125)n (%)	p Value*
Number of measures reported (mean±SD)	1.48±0.79	2.23±1.43	0.02
1	15 (65.2)	51 (40.8)
2	6 (26.1)	35 (28.0)
3	1 (4.3)	15 (12.0)
4	1 (4.3)	12 (9.6)
5	0	8 (6.4)
6	0	3 (2.4)
7	0	1 (0.8)
≥2	8 (34.8)	74 (59.2)	0.05
Percentage of time in range (INR=2–3)	15 (65.2)	102 (81.6)	0.14
Proportion of INR tests in range (INR=2–3)	4 (17.4)	20 (16.0)	0.89
Mean/median INR	7 (30.4)	31 (24.8)	0.76
Mean/median VKA dose	1 (4.3)	16 (12.8)	0.42
Frequency of INR monitoring	2 (8.7)	36 (28.8)	0.08
INR variability	0	8 (6.4)	0.46
INR testing interval	4 (17.4)	26 (20.8)	0.98
Proportion of patients in/out of range†	1 (4.3)	28 (22.4)	0.09
Other‡	0	12 (9.6)	0.26

*p Value for the comparison of studies published in 2000–2003 vs 2004–2013.

† For example, point prevalence, proportion of patients in range >50% of time or proportion of patients with ≥50% of INR measures <3.0.

‡Other measures include: number of dosage changes; INR measure after a previously subtherapeutic or supratherapeutic INR; proportion of patients with ≥1INR measure below range after reaching an adequate INR; number of days until the next INR measure after an extreme measure; proportion of days with treatment stability (two consecutive INR measures in range); days to reach a therapeutic INR; mean time until stable (6 months within target INR range); minimum and maximum INR values per patient.

INR, international normalised ratio; VKA, vitamin K antagonist.

Our assessment of the concomitant use of VKA control measures in identified studies suggested there was little consistency in their use (figure 2). TTR (the most frequently reported measure overall) was most often reported with mean INR, frequency of INR monitoring and INR testing interval.

Figure 2

Frequency of concomitant reporting of vitamin K antagonist (VKA) control measures in identified studies. The width of the line is proportional to the number of trials reporting each pair of VKA control measures. Each node is proportional to the total number of times the VKA control measure was reported. INR, international normalised ratio.

Discussion

We performed a systematic review to assess the degree of standardisation of VKA control measures reported in AF and VTE studies since 2000, and to determine the proportion of studies reporting ≥2 control measures. We found that while TTR was frequently reported in identified studies; other measures were more sporadically provided. Our analysis also demonstrated AF studies (compared with VTE studies), observational studies (compared with randomised trials) and more recently published studies (2004–2013) (compared with older ones) were more likely to report ≥2 VKA control measures per study and report a greater absolute number of measures per study. New oral anticoagulant studies utilised TTR quite frequently (>80% of the time), suggesting further standardisation in VKA control measure reporting. Finally, we observed little consistency in the combinations of measures used in identified studies.

We believe the results of our systematic review extend current knowledge regarding the frequency and consistency of VKA control measure reporting in anticoagulation studies, and can serve as a valuable tool for clinical trialists and systematic reviewers. The aforementioned review performed by Fitzmaurice et al15 included only 15 studies across varying indications (not just AF and VTE), making it only a fraction of the size of our own and suggesting that direct comparison between these systematic reviews should be made with caution. Fitzmaurice et al found 60% of VKA studies published between 1995 and 1999 reported ≥2 control measures (mean=1.93/study), but with a wide variation in the type of measures reported. TTR (47%), mean/median INR (33%), PINRR (40%) and mean/median warfarin dose (33%) were the most frequently reported VKA control measures identified in their review; however, none of their studies reported point prevalence despite its easy calculation and recommended use at the time.158

While our systematic review appears to confirm a number of findings of Fitzmaurice et al, our review also suggests that since 2000, additional measures of VKA control—including point prevalence—have become at least to some degree more common in the anticoagulation literature. Moreover, our findings of discrepancies in the number of control measures reported between AF and VTE studies and observational and randomised studies are novel.

There are a number of reasons why reporting multiple measures of VKA control in anticoagulation studies (as originally suggested by Fitzmaurice et al) seems wise. First, by reporting multiple measures of control, the likelihood that potentially comparable studies share at least one measure in common is increased. Furthermore, studies suggesting different VKA control measures may yield disparate findings even when utilised in the same patient population.16–18 In a retrospective cohort study of 633 patients undergoing anticoagulation with a VKA, Schmitt et al17 observed 24%, 24% and 22% absolute differences between TTR and PINRR estimates of INR control and 22%, 26% and 17% absolute differences between TTR and point prevalence (cross-sectional) estimates of INR control after 2-month, 3-month and 6-month time intervals of follow-up, respectively. Moreover, in a randomised trial of 367 patients receiving a VKA, Fitzmaurice et al16 demonstrated up to a 9% variance between TTR, point prevalence and PINRR estimates in the same population. Finally, a retrospective study of 1511 patients performed by Ansell et al18 provided yet more evidence to support population differences in VKA control estimates when different measures are used, demonstrating a 5–10% absolute difference between TTR and PINRR estimates for patients from five different countries.

A final reason for including multiple measures on VKA control in anticoagulation studies is that different VKA control measures have their own unique strengths and weaknesses.17 While TTR takes into account actual days in the target INR range (typically by assuming values vary linearly between two measures19), its calculation is more complex than other measures; it makes assumptions about INR values between actual tests and can be biased by extreme out-of-range INR values. In addition, while we were not able to assess this in our systematic review because of a lack of consistent reporting, there appears to be variability in what INR values are included in TTR calculations, with some studies excluding INR values occurring during the initiation phase (ie, first week) and/or around temporary interruptions of a VKA.6 8 PINRR is a simpler measure to calculate than TTR; it requires only one INR measurement per patient and is not influenced by the extent INRs are out of range; nevertheless, it fails to take into account actual days of anticoagulation like TTR and may underestimate control when more frequent INR testing occurs in unstable patients. Point prevalence is perhaps the simplest measure to calculate because it takes only one time point into consideration (a cross-sectional method), and like PINRR, it is not influenced by the extent an INR value is out of range; however, unlike the aforementioned methods, point prevalence takes individual patients into account. Finally, it is worth noting that VKA control measures may tend to stabilise over time, suggesting duration of study follow-up should be considered when interpreting a control measure.

On the basis of the results of our systematic review, we agree with the previous recommendation of Fitzmaurice et al of reporting at least two measures of VKA control. However, we would like to emphasise that while we recommend multiple measures be reported, we are by no means suggesting that the quantity of measures reported is more important than the quality of the measures. For this reason, we further suggest TTR be one of the measures because of its frequent study in the literature (use in studies and linkage to anticoagulation outcomes).

There are several limitations of our systematic review worth discussion. First, like any other systematic review, the possibility that we missed eligible studies could exist. However, we consider this risk to be minimal due to our systematic search strategy and manual backwards citation tracking. In addition, the large number of included studies in this review lessens the impact that missed studies might have on our overall conclusions. Next, it is reasonable to question the inclusion of mean/median warfarin dose as a true measure of VKA control, since unlike other measures, it does not consider INR values. However, we opted to include it as a measure in order to stay consistent with the methods of the prior review by Fitzmaurice et al.15 Finally, the possibility that journal word limits may have played some role in the under-reporting of VKA control measures should be considered.

Conclusions

VKA studies lack consistency in the types and combinations of control measures reported. A trend towards studies reporting greater numbers of VKA control measures over time was observed over our review time horizon, particularly, with AF and observational studies. The findings of this systematic review should be taken into consideration by researchers when performing future work in this area.