Long-term trends in death and dependence after ischaemic strokes: A retrospective cohort study using the South London Stroke Register (SLSR).

BACKGROUND: There have been reductions in stroke mortality over recent decades, but estimates by aetiological subtypes are limited. This study estimates time trends in mortality and functional dependence by ischaemic stroke (IS) aetiological subtype over a 16-year period. METHODS AND
FINDINGS: The study population was 357,308 in 2011; 50.4% were males, 56% were white, and 25% were of black ethnic backgrounds. Population-based case ascertainment of stroke was conducted, and all participants who had their first-ever IS between 2000 and 2015 were identified. Further classification was concluded according to the underlying mechanism into large-artery atherosclerosis (LAA), cardio-embolism (CE), small-vessel occlusion (SVO), other determined aetiologies (OTH), and undetermined aetiologies (UND). Temporal trends in survival rates were examined using proportional-hazards survival modelling, adjusted for demography, prestroke risk factors, case mix variables, and processes of care. We carried out additional regression analyses to explore patterns in case-fatality rates (CFRs) at 30 days and 1 year and to explore whether these trends occurred at the expense of greater functional dependence (Barthel Index [BI] < 15) among survivors. A total of 3,128 patients with first-ever ISs were registered. The median age was 70.7 years; 50.9% were males; and 66.2% were white, 25.5% were black, and 8.3% were of other ethnic groups. Between 2000-2003 and 2012-2015, the adjusted overall mortality decreased by 24% (hazard ratio [HR] per year 0.976; 95% confidence interval [CI] 0.959-0.993). Mortality reductions were equally noted in both sexes and in the white and black populations but were only significant in CE strokes (HR per year 0.972; 95% CI 0.945‒0.998) and in patients aged ≥55 years (HR per year 0.975; 95% CI 0.959‒0.992). CFRs within 30 days and 1 year after an IS declined by 38% (rate ratio [RR] per year 0.962; 95% CI 0.941‒0.984) and 37% (RR per year 0.963; 95% CI 0.949‒0.976), respectively. Recent IS was independently associated with a 23% reduced risk of functional dependence at 3 months after onset (RR per year 0.983; 95% CI 0.968-0.998; p = 0.002 for trend). The study is limited by small number of events in certain subgroups (e.g., LAA), which could have led to insufficient power to detect significant trends.
CONCLUSIONS: Both mortality and 3-month functional dependence after IS decreased by an annual average of around 2.4% and 1.7%, respectively, during 2000‒2015. Such reductions were particularly evident in strokes of CE origins and in those aged ≥55 years.

Introduction

Stroke continues to be a major public health concern affecting more than 10 million people each year around the globe [1]. It is the second most common cause of death and the third leading cause of long-term disability. Each year, stroke accounts for a loss of approximately 1,484 disability-adjusted life years per 100,000 people [2]. Approximately one-quarter of patients die within 1 month of stroke onset [3,4], and half of the survivors are left dependent on others for everyday activities [5,6]. Around 87% of all strokes are attributed to ischaemia and require timely management in order to improve their outcomes [1,7]. Therefore, quality improvement initiatives have included providing universal access to organised stroke services; reducing time to computed tomography (CT) scan; increasing the proportion of patients receiving intravenous thrombolysis; reducing time between arrival to emergency department and receiving the treatment (door-to-needle [DTN]); and, finally, participation in quality improvement registries such as the Sentinel Stroke National Audit Programme (SSNAP) in the United Kingdom, the Get with the Guidelines (GWTG)-Stroke registry in the United States, and the Riks-Stroke registry in Sweden [8-11]. Whether overall survival among stroke patients has improved with these efforts remains inconclusive. However, data from a number of emergency medical services suggested higher rates of survival among patients with stroke because of reduced stroke severity and advances in stroke care during recent decades. Between 1990 and 2013, global stroke mortality has reportedly declined from 113 to 67 deaths per 100,000 people [1].

Survival is also known to vary between different sectors of the population. It is well-documented that older patients have greater risk of death than younger patients [12]. Also, studies from the UK and the US have shown that people of black ethnic origin have better survival advantage than their white counterparts [3,12,13]. Suggested hypotheses attributed such ethnic disparity to the differences in receipt of effective interventions or the effect of healthy migrants, because individuals willing and able to migrate are usually healthier [14]. However, studies describing how survival trends might have changed across time in each of these demographic groups are lacking.

A recent study in England has calculated the 30-day case-fatality rate (CFR) for all strokes and reported a decline from 41.8% to 26.4% in men and 44.1% to 28.5% in women between 2001 and 2010 [15]. SSNAP data indicated a reduction in CFR by approximately 50% between 1998 and 2017 [16]. Although large and nationally representative, these studies used administrative data of those with stroke and may have excluded patients treated in the community or included patients without stroke (e.g., transient ischaemic attacks [TIAs], stroke mimics). Other investigators have reported either a decline or no change in stroke death rates [3,17-23]. Again, all these studies tended to present stroke as one category (i.e., total stroke). In addition, they have limitations because of a lack of ethnic diversity, selective inclusion of patients admitted to the hospital, or restricting analyses to younger age groups.

To date, no population-based study has examined secular trends in survival after ischaemic stroke (IS) by aetiological subtype among age, sex, and ethnic groups. Studying such trends is important for monitoring purposes, which may uncover hidden patterns that can help guide health programs, public policy, and the allocation of health services and research funding. The aims of this study were, first, to use population-based data to estimate long-term secular trends in death after IS by subtype among different patient groups in South London between 2000 and 2015. The second aim of the study was to examine the changes in dependence among IS survivors over time because a decline in mortality may translate into an increase in the proportion of survivors with a functional deficit.

Methods

Data source

The South London Stroke Register (SLSR) is a large, prospective, population-based, registry of patients with first-ever strokes in a defined population of inner London. The design of the registry has been previously described in detail [24-26]. Briefly, a multiple overlapping surveillance system was established in 1995 to identify all stroke patients in a study area comprising 22 electoral wards in the north of 2 London boroughs: Lambeth and Southwark. According to 2011 census data from the Office for National Statistics (ONS), the SLSR covered an area of 357,308 residents (50.4% males); 56% were white, 25% were black (14% black African, 7% black Caribbean, and 4% other black), and 18% were of other ethnic backgrounds [27]. The demographic composition of the SLSR area has changed significantly over time: during 2000–2003, people aged <65 years composed 91.1% of the population; 50.1% were females, 62.3 were white, 28.8% were black, and 9% were other. The corresponding figures during 2012–2015 were 93.4%, 54.7%, 24.7%, and 20.5%, respectively.

Patients were identified at 5 London hospitals—2 within and 3 outside (but close to) the SLSR area—in order to enhance case ascertainment. Additional community cases were notified by regular contact with all general practitioners (GPs) within and in the borders of the study area [28]. Notification sources included the accident and emergency records, hospital wards, radiology records, death certificates, coroner records, hospital stroke registries, GP computer records, hospital medical staff, GPs and practice staff, community therapists, and bereavement officers [25]. Completeness of case ascertainment was previously estimated in our population using the indirect methods of capture recapture—approximately 80% (between 75% and 88%) [26,29,30].

All data were collected prospectively by specially trained nurses, doctors, and fieldworkers who vouch for the completeness and accuracy of the data. Whenever possible, patients were assessed within 48 hours of referral to the SLSR, and data were checked against the patients’ GP and medical records [26]. Stroke diagnosis follows the World Health Organisation criteria [31]. Pathological classification was based on neuroradiology (CT/MRI scans), CSF analysis, or autopsy results, which was further verified by a study clinician. Accordingly, patients were classified into cerebral infarction, primary intracerebral haemorrhage, or subarachnoid haemorrhage, whereas cases without pathological confirmation of subtype were undefined. Subtype classification of IS was carried out—using the Trial of ORG 10172 in Acute Stroke Treatment (TOAST) criteria [32]—into (1) large-artery atherosclerosis (LAA), (2) cardio-embolism (CE), (3) small-vessel occlusion (SVO), (4) other determined aetiologies (OTH), and (5) undetermined aetiologies (UND). The proportion of IS patients who received any brain scan increased from 95% in 2000–2003 to 100% in 2012–2015; MRI uptake increased from 14% to 35%.

Information collected at initial assessment included the following:

Demographic variables: age (calculated as the difference between the date of birth and stroke onset), sex, and self-definition of ethnic origin (1991 census question [25]; stratified into black [African, Caribbean, and other], white, and others [Asian, Pakistani, Indian, Bangladeshi, Chinese, and other])

Pre-existing risk factors: smoking (current versus quitter/never), alcohol intake (≥21 units/wk for men, ≥14 units/wk for women), hypertension (general practice or hospital records of systolic blood pressure > 140 mmHg or diastolic > 90 mmHg), diabetes mellitus, hypercholesterolemia (total cholesterol concentration ≥ 6 mmol/L), myocardial infarction, TIAs, and atrial fibrillation (AF) (general practice or hospital records)

Clinical impairment indicators: urinary incontinence, swallow test (3-oz water swallow test), Glasgow coma scale (GCS), Barthel Index (BI) within 7 days of onset, and the National Institutes of Health Stroke Scale (NIHSS)

Processes of care in the acute phase: hospital admission (none, stroke unit, or other medical wards), brain imaging (CT, MRI, or both), antiplatelets, anticoagulants, and thrombolysis treatment

To evaluate functional abilities at follow-ups, BI was collected by means of face-to-face visits or postal questionnaires at 3 months and 1 year after the first IS.

Informed consent and assent, when appropriate, were obtained from all participants or from a next of kin for the individuals who were too impaired to provide written consent. Ethical approval for the study was obtained from the ethics committees of Guy’s and St Thomas’ Hospital Trust, King’s College Hospital, Queens Square, and Westminster Hospital (London).

Statistical analysis

Analysis was planned in October 2018. We included all first-ever IS cases between January 1, 2000, and December 31, 2015, and incorporated follow-up until March 31, 2016. Survival time was calculated from date of stroke to date of death, confirmed by the ONS. Patients with no record of death were censored at March 31, 2016. Continuous variables are summarised as mean (SD) and categorical data as count (percentage). To evaluate changes in baseline characteristics over time (by quadrennial cohorts), we used the Cochran-Armitage test of trend for categorical variables and linear regression for continuous variables.

Survival curves were constructed by time cohorts, IS subtype, age, sex, and ethnic group using the Kaplan-Meier method and the log rank tests (unadjusted). Multivariate Cox proportional-hazards models were performed to assess the independent effect of time on all-cause mortality. Variables of important prognostic value, as suggested by literature review, were included into the models, and backward elimination of the least significant variables was performed. The final models included—and were therefore adjusted to—demographic variables, prior risk factors, clinical impairment indicators, and processes of care in the acute phase. Time-by-ethnicity interaction was examined to account for the possible modification of the proportional effect of time on stroke outcomes by ethnicity.

To assess whether CFRs within 30 days and 1 year after the first-ever stroke had improved over time, multivariable Poisson regression models using generalised estimation equations were constructed for overall ISs and according to initial subtype and demographic group. We used Zou’s method to directly estimate relative risks instead of odds ratios by specifying a Poisson distribution and including a robust error variance in our models [33]. Our independent variable, time in 4-year periods, was included as a categorical variable, with 2000–2003 as the reference group. We multiplied the adjusted rate ratio (RR) for each period by the observed rate in the reference group to obtain adjusted rates for the study period. These rates represent the estimated risks in each period if the patient demographic characteristics were identical to that in the reference group. We also evaluated calendar year as a continuous variable to obtain adjusted changes in annual rates. The same methods were used to estimate trends in functional dependence (BI < 15) among survivors at 3 months post stroke. In addition, the modified Rankin Scale (mRS) was derived from BI [34], which was used to explore trends in dependence defined as mRS ≥ 3 (results are provided in Table G in S1 Appendix).

Data were complete for all covariates and outcomes, except smoking (7.9% missing); drinking (9.9%); hypertension (1.3%); diabetes (1.6%); hypercholesterolaemia (2.3%); AF (1.6%); myocardial infarction (2.3%); TIA (1.6%); hospital admission (0.7%); use of anticoagulants (28.1%), antiplatelets (27.9%), and thrombolysis (27.1%) in the acute phase; GCS (4.4%); NIHSS (16.9%); swallow test (13.3%); urinary incontinence (4.6%); BI at 7 days (13.6%); and BI at 3 months (31.9% after adjusting for death). Analysis of missing data—comparing those with complete BI at 3 months and those without—showed similar profiles with no significant differences in almost all baseline characteristics, including demography, risk factors, medication use, and stroke severity variables (Table F in S1 Appendix). Only age and BI (7 days) varied; those whose BI record was missing at 3 months tended to be 1.8 years younger (p = 0.002) and less dependent on presentation (p = 0.003). No significant trend in BI missingness was noted over the study period (p = 0.14). To minimise the potentially resulting bias from missing data, multiple imputation with chained equations was applied to generate 20 datasets. Each variable with incomplete values was imputed as a binomial using all variables in the study (including date of stroke onset). Parameter estimates were finally combined using Rubin’s principles [35]; these were not meaningfully different from the results of complete case analysis of nonimputed dataset.

All statistical analyses were conducted with the use of R software version 3.4.4 (Free Software Foundation). All hypothesis tests were two-sided, with a significance level of 0.05. This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist).

Results

Between 2000 and 2015, a total of 4,240 patients with a first-ever stroke were registered. Of these, 3,128 (73.8%) had an IS, 715 (16.9%) had a haemorrhagic stroke, and the remaining 397 (9.3%) were unspecified. Among all IS patients, 351 (11.2%) were LAA, 815 (26.1%) CE, 791 (25.5%) SVO, 1,089 (34.8%) UND, and 82 (2.6%) OTH. Table 1 shows temporal trends in characteristics of IS patients, grouped into 4 time periods. Although there was a time trend for younger age, fewer white ethnic groups, less smoking and drinking, and less baseline stroke severity, the prevalence of cardiovascular diseases, stroke unit admission, use of CT and MRI scans, and management with thrombolytic and anticoagulant agents in the acute phase increased significantly over time. Information on medication use prior to stroke is demonstrated in Table D and Table E in S1 Appendix. There has been a decline in the prior use of antihypertensive medications from 47.6% in 2000–2003 to 37% in 2012–2015 (p < 0.001). Among CE patients, 357 (44.6%) had a known history of AF. Of these, 58 (17.5%) reported having used anticoagulant medicines. The use of anticoagulants in CE patients who had had AF has increased significantly over time from 13.8% in 2000–2003 to 22.4% in 2012–2015.

Table 1

Trends in baseline characteristics in patients with first-ever IS.

Characteristic	Year group				p-Value for trend†
	2000–2003	2004–2007	2008–2011	2012–2015
	(N = 807)	(N = 1,117)	(N = 637)	(N = 567)
Demography
Age, mean (SD)	72.31 (13.20)	70.90 (14.27)	69.63 (15.81)	69.29 (15.35)	0.0002*
Female	426 (52.8)	524 (46.9)	325 (51.0)	260 (45.9)	0.06
Ethnic group
White	590 (73.1)	769 (68.8)	393 (61.7)	316 (55.7)	<0.0001*
Black	150 (18.6)	262 (23.5)	188 (29.5)	199 (35.1)	<0.0001*
Other/Unknown	67 (8.3)	86 (7.7)	56 (8.8)	52 (9.2)	0.44
TOAST subtype
LAA	67 (8.3)	132 (11.8)	102 (16.0)	50 (8.8)	0.15
CE	225 (27.9)	275 (24.6)	143 (22.4)	172 (30.3)	0.67
SVO	226 (28.0)	266 (23.8)	170 (26.7)	129 (22.8)	0.1
UND	265 (32.8)	407 (36.4)	215 (33.8)	202 (35.6)	0.51
OTH	24 (3.0)	37 (3.3)	7 (1.1)	14 (2.5)	0.13
Pre-existing risk factors
Current drinker	395 (52.2)	585 (58.6)	255 (45.9)	225 (44.4)	<0.0001*
Smoker	223 (29.9)	302 (29.2)	170 (29.2)	120 (23.1)	0.017*
Hypertension	483 (60.8)	757 (68.1)	415 (66.6)	397 (71.0)	0.0004*
Diabetes mellitus	154 (19.6)	249 (22.5)	151 (24.0)	143 (25.6)	0.006*
Hypercholesterolaemia	115 (14.9)	298 (26.8)	203 (33.0)	265 (47.7)	<0.0001*
AF	144 (18.1)	170 (15.3)	96 (15.5)	136 (24.5)	0.009*
Myocardial infarction	94 (11.8)	110 (9.9)	42 (6.9)	60 (11.2)	0.21
TIA	97 (12.2)	117 (10.5)	60 (9.7)	57 (10.3)	0.2
Stroke severity
Urinary incontinence	300 (40.8)	439 (39.8)	184 (30.4)	156 (28.9)	<0.0001*
Swallow test (fail)	278 (38.0)	311 (31.8)	130 (24.5)	99 (21.1)	<0.0001*
GCS < 13	169 (22.1)	230 (21.1)	122 (20.4)	105 (19.4)	0.22
BI < 15 (7 d)	362 (53.5)	523 (53.6)	237 (43.2)	186 (37.3)	<0.0001*
NIHSS severity
Mild (≤4)	175 (33.7)	415 (39.0)	232 (41.6)	217 (47.5)	<0.0001*
Moderate (5–20)	312 (60.0)	526 (49.4)	268 (48.0)	213 (46.6)	<0.0001*
Severe (>20)	33 (6.3)	123 (11.6)	58 (10.4)	27 (5.9)	0.53
Processes of care in the acute phase
Admission
None	125 (15.7)	73 (6.6)	42 (6.6)	31 (5.5)	<0.0001*
Stroke unit	438 (55.0)	843 (76.0)	509 (80.0)	477 (84.6)	<0.0001*
Other medical wards	234 (29.4)	193 (17.4)	85 (13.4)	56 (9.9)	<0.0001*
Brain imaging
None	43 (5.4)	6 (0.5)	3 (0.5)	0 (0.0)	<0.0001*
CT	630 (79.1)	823 (74.0)	296 (46.5)	339 (59.8)	<0.0001*
MRI	77 (9.7)	63 (5.7)	9 (1.4)	14 (2.5)	<0.0001*
CT and MRI	30 (3.8)	216 (19.4)	244 (38.4)	167 (29.5)	<0.0001*
CT or MRI	16 (2.0)	4 (0.4)	84 (13.2)	47 (8.3)	<0.0001*
Antiplatelet	31 (67.4)	914 (87.0)	484 (79.5)	487 (88.5)	0.33
Anticoagulant	4 (8.7)	172 (16.4)	76 (12.6)	123 (22.3)	0.006*
Thrombolysis	1 (2.1)	74 (7.0)	105 (16.9)	90 (16.1)	<0.0001*

Data are count (%) unless otherwise indicated.

†p-Values were obtained from the Cochran-Armitage tests of trend for categorical variables and linear regression for continuous variables.

*Denotes significant trends (p < 0.05).

Abbreviations: AF, atrial fibrillation; BI, Barthel Index; CE, cardio-embolism; CI, confidence interval; CT, computed tomography; GCS, Glasgow coma scale; IS, ischaemic stroke; LAA, large-artery atherosclerosis; NIHSS, National Institutes of Health Stroke Scale; OTH, other determined aetiologies; SVO, small-vessel occlusion; TIA, transient ischaemic attack; TOAST, Trial of ORG 10172 in Acute Stroke Treatment; UND, undetermined aetiologies

The median survival time for overall ISs was 5.99 years (interquartile range, 5.46–6.62). Survival rates has significantly improved over the 16-year study period (log rank test p < 0.0001) (Fig 1), and the risk of death declined by around 24% between 2000–2003 and 2012–2015 (hazard ratio [HR] 0.76; 95% confidence interval [CI] 0.62–0.94; p = 0.006) after adjusting for temporal changes in demographic characteristics, prior risk factors, clinical impairments at the outset, and processes of care. Stratified analyses showed similar reductions in both sexes and ethnic groups and all aetiological subtypes; however, these were only significant in white ethnicity (HR per year 0.978; 95% CI 0.959–0.995; p = 0.013), black ethnicity (HR per year 0.966; 95% CI 0.929–0.996; p = 0.028), CE subtype (HR per year 0.972; 95% CI 0.945–0.998; p = 0.034), and patients aged ≥65 years (HR per year 0.975; 95% CI 0.959–0.992; p = 0.003) (Fig 2).

Fig 1

Kaplan-Meier survival after IS by time cohorts.

IS, ischaemic stroke.

Fig 2

Adjusted* percent change in risk of death per year among IS patients by subtype, ethnic, sex, and age groups.

Adjusted risk ratios were determined with multivariate Cox models evaluating calendar year as a continuous variable. *Adjusted for demography, prior risk factors, stroke severity, and processes of care as appropriate (see Table A, Table B, and Table C in S1 Appendix for all model covariates). Full parameter estimates are available in Table A, Table B, and Table C in S1 Appendix. CE, cardio-embolism; IS, ischaemic stroke; LAA, large-artery atherosclerosis; SVO, small-vessel occlusion; UND, undetermined aetiologies.

Table 2 shows time trends in 30-day and 1-year CFRs over the study period (2000‒2015). The overall CFR within 30 days was 12.5% and within 1 year was 25.2% (391 and 789 of 3,128 patients, respectively). There was a significant trend toward decreased case fatality during the study period. After adjustment for temporal trends in demographic characteristics, the overall 30-day CFR decreased by 38% from 16.4% in 2000–2003 to 10.2% in 2012–2015 (adjusted RR per year 0.962; 95% CI 0.941‒0.984, p = 0.0006). Such trends were confirmed only in the white and male populations and in those who had CE and UND strokes (Table 2). Similarly, a 37% risk reduction in the 1-year CFR was noted, from 32.6% in 2000–2003 to 20.5% in 2012–2015 (Table 2), which was significant in all groups.

Table 2

Trends in CFRs within 30 days and 1 year after the first-ever IS.

	Adjusted CFRs†				Adjusted RR per year (95% CI)‡	p-Value for trend§
	2000‒2003	2004‒2007	2008‒2011	2012‒2015
Within 30 d after stroke
Overall ISs	16.4	13.4	12.3	10.1	0.962 (0.941‒0.984)	0.0006*
By ethnicity
White	18.3	15.5	12.7	11	0.956 (0.931‒0.981)	0.001*
Black	6	6.3	5.3	4.4	0.977 (0.918‒1.039)	0.42
By sex
Male	12.9	11.5	10.8	4.7	0.946 (0.914‒0.979)	0.0007*
Female	19.5	14.9	13.6	15.9	0.973 (0.945‒1.002)	0.12
By age group
<55 y	7.6	4.1	4.6	2.5	0.909 (0.843‒0.98)	0.027*
55+ y	19.4	16.6	14.6	12.9	0.968 (0.945‒0.991)	0.005*
By TOAST subtype
LAA**	3	7.6	8.9	6	1.044 (0.953‒1.143)	0.4
CE	25.3	24.2	15.9	10.7	0.929 (0.899‒0.96)	<0.0001*
SVO**	1.8	4.6	4.2	3.2	1.034 (0.954‒1.121)	0.44
UND	25.7	13.7	17.2	15.6	0.963 (0.929‒0.999)	0.019*
Within 1 y after stroke
Overall ISs	32.6	27.7	22.6	20.4	0.963 (0.949‒0.976)	<0.0001*
By ethnicity
White	37.1	30.8	22.7	24.3	0.959 (0.943‒0.974)	<0.0001*
Black	16.7	17.6	13.7	8.6	0.96 (0.927‒0.994)	0.009*
By sex
Male	25.2	21	19.1	15.6	0.968 (0.946‒0.989)	0.002*
Female	39.2	33.8	25.6	24.7	0.959 (0.942‒0.976)	<0.0001*
By age group
<55 y	11.4	8.4	7.1	6.2	0.939 (0.891‒0.99)	0.05*
55+ y	40	34.4	27.9	25.5	0.965 (0.952‒0.979)	<0.0001*
By TOAST subtype
LAA**	23.9	16.7	17.7	12	0.958 (0.904‒1.015)	0.15
CE	48	44.2	32	27.8	0.955 (0.937‒0.973)	<0.0001*
SVO**	11.5	10.1	10.6	6.2	0.964 (0.915‒1.015)	0.17
UND	41.1	31.3	26.4	21.9	0.956 (0.935‒0.978)	<0.0001*

†Unless otherwise indicated, figures are adjusted for demographic variables as appropriate (see Table 1). Adjusted rates for each time cohort were obtained by multiplying the observed rate for the reference period (2000‒2003) by the corresponding RRs for the later periods from a model evaluating time cohorts as a categorical variable.

‡Adjusted risk ratios were determined with a model evaluating calendar year as a continuous variable.

§p-Values were obtained from the Cochran-Armitage tests for trend.

*Denotes significant trends (p < 0.05).

**Unadjusted because of small number of events.

Abbreviations: CE, cardio-embolism; CFR, case-fatality rate; CI, confidence interval; IS, ischaemic stroke; LAA, large-artery atherosclerosis; RR, rate ratio; SVO, small-vessel occlusion; TOAST, Trial of ORG 10172 in Acute Stroke Treatment; UND, undetermined aetiologies

Although rates of survival increased, proportions of functionally dependent patients (BI < 15) among IS survivors decreased. Fig 3 illustrates the proportion of survivors with functional dependency (BI < 15) across time at several points after IS onset. Most of the improvements in functional abilities were noted in the acute stage of (7 days) and early phase of (3 months)—but not 1 year after—IS onset. A 23% risk reduction of functional dependence at 3 months was estimated over the study period (demography adjusted rate 34.7% in 2000–2003 and 26.7% in 2012–2015; adjusted RR per year 0.983; 95% CI 0.968–0.998; p = 0.002 for trend) (Table 3). Reductions in disability levels were observed in all groups but did not reach statistical significance in females, LAA, SVO, or UND subtypes.

Fig 3

Changes in the proportion of dependent (BI < 15) patients among IS survivors over time.

BI, Barthel Index; IS, ischaemic stroke.

Table 3

Trends in functional dependence (BI < 15) at initial assessment and 3 months after the first-ever IS.

	Adjusted disability rates (BI < 15)†				Adjusted RR per year (95% CI)‡	p-Value for trend§
	2000‒2003	2004‒2007	2008‒2011	2012‒2015
At initial assessment (7 d)
Overall ISs	53.5	52.2	43.7	39.6	0.976 (0.964‒0.988)	<0.0001*
By ethnicity
White	55.8	53.6	45.7	40	0.974 (0.959‒0.989)	<0.0001*
Black	47	47.7	34.6	36.7	0.977 (0.953‒0.999)	0.005*
By sex
Male	46.2	44.4	33.3	26.1	0.959 (0.941‒0.978)	<0.0001*
Female	59.9	58.5	53.1	53.9	0.988 (0.973‒0.999)	0.043*
By age group
<55 y	33.1	29.8	23.2	24.2	0.968 (0.942‒0.996)	0.013*
55+ y	60.9	60	50.8	44.7	0.977 (0.964‒0.990)	<0.0001*
By TOAST subtype
LAA	72.4	55.7	55.3	46.6	0.975 (0.939‒0.999)	0.009*
CE	65	65.3	58.6	49.7	0.979 (0.959‒0.999)	0.0009*
SVO	34.6	33.4	24.9	22.5	0.964 (0.934‒0.994)	0.004*
UND	57.4	54.3	43.5	41	0.973 (0.953‒0.994)	<0.0001*
At 3 mo post stroke
Overall ISs	34.7	32	30.8	26.8	0.983 (0.968‒0.998)	0.002*
By ethnicity
White	35.7	31.5	30.2	28.1	0.982 (0.964‒0.999)	0.013*
Black	31.8	33.6	27.9	24.3	0.981 (0.951‒0.999)	0.043*
By sex
Male	28.6	25.3	23.1	18.9	0.972 (0.948‒0.996)	0.002*
Female	40.5	38.1	38.1	34.9	0.991 (0.972‒1.01)	0.17
By age group
<55 y	23.3	17.1	19.1	14	0.967 (0.932‒0.999)	0.036*
55+ y	39.3	37.4	34.8	31.7	0.985 (0.969‒0.999)	0.013*
By TOAST subtype
LAA**	59	31.5	42.3	33.4	0.974 (0.929‒1.021)	0.05
CE	42	42.3	36.7	32.8	0.982 (0.958‒0.999)	0.036*
SVO**	22	21.3	20	17.5	0.988 (0.94‒1.017)	0.17
UND	36.7	33	32.6	27.8	0.982 (0.958‒1.007)	0.05

†Unless otherwise indicated, figures are adjusted for demographic variables as appropriate (see Table 1). Adjusted rates for each time cohort were obtained by multiplying the observed rate for the reference period (2000‒2003) by the corresponding RRs for the later periods from a model evaluating time cohorts as a categorical variable.

‡Adjusted risk ratios were determined with a model evaluating calendar year as a continuous variable.

§p-Values were obtained from the Cochran-Armitage tests for trend.

*Denotes significant trends (p < 0.05).

**Unadjusted because of small number of events.

Abbreviations: BI, Barthel Index; CE, cardio-embolism; CI, confidence interval; IS, ischaemic stroke; LAA, large-artery atherosclerosis; RR, rate ratio; SVO, small-vessel occlusion; TOAST, Trial of ORG 10172 in Acute Stroke Treatment; UND, undetermined aetiologies

Discussion

In this population-based analysis of patients with first-ever IS, from an urban multi-ethnic population of south London, we found that risk of death after IS decreased by 24% (annual percent change [APC], −2.4%) between 2000 and 2015 after accounting for demography and other time-varying variables (Fig 2). Disproportionate trends were observed among aetiological subtypes for which reductions were mainly identified in strokes of CE origins (APC, −2.8%). Moreover, the rate of functional dependence at 3 months after IS declined by 23% (Table 3). Using a conservative estimate of 52,000 incident ISs per year in the UK [36], we estimate additional survival beyond 30 days and 1 year of 3,200 and 6,300 patients, respectively (based on 6.2% and 12.1% absolute improvements in the adjusted 30-day and 1-year CFRs). We also estimate aversion of more than 3,200 cases of significant disability among those who survived up to 3 months after IS. These improved mortalities and morbidities have been accompanied by higher admission rates to hospital, increased use of CT and MRI scans, and more frequent treatment with thrombolytic and anticoagulant medications in the acute phase of stroke. However, a shift in IS characteristics toward less severe strokes was observed.

Unlike previous studies that explored trends in stroke death as a whole or by pathological types, the current analysis investigated trends in IS and its aetiological groups. The Oxford Vascular Study (OXVASC) used a modest-sized cohort of predominantly white patients and did not detect trends in early CFRs after stroke. In contrast, we estimated a significant 3.8% annual decline in the 30-day CFR after IS. Our finding corroborates national data by Seminog and colleagues suggesting a reduction of 4.7% per year in men and 4.4% per year in women (2001–2010). The steeper declines in the later study are due in large part to investigating all strokes, including haemorrhagic events, rather than conducting subtype-specific analyses. Another study in the US reported a decrease in the overall mortality after stroke by 20% for each 10-year period between 1987 and 2011 [37], which is comparable to our estimate of 24% after IS during a 16-year period. None of the previous investigations, however, were able to characterise the trends by stroke subtype due to lack of reliable diagnostic information.

Published literature from previous studies like OXVASC and the Minnesota Stroke Survey (MSS) is limited to data collected before 2005—before many current strategies for the prevention and management of stroke were implemented more widely [5,20]. For instance, national UK data from SSNAP showed an increase in thrombolysis administration rate from 1.4% in 2008 to approximately 12% in 2016 [38,39]. Similarly, our IS cohort showed increased proportion of thrombolysed patients from 2.1% in 2000–2003 to 16.1% in 2012–2015 (Table 1), approaching the estimated 20% eligibility level among acute IS cases. The treatment has a strong evidence base suggesting a 34% reduced risk of death or dependence if administered within a 3-hour time window and could therefore have contributed to the observed trends [40]. Also, the first decade of the 21st century witnessed the launch of several national awareness campaigns, including the Act FAST campaign (Face, Arms, Speech, and Time), which presumably might have had an influence on reducing time to receiving care and DTN times and subsequent mortalities [41]. Finally, stroke services in London underwent major reconfiguration in 2010 in that 8 hyperacute stroke units (HASUs)—optimally equipped and staffed—replaced 30 hospitals used to provide acute stroke care across the city. The hospitals serving residents living in the SLSR area include 2 HASUs, while the remaining 3 hospitals have standard stroke units. This centralised model of service provision has been shown to reduce mortality [42-44].

The improved prognosis of ISs over time is likely owing to multiple factors, including not only advances in acute therapies and management but also decreased severity on presentation—genuinely driven by better primary prevention and artefactually by increased use of enhanced diagnostics methods (e.g., CT/MRI scans) enabling better detection of milder strokes [17,37,45]. Increasing emphasis has been put on measures to improve cardiovascular health at the individual and community levels [46], including smoking cessation programs and lower target levels of cholesterol and blood pressure. We have previously demonstrated increasing use of statins in our study population [36]. These agents may have beneficial effects beyond their impact on incidence and risk factors and may contribute to lower IS severity and lower subsequent morbidity and mortality [47]. Despite these efforts, countervailing trends such as the increasing prevalence of diabetes and obesity could have the opposite effect [46]. Our results show that the net effect has been a lower CFR and dependency, and that—after adjusting for the changes in demography, stroke severity, and acute phase management—there was a residual, unexplained improvement in survival after ISs during 2000–2015. Future studies are needed to better understand which specific factors have driven CFRs down so that survival gains can be consolidated and expanded.

Several issues also merit further discussion. Our results show that, for all IS subtypes, dependency levels at 3 months after the index stroke were lower in the later time cohorts than the earlier ones (Table 3). However, regression analysis showed that certain groups had steeper declines than others, which was particularly evident in CE strokes (p = 0.036). The unequal declines in dependency (BI < 15) among IS subtypes may indicate differential capacity to improve and respond to interventions. CE strokes may constitute functionally disabled yet structurally intact areas the recovery of which is largely time-dependent on restoring recanalization and reperfusion. It is therefore possible that CE stroke might have benefited more from the overall advances in stroke prevention and management during recent decades [48]. However, in certain IS subtypes (e.g., LAA), power might have not been sufficient to detect significant trend because of a smaller number of events.

This population-based study describes time changes in death and dependence after IS by aetiological subtypes and demographic groups. The study was carried out in a well-defined urban area where diverse sociodemographic and economic characteristics allow generalisability of the results to other settings. Moreover, our survival trends were adjusted for—and were therefore independent of—the corresponding changes in demography, prestroke risk factors, case mix variables, and processes of care. However, our findings should be interpreted considering the following potential limitations. First, the observed trends could be in part attributable to secular changes in case ascertainment and detection rates. Increased sensitivity of surveillance methods over time would particularly improve the detection of less severe strokes. Such cases are expected to have a lower risk of death, which could have thus contributed to the observed declines in death rates. In an attempt to address the issue, we controlled for the changes in stroke severity variables. Nevertheless, certain analyses (e.g., 30-day CFR) and subgroups (e.g., LAA) had a limited number of events, and thus we were unable to account for the variations in stroke severity in these groups. Second, there are several factors that we could not adjust for in our analysis, e.g., time to hospital admission, secondary prevention, and rehabilitation received, which are also important determinants of post-IS outcomes and might explain the observed trends. Third, BI might have had suboptimal sensitivity in measuring extremes of functional deficit (ceiling and floor effects) [49]. Finally, the incomplete record of BI at 3 months may have introduced bias in the analysis of functional outcomes. In a highly dynamic population such as inner London, migration can lead to higher rates of loss to follow-up [26]. In addition, our study period covers times of great economic recession in which participants were even more reluctant to engage in research activities. However, every effort was made to keep track of all patients. Address details were updated regularly from hospital data, GP records, and through direct contact with patients or their next of kin. If patients had moved overseas, postal questionnaires were often sent and returned (S2 Appendix). No systematic loss of data was recognised; therefore, the collected information was adequate for statistical analyses. Comparing those with complete BI at 3 months and those without, both groups had similar baseline characteristics except for age and BI (7 days). Incomplete record of BI at follow-up was associated with younger age and less impaired activities of daily living on presentation. Since these variables (age and BI at 7 days) are associated with improved functional activities at follow-up, our figures regarding the prevalence of dependency among IS survivors could be an overestimate. However, our time trend estimates are unlikely to have been remarkably distorted, particularly because no significant pattern of missingness was identified over time. Furthermore, multiple imputation analysis was performed to minimise any potential bias attributed to missing data.

There is limited information on factors driving IS fatalities by aetiological subtype, including the impact of HASUs, centralised stroke care, and early rehabilitation. Exploring these would help in tailoring interventions and developing optimal care paths for each subtype of IS. In the UK, the National Health Service (NHS) has set milestones for improving stroke care over the next 10 years [50]. These include delivering a 10-fold increase in the proportion of patients who receive thrombectomy after a stroke (currently only 1% of stroke patients receive the treatment). Therefore, further reductions are expected through expanding the use of novel interventions (such as thrombectomy) and rolling out efficient models of service provision and rehabilitation. Future planning could benefit from more detailed information by IS aetiologies to develop more targeted interventions and conserve resources.

As evidenced by the results of this study, survival after IS has improved significantly in South London over the past 16 years from 2000 to 2015, but our data were not adequate to settle the issue for certain groups (e.g., LAA). Moreover, rates of functional dependence on presentation and follow-up have declined. Most of the improvements were noted in CE strokes and in patients aged ≥65 years old.

Completed STROBE checklist.

STROBE, Strengthening the Reporting of Observational Studies in Epidemiology.

(PDF)

Click here for additional data file.

Supplementary tables and figures.

(PDF)

Click here for additional data file.

Postal questionnaire used for collecting information at follow-ups.

(PDF)

Click here for additional data file.

29 Nov 2019

Dear Dr. Wafa,

Thank you very much for submitting your manuscript "Long-term trends in death and dependence after ischaemic strokes: Prospective population study of the South London Stroke Register (SLSR)" (PMEDICINE-D-19-03587) for consideration at PLOS Medicine.

Your paper was evaluated by a senior editor and discussed among all the editors here. It was also discussed with an academic editor with relevant expertise, and sent to independent reviewers, including a statistical reviewer. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, I am afraid that we will not be able to accept the manuscript for publication in the journal in its current form, but we would like to consider a revised version that addresses the reviewers' and editors' comments. Obviously we cannot make any decision about publication until we have seen the revised manuscript and your response, and we plan to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the 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. 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 us at PLOSMedicine@plos.org.

We expect to receive your revised manuscript by Dec 20 2019 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see http://journals.plos.org/plosmedicine/s/submission-guidelines#loc-methods.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

We look forward to receiving your revised manuscript.

Sincerely,

Adya Misra, PhD

Senior Editor

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

Title- it appears the register contains prospectively collected data but the analysis retrospective? As such the title ought to be modified to reflect this as well as include a study descriptor to adhere to PLOS Medicine style. You may consider “Long-term trends in death and dependence after ischaemic strokes: a retrospective cohort study using the South London Stroke Register (SLSR)"

Data availability statement- please provide the link for data requests and/or an accession number to identify the data set used in the study

Author summary "At this stage, we ask that you include a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract. Please

see our author guidelines for more information: https://journals.plos.org/plosmedicine/s/revising-your-manuscript#loc-author-summary"

Did your study have a prospective protocol or analysis plan? Please state this (either way) early in the Methods section.

a) If a prospective analysis plan (from your funding proposal, IRB or other ethics committee submission, study protocol, or other planning document written before analyzing the data) was used in designing the study, please include the relevant prospectively written document with your revised manuscript as a Supporting Information file to be published alongside your study, and cite it in the Methods section. A legend for this file should be included at the end of your manuscript.

b) If no such document exists, please make sure that the Methods section transparently describes when analyses were planned, and when/why any data-driven changes to analyses took place.

c) In either case, changes in the analysis—including those made in response to peer review comments—should be identified as such in the Methods section of the paper, with rationale.

Please ensure that the study is reported according to the STROBE guideline, and include the completed STROBE checklist as Supporting Information. When completing the checklist, please use section and paragraph numbers, rather than page numbers. Please add the following statement, or similar, to the Methods: "This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology STROBE guideline (S1 Checklist)."

Abstract- please include brief patient demographics in the methods and findings section

Abstract- in the methods and findings section please include the limitations of your study as the last sentence

References must be provided within square brackets and there must be a space between the text and the square bracket followed by a full stop.

Introduction- please briefly explain door- to needle times

Methods- please clarify if any data were excluded for analysis, providing reasons

Results- please do not use “less white patients” as an ethnic descriptor and revise as appropriate

Figure 1 requires a better explanation perhaps, specifically what are the numbers along the X-axis

Figure 3 could use X-axis labels

Discussion- please comment on why you only studied IS trends

Line 312 “Our results show…”

Line 333- please avoid assertions of primacy

Comments from the reviewers:

Reviewer #1: Long-term trends in death and dependence after ischaemic strokes: Prospective population study of the South London Stroke Register (SLSR)

This study explores death and dependence among ischemic stroke patients in South London using a population-based, registry of patient with first-ever strokes in a defined population of inner London. The aims of this study were to estimate secular trends in death after IS by subtypes among black and white patients and to examine the changes in dependence among IS survivors over time.

1. I think the novel part about this study is the description of prognosis and mortality trends for the different subtypes of ischemic stroke. But this needs to be better explained in the background. Wy do the authors do this? Why is this important? I also lack an explanation to the aim to describe trends separately for black and white patients? This is not brought up in the background or put in a context. Do the authors expect the trends to be different? Is it socioeconomic or ethnic differences that the authors want to explore?

2. Why only IS stroke?

3. The authors write in the background when they mention previous studies on the topic that "Although large and nationally representative, these studies used administrative data and may have excluded patients treated in the community or included patients without stroke (e.g. transient ischemic attacks, stroke mimics)."

But it is not clear to me in what way the register used in this study differ from other administrative registers and thus avoids the limitations brought up in the previous studies? For example, TIA and mimics? The authors state themselves that completeness of case ascertainment has been estimated to 80%, isn't that lower than in many administrative registers?

4. In row 67-69 the authors write "Other investigations reported conflicting results, showing either no changes or declines in death rates.5,12,15-20" But as far as I can see there are no conflicting results in these studies regarding case fatality or stroke mortality, which is what the authors refer to. All of them show improved survival after stroke over time, right?

5. The authors report that "Between 2000 and 2015, a total of 3,843 patients with stroke was registered. Of these, 3,128 180 81.4%) had an IS, and the remaining 715 (18.6%) had haemorrhagic strokes." Were there no unclassified cases?

6. It is not completely clear if only first strokes are included, or whether there is a mix of first and recurrent? The authors sate in the methods that they included all index IS cases between Jan 1, 2000 and Dec 31, 2015. And in the discussion they say it is "first-ever" strokes. How was the procedure in order to ensure that they included first ever strokes only? This needs to be described in the manuscript.

7. Why and when were the confounders used? Many variables are brought up in the methods section but not shown in the models. No stepwise inclusion in models are shown. Did the authors use any exclusion criteria?

8. The authors communicate the result as if they only find improvements for CE and discuss that the unequal declines among IS subtypes may indicate differential capacity to improvement and response to interventions. However, they find declines for all strokes, even if it is statistically significant only for CE. Therefore I think it should not be over-emphasized that only CE improved, but also be discussed that the others may also have improved.

9. Seems strange to me that the mean age of stroke declined over time. The incidence of stroke has come down substantially over the past decade, which to some extent must be explained by a shift upward in the age at onset. Can this be related to how the patients were selected? I also think the mean age is quite low. Is this representative to UK overall?

10. I understand that the panels in figure 2 cannot have the exact same units at the y-axis. But if they can be made a little more similar would be helpful.

11. Discussion row 325 to 329, can the authors please expand on this? I don't understand this reasoning.

12. Can the authors elaborate a little on the generalizability of the risk factors and the survival results? (Also related to comment 9)

13. There are result tables and figures in the discussion, which is a bit confusing to me. I think all results should be presented in the result section. And in the discussion the interpretation of them can be brought up.

14. Do the authors want to describe survival trends or explain them? A lot of the discussion and limitation section is about not being able to explain the trends, but I don´t think that is the primary aim of the paper. I think the authors should focus more on discussing comparability over the time periods such as difference in detection, that could limit comparability over the periods. That the treatments have become better is already known and one of the primary factors behind the survival improvement.

15. How to read figure 3, left panel? What does it show? And what is the Barthel index and the interpretation of it?

16. Last paragraph not really related to the paper

17. Language editing is recommended.

Bets wishes,

Karin Modig

Associate Professor of Epidemiology, Karolinska Institutet

Reviewer #2: Based on 3128 patients with an ischemic stroke registered in the population-based South London Stroke Register (SLSR), secular trends in death after ischemic stroke by subtypes among black and white patients over a period of 16 years were estimated. Furthermore, the changes in dependence among ischemic stroke survivors over time were investigated. It was found that both mortality and functional dependence after an ischemic stroke decreased by an annual average of about 2% between 2000 and 2015.

I have the following comments:

The Barthel index (BI) was assessed as an indicator of clinical impairment. By whom was the BI assessed at baseline examination? At follow-up, BI was collected by means of face-to-face visits or postal questionnaires. Were the information collected by a face-to-face visit comparable to the self-reported information via questionnaire? It is well known that the BI does not define sufficiently sharp classification criteria in many items, despite its widespread distribution. Thus, a course assessment by BI can be difficult.

Did the authors also assess the National Institute of Health Stroke Scale (NIHSS) and the modified rankin scale? The NIHSS is used to assess the severity, i.e. the extension of an ischemic insult. The score can be used on the one hand as a course parameter, but on the other hand also for the evaluation of a possible therapy. Furthermore, the degree of disability or dependence in daily activities is most widely used as clinical outcome measure after stroke.

The authors stratified the analyses by ethnicity and by TOAST subtypes. It would be of interest whether there are differences due to sex and age. Did the authors conduct formal tests on the interaction with sex and age? If not, this should be carried out. If there would be a significant interaction further stratified analyses should be conducted.

How was confounder selection done by the authors?

As already mentioned by the authors, the number of events in some TOAST subgroups were very low. Thus, statements about the mortality trends in these groups were not possible.

An age-range of the study population and the proportions of males/females included in the analysis should be given in the abstract and methods section.

Reviewer #3: This is an interesting population based study on time trends in survival and outcome in subtypes of ischemic stroke. Even though the material is limited it is a well defined population and the work up of the cases seems thorough, thus the reliability on TOAST score should be good. The paper is over all well written. However, the introduction could do with some proof reading of the grammer.

In the introduction, some of the references on stroke epidemiology are rather old. A lot has changed in the last 15 years, as the present paper shows. If 15 years old references, such as 4 and 6 are used, it could be an idea to also state that they are old, particularly as the change over time is the focus of the paper.

There are a couple of issues that I fell need clarification

1) In the methods section: for readers unfamiliar with the demographics of London, could a sentence be added about changed (or not changed) demographic in the catchment area for the study. Socioeconomic factors are important in stroke and thus this is of interest. I am a bit unclear if the data was adjusted for changed demographics over all or for the demographics of the stroke cohort. Could the authors please clarify.

2) For hypertension, was it one registered blood pressure at any time and in any situation above 140/90 mmHg that was used?

3) In the statistics section it is stated that data was complete except for smoking (7.9%). Does that mean that there was data for 7.9% of the paritcipants or that data was missing for 7.9% of the stroke cases? Did the ompleteness of dat chage over time. Please clarify.

4) In the results: How many patients declined to participate in the study?

5) I miss information on medication at time and before stroke onset. Was that not available? Of the CE strokes, how many had Afib known before stroke onset and were treated?In many countries, AFib is much more treated after the introduction of NOAC. Is this the case also in Southern London? Also antihypertensive medication at stroke onset would be informative. Pharmacological treament options has evolved over time and thus, the information is very relevant to the study. If not present, this is a major limitation that needs to be discussed.

6) Do you have access to mRS? If so, it would be nice to have that included in the paper.

Except for this, I find the discussion well written and in line with the results.

Reviewer #4: I confine my remarks to statistical aspects of this paper. The general approach is fine but I have some issues to resolve before I can recommend publication.

Line 35 and many other places. Maybe give the HR (or RR) per decade rather than per year or, alternatively, give a 3rd decimal place. What's here isn't wrong, and it matches e.g. the APA style guide, but there's a big difference, over 15 years, between, e.g., 0.975 and 0.985 per year. The former gives 0.68 and the latter gives 0.80.

Lines 47-49: Instead of (or in addition to) giving ranks of causes of death or disability, give estimates of the actual rates. Ranks have a couple problems in this context. First, *something* has to be the leading cause. Second, such rankings are very dependent on how causes are broken down. E.g. is "cancer" one cause or is "lung cancer" one cause?

Line 67-69: Did you compare effect sizes or statistical significance? If it was the latter, please reconsider. Andrew Gelman wrote an article "The Difference between 'Significant' and 'Not Significant' is not Itself Statistically Significant" https://amstat.tandfonline.com/doi/abs/10.1198/000313006X152649#.Xc1Mu9V7lPY

Line 118 and other places in that paragraph: Categorizing continuous independent variables is almost always a mistake. Frank Harrell, in *Regression Modeling Strategies* listed 11 problems with this and summed up "Nothing could be more disastrous". I wrote a blog post showing, graphically, some of the issues. https://medium.com/@peterflom/what-happens-when-we-categorize-an-independent-variable-in-regression-77d4c5862b6c

Line 149 Insert "logistic" between multivariable and regression

Line 153: Another IV that got categorized. This may be necessary for a figure or a table, but it shouldn't be used in the analysis. Use year as year (and maybe look for nonlinear effects via a spline). However, in the footnote to table 2 it says year was used as a continuous variable, so I am a little confused.

Line 160 and other places: What are these %s? Are they the amount of missing? Why was data missing? Was it MCAR, MAR, or NMAR?

Line 165-168 Significance is not really the issue here. The effect size is more important, but the missing data mechanism is critical.

Line 203-204 Although some were stat sig and some not, the effect sizes were similar and that is more important.

Figure 1: I'm not completely happy with this figure in general, but I'm not going to object as I don't have a better suggestion. However, please make the lines easier to distinguish. One option is to put the labels to the right of the lines, rather than on the upper right. Another would be to vary line type (e.g. use .... , ---, and so on) or use colors other than shades of blue (or, best yet, some combination)

Figure 2: I like this represenation.

Figure 3a has several problems. First, the gaps on the x-axis are wrong. The first gap (7 days to 3 months) is 1/3 as long as the second gap (3 months to 1 year) but is shown as the same size. Second, it is very hard to read. In addition to the suggestions for figure 1, you could jitter the points so that they don't completely overlap.

Figure 3b is what's known as a "dynamite plot". They aren't a great graphical method. See http://biostat.mc.vanderbilt.edu/wiki/Main/DynamitePlots

Peter Flom

Any attachments provided with reviews can be seen via the following link:

[LINK]

20 Jan 2020

Dear Dr. Wafa,

Thank you very much for re-submitting your manuscript "Long-term trends in death and dependence after ischaemic strokes: a retrospective cohort study using the South London Stroke Register (SLSR)" (PMEDICINE-D-19-03587R1) for review by PLOS Medicine.

I have discussed the paper with my colleagues and the academic editor and it was also seen again by 3 reviewers. I am pleased to say that provided the remaining editorial and production issues are dealt with we are planning to accept the paper for publication in the journal.

The remaining issues that need to be addressed are listed at the end of this email. Any accompanying reviewer attachments can be seen via the link below. Please take these into account before resubmitting your manuscript:

[LINK]

Our publications team (plosmedicine@plos.org) will be in touch shortly about the production requirements for your paper, and the link and deadline for resubmission. DO NOT RESUBMIT BEFORE YOU'VE RECEIVED THE PRODUCTION REQUIREMENTS.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

In revising the manuscript for further consideration here, please ensure you address the specific points made by each reviewer and the editors. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments and the changes you have made in the manuscript. Please submit a clean version of the paper as the main article file. A version with changes marked must also be uploaded as a marked up manuscript file.

Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. If you haven't already, we ask that you provide a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract.

We expect to receive your revised manuscript within 1 week. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

We ask every co-author listed on the manuscript to fill in a contributing author statement. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

If you have any questions in the meantime, please contact me or the journal staff on plosmedicine@plos.org.

We look forward to receiving the revised manuscript by Jan 27 2020 11:59PM.

Sincerely,

Adya Misra, PhD

Senior Editor

PLOS Medicine

plosmedicine.org

------------------------------------------------------------

Requests from Editors:

Prospective data analysis plan- if no such plan exists please make sure that the Methods section transparently describes when analyses were planned, and when/why any data-driven changes to analyses took place. Changes in the analysis—including those made in response to peer

review comments—should be identified as such in the Methods section of the paper,

with rationale.

Abstract methods and findings- study area was 357,308- presumably this is the study population and not area? Please correct as needed

Abstract- please explicitly state “the limitations of this study are ….”

Line 388 this sentence requires revision for grammar and usage “Insofar as such patients are expected to have a lower risk of death, this would contribute to the observed declines in death rates” and Line 408 “Inasmuch as these differences are associated with improved functional activities at follow up, our figures regarding…”

Line 95 - what is 'healthy migrant effect '? Please rephrase for clarity and to ensure non stigmatising labels are used

Line 131 – Please state if it is possible to disclose what the 5 hospitals are? and why 3 outside the study area and how far out - still in london? - and what was the basis for their selection / inclusion? Also line 348 states how london underwent a reconfiguration in 2010 for stroke care and 8 new units were re-equipped and replaced 30 hospitals that had previously provided stroke care. Please clarify if the 5 included in this study are part of this new structure.

Line 403-Please ensure all questionnaires are provided as SI files or cited as a reference if previously published

Please ensure p values and 95% CI are provided throughout the main text. For example around Line 254

Comments from Reviewers:

Reviewer #2: The authors have satisfactorily answered to my comments and revised the manuscript accordingly. I have no further comments.

Reviewer #3: If it would it be possible to do an estimated mRS based on the Barthel index (for example using Wolfe et al) I Think this information could be helpful as mRS is the most used outcome measure today. Otherwise, I find my questions and concerns met

Reviewer #4: The authors have addressed my concerns and I now recommend publication

Peter Flom

Any attachments provided with reviews can be seen via the following link:

[LINK]

10 Feb 2020

Dear Mr Wafa,

On behalf of my colleagues and the academic editor, Dr. Christa Meisinger, I am delighted to inform you that your manuscript entitled "Long-term trends in death and dependence after ischaemic strokes: a retrospective cohort study using the South London Stroke Register (SLSR)" (PMEDICINE-D-19-03587R2) has been accepted for publication in PLOS Medicine.

PRODUCTION PROCESS

Before publication you will see the copyedited word document (in around 1-2 weeks from now) and a PDF galley proof shortly after that. The copyeditor will be in touch shortly before sending you the copyedited Word document. We will make some revisions at the copyediting stage to conform to our general style, and for clarification. When you receive this version you should check and revise it very carefully, including figures, tables, references, and supporting information, because corrections at the next stage (proofs) will be strictly limited to (1) errors in author names or affiliations, (2) errors of scientific fact that would cause misunderstandings to readers, and (3) printer's (introduced) errors.

If you are likely to be away when either this document or the proof is sent, please ensure we have contact information of a second person, as we will need you to respond quickly at each point.

PRESS

A selection of our articles each week are press released by the journal. You will be contacted nearer the time if we are press releasing your article in order to approve the content and check the contact information for journalists is correct. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact.

PROFILE INFORMATION

Now that your manuscript has been accepted, please log into EM and update your profile. Go to https://www.editorialmanager.com/pmedicine, log in, and click on the "Update My Information" link at the top of the page. Please update your user information to ensure an efficient production and billing process.

Thank you again for submitting the manuscript to PLOS Medicine. We look forward to publishing it.

Best wishes,

Adya Misra, PhD

Senior Editor

PLOS Medicine

plosmedicine.org