Article Text

Download PDFPDF

Original research
Longer term stroke risk in intracerebral haemorrhage survivors
  1. Gargi Banerjee1,
  2. Duncan Wilson1,
  3. Gareth Ambler2,
  4. Isabel Charlotte Hostettler1,
  5. Clare Shakeshaft1,
  6. Hannah Cohen3,
  7. Tarek Yousry4,
  8. Rustam Al-Shahi Salman5,
  9. Gregory Y H Lip6,7,
  10. Henry Houlden8,
  11. Keith W Muir9,
  12. Martin M Brown1,
  13. Hans Rolf Jäger4,
  14. David J Werring1
  15. on behalf of the CROMIS-2 collaborators
    1. 1 Stroke Research Centre, Department of Brain Repair and Rehabilitation, UCL Queen Square Institute of Neurology and the National Hospital for Neurology and Neurosurgery, London, United Kingdom
    2. 2 Department of Statistical Science, University College London, London, United Kingdom
    3. 3 Haemostasis Research Unit, Department of Haematology, University College London, London, United Kingdom
    4. 4 Lysholm Department of Neuroradiology and the Neuroradiological Academic Unit, Department of Brain Repair and Rehabilitation, UCL Queen Square Institute of Neurology, London, United Kingdom
    5. 5 Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, United Kingdom
    6. 6 Liverpool Centre for Cardiovascular Science, University of Liverpool and Liverpool Heart & Chest Hospital, Liverpool, United Kingdom
    7. 7 Aalborg Thrombosis Research Unit, Department of Clinical Medicine, Aalborg University, Aalborg, Denmark
    8. 8 Department of Molecular Neuroscience, UCL Queen Square Institute of Neurology and National Hospital for Neurology and Neurosurgery, London, United Kingdom
    9. 9 Institute of Neuroscience & Psychology, University of Glasgow and Queen Elizabeth University Hospital, Glasgow, United Kingdom
    1. Correspondence to Dr David J Werring, Stroke Research Centre, Department of Brain Repair and Rehabilitation, UCL Queen Square Institute of Neurology, London, UK; d.werring{at}ucl.ac.uk

    Abstract

    Objective To evaluate the influence of intracerebral haemorrhage (ICH) location on stroke outcomes.

    Methods We included patients recruited to a UK hospital-based, multicentre observational study of adults with imaging confirmed spontaneous ICH. The outcomes of interest were occurrence of a cerebral ischaemic event (either stroke or transient ischaemic attack) or a further ICH following study entry. Haematoma location was classified as lobar or non-lobar.

    Results All 1094 patients recruited to the CROMIS-2 (Clinical Relevance of Microbleeds in Stroke) ICH study were included (mean age 73.3 years; 57.4% male). There were 45 recurrent ICH events (absolute event rate (AER) 1.88 per 100 patient-years); 35 in patients presenting with lobar ICH (n=447, AER 3.77 per 100 patient-years); and 9 in patients presenting with non-lobar ICH (n=580, AER 0.69 per 100 patient-years). Multivariable Cox regression found that lobar ICH was associated with ICH recurrence (HR 8.96, 95% CI 3.36 to 23.87, p<0.0001); similar results were found in multivariable completing risk analyses. There were 70 cerebral ischaemic events (AER 2.93 per 100 patient-years); 29 in patients presenting with lobar ICH (AER 3.12 per 100 patient-years); and 39 in patients with non-lobar ICH (AER 2.97 per 100 patient-years). Multivariable Cox regression found no association with ICH location (HR 1.13, 95% CI 0.66 to 1.92, p = 0.659). Similar results were seen in completing risk analyses.

    Conclusions In ICH survivors, lobar ICH location was associated with a higher risk of recurrent ICH events than non-lobar ICH; ICH location did not influence risk of subsequent ischaemic events.

    Trial registration number NCT02513316.

    View Full Text

    Statistics from Altmetric.com

    Introduction

    Intracerebral haemorrhage (ICH) is associated with high rates of mortality (1-year and 5-year survival estimated at 46% and 29%, respectively1), and consequently data on subsequent stroke events in ICH survivors are limited. Recent data2 3 have challenged the prevailing view4 that antiplatelet and anticoagulant medications increase the risk of further ICH to an extent that outweighs any potential benefits with regard to ischaemic risk, suggesting that the risk of ischaemic events is underestimated in this population.

    One baseline feature that might help stratify future stroke risk is the dominant underlying cerebral small vessel disease. Lobar ICH has a higher recurrence rate,1 which is thought to reflect its association with the bleeding-prone cerebral amyloid angiopathy (CAA).5 6 Hypertensive arteriopathy, also termed deep perforator arteriopathy, is thought to be responsible for non-lobar or ‘deep’ ICH; it is associated with cardiovascular risk factors and lacunar infarction5 and therefore might confer greater ischaemic risks, in addition to lower ICH risks. Data for relative ischaemic and haemorrhagic risks based on index ICH location could therefore also be useful for future decision making, but there are limited data available.1 7

    Our aim was to provide new data on stroke risk following spontaneous ICH in a large cohort of ICH survivors (ie, patients with ICH who survived the index event for a period of time that allowed for study enrolment). The specific objectives are: (1) to describe the incidence of recurrent ICH and cerebral ischaemic events in the longer term (up to 3 years) following ICH and (2) to evaluate the influence of ICH location on stroke outcomes.

    Methods

    Standard protocol approvals, registrations and patient consents

    We included patients recruited to a multicentre observational cohort study of adults with imaging-confirmed symptomatic ICH (CROMIS-2 (Clinical Relevance of Microbleeds in Stroke) ICH; https://clinicaltrials.gov; NCT02513316). Patients with capacity gave informed written consent; in those without capacity, written consent was obtained from a proxy, as defined by relevant local legislation. The study was approved by the National Research Ethics Service (IRAS reference 10/H0716/61).

    Participants

    Full details of the CROMIS-2 ICH study protocol have been published previously.8 Briefly, participants were adults (aged 18 years or above) with spontaneous ICH confirmed on brain imaging (CT or MRI) within the preceding month, with or without a history of anticoagulant use prior to the index event. Patients with ICH secondary to a known structural cause or major head trauma were excluded. Patients were considered to have pre-existing cognitive impairment if they had a formal diagnosis of dementia or cognitive impairment at study entry, or if they scored more than 3.3 on the 16-item IQCODE (Informant Questionnaire on Cognitive Decline in the Elderly), in accordance with previous work.9 History of coronary artery disease was defined as a prior history of angina, myocardial infarction or cardiac revascularisation (percutaneous coronary intervention or coronary artery bypass grafting). APOE genotype was established from peripheral blood samples; the method for this has been previously described.10

    Outcomes

    For the first 6 months after the index event, outcomes were collected using multiple ascertainment methods, as detailed in the previously published study protocol.8 Briefly, these methods included postal questionnaires sent to patients and their general practitioners and notifications from NHS Digital (previously the Health and Social Care Information Centre).8 NHS Digital is a national centralised body that collects data on health and social care in the UK, including ‘hospital episode statistics’ (HES; records of all NHS patient admissions) and information on registered deaths from the Office of National Statistics (death registration is a legal requirement in the UK).

    Outcome data from 6 months to 3 years were compiled from notifications from NHS Digital. HES for all admitted patient care (APC) events were reviewed using the NHS Digital HES Data Dictionary for APC episodes.11 An ‘admission’ was defined as one or more individual episodes, which ended with the patient being discharged to a ‘home destination’ (DISDEST codes 19, 29, 30, 49, 50, 54, 65 and 85) or hospice (DISDEST code 88), or with the death of the patient (DISDEST code 79). The primary diagnosis (DIAG_01 code) was determined using the online version of the WHO International Statistical Classification of Diseases and Related Health Problems.12 A cerebrovascular event was defined as an admission due to a cerebral ischaemic event (G459, I632, I633, I634, I635, I638, I639 and I663), ICH (I610, I611, I612, I614, I615, I616, I618 and I619), other non-traumatic intracranial bleeding events (I609, I620 and I629) or unspecified stroke event (I64X and I678). Outcome events were diagnosed locally and not adjudicated centrally.

    The outcomes of interest were occurrence of a cerebral ischaemic event (either stroke or transient ischaemic attack, TIA) or a further ICH following study entry. Follow-up time was defined as time to first cerebrovascular event, and for those patients who did not have a subsequent cerebrovascular event, follow-up time was defined as time to death. For patients who did not have a cerebrovascular or mortality event, follow-up time was defined as either 3 years following the index event or at the time of the study’s last notification from NHS Digital (31 March 2017), with the earlier date used in these cases. For each analysis (except for competing risk analyses), patients were considered censored if they did not have the event of interest.

    Imaging

    Brain CT imaging was acquired acutely at the time of the index event as part of the patient’s routine clinical care. Imaging analysis was carried out by a clinical research associate trained in neuroimaging rating and blinded to the participant clinical details. Haematoma location was classified using the CHARTS scale13 as lobar (including convexity subarachnoid haemorrhage), deep (involving the basal ganglia or thalamus), cerebellar or brainstem. Non-lobar was defined as the presence of either deep or brainstem haemorrhage; cerebellar haemorrhage was excluded from this definition as this does not have a clear small vessel disease association. CT images were also rated for the presence of lacunes, which were defined in accordance with STRIVE criteria as a ‘round or ovoid, subcortical, fluid-filled cavity (signal similar to CSF) of between 3 mm and about 15 mm in diameter, consistent with a previous acute small subcortical infarct or haemorrhage in the territory of one perforating arteriole’.14 White matter changes were rated on CT images using the Van Swieten score; the highest scores for anterior and posterior regions were combined in order to generate a ‘total’ score (range 0–4).15 Haematoma volume was rated using a previously described semiautomated planimetric method.16 17

    Statistics

    Statistical analysis was performed using Stata (V.11.2). Univariable Cox regression was used to investigate which clinical and imaging variables were associated with the occurrence of an outcome of interest. Multivariable Cox regression analysis was then performed; adjustments were made for all variables with p<0.10 in univariable analyses, in addition to the primary variable of interest (ICH location). The proportional hazards assumption test based on Schoenfeld residuals was applied to all Cox models (univariable and multivariable). Univariable and multivariable competing risk analyses (using the Fine-Gray subdistribution hazard model18 19) were also performed; subdistribution hazard ratios (HRs) are provided. Figures for the cumulative incidence of outcome events were generated using Kaplan-Meier survival analyses.

    Data availability

    Analyses for the CROMIS-2 study are ongoing; once all of these analyses are completed, the CROMIS-2 Steering Committee will consider applications from other researchers for access to anonymised source data.

    Results

    All 1094 patients recruited to CROMIS-2 ICH were included (baseline characteristics are shown in table 1); 447 (40.9%) were lobar ICH, 546 (50.0%) were deep, 65 (6.0%) were cerebellar and 34 (3.1%) occurred in the brainstem. Follow-up was for a total of 2391 patient-years (median 3.00 years, IQR 1.48–3.00 years).

    Table 1

    Baseline characteristics

    Recurrent ICH events

    There were 45 recurrent ICH events (absolute event rate 1.88 per 100 patient-years, 95% CI 1.41 to 2.52 per 100 patient-years); 35 were in patients whose index event was lobar; and 9 in patients presenting with non-lobar ICH. Absolute event rates are provided in table 2.

    Table 2

    Absolute rates for recurrent ICH and cerebral ischaemic events

    In univariable Cox regression analyses (table 3; online supplementary table 1), the following predictors showed associations with recurrent ICH events (p<0.10): increasing age, history of previous cerebral ischaemic events, ICH prior to study entry, presence of at least one APOE ε2 allele and antiplatelet use prior to study entry. There were also associations with the severity of white matter disease (as measured by increasing Van Swieten score), ICH volume and lobar ICH location on baseline imaging (figure 1A). In univariable competing risk regression for recurrent ICH events (online supplementary table 1), where occurrence of a cerebral ischaemic event or death was the competing risk, a similar association with lobar ICH location was observed.

    Supplementary data

    Table 3

    Cox regression analyses for recurrent ICH events

    Figure 1

    Unadjusted Kaplan-Meier failure estimates, comparing patients with lobar and non-lobar ICH, for (A) recurrent ICH and (B) subsequent cerebral ischaemic events. P values are from univariable Cox regression analyses.

    Multivariable Cox regression including variables of interest identified in univariable Cox analyses (listed previously) found that lobar ICH location at presentation remained associated with subsequent ICH occurrence (HR 8.96, 95% CI 3.36 to 23.87, p<0.0001; table 3). A history of cerebral ischaemic events (HR 2.86, 95% CI 1.37 to 5.96, p=0.005), previous ICH (HR 4.86, 95% CI 1.41 to 16.67, p=0.012) and antiplatelet use (HR 2.22, 95% CI 1.09 to 4.50, p=0.028) prior to the index event were also associated with subsequent ICH occurrence. Multivariable completing risk analyses (including the same variables as the adjusted Cox regression) with either subsequent cerebral ischaemic events or death as the competing risk, showed similar results for the association with lobar ICH location (online supplementary table 2).

    Cerebral ischaemic events

    There were 70 cerebral ischaemic events (absolute event rate 2.93 per 100 patient-years, 95% CI 2.32 to 3.70 per 100 patient-years), of which 29 occurred in patients presenting with lobar ICH and 39 in patients with non-lobar ICH. Absolute event rates are provided in table 2.

    In univariable Cox regression analyses (table 4; online supplementary table 3), subsequent cerebral ischaemic events were associated with increasing age, hypercholesterolaemia, atrial fibrillation, alcohol use at study entry, history of previous cerebral ischaemic events, anticoagulant use prior to ICH and increasing Van Swieten Score; there was no association with ICH location (figure 1B). Univariable competing risk regression for subsequent cerebral ischaemic events, with occurrence of recurrent ICH or death as the competing risk, found similar results (online supplementary table 3); the association with age was no longer observed.

    Table 4

    Cox regression analyses for subsequent cerebral ischaemic events

    Multivariable Cox regression (table 4) found significant associations with age (HR 1.04, 95% CI 1.01 to 1.06, p=0.016), alcohol use at study entry (HR 2.16, 95% CI 1.20 to 3.91, p=0.011) and a history of previous ischaemic events (HR 2.17, 95% CI 1.24 to 3.79, p=0.007). There was no association with ICH location (HR 1.13, 95% CI 0.66 to 1.92, p=0.659). Similar results were seen in multivariable completing risk analyses with recurrent ICH or death as the competing event (online supplementary table 4).

    Discussion

    Our main findings are: (1) at 3-year follow-up, there were fewer ICH events than cerebral ischaemic events (45 vs 70); (2) there was a difference in absolute event rates for recurrent ICH events for patients with lobar and non-lobar ICH (3.77 vs 0.69 per 100 patient-years) and lobar ICH location was independently associated with a higher risk of recurrent ICH events; and (3) absolute event rates for subsequent cerebral ischaemic events were similar for lobar and non-lobar groups (3.12 vs 2.97 per 100 patient-years), and there was no association between ICH location and the risk of subsequent cerebral ischaemic events. In addition to ICH location, recurrent ICH events were associated with a history of previous cerebral ischaemic events, previous ICH and antiplatelet use prior to study entry, whereas cerebral ischaemic events were associated with age, alcohol use at study entry and a history of previous cerebral ischaemic events.

    Our results support two recent studies that challenge the idea that the risks of antiplatelet and anticoagulant medications in ICH patients outweigh the benefits. An individual patient data meta-analysis2 of 1012 patients who resumed treatment with oral anticoagulant therapy following spontaneous ICH found that resumption was associated with reduced mortality and all-cause stroke incidence, as well as more favourable outcomes, at 1 year. RESTART3 (the REstart or STop Antithrombotics Randomised Trial) was a prospective multicentre randomised trial of patients taking antiplatelet medications at the time of their ICH; patients were randomised to either restart or discontinue antiplatelet medications following their ICH. RESTART did not find significant changes in either ICH or major vaso-occlusive events with antiplatelet treatment, even in subgroup analyses where these events were considered by index ICH location.20 Our finding that cerebral ischaemic events are more frequent than ICH events suggests that the ischaemic risk in patients with ICH, particularly those with lobar ICH, is underestimated. These data support the argument for further randomised trials of antiplatelet or anticoagulant treatments in ICH patients; there has previously been little appetite for this, due to a perceived lack of clinical equipoise.

    The finding that lobar ICH is associated with a higher ICH recurrence rate is in keeping with previous work,1 5 6 and this is believed to reflect the association of lobar ICH with CAA. We also observed an association between APOE ε2 and recurrent ICH events, and although this association did not reach statistical significance in univariate analyses, it was of reasonable magnitude (HR 1.84); this association was less apparent in the multivariable analysis. This association could reflect the association of APOE ε2 with lobar ICH and CAA, and in particular CAA with vasculopathic ‘haemorrhagic’ changes.21–23 However, lobar haemorrhage is not only due to CAA, with one recent clinico-pathological study observing that of 62 patients with lobar ICH, 26 had absent or mild CAA.24 Given that CAA frequently coexists with other small vessel pathologies in patients with lobar ICH (the same study found that of 36 patients with moderate or severe CAA, 26 also had evidence of deep perforator arteriopathy),24 the increased recurrent ICH risk in these patients might represent more ‘severe’ small vessel disease—be it CAA, deep perforator arteriopathy, or both. We also found an association with a prior history of ICH was associated with subsequent ICH occurrence, suggesting that some individuals are particularly ‘bleeding-prone’, independent of ICH location. The observed association with previous cerebral ischaemic events might reflect that these events (in particular lacunar infarction) are associated with more severe small vessel disease (specifically, deep perforator arteriopathy); the association with prior antiplatelet use might be a surrogate marker for this (although other explanations are possible; this observation could also reflect that those taking antiplatelet medications prior to ICH are more likely to be restarted on them, following discharge). When considering subsequent cerebral ischaemic risk, we did not see an increased number of events in those with non-lobar ICH, which might suggest that the ischaemic risk of deep perforator arteriopathy is overestimated. Taken together, this suggests that factors beyond ICH location are important for identifying those at highest risk of subsequent stroke events and that severe small vessel disease, regardless of subtype, is important.

    The strengths of this study are its large size, its multicentre design and the detailed clinical and imaging data available for participants. Limitations include those inherent to the coding of hospital episodes (with regard to accuracy) and the lack of central adjudication of events. This method of ascertainment could result in some events being missed, for example, if patients were treated in non-NHS facilities (such as those outside the UK or private hospitals), or in the case of minor events, which might not have resulted in a hospital attendance. New neurological symptoms in ICH survivors can represent a recrudescence of previous stroke symptoms, which could be misdiagnosed as a new cerebrovascular event, particularly in the absence of appropriate investigations including MRI brain with diffusion-weighted sequences.25 Additionally, in order to complete competing risk analyses, only the first cerebrovascular event was considered; this work did not explore repeated events, and in patients who had both ischaemic and haemorrhagic events, only the first event was included. The number of outcome events was relatively low, and as a consequence, we were unable to explore the role of ICH location in further detail (ie, in those with cerebellar or brainstem ICH); additionally, we were unable to comment on the clinical severity of the outcome events, as this information was not available. MR data were not available for all patients, and as a consequence we were unable to provide more detailed information on the nature and severity of any underlying cerebral small vessel disease (including the number, presence and distribution of cerebral microbleeds); we acknowledge that CT-based quantification methods of white matter changes and lacunes are less sensitive than equivalent MR measures. We acknowledge that our results might be subject to selection bias, as our cohort only included ICH survivors. We also acknowledge that our cohort is older and has a higher rate of pre-event antiplatelet and anticoagulant use than other cohorts26 (likely to reflect higher rates of comorbidities in our cohort, including atrial fibrillation), which again might limit its generalisability. Finally, as noted above, we did not have information on the prescription of antiplatelet or anticoagulant medications following discharge, or the indications for their use on discharge, which could have influenced our results.

    In conclusion, in this large cohort study, there were fewer ICH events than cerebral ischaemic events at 3 years in ICH survivors. Lobar ICH location is associated with a higher risk of recurrent ICH events than non-lobar ICH, as are some features that may reflect cerebral small vessel disease severity. Outstanding questions remain about whether these associations with lobar ICH occur reflect a single small vessel pathology or the severity of small vessel disease more generally. Further work is needed to disentangle the complex interaction between these small vessel diseases, and their impact on an individual’s future stroke risk.

    Acknowledgments

    We would like to acknowledge Surabhika Lunawat for her assistance in quantifying intracerebral haemorrhage (ICH) volumes for this study.

    References

    View Abstract

    Footnotes

    • Twitter @DrGargiBanerjee, @BleedingStroke, @UCLStrokeRes

    • Collaborators The CROMIS-2 collaborators: Louise Shaw, MD; Kirsty Harkness MD; Jane Sword, MD; Azlisham Mohd Nor, MD; Pankaj Sharma, PhD; Deborah Kelly, MD; Frances Harrington, MD; Marc Randall, MD; Matthew Smith, MD; Karim Mahawish, MD; Abduelbaset Elmarim, MD; Bernard Esisi, MD; Claire Cullen, MD; Arumug Nallasivam, MD; Christopher Price, MD; Adrian Barry, MD; Christine Roffe, MD: John Coyle, MD; Ahamad Hassan, MD; Caroline Lovelock, DPhil; Jonathan Birns, MD; David Cohen, MD; L Sekaran, MD; Adrian Parry-Jones, PhD; Anthea Parry, MD; David Hargroves, MD; Harald Proschel, MD; Prabel Datta, MD; Khaled Darawil, MD; Aravindakshan Manoj, MD; Mathew Burn, MD; Chris Patterson, MD; Elio Giallombardo, MD; Nigel Smyth, MD; Syed Mansoor, MD; Ijaz Anwar, MD; Rachel Marsh, MD; Sissi Ispoglou, MD; Dinesh Chadha, MD; Mathuri Prabhakaran, MD; Sanjeevikumar Meenakishundaram, MD; Janice O’Connell, MD; Jon Scott, MD; Vinodh Krishnamurthy, MD; Prasanna Aghoram, MD; Michael McCormick, MD; Paul O’Mahony, MD; Martin Cooper, MD; Lillian Choy, MD; Peter Wilkinson, MD; Simon Leach, MD; Sarah Caine, MD; Ilse Burger, MD; Gunaratam Gunathilagan, MD; Paul Guyler, MD; Hedley Emsley, MD; Michelle Davis, MD; Dulka Manawadu, MD; Kath Pasco, MD; Maam Mamun, MD; Robert Luder, MD; Mahmud Sajid, MD; Ijaz Anwar, MD; James Okwera, MD; Julie Staals, PhD; Elizabeth Warburton, MD; Kari Saastamoinen, MD; Timothy England, MD; Janet Putterill, MD; Enrico Flossman, MD; Michael Power, MD; Krishna Dani, MD; David Mangion, MD; Appu Suman, MD; John Corrigan, MD; Enas Lawrence, MD; and Djamil Vahidassr, MD.

    • Contributors GB contributed to the design of the project, analysed the data and drafted and revised the manuscript. DW had a major role in the acquisition of data and revised the manuscript for intellectual content. GA contributed to the statistical analysis and interpretation of the data and revised the manuscript for intellectual content. ICH, CS and HH had a major role in the acquisition of data and revised the manuscript for intellectual content. HC, TAY, RA-SS, GL, KM, MMB and HRJ were involved in the design and execution of the CROMIS-2 study (as members of the Study Steering Committee) and revised the manuscript for intellectual content. DJW designed and conceptualised the study, and revised the manuscript for intellectual content.

    • Funding The CROMIS-2 study is funded by the Stroke Association and British Heart Foundation. GB holds a National Institute for Health Research (NIHR) Academic Clinical Fellowship and received funding from the Rosetrees Trust. GA receives funding from the National Institute for Health Research University College London Hospitals (UCLH) Biomedical Research Centre. MMB’s Chair in Stroke Medicine is supported by the Reta Lila Weston Trust for Medical Research. DJW receives research support from the Stroke Association, the British Heart Foundation and the Rosetrees Trust. This work was undertaken at University College London Hospitals and Univeristy College London, which receives a proportion of funding from the Department of Health’s NIHR Biomedical Research Centres funding scheme.

    • Competing interests HC has received institutional research support from Bayer; honoraria for lectures and an Advisory Board from Bayer, diverted to a local charity; and travel/accommodation expenses for participation in scientific meetings covered by Bayer. GL acts as a consultant for Bayer/Janssen, BMS/Pfizer, Medtronic, Boehringer Ingelheim, Novartis, Verseon and Daiichi-Sankyo, and as a speaker for Bayer, BMS/Pfizer, Medtronic, Boehringer Ingelheim and Daiichi-Sankyo; no fees are directly received personally. DJW has received honoraria for consultancy and lectures from Bayer, Portola and Alnylam. The remaining authors report no disclosures or conflicts of interest relevant to the manuscript.

    • Patient consent for publication Not required.

    • Ethics approval The study was approved by the National Research Ethics Service (IRAS reference 10/H0716/61).

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Data availability statement Data are available on reasonable request. Analyses for the CROMIS-2 study are ongoing; once all of these analyses are completed, the CROMIS-2 Steering Committee will consider applications from other researchers for access to anonymised source data.

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.