Long term follow up of direct oral anticoagulants and warfarin therapy on stroke, with all-cause mortality as a competing risk, in people with atrial fibrillation: Sentinel network database study.

BACKGROUND: We investigated differences in risk of stroke, with all-cause mortality as a competing risk, in people newly diagnosed with atrial fibrillation (AF) who were commenced on either direct oral anticoagulants (DOACs) or warfarin treatment. METHODS AND
RESULTS: We conducted a retrospective cohort study of the Oxford Royal College of General Practitioners (RCGP) Research and Surveillance Centre (RSC) database (a network of 500 English general practices). We compared long term exposure to DOAC (n = 5,168) and warfarin (n = 7,451) in new cases of AF not previously treated with oral anticoagulants. Analyses included: survival analysis, estimating cause specific hazard ratios (CSHR), Fine-Gray analysis for factors affecting cumulative incidence of events occurring over time and a cumulative risk regression with time varying effects.We found no difference in CSHR between stroke 1.08 (0.72-1.63, p = 0.69) and all-cause mortality 0.93 (0.81-1.08, p = 0.37), or between the anticoagulant groups. Fine-Gray analysis produced similar results 1.07 (0.71-1.6 p = 0.75) for stroke and 0.93 (0.8-1.07, p = 0.3) mortality. The cumulative risk of mortality with DOAC was significantly elevated in early follow-up (67 days), with cumulative risk decreasing until 1,537 days and all-cause mortality risk significantly decreased coefficient estimate:: -0.23 (-0.38-0.01, p = 0.001); which persisted over seven years of follow-up.
CONCLUSIONS: In this large, contemporary, real world primary care study with longer follow-up, we found no overall difference in the hazard of stroke between warfarin and DOAC treatment for AF. However, there was a significant time-varying effect between anti-coagulant regimen on all-cause mortality, with DOACs showing better survival. This is a key methodological observation for future follow-up studies, and reassuring for patients and health care professionals for longer duration of therapy.

Introduction

Atrial fibrillation (AF), for all but low risk patients, is managed using anticoagulants, either direct oral anticoagulants (DOACs) or warfarin, to reduce the risk of stroke [1, 2]. DOACs are increasingly prescribed instead of warfarin in routine clinical practice, due to their fixed dosing, rapid onset, fewer dietary and drug interactions and no requirement of haematological monitoring [3-6]. However, an increased risk of mortality with the use of DOACs compared with warfarin has been reported in a recent large observational study in primary care [7].

Longer term data on stroke and mortality with DOAC use are important because meta-analyses of randomised controlled trials (RCTs) and subsequent studies are not clear about their relative safety. Meta-analysis of DOAC RCTs have shown either non-inferiority or benefit compared to warfarin therapy in the prevention of stroke in patients with non-valvular AF [2, 8–11]. However, many RCTs assessed an individual DOAC drug and importantly had limited follow-up periods of up to three years [12]. In subsequent prospective and retrospective real world studies, some over a slightly longer follow-up period, there were mixed results with respect to a greater reduction in all-cause mortality compared to ischaemic stroke (IS) [13-Am J Cardiol.. 2018 ">15].

The aims of the current study to evaluate with a longer duration of follow-up for any differences between DOAC and warfarin use in newly diagnosed cases of AF on stroke and all-cause mortality. Additionally, to highlight methodological issues in assessing outcomes over a longer follow-up period, we flagged stroke and all-cause mortality and as competing outcomes.

Methods

Overview

We identified 12,619 patients with incident AF between 1st January 2008 - 31st July 2019 in the Oxford Royal College of General Practitioners (RCGP) Research and Surveillance Centre (RSC) network database, a nationally representative sample of 3.5 million people [16]. We followed up cases treated continuously with a single anticoagulant treatment type (either DOAC or warfarin), until an event of interest occurred. Drug choice were made according to local guidelines or NICE recommendations. The primary outcome event of stroke or all-cause mortality was treated as a competing risk in order to avoid potential biasing of risk estimates when mortality is treated as a censoring event [17]. We utilised an incident user design in order to reduce confounding, in particular survivor bias [18].

Population

The RCGP RSC primary care network is one Europe’s oldest sentinel systems collecting data from its member general practices for over 50 years [19]. Historically it has largely been involved in the monitoring of infectious disease [19, 20] but since 2015 has been involved in a wider range of research including AF, [21] diabetes, [22] and cancer [23]. UK primary care lends itself to this type of research because: (1) It is a registration-based system, each citizen registers with a single practice, (2) Individuals have a unique healthcare number (NHS number) that links to a “deaths and leavers table” that ensures reliability of years of exposure and all-cause mortality, (3) Computerised medical record (CMR) systems have been in place since the 1990s, [24] and (4) Prescribing is electronic, ensuring prescription numbers are accurate and it is therefore possible to measure persistence [25].

Case ascertainment

We used clinical codes in primary care CMR systems, in the UK Read codes, [26] to identify cases of AF. Whilst key conditions such as AF have been recorded well for some years, [27] pay-for-performance (P4P) for chronic disease management, in place since 2004, have further raised data quality [28]. We excluded patients (see S1 Fig in S1 File) who had had a previous stroke and those who were not on British National Formulary recommended dose of DOAC to reduce the impact of indication bias [29]. Censoring also took place if patients de-registered from a practice or the study period ended.

Exposures and outcomes

Exposures were to continuous anticoagulant prescription for the first time after receiving an AF diagnosis. Almost all anticoagulant prescriptions in the UK are issued from primary care and are, therefore, captured by the RCGP RSC database. Sometimes an anticoagulant is started in hospital; we measured exposure from the first GP prescription. We excluded patients (see S1 Fig in S1 File) who had an interval of greater than 90 days between anticoagulant prescriptions,

Outcomes were the first record of stroke and all-cause mortality, using previously published Read codes [30, 31]. Stroke was included regardless of aetiology, an approach used in other studies, [21, 32, 33] Study participants were followed for stroke and all-cause mortality up to 31 Jul 2019.

We found examples of patients receiving warfarin anti-coagulation with follow-up times exceeding seven years, longer than the DOAC cohort, for whom no examples exceed seven years of follow-up. We therefore truncated event times at seven years of follow-up, censoring events in the warfarin group that occurred after this length of follow-up.

Covariates

We included in our study clinical, variables likely to be used as indicators in prescribing anticoagulants: age-band, gender, and deprivation reporting index of multiple deprivation (IMD) quintile. IMD is a national measure of socioeconomic status which can be derived at individual level from first part of postcode, we divided IMD into five quintiles where Q1 is the most and Q5 the least deprived. We included comorbities from the CHADVASC score and we counted the number of such risks into a cumulative score. We also adjusted for comorbidities that form part of the stroke risk score (CHA2DS2-VASc), [34] namely heart failure, hypertension, stroke or transient ischaemic attack, myocardial infarction or peripheral vascular disease at baseline. We categorized this variable by the number of such comorbidities into low (< = 1), mild (2–3) and high (>4) in our model. We categorised smoking into current smoker, ex-smoker or never smoker.

To control for potential confounding by year of entry study, we included a binary covariate indicating year of entry into the study (before and including 2014 and after 2014) in all multivariate analyses. 2014 was also the year that national guidance was published and may have affected the quality of prescribing [35]. Uptake of this guidance was reported (using P4P data about AF management) to be around 94%, for stroke risk assessment, and for those at risk anticoagulants being offered to 78% in 2017 rising to 86% in 2019 [36].

Finally, we reported renal function, using estimated glomerular filtration rate (eGFR). Whilst creatinine clearance is the recommended measure, [37] this is rarely calculated in primary care records and contemporary eGFR data was available for nearly all patients.

Statistical methods

We studied the influence of anti-coagulation regimen by evaluating the cause-specific hazard ratio and the subdistribution hazard ratios of both events. Additionally, we estimated cumulative incidence for both events by direct regression, utilising the inverse of the probability of censoring weights (IPCW) method [15] with time-varying effects [38].

As recommended, [39] we report and interpret both cause specific and subdistribution analyses.

The cause specific hazard ratio (CSHR) is often interpreted as estimating aetiological association, estimating associations between covariates and the rate at which events occur in those subjects who are event-free. The cause‐specific hazard ratio can be interpreted as a rate ratio. Cox proportional hazards models are employed to estimate such hazards for both events.

The subdistribution hazard ratio, evaluated by the Fine-Gray (FG) methodology, may be thought of as a measure of prognostic association, summarising predictive relationships. In one interpretation, the exponentiated regression coefficient from a subdistribution hazards model indicates the relative effect of a covariate on the instantaneous rate of occurrence in subjects who are either event‐free or who have experienced a competing event. This may be an unpalatable interpretation for many as it includes subjects who have experienced the competing event and are unable to suffer the primary event. Their inclusion in the risk set after the competing event is therefore immortal time [40]. In short, it is not possible to interpret a subdistribution hazard as an epidemiological rate.

Cumulative incidence models were tested by a Kolmogorov-Smirnov type test-statistic and a Cramer von Mises type test-statistic (see S2 Table in S1 File) as well as by inspection of the Schoenfeld residuals. Such tests revealed time-varying effects for the anti-coagulation regimen for the cause-specific hazard and sub-distribution hazards analysis with respect to all-cause mortality. Despite the time-varying nature of some covariates and therefore the non-proportionality of hazards we report the CSHR and Fine-Gray analysis but interpret the hazards as time-averaged effects [40].

We carried out a sensitivity analysis on a propensity score matched (1–1) cohort after multiply imputing missing covariate values by the chained equations method. In addition to the covariates above, we included ethnicity and urban-rural and matched on these characteristics (likely to be associated with anti-coagulation prescription) using a propensity score derived from a multivariate model to perform survival analysis on the full cohort (i.e. including those with missing demographic status variables)

All analyses employed the statistical software R, version 3.5.3, additionally using the R libraries: survival, version 3.1–8, cmprsk, version 2.2–9 for estimating cause-specific hazard ratios (for right censored data and large samples), and riskRegression version 2019.11.3 for the subdistribution hazards estimation and the binomial cumulative incidence regression. In the sensitivity analysis, we used the mice library, version 3.7.0, for the imputation and the MatchIt library, version 3.0.2 for the propensity score matching.

Ethical considerations

Study approval was granted by the Research Committee of the RCGP RSC. The study did not meet the requirements for formal ethics board review as defined using the NHS Health Research Authority research decision tool (). The study was conducted in line with the Reporting of studies conducted using observational routinely collected data (RECORD) guidelines; [41] the cohort study diagram is included in S1 Fig of

Results

Baseline characteristics of study cohort

The incidence of AF slowly increased over the study period from 2.11 per 1,000 in 2008, to 2.99 in 2018. The incidence was consistently higher in men than women, the overall incidence rates for the period were 2.98 and 2.55/1,000 respectively. Men were generally diagnosed a decade younger (mean age 67.3) than women (mean age 73 years) over the observation period (The baseline characteristics of the study cohort with atrial fibrillation treated with either Warfarin or DOAC are shown in Table 1.

Table 1

Baseline characteristics of study cohort (n, %) showing the probability of any differences between event free, stroke and all-cause mortality groups between those exposed to Warfarin or DOACs.

		Warfarin (n = 7451)			DOAC (n = 5168)			Difference in proportion (p) comparing columns
	Status	Event Free	stroke	All-cause mortality	Event Free	stroke	All-cause mortality
Sex	Female	2561 (41.90)	70 (45.20)	493 (41.30)	2088 (44.20)	31 (54.40)	178 (46.20)
	Male	3542 (58.00)	85 (54.80)	700 (58.70)	2638 (55.80)	26 (45.60)	207 (53.80)	<0.01	0.30	0.10
Age Band	< = 65 years	1148 (18.80)	14 (9.00)	72 (6.0)	750 (15.90)	4 (7.00)	25 (6.50)
	65–75 years	2025 (33.20)	40 (25.80)	269 (22.50)	1756 (37.20)	14 (24.60)	85 (22.10)
	>75 years	2930 (48.10)	101 (65.20)	852 (71.40)	2220 (46.90)	39 (68.40)	275 (71.40)	<0.00	0.86	0.94
IMD Quintile (Q1 Least deprived) (Q5 Most deprived)	Q1	699 (11.50)	12 (7.70)	140 (11.70)	451 (9.50)	8 (14.00)	51 (13.20)
	Q2	850 (13.90)	19 (12.30)	192 (16.10)	630 (13.30)	9 (15.80)	53 (13.80)
	Q3	1357 (22.20)	32 (20.60)	291 (24.40)	1021 (21.60)	14 (24.60)	88 (22.90)
	Q4	1557 (25.50)	40 (25.80)	321 (26.90)	1306 (27.60)	17 (29.80)	105 (27.30)
	Q5	1640 (26.90)	52 (33.50)	249 (20.80)	1318 (27.90)	9 (15.80)	88 (22.90)	0.00	0.12	0.66
Comorbidities	< = 3	5355 (87.70)	126 (81.30)	987 (82.70)	4340 (91.80)	50 (87.70)	335 (87.00)
	>3	748 (12.30)	29 (18.70)	206 (17.30)	386 (8.20)	7 (12.30)	50 (13.0)	<0.00	0.37	0.06
Smoking Status	Active Smoker	666 (10.90)	18 (11.60)	144 (12.10)	412 (8.90)	6 (10.50)	40 (10.4)
	Ex-Smoker	3637 (59.60)	94 (60.60)	747 (62.60)	2936 (62.10)	32 (56.10)	250 (64.90)
	Never-Smoker	1800 (29.50)	43 (27.70)	302 (25.30)	1378 (29.20)	19 (33.30)	97 (24.70)	<0.00	0.73	0.62
Year of Study Entry	2014 or before	25067 (79.70)	995 (88.20)	3705 (86.60)	367 (5.60)	15 (9.80)	42 (8.80)
	After 2014	6367 (20.30)	133 (11.80)	574 (13.40)	6139 (94.40)	138 (90.20)	435 (91.20)	<0.00	<0.00	<0.00
eGFR *		71.8 (57.30–84.70)	65.3 (50.90–80.10)	63.9 (49.10–79.50)	77.6 (64.40–87.70)	71.0 (51.90–77.60)	70.6 (55.40–83.30)	<0.00	0.97	<0.00

DOAC, direct oral anticoagulant; IMD, index of multiple deprivation; eGFR, estimated glomerular filtration rate

* median and inter-quartile range

Unadjusted rates of stroke and all-cause mortality

The crude incidence rates for stroke and all-cause mortality were 0.59 (0.52–0.67) and 4.39 (4.2–4.6) per 100 person years, respectively (Table 2, Fig 1). There were no differences in the crude incident rates of stroke between warfarin 0.59 (0.51–0.69) and DOAC 0.58 (0.44–0.75) however there were for all-cause mortality: warfarin prescription was associated with a higher incidence 4.56 (4.33–4.81) compared with DOACs 3.89 (3.52–4.29, p<0001, Table 2, Fig 2).

Table 2

Crude incidence rates of stroke and all-cause mortality.

	Event	Events	Person years at risk	Incident rates /100 person years (95% Cl)
Warfarin	Stroke	171	28,878.22	0.59 (0.51–0.69)
	all-cause mortality	1317		4.56 (4.33–4.81)
DOAC	Stroke	57	9905.04	0.58 (0.44–0.75)
	all-cause mortality	385		3.89 (3.52,4.29)

Fig 1

Unadjusted cumulative incidence of stroke and all-cause mortality.

Fig 2

Unadjusted cumulative incidence of stroke and all-cause mortality by type of anti-coagulant.

Overall test of curve separation p<0.0001 for both outcomes.

Cause specific hazard ratios: Multivariate analysis

Among subjects who have not experienced any event, multivariate analysis suggested that the instantaneous rate of occurrence of stroke was associated with age over 75 years, CSHR 2.41 (1.44–4.00, p<0.001), eGFR, CSHR 0.61(0.41–4.00, p = 0.02) and five or more CHA2DS2-VASc listed comorbidities CSHR 2.32 (1.27–4.20, p = 0.01), Table 3, S2 Table in S1 File

Table 3

Multivariate analysis of the cause-specific hazard ratio (CSHR) for stroke.

Variable		Ref	HR	95% CI		p
Anticoagulation type	DOAC	Warfarin	1.08	0.72	1.63	0.67
Gender	Male	Female	0.86	0.65	1.13	0.28
Age band	>65 -< = 75yrs	≤65yrs	1.37	0.81	2.35	0.25
	>75yrs		2.41	1.44	4.00	<0.001
IMD Quintile	Q2	Q1	1.07	0.61	1.89	0.82
	Q3		1.08	0.64	1.83	0.77
	Q4		1.17	0.70	1.95	0.55
	Q5		1.20	0.72	1.99	0.48
log GFR			0.61	0.41	0.92	0.02
Comorbidities	2–4	≤1	1.10	0.77	1.57	0.60
(from CHA2DS2-VASc)	> = 5		2.32	1.27	4.20	0.01
Baseline smoking status	Ex-smoker	Smoker	0.77	0.50	1.20	0.29
	Never		0.78	0.48	1.25	0.30
Year of Entry	after 2014	≤2014	0.91	0.62	1.34	0.64

Ref, reference; HR, hazard ratio; CI, confidence interval; IMD, index of multiple deprivation; GFR, glomerular filtration rate; CHA2DS2-VASc, stroke risk score

The all-cause mortality rate was similarly associated with five or more CHA2DS2-VASc listed comorbidities CSHR 1.34 (1.05–1.71, p = 0.02), with age over 75years CSHR 4.12 (3.32–5.1, p<0.001) as well as age between 65 and 75 years CSHR 1.77 (1.41–2.22, p<0.001). Further instantaneous occurrence was associated with year of entry after 2014 CSHR 1.40 (1.22–1.61, p<0.001) and being in deprivation quintile 5 (the least deprived) CSHR 0.65 (0.54–0.77, p<0.001) in addition to never have smoked CSHR 0.65 (0.55–0.78, p<0.001), Table 4,

Table 4

Multivariate analysis of the cause-specific hazard ratio (CSHR) for all-cause mortality, interpreted as time-averaged effects.

Variable		Ref	HR	95% CI		p
Anticoagulation type	DOAC	warfarin	0.93	0.81	1.08	0.37
Gender	Male	Female	1.10	0.99	1.22	0.07
Age band	>65 -< = 75yrs	≤65yrs	1.77	1.41	2.22	<0.00
	>75yrs		4.12	3.32	5.10	<0.00
IMD Quintile	Q2	Q1	0.96	0.79	1.15	0.60
	Q3		0.88	0.74	1.05	0.16
	Q4		0.87	0.73	1.03	0.11
	Q5		0.65	0.54	0.77	<0.00
log GFR			0.52	0.45	0.60	<0.00
Comorbidities	2–4	≤1	0.96	0.85	1.10	0.55
(from CHA2DS2-VASc)	> = 5		1.34	1.05	1.71	0.02
Baseline smoking status	Ex-smoker	Smoker	0.79	0.67	0.92	0.00
	Never		0.65	0.55	0.78	<0.00
Year of Entry	after 2014	≤2014	1.40	1.22	1.61	<0.00

Ref, reference; HR, hazard ratio; CI, confidence interval; IMD, index of multiple deprivation; GFR, glomerular filtration rate; CHA2DS2-VASc, stroke risk score

Comparison between subdistribution hazards and cause-specific hazard ratios

There was no difference in the cause-specific hazard ratio for stroke 1.08 (0.72–1.63, p = 0.69) or in all-cause mortality 0.93 (0.81–1.08, p = 0.37); comparing the warfarin (reference) group with the DOAC group. A Fine-Gray analysis produced very similar adjusted hazard ratios of 1.07 (0.71–1.6 p = 0.75) for stroke and 0.93 (0.8–1.07, p = 0.3) for mortality. These initial results suggest no difference between anti-coagulation by warfarin or by DOACs from a time averaging approach. However, whilst there was no time-varying effect for stroke, the proportional hazards assumption was not violated (p = 0.98, using Schoenfeld residuals), there was for mortality (p<0.01, Table 5).

Table 5

Summary of multivariate adjusted cause-specific hazards and the subdistribution hazards for comparison.

	Stroke	95% CI	All-cause mortality	95% CI
CSHR: DOAC -v-warfarin	1.08	0.72–1.63	0.93	0.81–1.08
Sub-distribution HR: DOAC -v- warfarin	1.07	0.71–1.60	0.93	0.80–1.08

CI, confidence interval; CSHR, cause specific hazard ratio; DOAC, direct oral anticoagulant; IS, ischaemic stroke

Risk-regression to elucidate the time varying effect

The time-varying nature of the exposure was explored by investigation of the covariate effect by risk regression (using a proportional link function), estimating sub-distribution hazards ratios with time-varying effects for anti-coagulation regimen. This allowed us to plot the time-varying effect of the anti-coagulation regime. The time-varying effect is non-significant for ischaemic stroke (Fig 3).

Fig 3

Time-Varying effect of anti-coagulation regimen on sub-distribution hazards for stroke.

The effect of anti-coagulation regimen on mortality was found to be significantly time-varying: in early follow-up, around 67 days, a significantly elevated cumulative risk of mortality was present in the cohort of people prescribed a DOAC rather than warfarin. Subsequently the DOAC cumulative risk decreases until 1,537 days of follow-up when there is a significant decrease in the risk of all-cause mortality in the DOAC group: coefficient estimate: -0.23 (-0.38, -0.01, p = 0.01, Fig 4); this decreased risk persists for the follow-up period investigated in this study. In S3 Table in S1 File, we show subgroup examples.

Fig 4

Time-varying effect of anti-coagulation regime on sub-distribution hazards for all-cause mortality.

For completeness we note that in early follow-up there is a statistically significantly elevated cumulative incidence of all-cause mortality in the DOAC group that lasts until around 1220 days (see S3 Table in S1 File). By 1537 days this coefficient estimate has become negative and decreases monotonically until end of the follow-up period.

Sensitivity analysis

The sensitivity analysis (see S1 Table in S1 File) supports the main findings: no statistically significant differences between anti-coagulation regimen and the risk of stroke and all-cause mortality are found in the cause-specific hazard ratios and the cumulative incidence regression. Furthermore, tests reveal that the proportional hazards assumption is violated for both mortality models with evidence of the time-varying nature of anti-coagulation group present in the matched analysis.

Discussion

Principal findings

This real world study, showed no difference between DOAC and warfarin treatment with respect to stroke, but did show a reduction in all-cause mortality with DOACs, with a key observation of longer follow-up compared to previous RCTs and real world studies. Importantly, the difference in favour of DOAC benefit only emerged at 1,537 days (around 4.2 years) into treatment. Given that the mean age of diagnosis of AF in men was 67 years and for women 73 years, these long-term benefits are likely to be of clinical significance and highlights the importance of staying on treatment for an extended period.

These finding are reassuring for both patients and prescribers, as the largest and previous large real world study, albeit over a shorter follow-up, had indicated an increase in all-cause mortality with DOAC use [7].

The DOAC group was associated with elevated mortality in early follow-up. However, over time, this effect significantly lowered over the warfarin group until end of follow-up at seven years. This may have been due to a ‘learning curve’ for general practice in using this new group of medicines (DOACs). We should interpret this as a population effect on the cumulative incidence rather than an individual risk effect and therefore is possibly due to a selection bias based on individual frailties, with the higher mortality in higher risk patients in the DOAC group which leaves leave lower risk patients, resulting in a decreasing relative cumulative incidence.

Our approach has highlighted the importance of competing risks when analysing prospective data with interlinked or interdependent clinical endpoints.

The lack of evidence of treatment effect on the cumulative incidence of stroke between people prescribed DOAC and warfarin was further underlined by significant variation over time in the exposure effect on the cumulative incidence all-cause mortality

Strengths and limitations

The Oxford RCGP RSC network comprises a large nationally representative sample of people attending general practice throughout England. UK general practice lends itself to this type of research because it is a registration based system, providing an accurate denominator. General Practitioners have been recording data about AF for many years [42] with a high level of data completeness, [16] which has enabled the network to be an important resource for real world evidence-based research [43]. However, there is an element of selection bias since practices volunteered to join the network, with a marginal increase in affluent areas than the national population as a whole. Practices within the network have access to dashboard to improve data quality and the quality of care [44].

Financial incentives provided in 2014 in English general practice maybe important and a useful tool for other health systems wishing to promote more recognition and anticoagulation in AF [36, 37]. This change in anticoagulant use enabled us to compare DOACs and warfarin (Fig 5).

Fig 5

Change in use of Warfarin and DOACs over time, financial incentives to encourage increased detection in the management of AF introduced in 2014.

A strength of this study was the focus on a rigorous methodological approach with regards to appropriate statistical analyses. The cause-specific hazards and the subdistribution hazards are summarised for comparison in Table 5. The concordance between the adjusted cause-specific and sub-distribution hazards ratios is to be expected since the cumulative incidence of stroke is small compared to the incidence of all-cause mortality [27].

We found a non-significant exposure effect on the cumulative incidence of stroke, which further underlined the lack of evidence in variation in time to stroke between people prescribed DOAC and warfarin. However, there we a significant variation over time in the exposure effect on the cumulative incidence all-cause mortality.

Possible selection bias may also limit our study as some practices may use one of the medicines more than another according to specific clinical characteristics, limitations or risks. We did not utilise the CHA2DS2VASc score in our analysis due to the potential problem of calculator implementation on GP computer system already highlighted in previous publications [45, 46]. In future studies we could consider data linkage to hospital and registry data be able to more precisely differentiate ischaemic from haemorrhagic stroke. This may also allow us to identify patients who have suffered haemorrhages, and those that required hospital admission.

Further research

An additional study over a longer period of time and using the enlarged Oxford RCG RSC network would be supported by the enthusiasm among general practice colleagues during-pandemic research which has grown our network to over 1,600 practices [47]. Further research should explore precise stratification of the population by CHADVASC score, risks of haemorrhage identify causes of mortality and level of frailty and evaluate if there is any differential effect between drugs in class It is also possible that with longer follow-up there might be an emergent difference in stroke.

Conclusions

In this real world study with a longer follow-up compared to previous studies, we found no difference between DOAC and warfarin treatment for atrial fibrillation with respect to stroke reduction. Taking into account that stroke and mortality are competing endpoints, we found a significant time-varying effect for specific anti-coagulation drug on all-cause mortality. People prescribed DOACs had elevated mortality in early follow-up, however over time, this was significantly lowered compared with warfarin right through until end of follow-up at seven years. This is a key methodological observation for future follow-up studies, but additionally reassuring from a therapeutic viewpoint for patients and health care professionals for long duration of therapy

(DOCX)

Click here for additional data file.

Transfer Alert

This paper was transferred from another journal. As a result, its full editorial history (including decision letters, peer reviews and author responses) may not be present.

18 May 2021

PONE-D-21-07035

Long term follow up of direct oral anticoagulants and warfarin therapy on stroke, with all - cause mortality as a competing risk, in people with atrial fibrillation: sentinel network database study.

PLOS ONE

Dear Dr. de Lusignan,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Jun 25 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see:  http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at  https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols .

We look forward to receiving your revised manuscript.

Kind regards,

Michael Nagler, M.D., Ph.D., MSc

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. In ethics statement in the manuscript and in the online submission form, please provide additional information about the patient records/samples used in your retrospective study. Specifically, please ensure that you have discussed whether all data/samples were fully anonymized before you accessed them and/or whether the IRB or ethics committee waived the requirement for informed consent. If patients provided informed written consent to have data/samples from their medical records used in research, please include this information.

3.Thank you for stating the following in the Acknowledgments Section of your manuscript:

"FDRH acknowledges part-funding from the National Institute for Health Research (NIHR)

428 School for Primary Care Research, the NIHR Collaboration for Leadership in Health Research and

429 Care (CLARHC) Oxford, the NIHR Oxford Biomedical Research Centre (BRC, UHT), and the NIHR

430 Oxford Medtech and In-Vitro Diagnostics Co-operative (MIC)."

We note that you have provided funding information that is not currently declared in your Funding Statement. However, funding information should not appear in the Acknowledgments section or other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form.

Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows:

"The work was supported by an unconditional grant to SdeL. Grant number DSJP3700 by Daiichi Sankyo Limited. URL: https://www.daiichi-sankyo.co.uk.The funder had no role in  study design, data collection and analysis, decision to publish, or preparation of the manuscript"

Please include your amended statements within your cover letter; we will change the online submission form on your behalf.

4.Thank you for stating the following in the Competing Interests section:

"I have read the journal's policy and the author of this manuscript has the following competing interest: Simon de Lusignan is the Director of the Oxford RCGP RSC and has received funding through his University for studies from Eli Lilly, Astra-Zeneca, Sanofi, GSK, Seqirus and Takeda; and been member of advisory boards for influenza for Seqirus and Sanofi. FDRH has received occasional fees from Bayer and Boehringer Ingelheim for speaking or consulting on atrial fibrillation related stroke risk. All other authors have declared no financial relationships with any organisations that might have an interest in the submitted work in the previous three years, no other relationships or activities that could appear to have influenced the submitted work."

Please confirm that this does not alter your adherence to all PLOS ONE policies on sharing data and materials, by including the following statement: "This does not alter our adherence to  PLOS ONE policies on sharing data and materials.” (as detailed online in our guide for authors http://journals.plos.org/plosone/s/competing-interests).  If there are restrictions on sharing of data and/or materials, please state these. Please note that we cannot proceed with consideration of your article until this information has been declared.

Please include your updated Competing Interests statement in your cover letter; we will change the online submission form on your behalf.

Please know it is PLOS ONE policy for corresponding authors to declare, on behalf of all authors, all potential competing interests for the purposes of transparency. PLOS defines a competing interest as anything that interferes with, or could reasonably be perceived as interfering with, the full and objective presentation, peer review, editorial decision-making, or publication of research or non-research articles submitted to one of the journals. Competing interests can be financial or non-financial, professional, or personal. Competing interests can arise in relationship to an organization or another person. Please follow this link to our website for more details on competing interests: http://journals.plos.org/plosone/s/competing-interests

5. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

6.We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For more information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially sensitive information, data are owned by a third-party organization, etc.) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: No

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The authors conducted a large study investigating mortality and stroke in patients with newly diagnosed atrial fibrillation treated either with DOAC or warfarin. Routinely obtained data of the RCGP database were used. This is an interesting study. I have three major comments which should be addressed to improve the manuscript before publication.

1. While reading the manuscript, it remains unclear why the study was conducted and what the results mean (in the context of previous literature). The clinical problem or the unsolved scientific question leading to this study is not clearly stated. Thus, it is also not clear how to interpret the results.

2. The authors used elaborated statistical techniques to analyze the data. However, it is not explained how these techniques help to answer the research questions raised. This issue increases the problem of how to interpret the results of this study. Besides, please reword the results (results section, table headings, figure legends) in a way that the results answer the research question.

3. The authors used a database of routinely obtained data to do the analysis. A large number of pitfalls are possible in this kind of study that might lead to biased results or wrong interpretations. The authors already spend some efforts to convince the reader that the results are valid. But I strongly believe that the authors must give more details how the outcomes (stroke, death), and co-variates are recorded and how completeness of these measurements is ensured. Besides, incomplete data are a major risk of these kind of studies and the authors must elaborate on this in the limitations part of the manuscript.

4. The authors did not discuss the results in context of previous literature. What did other studies find? Why are the results different? It is the methodology, I guess. Please add a comprehensive paragraph on this issue referring to all key studies comparing the risk of stroke and mortality between DOAC and Warfarin in patients with atrial fibrillation.

Some detailed comments as examples:

1. Title: What is a ‘sentinel network database study’? Please describe early in the manuscript or even in the abstract.

2. Abstract: It appears that the abstract is not formatted according to PLOS one.

3. Abstract/ Background: Please describe the clinical problem or the unresolved scientific question that leads to the present study.

4. Abstract/ Methods: The population is not fully described. Please add more information on the inclusion criteria.

5. Abstract/ Results: You state the statistical methods and results, but it is hard to follow what this means in terms of the research question. I suggest rewording it in a way that a reader can follow who is less familiar with statistical concepts.

6. Abstract/ Results: what does long-term mean? Please state the observation time.

7. Abstract/ Results: Patient characteristics are not given. Please state at least age and sex.

8. Abstract/ Results: Outcomes (incidence rates of stroke, death are not given)

9. Abstract/Conclusions: You did not find a difference in hazards of stroke between DOAC and Warfarin. This is in contrast to a number of previous studies. Why? It’s the methodology I guess. Please clarify this already in the abstract.

Reviewer #2: In this study, the authors analyzed a retrospective cohort of 12619 primary care patients with a first-time diagnosis of atrial fibrillation who were treated with either DOAC or warfarin and performed a competing risk analysis between stroke and all-cause mortality. They conclude that there is no significant difference of hazard between anticoagulants for stroke but all-cause mortality DOACs performed better. While this study is certainly of interest some points need to be addressed before publication:

Major Comments:

(Page 4, Line 3): It is appreciated that the authors thoroughly describe all analysis, however, the background section of the abstract is lacking a sentence about the context of the study (Why did you perform this study?)

(Page 5, Lines 54-57): While reading this part of the introduction, I did not clearly understand the authors' research question. The authors might want to change the wording here

(Page 6,) Thank you for describing the database in such detail. However, some information would still improve the understandability. Are comorbidities entered in the database by the physician or derived from the UK read codes? Is the smoking status also recorded by the physician?

(Page 8): While the statistical analysis is described in-depth, I still wonder how you handled missing data for the primary analysis. You write that you performed a sensitivity analysis with multiple imputations, but did you perform a complete-case analysis for the primary analysis? Please consider describing this more in detail.

(Page 10): The authors might want to include the median observation time of the two groups (DOAC vs. Warfarin).

(Figure 2): I am not sure if the figure legend here is correct. The legend text suggests that there should be more than one curve (e.g. “curve separation”) but the figure just shows one curve.

Minor comments:

(Page 5, Line 54-57) Although I am not a native English speaker, the first sentence seems to miss a verb. The second sentence seems to have an “and” after all-cause mortality too much.

(Page 7): The authors might want to consider moving the flow chart from Figure S1 from the supplementary material to the main article. In my opinion, it would improve readability.

(Page 7, Line 90): A small typo. I think the “to” is too much here.

(Page 8 Lines 109-113): The authors might want to revise the wording in these sentences. It remains unclear to me.

(Table 1): You present here an expansive table. In my opinion, including the number of missing values per covariate would further improve the table. Also, the third column (“Difference in proportion”) is quite hard to understand. You might want to write in the table legend that the p-values are for the comparison of Warfarin to DOAC patients.

(Page 15, Line 238): Probably just a typo but the upper confidence interval of the eGFR is according to table 3 0.92, not 4.00.

(Figure 3 & Figure 4): It might be helpful to show the sub-distribution in one figure to make the graphs easier to compare.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Michael Nagler

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

15 Feb 2022

Comments from Reviewer #1

1. While reading the manuscript, it remains unclear why the study was conducted and what the results mean (in the context of previous literature). The clinical problem or the unsolved scientific question leading to this study is not clearly stated. Thus, it is also not clear how to interpret the results.

Response

For clarity we have re-phrased the aims of the study, conforming to the reviewer's comments. We hope that this aids the interpretation of the statistical results.

2. The authors used elaborated statistical techniques to analyze the data. However, it is not explained how these techniques help to answer the research questions raised. This issue increases the problem of how to interpret the results of this study. Besides, please reword the results (results section, table headings, figure legends) in a way that the results answer the research questions.

Response

A senior statistician was part of the research team and conducted an appropriate statistical analysis for a time to event analysis, with outcome stroke (and a competing risk of all-cause mortality). In the light of the updated research aim, to provide real-world evidence re stroke outcomes, we anticipate that the results will be easier to interpret.

3. The authors used a database of routinely obtained data to do the analysis. A large number of pitfalls are possible in this kind of study that might lead to biased results or wrong interpretations. The authors already spend some efforts to convince the reader that the results are valid. But I strongly believe that the authors must give more details how the outcomes (stroke, death), and co-variates are recorded and how completeness of these measurements is ensured. Besides, incomplete data are a major risk of these kind of studies and the authors must elaborate on this in the limitations part of the manuscript.

Response

We have amended the "Case ascertainment section": We refer the reviewer to lines 88-95 in revised m/s

4. The authors did not discuss the results in context of previous literature. What did other studies find? Why are the results different? It is the methodology, I guess. Please add a comprehensive paragraph on this issue referring to all key studies comparing the risk of stroke and mortality between DOAC and Warfarin in patients with atrial fibrillation.

Response

We refer the reviewer to the m/s lines 47-54 and references 4-7.

Reviewer #1: Detailed Comments

4. Abstract/ Methods: The population is not fully described. Please add more information on the inclusion criteria.

Response

We would refer the reviewer to the para. 2 of the introduction (beginning line 45) for discussion around comparison with previous studies.

1. Title: What is a ‘sentinel network database study’? Please describe early in the manuscript or even in the abstract.

Response

We have updated the abstract with a brief explanation of a sentinel primary care network.

2. Abstract: It appears that the abstract is not formatted according to PLOS

Response

As per PLOS ONE requirements we now have the sections: Title (affiliations), Abstract, Introduction, followed by the Middle section, etc. The original Methods and Conclusions section have migrated to the Middle section. We have also updated formatting.

3. Abstract/ Background: Please describe the clinical problem or the unresolved scientific question that leads to the present study

Response

We have updated the background section in the manuscript to address this issue.

4. Abstract/ Methods: The population is not fully described. Please add more information on the inclusion criteria

Response

See Supplementary Fig. S1 has been included in the main m/s.

5. Abstract/ Results: You state the statistical methods and results, but it is hard to follow what this means in terms of the research question. I suggest rewording it in a way that a reader can follow who is less familiar with statistical concepts

Response

Please see our response to Point 2 in the section preceding the Detailed comments section.

6. Abstract/ Results: what does long-term mean? Please state the observation time.-

Response

We have written this more clearly in the m/s that the follow-up time was up to 7 years.

7. Patient characteristics are not given. Please state at least age and sex

Response

Please refer to Table 1: Baseline characteristics of study cohort

8. Abstract/ Results: Outcomes (incidence rates of stroke, death are not given)

Response

Please refer to section: "Unadjusted rates of stroke and all-cause mortality"

9. Abstract/Conclusions: You did not find a difference in hazards of stroke between DOAC and Warfarin. This is in contrast to a number of previous studies. Why? It’s the methodology I guess. Please clarify this already in the abstract.

Response

In the Strengths and limitations section we have highlighted the possibility of selection bias re utilisation of drugs in primary care units and also the potential problem of fully recording the CHA2DS2VASc score during primary care consultations. RCTs also trial one drug. For these reasons it may be speculated that our results differ from those of RCTs.

Comments from Reviewer #2

Major Comments:

It is appreciated that the authors thoroughly describe all analysis, however, the background section of the abstract is lacking a sentence about the context of the study (Why did you perform this study?)

While reading this part of the introduction, I did not clearly understand the authors' research question. The authors might want to change the wording here.

Response

For our response to these comments we refer the reviewer to the updated abstract

Are comorbidities entered in the database by the physician or derived from the UK read codes? Is the smoking status also recorded by the physician?

READ codes are recorded by the primary care physician during all patient consultation. These codes are entered electronically during consultation and are then batch-uploaded to a central repository which then supplies the RCGP RSC database used in this study. Smoking is recorded by the physician.

(Page 8): While the statistical analysis is described in-depth, I still wonder how you handled missing data for the primary analysis. You write that you performed a sensitivity analysis with multiple imputations, but did you perform a complete-case analysis for the primary analysis? Please consider describing this more in detail.

Response

A compete cases analysis was considered as the primary mode of analysis. We have updated the manuscript to emphasise this point.

(Page 10): The authors might want to include the median observation time of the two groups (DOAC vs. Warfarin).

Fig. 2 what's gone wrong??

Response

We have inserted the correct Figure in the updated m/s.

missing verb:

Thank you for the comment: appropriately updated.

(Page 7): The authors might want to consider moving the flow chart from Figure S1 from the supplementary material to the main article. In my opinion, it would improve readability.

Response

Cohort diagram has been included in m/s as Figure 1 -other Fig.s have numbering updated.

(Page 7, Line 90): A small typo. I think the “to” is too much here.

Response

Deleted agreed

(Table 1): You present here an expansive table. In my opinion, including the number of missing values per covariate would further improve the table. Also, the third column (“Difference in proportion”) is quite hard to understand. You might want to write in the table legend that the p-values are for the comparison of Warfarin to DOAC patients.

Response

We have updated the bottom legend of this Table

(Page 15, Line 238): Probably just a typo but the upper confidence interval of the eGFR is according to table 3 0.92, not 4.00.

Response

Corrected - thank you.

IMD is a UK-nationally accepted index of SES; similarly, CHA2DS2-VASc is an international index of cardiovascular risk (ref 34 in m/s).

(Figure 3 & Figure 4): It might be helpful to show the sub-distribution in one figure to make the graphs easier to compare.

Response

We have relabelled the figures and trust that this suffices

14 Mar 2022

Long term follow up of direct oral anticoagulants and warfarin therapy on stroke, with all-cause mortality as a competing risk, in people with atrial fibrillation: sentinel network database study.

PONE-D-21-07035R1

Dear Dr. de Lusignan,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Michael Nagler, M.D., Ph.D., MSc

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

23 Aug 2022

PONE-D-21-07035R1

Long term follow up of direct oral anticoagulants and warfarin therapy on stroke, with all-cause mortality as a competing risk, in people with atrial fibrillation: sentinel network database study.

Dear Dr. de Lusignan:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Prof. Dr. Michael Nagler

%CORR_ED_EDITOR_ROLE%

PLOS ONE