Article Text

Original research
Haptoglobin genotype and outcome after spontaneous intracerebral haemorrhage
  1. Isabel Charlotte Hostettler1,
  2. Matthew J Morton2,
  3. Gareth Ambler3,
  4. Nabila Kazmi4,
  5. Tom Gaunt4,
  6. Duncan Wilson1,
  7. Clare Shakeshaft1,
  8. H R Jäger5,
  9. Hannah Cohen6,
  10. Tarek A Yousry5,
  11. Rustam Al-Shahi Salman7,
  12. Gregory Lip8,
  13. Martin M Brown1,
  14. Keith Muir9,
  15. Henry Houlden10,
  16. Diederik O Bulters11,
  17. Ian Galea12,
  18. David J Werring1
  19. On behalf of the CROMIS-2 collaborators
    1. 1 Stroke Research Centre, University College London, Queen Square Institute of Neurology, London, UK
    2. 2 Clinical Neurosciences, Clinical and Experimental Sciences, Faculty of Medicine, University of Southampton, Southampton, UK
    3. 3 Department of Statistical Science, University College London, London, UK
    4. 4 MRC Integrative Epidemiology Unit (IEU), Faculty of Health Sciences, University of Bristol, Bristol, UK
    5. 5 Neuroradiological Academic Unit, Department of Brain Repair and Rehabilitation, University College London, Queen Square Institute of Neurology, London, UK
    6. 6 Department of Haematology, University College London, London, UK
    7. 7 Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, UK
    8. 8 Liverpool Centre for Cardiovascular Science, University of Liverpool and Liverpool Heart and Chest Hospital, Liverepool, UK
    9. 9 Institute of Neuroscience and Psychology, Queen Elizabeth University Hospital, University of Glasgow, Glasgow, UK
    10. 10 MRC Centre for Neuromuscular Diseases, Department of Neuromuscular Diseases, UCL Queen Square Institute of Neurology and National Hospital for Neurology and Neurosurgery, London, UK
    11. 11 Department of Neurosurgery, University Hospital Southampton NHS Foundation Trust, Southampton, UK
    12. 12 Faculty of Medicine, University of Southampton, Southampton, UK
    1. Correspondence to Dr David J Werring, Stroke Research Centre, University College London, Institute of Neurology, London WC1N 3BG, UK; d.werring{at}ucl.ac.uk

    Abstract

    Objective Haptoglobin is a haemoglobin-scavenging protein that binds and neutralises free haemoglobin and modulates inflammation and endothelial progenitor cell function. A HP gene copy number variation (CNV) generates HP1 and HP2 alleles, while the single-nucleotide polymorphism rs2000999 influences their levels. The HP1 allele is hypothesised to improve outcome after spontaneous (non-traumatic) intracerebral haemorrhage (ICH). We investigated the associations of the HP CNV genotype and rs2000999 with haematoma volume, perihaematomal oedema (PHO) volume, functional outcome and mortality after ICH.

    Methods We included patients with neuroimaging-proven ICH, available DNA and 6-month follow-up in an observational cohort study (CROMIS-2). We classified patients into three groups according to the HP CNV: 1–1, 2–1 or 2–2 and also dichotomised HP into HP1-containing genotypes (HP1-1 and HP2-1) and HP2-2 to evaluate the HP1 allele. We measured ICH and PHO volume on CT; PHO was measured by oedema extension distance. Functional outcome was assessed by modified Rankin score (unfavourable outcome defined as mRS 3–6).

    Results We included 731 patients (mean age 73.4, 43.5% female). Distribution of HP CNV genotype was: HP1-1 n=132 (18.1%); HP2-1 n=342 (46.8%); and HP2-2 n=257 (35.2%). In the multivariable model mortality comparisons between HP groups, HP2-2 as reference, were as follows: OR HP1-1 0.73, 95% CI 0.34 to 1.56 (p value=0.41) and OR HP2-1 0.5, 95% CI 0.28 to 0.89 (p value=0.02) (overall p value=0.06). We found no evidence of association of HP CNV or rs200999 with functional outcome, ICH volume or PHO volume.

    Conclusion The HP2-1 genotype might be associated with lower 6-month mortality after ICH; this finding merits further study.

    • stroke
    • MRI
    • congnition
    • cerebrovascular disease
    View Full Text

    Statistics from Altmetric.com

    Introduction

    Spontaneous (non-traumatic) intracerebral haemorrhage (ICH) is the most devastating form of stroke with a mortality of about 40% at 1 month, and 65% at 1 year.1–3 Patients who survive frequently remain severely disabled.4 Moreover, incidence of ICH is increasing in the elderly population,5–7 in part due to increasing use of oral anticoagulation (OAC).5–7

    Spontaneous ICH results from bleeding into the brain parenchyma arising from the rupture of an arterial vessel, most often (>80%) a small arteriole affected by cerebral small vessel diseases (SVD). The most common sporadic SVD that cause ICH are deep perforator arteriopathy (also termed hypertensive arteriopathy or arteriolosclerosis) and cerebral amyloid angiopathy (CAA). A minority of ICH (<20%) is caused by structural or macrovascular bleeding sources such as tumours, arteriovenous malformations, cavernomas or fistulas. Deep perforator arteriopathy is associated with hypertension and is a frequent cause of deep ICH (in the basal ganglia, thalamus, and brainstem); CAA is caused by amyloid beta deposition in cortical and leptomeningeal blood vessels and is a key cause of lobar ICH.

    Haptoglobin is an acute-phase protein which neutralises free haemoglobin by binding it, and in doing so targets haemoglobin to the CD163 receptor for clearance.8–15 Haptoglobin prevents the toxic and inflammatory effects of haemoglobin by shielding its iron-containing pocket, and preventing its breakdown into haem and iron, which consequently cause cytotoxicity and brain oedema.8–15 The HP gene has a copy number variation (CNV), which leads to two co-dominant alleles: HP1 and HP2. Three different HP CNV genotypes exist: HP1-1, HP2-1 and HP2-2, and their respective protein products differ in molecular size and haemoglobin-binding capacity.15–17 A previous study demonstrated some evidence that patients with the HP2 allele have a larger haematoma volume, though the underlying mechanisms remain unknown.18 An increase in haematoma volume may be accompanied by more perihaematomal oedema (PHO).18 19 ICH and PHO volume have been demonstrated to influence functional outcome.18 19 A previous study reported worse functional outcome for patients with HP2 allele (HP2-1 or 2–2) compared with HP1-1 patients as well as some evidence for increased mortality for each HP2 allele.18 The HP CNV might be associated with functional outcome after ICH through differences in haemoglobin clearance and protection from the cytotoxic and inflammatory effects of haemoglobin breakdown products. However, most previous studies investigating haptoglobin in ICH are based on investigations in rodents.

    The single-nucleotide polymorphism (SNP) rs2000999 accounts for up to 50% of variation in circulating haptoglobin levels in the blood independently of the HP CNV.20 The combined use of the HP CNV and rs2000999 has been suggested as an important genetic tool to discriminate between two potential mechanisms underlying differences between HP1 and HP2 alleles: haptoglobin expression level and functional differences in haptoglobin protein products.21

    We performed a comprehensible multivariable study investigating the influence of the HP CNV and rs2000999 SNP on functional outcome and mortality after ICH. We also aimed to assess the influence of the HP CNV and the rs2000999 SNP on ICH volume and oedema extension distance (OED).

    Methods

    Data collection

    We considered patients, of predominantly Caucasian descent, with spontaneous ICH and available blood samples recruited into the Clinical Relevance of Microbleeds in Stroke ICH study.22 We defined spontaneous ICH as a non-traumatic haemorrhage into the brain parenchyma, presumed due to cerebral SVD after the exclusion of patients with an underlying structural or macrovascular cause.

    We collected detailed information on demographics, risk factors, medication, clinical presentation and radiological data. A diagnosis of hypertension, hypercholesterolaemia and diabetes mellitus was present if reported by the patient, stated on medical records or if either drug treatment or any other form of advice (including lifestyle changes) was given. Smoking was defined as current and previous use. All patients had acute brain imaging with CT. Written informed consent was obtained from all participants, or a relative or representative. We excluded patients <18 years, patients without available or adequate CT scan. Patients with a CT scan after 72 hours from symptom onset were excluded from the primary ICH and PHO volume analysis.18 23 24 We classified ICH location into lobar, deep (basal ganglia, thalamus), cerebellar and brainstem according to a validated rating scale.25 Our outcomes were death and functional outcome at 6 months (measured by the modified Rankin Scale score (mRS) dichotomised into favourable (mRS 0–2) or unfavourable (mRS 3–6) categories).

    Haptoglobin genotyping

    To determine the HP CNV, we optimised a high-throughput qPCR genotyping assay as described previously.26 The assay amplified a region in the 5′ terminal of the HP gene’s first exon as an internal control (HP5′), and the breakpoint of the HP duplication (HP2). The HP2/HP5′ ratio (theoretically either 0, 1 or 2) was used to determine the genotype as HP1-1, HP2-1 or HP2-2, respectively. Samples were run in triplicates; triplicates with a HP2/HP5′ ratio coefficient of variation >10% were reassayed. A second method of HP genotyping by PCR27 was performed on samples with HP2/HP5′ ratio values between 0.46 and 077, in order to confirm the HP CNV genotype. Rs2000999 was genotyped using Kompetitive allele specific PCR assay technology28 (LGC Genomics Limited, Hertfordshire, UK); the call rate was 97.3%.

    Measurement of ICH and PHO volume

    We measured ICH and PHO volume as previously described via a semi-automated, threshold-based approach.29 PHO was measured by the OED using a previously described formula19; the rationale behind using OED is that PHO extends a consistent mean linear distance from the border of the ICH, independently of its volume.

    Statistical analysis

    We present categorical variables using frequency and percentages, continuous variables using mean±SD. We transformed ICH and PHO volume with cube root transformation to satisfy statistical normal distribution assumptions. We conducted a post hoc sensitivity analysis comparing patients with ICH volume and OED before and after 72 hours.

    We assessed the distribution of the HP CNV and rs2000999 SNP in the CROMIS-2 cohort compared with ALSPAC (Avon Longitudinal Study of Parents and Children) cohort of healthy individuals, which we used as controls. ALSPAC is a general population cohort study;30 31 HP genetic data and rs2000999 SNP data were available from 927 and 748 participants respectively. The ALSPAC study website (http://www.bristol.ac.uk/alspac/researchers/our-data/) contains details of all the data available through a fully searchable data dictionary and variable search tool. To evaluate the HP1 allele, we also assessed the HP CNV as a dichotomised variable (HP1-1 and HP2-1 vs HP2-2) according to our prespecified analysis plan.

    We first performed univariable analyses for each of the four outcomes separately with demographic, clinical and radiological variables of interest. We subsequently fitted multivariable logistic regression models with significant variables from the univariable analysis in addition to prespecified variables. For the analysis of ICH and OED volume, we adjusted the models with the prespecified variables: time from event to imaging, location of ICH, systolic blood pressure (SBP), HP CNV and rs200999 SNP. For functional outcome and mortality analysis, we fitted the multivariable model with the prespecified variables: age, sex, hypertension, OAC, HP CNV and rs200999 SNP. Additionally, we fitted the multivariable models with variables that were statistically significant at the 20% level in the univariable analysis.

    We investigated whether there were interactions between different variables. However, no interaction reached our prespecified significant threshold for interactions of p<0.001 (chosen to guard against overfitting) and were therefore not included in the models.32

    Statistical analysis was performed using STATA V.15.

    Ethical approval

    The CROMIS-2 study was approved by the local Ethics Committee (reference: 10/H0716/64).

    Results

    For the primary analysis of functional outcome at 6 months, we included 732 patients. One DNA sample was uncallable for the HP CNV and 20 uncallable for the rs2000999 SNP. For the secondary analyses of ICH volume and PHO, we included 709 patients with an available CT scan (figure 1). OED mas measured at a mean of 10 hours from ICH onset. Patients who were genotyped (n=844) were not different to those without DNA (n=250) with regard to baseline characteristics and risk factor profile (data not shown). The rs2000999 genotype frequency in CROMIS-2 was as expected when compared with ALSPAC (online supplementary table 1). However, compared with ALSPAC, CROMIS-2 patients less often had the HP2-2 CNV. We found no systematic difference in demographics, comorbidities and ICH characteristics between those with and without available outcome variable (data not shown).

    Supplementary data

    Figure 1

    Patient selection flow diagram.

    Mortality

    Of 731 patients with available follow-up and genotype data, 318 (43.5%) were female, and 112 died within 6 months (15.3%).

    The distribution of the HP CNV was: 132 HP1-1 (18.1%); 342 HP2-1 (46.8%); and 257 HP2-2 (35.2%). The sistribution of the SNP allele was: 27 A:A (3.8%); 234 A:G (32.9%); and 451 G:G (63.3%). 20 samples were not callable (2.7%).

    Patients who died were older, more frequently female, more frequently on OAC, had a lower GCS on admission (GCS<8),higher ICH and PHO volumes, and intraventricular (IV) extension. Results of the univariable analysis are shown in online supplementary table 2.

    The mortality according to HP CNV was as follows: HP1-1 18.2%; HP2-1 12.6%; HP2-2 17.5%. In the multivariable model (n=608), mortality comparisons between the HP groups, with HP2-2 as a reference group, were as follows: OR HP1-1 0.73, 95% CI 0.34 to 1.56 (p value=0.41) and OR HP2-1 0.5, 95% CI 0.28 to 0.89 (p value=0.02) (overall p value=0.06, table 1).

    Table 1

    Factors associated with 6 month mortality after ICH in an adjusted multivariable logistic regression model

    When dichotomising HP into HP1-1/2-1 versus HP2-2, there was evidence for lower mortality with the HP1 allele compared with HP2-2 (OR 0.55, 95% CI 0.31 to 0.95, p=0.03, online supplementary table 3). As expected, there was also evidence for higher mortality with increasing age (OR 1.11, 95% CI 1.07 to 1.14, p<0.001), lower GCS on admission <9 (OR 4.37, 95% CI 1.39 to 13.73, p=0.01) and increasing cube root ICH volume in mL (OR 1.99, 95% CI 1.45 to 2.74, p<0.001).

    We further investigated the association between mortality and HP CNV across tertiles of all the covariates included in the multivariable model as a post hoc analysis. Mortality differed between the HP groups for older patients (>80 years) with lower (<12.2 mL) ICH volume: in this subgroup, mortality was 26% for HP1-1, 14% for HP2-1% and 42% for HP2-2. Patients died at a median of 3.8 months after ICH. There was no difference (early vs late death) in the time of death after ICH across HP CNV or rs2000999 groups, in the overall cohort or the subgroup of >80 years and <12.2 mL ICH volume (regression data not shown, online supplementary figure 1). The mortality rate was similar across the HP groups for the remaining patients: 15% for HP1-1, 12% for HP2-1% and 12% for HP2-2. The association between mortality and HP CNV was confirmed across tertiles of all the other covariates. Finally, we investigated covariates not included in the multivariable model, to see whether they differed across HP genotypes, but found no bias to explain the association between mortality and HP CNV (data not shown).

    Supplementary data

    Functional outcome

    Of 731 patients, 444 (60.7%) suffered an unfavourable outcome (mRS 3–6). Dichotomised unfavourable mRS according to HP CNV was as follows: HP1-1 64.4%; HP2-1 59.7%; HP2-2 60.3%.

    Patients with an unfavourable outcome were older, more frequently female, on OAC, more frequently had hypertension, hypercholesterolaemia, presented with a lower GCS (GCS of 3–8), had a higher ICH and PHO volume and IV extension. See online supplementary table 2 for univariable analysis.

    In the multivariable model (n=623) age (OR 1.04, 95% CI 1.02 to 1.06, p<0.001), female sex (OR 2.31, 95% CI 1.58 to 3.37, p<0.001) and the cube root of the ICH volume (OR 1.5, 95% CI 1.22% to 1.85, p<0.001) were significantly associated with functional outcome (table 2). Neither HP CNV nor rs2000999 SNP were associated with functional outcome.

    Table 2

    Factors associated with unfavourable outcome after ICH in an adjusted multivariable regression model

    ICH volume and OED

    Of the 731 patients included in the functional analysis, 709 had a CT scan available, and of these 68 were >72 hours after symptom onset (figure 1). Of the remaining 641 individuals, 453 (70.7%) had a scan <24 hours, 172 (26.8%) between 24 and 48 hours and 16 (2.5%) between 48 and 72 hours.

    See figure 2 for the association of the HP CNV and SNP with OED and ICH volume.

    Figure 2

    (A) Differences in OED in haptoglobin genotype and SNP. (B) Differences in ICH volume in haptoglobin genotype and SNP. ICH, intracerebral haemorrhage; OED, oedema extension distance; SNP, single-nucleotide polymorphism.

    The mean ICH volume was 13.8 mL (±18.82 SD), the mean PHO volume 19.54 mL (±20.56 SD), and the mean OED 0.51 cm (±0.23 SD). Variables significantly associated with ICH volume in the univariable analysis are listed in the online supplementary table 3.

    In the fitted multivariable model (n=604), ICH location (overall p<0.001) and intraventricular extension (coefficient 0.53, 95% CI 0.37 to 0.68, p<0.001) were associated with greater ICH volume (table 3). Neither HP CNV nor the SNP rs2000999 were associated with ICH volume.

    Table 3

    Factors associated with the cube root ICH volume in an adjusted multivariable regression model

    After dichotomising the HP CNV into HP1-1/2-1 versus HP2-2, we did not observe any evidence of an association in univariable or multivariable analyses (p=0.39 (online supplementary table 4) and p=0.6, respectively (data not shown)). Similar results were observed when dichotomising HP CNV into HP1-1 versus HP2-1/2-2 (online supplementary table 4).

    Oedema extension distance

    Variables significantly associated with OED in the univariable analysis are listed in online supplementary table 4. For comparison of HP CNV and SNP for ICH volume and OED, see figure 2.

    In the multivariable linear regression model (n=623), ICH location (with lobar and deep ICH locations featuring a longer OED and with a brainstem location featuring a shorter OED, compared with the reference group of cerebellar location, overall p<0.001) and antihypertensive medication (coefficient −0.09, 95% CI −0.16 to −0.02, p=0.01) were significantly associated with OED (table 4). Neither the univariable nor multivariable analysis showed evidence of association of HP CNV or rs2000999 SNP with OED.

    Table 4

    Factors associated with size of oedema extension distance in an adjusted multivariable regression model

    Similar to the ICH volume model, dichotomising HP did not yield any evidence of association in univariable and multivariable models (data not shown).

    Discussion

    In this large prospective, multicentre cohort study, HP was not associated with 6-month functional outcome after ICH, as assessed by the dichotomised mRS. The HP CNV distribution was comparable to that reported in a previous study, apart from a slightly higher proportion of HP1-1 patients and lower proportion of HP2-2.18 Despite the larger sample size, we could not replicate this previous study’s finding of an association of the HP2 allele with functional outcome.18

    However, we found evidence that mortality was lower in HP2-1 patients compared with HP2-2 homozygotes; our post hoc analyses suggest that this observation is mostly driven by older patients with low ICH volumes. No association with mortality was found for the rs2000999 SNP (which is associated with haptoglobin expression level).21 This suggests that any link between the HP CNV and mortality is mediated by factors other than haptoglobin expression.

    While the HP CNV’s association with mortality could have been confounded by bias in a variable not included the model, we did not find any evidence for this in any variables we measured. Such a factor could still remain unidentified, but a more likely explanation is that patients who died did not contribute to functional outcome analysis. We found evidence of HP2-2 missingness (of subjects of a particular genotype, in this case HP2-2), when comparing CROMIS-2 with ALSPAC cohorts, which might suggest that the HP2-2 genotype confers a mortality risk.

    We confirmed previous results showing evidence towards increased mortality with HP2-2,18 but did not observe a unidirectional dose response of HP alleles in a direction of increasing or decreasing mortality across HP genotypes (mortality: HP1-1 18.2%; HP2-1 12.6%; HP2-2 17.5%). The lower mortality in HP2-1 individuals could be a chance finding. Another possible but unlikely explanation is heterozygote advantage or heterosis.33 At a molecular level, the HP1 allele might protect against the deleterious effect of the HP2 allele only when the two alleles are present together in HP2-1 individuals. Both HP1 and HP2 alleles scavenge haemoglobin, with HP2 being superior,34 35 and this confers a beneficial effect. However, HP2 has additional off-target effects which are deleterious, mostly proinflammatory.36 In HP2-2 individuals, the better haemoglobin scavenging potential of HP2 versus HP1 is offset by its proinflammatory effects, so that mortality is similar in HP1-1 and HP2-2 individuals. In HP2-1 individuals, the HP1 allele may be negating the deleterious effect of HP2, so that a greater benefit is observed in HP2-1 individuals than is expected by simple codominance of the two alleles.

    We did not confirm previous findings of worse functional outcome in patients with HP2 allele, which could be due to the significantly smaller cohort size and statistical power of the previous study, with potential for a chance finding.18

    PHO develops over a continuous period of time in three main stages. It peaks after 2 weeks, but its evolution is most rapid in the first 2–3 days.37 PHO is thought to be mediated by a process of toxicity and inflammation.19 37 We hypothesised that by modulating neurotoxicity and inflammatory processes haptoglobin might have influenced PHO and functional outcome.38 However, we did not find any association of HP genetic variants (CNV or the rs2000999 SNP) with OED. Similarly, HP genetic variants were not associated with ICH volume, which, like haemtoma expansion, is more likely to be driven by other factors including hydrostatic pressure at the bleeding point.18

    Despite having a large cohort available, we could not replicate the previous study’s reported finding of an association of the HP2 allele with larger ICH volumes and IV extension.18 Since ICH volume and OED was assessed on CT scans performed within 72 hours of symptom onset, we cannot exclude an association of HP with ICH volume or OED after this timepoint, although our exploratory analysis of scans beyond 72 hours (n=68) and found no difference in ICH volume and OED across HP genotypes (for both CNV and rs2000999 SNP) (data not shown).

    We found that long-term antihypertensive medication prior to ICH event is independently associated with decreased OED, even after correcting for SBP. It is possible that patients on antihypertensive medication could have reduced sympathetic activity and inflammatory response when ICH occurs,39 a hypothesis that merits further study. As we did not collect follow-up scans, we cannot comment on a potential influence of SBP on haematoma growth.

    Our study has strengths. Our prospective, multicentre study is the largest on HP and ICH to date, and should be generalisable to Caucasian populations. We collected detailed baseline clinical and brain imaging data and undertook multivariable regression analysis adjusting and correcting for important predictors of all four outcomes, and took exceptional care to control for covariates.

    However, our study also has limitations. Since we obtained informed or proxy consent, our study is biassed towards ICH survivors with less severe ICH than would be included in an unselected incident ICH population. However, it is likely that any protective effect of HP is most relevant in ICH patients who survive the acute period. Additionally, CT scans at multiple timepoints were not available and therefore we could not assess the influence of HP CNV and rs200999 SNP on ICH, PHO or OED expansion over time. We also did not have data on the exact time interval between the ICH and CT scan. However, in a post hoc sensitivity analysis, ICH volume before and after 72 hours was very similar, although OED was larger in patients with first imaging after 72 hours. As PHO increases beyond 72 hours further studies are needed to assess an influence of the HP CNV and rs2000999 SNP on later oedema expansion. Although we excluded patients without blood samples available for genetic analysis, there were no systematic differences in demographics, comorbidities and ICH characteristics between those with and without genetic data available. Finally, it would have been interesting to study plasma and cerebrospinal fluid haptoglobin levels in relation to HP genetic variants, but unfortunately these were not available.

    Conclusion

    We investigated the association of HP genetic variation (the HP CNV and the rs2000999 SNP) in a large prospective cohort study of 731 ICH patients. We found evidence in support of a lower mortality with the HP2-1 genotype, but not functional outcome, ICH volume or OED. While HP genotype may not have impact on 6-month functional outcome, upregulating or supplementing haptoglobin may still be of benefit, as demonstrated in animal studies,40 so better understanding how different haptoglobin types associate with mortality and functional outcome remains important. A future meta-analysis may be appropriate to confirm our observations, and longer follow-up may be needed in case there is an association with longer term outcome.

    Acknowledgments

    We are extremely grateful to all patients, hospital staff and researcher who took part in this study. We also want to thank the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists and nurses.

    References

    View Abstract

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Footnotes

    • IG and DJW are joint senior authors.

    • Twitter @Isabel Hostettler, @BleedingStroke

    • Contributors ICH: design and conceptualised study; acquisition of data; performed laboratory work; analysed the data; drafted the manuscript; revised the manuscript. MJM: performed laboratory work; analysed the data; drafted the manuscript; revised the manuscript. GA: design and conceptualised study; analysed the data; drafted the manuscript; revised the manuscript. NK and TG: analysed the data; drafted the manuscript; revised the manuscript. DW: acquisition of data; analysed the data; drafted the manuscript; revised the manuscript. CS and HRJ: design and conceptualised study; acquisition of data; revised the manuscript. HC, RA-SS, GL, MMB, KM and HH: design and conceptualised study; revised the manuscript. TY: revised the manuscript. DOB: design and conceptualised study; acquisition of data; analysed the data; drafted the manuscript; revised the manuscript. IG: design and conceptualised study; interpreted the data; revised the manuscript for intellectual content. DJW: design and conceptualised study; interpreted the data; drafted the manuscript; revised the manuscript; obtained funding for the study.

    • Funding DJW and DW received funding from the Stroke Foundation/British Heart Foundation. This work was undertaken at UCLH/UCL which receives a proportion of funding from the Department of Health’s National Institute for Health Research (NIHR) Biomedical Research Centers funding scheme. MJM and IG received funding from the Medical Research Council (MR/L01453X/1). NK received funding from Cancer Research UK program grant C18281/A19169. The UK Medical Research Council (MRC) and Wellcome Trust (Grant ref: 102215/2/13/2) and the University of Bristol provide core support for ALSPAC. This publication is the work of the authors and ICH will serve as guarantor of the contents of this paper. A comprehensive list of grant funding is available on the ALSPAC website (http://www.bristol.ac.uk/alspac/external/documents/grant-acknowledgements.pdf).

    • Competing interests None declared.

    • Patient consent for publication Not required.

    • Ethics approval Ethical approval for the study was obtained from the ALSPAC Ethics and Law Committee and the Local Research Ethics Committees.

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

    • Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information.

    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.