NULL HYPOTHESIS

Haptoglobin genotype and aneurysmalsubarachnoid hemorrhageIndividual patient data analysis

Ben Gaastra MRCS Dianxu Ren PhD Sheila Alexander PhD Ellen R Bennett PhD

Dawn M Bielawski PhD Spiros L Blackburn MD Mark K Borsody MD PhD Sylvain Dore PhD

James Galea FRCS(SN) PhD Patrick Garland PhD Tian He MS Koji Iihara MD PhD

Yoichiro Kawamura MD PhD Jenna L Leclerc MD PhD James F Meschia MD Michael A Pizzi DO PhD

Rafael J Tamargo MDWuyang Yang MDMS Paul A Nyquist MDMPHdaggerDiederik O Bulters FRCS(SN)dagger and

Ian Galea FRCP PhDdagger

Neurologyreg 201992e2150-e2164 doi101212WNL0000000000007397

Correspondence

Dr Galea

IGaleasotonacuk

AbstractObjectiveTo perform an individual patientndashlevel data (IPLD) analysis and to determine the relationshipbetween haptoglobin (HP) genotype and outcomes after aneurysmal subarachnoid hemor-rhage (aSAH)

MethodsThe primary outcome was favorable outcome on the modified Rankin Scale or GlasgowOutcome Scale up to 12 months after ictus The secondary outcomes were occurrence ofdelayed ischemic neurologic deficit radiologic infarction angiographic vasospasm and trans-cranial Doppler evidence of vasospasm World Federation of Neurological Surgeons (WFNS)scale Fisher grade age and aneurysmal treatment modality were covariates for both primaryand secondary outcomes As preplanned a 2-stage IPLD analysis was conducted followed bythese sensitivity analyses (1) unadjusted (2) exclusion of unpublished studies (3) all per-mutations ofHP genotypes (4) sliding dichotomy (5) ordinal regression (6) 1-stage analysis(7) exclusion of studies not in Hardy-Weinberg equilibrium (HWE) (8) inclusion of studieswithout the essential covariates (9) inclusion of additional covariates and (10) including onlycovariates significant in univariate analysis

ResultsEleven studies (5 published 6 unpublished) totaling 939 patients were included Overall thestudy population was in HWE Follow-up times were 1 3 and 6 months for 355 516 and 438patients HP genotype was not associated with any primary or secondary outcome No trendswere observed When taken through the same analysis higher age and WFNS scale wereassociated with an unfavorable outcome as expected

ConclusionThis comprehensive IPLD analysis carefully controlling for covariates refutes previous studiesshowing that HP1-1 associates with better outcome after aSAH

RELATED ARTICLE

EditorialHaptoglobin andhemoglobin insubarachnoid hemorrhageA tale of 2 globins

Page 831

These authors contributed equally to this work as first authors

daggerThese authors contributed equally to this work as senior authors

From the Wessex Neurological Centre (BG DOB IG) University Hospital Southampton NHS Foundation Trust UK School of Nursing (DR SA) and Department of Biostatistics(DR TE) University of Pittsburgh PA Department of Neurology (ERB) Duke University School of Medicine Durham NC NeuroSpring (DMB MKB) Dover DE Department ofNeurosurgery (SLB) University of Texas Health Science Center at Houston Department of Anesthesiology Neurology Psychiatry Psychology Pharmaceutics and Neuroscience(SD JLL) College of Medicine Center for Translational Research in Neurodegenerative Disease McKnight Brain Institute University of Florida Gainesville Brain Injury ResearchGroup (JG) Division of Cardiovascular Sciences (University of Manchester) Salford Royal NHS Foundation Trust UK Clinical Neurosciences Clinical amp Experimental Sciences (PGIG) Faculty of Medicine University of Southampton UK Department of Neurosurgery (KI YK) Graduate School of Medical Sciences Kyushu University Fukuoka Japan De-partment of Neurology (JFM MAP) Mayo Clinic Jacksonville FL and Division of Cerebrovascular Neurosurgery (RJT) and Departments of Neurology AnesthesiaCritical CareMedicine and Neurosurgery (WY PAN) Johns Hopkins University School of Medicine Baltimore MD

Go to NeurologyorgN for full disclosures Funding information and disclosures deemed relevant by the authors if any are provided at the end of the article

e2150 Copyright copy 2019 American Academy of Neurology

Copyright copy 2019 American Academy of Neurology Unauthorized reproduction of this article is prohibited

Aneurysmal subarachnoid hemorrhage (aSAH) survivors ex-perience significant morbidity1 The strongest predictor oflong-term outcome is the World Federation of NeurologicalSurgeons (WFNS) grade2 but it explained only 12 of thevariance in outcome as determined by the Glasgow OutcomeScale (GOS)3 in the largest study of outcome prediction inaSAH4 Some studies have suggested that haptoglobin(HP) genotype may also influence outcome5ndash9 but resultsare conflicting10 A recent meta-analysis found that theHP2 allele was associated with a worse short-term but notlong-term outcome11 This meta-analysis had limitationsincluding small study sizes heterogeneity in outcomeclassification and inability to control for covariates Therelative contribution of HP genotype to outcome afteraSAH in a large cohort relative to the WFNS and othercovariates remains unknown

In humans there are 2 HP alleles HP1 and HP2 and anindividual can be 1 of 3 genotypes HP1-1 HP2-1 andHP2-2The main potential mechanism of theHP effect is through thefunction of its protein (Hp) as a scavenger of extracellularhemoglobin (Hb)10 Because the Hp-Hb scavenging system isactive in the CNS after aSAH12 there is a strong biologicalrationale to hypothesize that HP genetic variation influencesaSAH outcome We therefore conducted an individualpatientndashlevel data (IPLD) analysis of all identified publishedand unpublished studies to investigate the relationship be-tween HP genotype and outcome after aSAH

MethodsThe IPLD analysis was conducted in accordance withPreferred Reporting Items for Systematic Reviews andMeta-AnalysesndashIndividual Participant Data (PRISMA-IPD)guidelines13 To ensure rigor we enrolled early pro-fessional statistical input publicly deposited a protocol de-fining a priori the outcomes and analytic strategy in June201714 and included all published and unpublished studiesidentified before analysis All individual studies had approvalfrom the respective institutions and the overall IPLD anal-ysis had institutional ethics approval from the University ofSouthampton

Search strategyPublished studies were identified by PubMed and Web ofScience searches conducted in January 2017 with the keywords subarachnoid hemorrhage or subarachnoid

hemorrhage and haptoglobin including reference lists withinpublications Abstracts were screened and then full articleswere reviewed for eligibility according to the inclusionexclusion criteria Two study investigators (BG and IG)conducted the search Unpublished studies were identifiedwith the same search terms in Google and via the professionalnetwork of the authors in 3 continents (United KingdomUnited States and Japan)

Inclusionexclusion criteriaPublished and unpublished studies were eligible for inclusionand there were no restrictions on study design Inclusioncriteria consisted of (1) confirmed aSAH at age gt18 years (2)HP genotype or phenotype available (3) all essential cova-riates available (see below) (4) data available and contractualagreement reached by March 31 2017 and (5) primaryoutcome measures available at 1 month (plusmn2 weeks) andor 3months (plusmn15 months) andor 6 months (45ndash12 months) ofaSAH If gt1 outcome was available within each of these timeframes the one closest to 1 3 or 6 months was used The onlyexclusion criterion was non-aSAH

Data collection and managementThe lead authors of all published and unpublished studiesidentified with the search strategy were contacted verbally orby e-mail and were invited to join the study IPLD for patientsmeeting the inclusionexclusion criteria was requested inspreadsheet format and was stored on a secure server at theUniversity of Southampton UK The type of data requestedhas been published14 (additional Methods available fromEprints eprintssotonacuk426525) Data were collatedaccording to study center and encoded with study and patientidentifiers to blind the statistical team

Quality controlStudies were assessed for risk of bias with the Newcastle-Ottawa Scale The data underwent a number of quality con-trol checks Automated screens were conducted to identifynonsensical values (eg out-of-range modified Rankin Scale[mRS] and GOS scores impossible age) and to check internalconsistency (eg mRS and GOS scores) Data descriptiveswere used to compare with the expected norm and therebyidentify potential errors Hardy-Weinberg equilibrium(HWE)was assessed as a marker for case missingness Manualdata checks were performed by 4 authors (BG DOB DRand IG) independently If any inconsistency or missing datawere identified or if further clarification was required theindividual study lead was contacted

GlossaryaSAH = aneurysmal subarachnoid hemorrhage CI = confidence interval DCI = delayed cerebral ischemia GOS = GlasgowOutcome ScaleHb = hemoglobinHp = haptoglobinHP = haptoglobin geneHWE = Hardy-Weinberg equilibrium IPLD =individual patientndashlevel data mRS = modified Rankin Scale OR = odds ratio PRISMA-IPD = Preferred Reporting Items forSystematic Reviews and Meta-AnalysesndashIndividual Participant Data SAH = subarachnoid hemorrhage TCD = transcranialDoppler WFNS = World Federation of Neurological Surgeons

NeurologyorgN Neurology | Volume 92 Number 18 | April 30 2019 e2151

Copyright copy 2019 American Academy of Neurology Unauthorized reproduction of this article is prohibited

Primary outcomeThe per-protocol primary outcome was the mRS15 or GOSscore3 dichotomized into favorable and unfavorable Thisoutcome was selected because it was deemed to be the mostclinically relevant outcome and was the most consistentlyrecorded among the studies analyzed If both mRS and GOSscores were available for the same dataset the mRS score wasused The GOS score was dichotomized into favorable (GOSscore 4ndash5 [good recovery or moderate disability]) or un-favorable (GOS score 1ndash3 [severe disability vegetative stateor death]) The mRS score was dichotomized into favorable(mRS score 0ndash2 [good recovery no significant disabilityslight disability]) or unfavorable (mRS score 3ndash6 [moderatedisability moderately severe disability severe disability ordeath])

Secondary outcomesThe secondary outcomes were delayed cerebral ischemia(DCI) radiologic infarction angiographic evidence of vaso-spasm and transcranial Doppler (TCD) evidence of vaso-spasm defined as velocity gt200 cms The secondaryoutcomes could have been at any time during admission foraneurysm rupture If a secondary outcome variable was notcollected in a specific study that study was not included in themeta-analysis of that secondary outcome

CovariatesAminimum set of essential covariates was identified a priori tominimize sample size attrition while retaining the strongestknown predictors of outcome after aSAH16 These essentialcovariates were used for both primary and secondary out-comes (1) age (2) Fisher grade dichotomized into grades 1+ 2 and 3 + 4 (3) admission WFNS or Hunt and Hess gradedichotomized into good (grades 1ndash3) and poor (grades 4ndash5)and (4) treatment categorized into endovascular and surgicalAneurysms treated conservatively were excluded due to smallnumbers If both WFNS and Hunt and Hess grades wereavailable the WFNS grade was used For the primary out-come follow-up time was used as a covariate using longitu-dinal modeling During the analysis subsequent topublication of the protocol it became apparent that some ofthese essential covariates resulted in highly sparse cells es-pecially in small studies Because WFNS grade is by far themost influential predictor among the minimum essentialcovariates4 studies in which WFNS grade could not be in-cluded as a covariate were excluded from the analysis Forboth primary and secondary outcomes when possibleallowing for sample size and data availability (tablee-1available from Eprints eprintssotonacuk426525) ad-ditional covariates were used and prioritized in the followingorder diabetes mellitus hypertension race and aneurysmsite If additional covariates were used the same covariateswere used in all studies

Data analysisThe primary analytic strategy was a 2-stage IPLD study de-sign we also planned a secondary 1-stage IPLD analysis

because it is thought that both designs have their individualstrengths17 The primary study statistician (DR) mainlyconducted the statistical analysis and a second statistician inthe team (TH) confirmed the results The statisticians wereblinded to the identification of the studies and patientsthroughout the analysis HWE was assessed for all studiesincluded Two primary comparisons were planned a binarycomparison of HP2-2 vs HP2-1 and HP1-1 (because theprevious meta-analysis showed that HP2-1 is similar to HP1-1with respect to outcome) and a multicategorical comparisonof HP1-1 vs HP2-1 vs HP2-2 The primary 2-stage IPLDanalysis was performed for the primary and all secondaryoutcomes as follows In the first stage the analysis was con-ducted for each individual study to estimate the association ofoutcome with Hp adjusting for the baseline covariates andtime point Given the binary nature of primary outcomesmeasured at time points of 1- 3- and 6-month follow-up(with studies having 1 2 or 3 of these time points) gener-alized estimating equation models with logit link wereimplemented to account for the correlation between differenttime points within the same subject for each individual studyThe interaction between time point andHP was also checkedBinary logistic regression models were used for all the sec-ondary outcomes because they were associated with only 1time point Odds ratios (ORs) and 95 confidence intervals(CIs) from each individual study were estimated from theabove models Cases with missing data were excluded that isdata were not imputed In the second stage ORs from theindividual studies were combined by the use of random-effectsmeta-analysis Results are reported as ORs with their 95 CIsand corresponding p values Heterogeneity was assessed withthe Higgins and Thompson I2 statistic and Cochrane Q testand publication bias (small-study effects) was examined withfunnel plots and the Egger test All hypotheses were tested ata nominal significance level of 005 that is the probability ofa type I error (α) was 5 SAS (version 94 SAS Institute IncCary NC) and STATA (version 14 StataCorp CollegeStation TX) were used for all the analyses

To assess the robustness of results from the above primaryanalysis extensive sensitivity andor subgroup analyses wereconducted We first replicated the analysis using the 1-stageapproach with random-effect logistic regression modeling forall outcomes including site as random effect Next severalsubgroup analyses were performed including exclusion ofthose studies not in HWE or at high risk of bias inclusion ofstudies that do not have all the essential covariates exclusionof essential covariates and inclusion of additional covariatesas defined above We also coded GOS or mRS scores in 2other ways In the above analysis we used the traditionalapproach of dichotomizing of GOS and mRS scores into 2categories namely favorable and unfavorable (ie GOSscores 1ndash3 vs 4ndash5 mRS scores 0ndash2 vs 3ndash6) to render sta-tistical analysis and interpretation of results more straight-forward and identical to most other published studiesHowever this approach poses disadvantages it discardsvaluable information of the full ordinal nature of outcome

e2152 Neurology | Volume 92 Number 18 | April 30 2019 NeurologyorgN

Copyright copy 2019 American Academy of Neurology Unauthorized reproduction of this article is prohibited

measures and ignores initial prognostic risk of patients18ndash21

Hence 2 alternative approaches were used as part of a sensi-tivity analysis for the primary outcome First in the slidingdichotomy method the cutoff point for binarization of GOSand mRS scores is differentiated by the predicted baselineprognosis risk1821 Instead of defining good or bad outcomefor all patients using a single dichotomization point thesliding dichotomy approach customizes the definition of goodoutcome according to the baseline prognosis risk of eachpatient (additional Methods available from Eprints eprintssotonacuk426525) Second the proportional odds model(also referred to as shift analysis or ordinal logistic regression)was used to analyze the ordinal outcomes This method issensitive for detecting a shift of the entire ordinal outcomedistribution and estimates a common OR for each of thepossible cutoff points of the outcome scale The common ORis formally valid if the ORs for each cut point are the same (theproportional odds assumption)

Data availabilityAnonymized aggregate data will be shared after formal requestto the corresponding author in accordance with the Univer-sity of Southamptonrsquos data-sharing policies and contracts withthe coauthors and their institutions

ResultsStudy inclusionAPRISMA-IPD flow diagram details how studies were identified(figure 1) Of 18 published studies identified in the literaturesearch 5 were eligible for inclusion in the meta-analysis5ndash922

(table e-2 available from Eprints eprintssotonacuk426525)A further 6 unpublished studies were identified through theHaemoglobin after Intracranial Haemorrhage (HATCH) Con-sortium (wwwsouthamptonacukhatch) The lead authors ofthese 11 eligible studies were invited to join the study and allprovided IPLD The study characteristics of published and

Figure 1 Preferred Reporting Items for Systematic Reviews and Meta-AnalysesndashIndividual Patient Data (IPD) flow diagram

NeurologyorgN Neurology | Volume 92 Number 18 | April 30 2019 e2153

Copyright copy 2019 American Academy of Neurology Unauthorized reproduction of this article is prohibited

Table 1 Demographic information on studies included in the IPLD labeled by study identifier

Study identifier

A B C D E F G H I J K

Country of origin UK Japan US US UK US US US US US US and Italy

Published sample size n mdash 95 74 mdash mdash 133 mdash mdash 193 32 mdash

Cases excluded from published data n

Conservative treatment mdash mdash mdash mdash mdash mdash mdash mdash 1 2 mdash

Nonaneurysmal mdash mdash 2 mdash mdash mdash mdash mdash mdash mdash mdash

Unpublished cases n 44 mdash 16 57 214 10 47 59 mdash mdash 55

Cases excluded from unpublished data n

Conservative treatment 2 mdash 2 2 11 mdash mdash mdash mdash mdash 1

Nonaneurysmal 5 mdash mdash mdash 26 mdash mdash 21 mdash mdash mdash

No HP status mdash mdash mdash mdash 8 mdash 1 1 mdash mdash 5

Sample size for analysis n 37 95 86 55 169 143 46 37 192 30 49

HP typing Westernblot

Westernblot

Westernblot

Westernblot

Westernblot

Westernblot

Westernblot

Westernblot

PCR Western blot Westernblot

Outcomes available Primaryoutcome

DCIRadiologicinfarction

TCD VSa

Primaryoutcome

DCIRadiologicinfarction

AngiographicVS

Primaryoutcome

DCIRadiologicinfarction

AngiographicVS

Primaryoutcome

DCIRadiologicinfarction

AngiographicVS

TCD VS

Primaryoutcome

TCD VS

Primaryoutcome

DCIa

AngiographicVS

TCD VS

Primaryoutcome

DCIRadiologicinfarction

AngiographicVSa

TCD VS

Radiologicinfarction

TCD VSa

Primaryoutcome

DCIAngiographicVS

TCD VS

AngiographicVSb

DCIb

AngiographicVSb

Outcomes entered into analysis(percentage of sample size available used inanalysis)c

Primaryoutcome(100)

DCI (100)Radiologicinfarction(100)

Primaryoutcome(100)

DCI (100)Radiologicinfarction(100)

AngiographicVS (99)

Primaryoutcome(988)

DCI (977)Radiologicinfarction(919)

AngiographicVS (686)

Primaryoutcome(764)

DCI (100)Radiologicinfarction(982)

AngiographicVS (100)

TCD VS (746)

Primaryoutcome(953)

TCD VS (55)

Primaryoutcome(776)

AngiographicVS (811)

TCD VS (811)

Primaryoutcome(891)

DCI (935)Radiologicinfarction(935)

TCD (891) VS

mdash Primaryoutcome(953)

DCI (953)AngiographicVS (557)

TCD VS (906)

mdash mdash

Follow-up time for primary outcome n

1 mo 0 95 86 42 0 111 38 mdash 0 mdash mdash

Continued

e2154Neu

rology

|Vo

lume92N

umber

18|

April302019

NeurologyorgN

Copyright

copy2019

American

Academ

yof

Neurology

Unauthorized

reproductionof

thisarticle

isprohibited

Table 1 Demographic information on studies included in the IPLD labeled by study identifier (continued)

Study identifier

A B C D E F G H I J K

3 mo 37 95 84 42 0 58 25 mdash 192 mdash mdash

6 mo 34 61 0 23 162 6 2 mdash 161 mdash mdash

Age mean (SD) y 595 (130) 621 (137) 544 (143) 541 (129) 508 (116) 537 (139) 535 (129) 525 (157) 5442 (112) 5153 (128) 5320 (112)

Fisher grade n ()

1 + 2 0 9 (96) 7 (81) 12 (218) 45 (266) 50 (365) 3 (68) 2 (25) 54 (281) 0 0

3 + 4 37 (100) 85 (904) 79 (919) 43 (782) 124 (734) 87 (635) 41 (932) 6 (75) 138 (719) 30 (100) 49 (100)

WFNS grade n ()

Good (1ndash3) 12 (324) 57 (60) 55 (64) 34 (618) 155 (917) 87 (626) 31 (705) 17 (85) 152 (792) mdash mdash

Poor (4ndash5) 25 (676) 38 (40) 31 (36) 21 (382) 14 (83) 52 (374) 13 (295) 3 (15) 40 (208) mdash mdash

Hunt and Hess grade n ()

Good grade (1ndash3) mdash 55 (579) 66 (767) 35 (636) mdash 105 (755) 30 (667) 5 (100) 152 (792) 15 (682) mdash

Poor grade (4ndash5) mdash 40 (421) 20 (233) 20 (364) mdash 34 (245) 15 (333) 40 (208) 7 (318) mdash

Treatment n ()

Endovascular 32 (865) 22 (232) 41 (477) 35 (636) 143 (851) 25 (20) 31 (674) 3 (50) 115 (60) 13 (433) 17 (347)

Surgical 5 (135) 73 (768) 45 (523) 20 (364) 25 (149) 100 (80) 15 (326) 3 (50) 77 (40) 17 (567) 32 (653)

Haptoglobin status n ()

HP1-1 8 (216) 7 (74) 11 (128) 13 (236) 25 (148) 34 (238) 15 (326) 6 (162) 25 (13) 9 (30) 7 (143)

HP2-1 16 (433) 39 (41) 45 (523) 30 (546) 80 (473) 62 (433) 24 (522) 22 (595) 109 (568) 10 (333) 26 (531)

HP2-2 13 (351) 49 (516) 30 (349) 12 (218) 64 (379) 47 (329) 7 (152) 9 (243) 58 (302) 11 (367) 16 (326)

HWE χ2 (p value) 052 (0471) 004 (0841) 086 (0354) 046 (0498) 0 (1) 226 (0133) 027 (0603) 144(0230)

555 (0018) 327 (0071) 047 (0493)

Abbreviations DCI = delayed cerebral ischemia HP = haptoglobin HWE = Hardy-Weinberg equilibrium IPLD = individual patientndashlevel data TCD = transcranial Doppler VS = vasospasmMissing data are signified by mdasha Excluded from analysis because of the small number of observations availableb Excluded from analysis because core covariates were not availablec Includes only studies entered into primary and secondary analyses

Neurolo

gyorgN

Neurology

|Volum

e92N

umber

18|

April302019

e2155

Copyright

copy2019

American

Academ

yof

Neurology

Unauthorized

reproductionof

thisarticle

isprohibited

unpublished studies are summarized in table 1 For blindingpurposes each study was allocated a study identifier and is re-ferred to by this identifier throughout the article Across the 11studies 939 patients had IPLD available Published studies wereassessed for risk of bias and attained a score of 5 (of a total of 9maximum points) with the Newcastle-Ottawa Scale

DemographicsTable 1 summarizes the demographics of the 11 studies includedin the meta-analysis Only 1 study (study I) was significantly outof HWE (p= 0018) Overall the IPLDwas inHWE (p= 0663)

Primary outcomeEight studies (AndashG and I) with a total sample size of 755were included in the primary outcome analysis due to dataavailability (table e-1 available from Eprints eprintssotonacuk426525) Age Fisher grade WFNS grade treatment andhypertension were controlled for in all studies except studiesB and G in which inclusion of Fisher grade resulted in highlysparse cells Follow-up times were as follows 1 3 and 6months for 355 516 and 438 patients (table 1 for individualstudies) Using the 2-stage approach we identified no asso-ciation betweenHP genotype and unfavorable outcome whencomparing HP2-2 vs HP2-1 and HP1-1 (OR 0977 95 CI0672ndash1421 p = 0905) (figure 2A) When taken through thesame multivariate analysis higher age (OR 105 95 CI102ndash108) andWFNS grade (OR 84 95CI 44ndash159) wereassociated with an unfavorable outcome Higher Fisher gradewas associated with poor outcome (OR 26 95 CI 13ndash53)and a trend for hypertension was seen (OR 15 95 CI09ndash24) in univariate analysis only Treatment was not sig-nificant (table e-3 available from Eprints eprintssotonacuk426525) No significant association was identified in sub-group analysis of HP2-2 vs HP1-1 HP2-1 vs HP1-1 HP2-2 vsHP2-1 and HP2-2 and 2-1 vs HP1-1 (table 2 and figure e-1available from Eprints eprintssotonacuk426525)

Because not all patients were included in the primary outcomeanalysis due to data availability we assessed selection bias inthese patients (ie only those included in this analysis) Withrespect to HP genotype patients were overall in HWE (p =0671) Compared to a typical hospital population23 patientsin the primary outcome analysis of this IPLD had similarWFNS grades (grades 4ndash5 273 vs 247 p = 0143)similar age (median 549 vs 55 years) lower coiling ratecompared to clipping (558 vs 814 p lt 0001) anda higher incidence of DCI (372 vs 217 p lt 0001)Compared to a typical randomized controlled trial pop-ulation24 patients in the primary outcome analysis of thisIPLD had similar WFNS grades (grades 4ndash5 273 vs 229p = 005) lower Fisher grade (grades 3ndash4 773 vs 842 p =0001) higher age (mean 547 vs 50 years p lt 0001) lowercoiling rate compared to clipping (558 vs 669 p lt 0001)higher incidence of DCI (372 vs 16 p lt 0001) andsimilar favorable dichotomized mRS score at 6 months (74vs 717 p = 0353)

Secondary outcomesThe preplanned secondary outcomes were DCI radiologicinfarction angiographic vasospasm and TCD evidence ofvasospasm All analyses were conducted comparing HP2-2 vsHP2-1 and HP1-1 HP2-2 vs HP1-1 HP2-1 vs HP1-1 HP2-2vs HP2-1 and HP2-2 and 2-1 vs HP1-1 Only results for HP2-2 vs HP2-1 andHP1-1 are detailed in the text below the othercomparisons are summarized in table 2 (and their forest plotdata figures e-2ndashe-5 available from Eprints eprintssotonacuk426525)

Six studies (AndashC D G and I) with a total sample size of 497were included in the secondary outcome analysis for DCIAge Fisher grade WFNS grade treatment and hypertensionwere controlled for in all studies except studies B C and GFisher grade was not controlled for in studies B C and G dueto highly sparse cells Using the 2-stage approach we identi-fied no association between HP genotype and DCI whencomparing HP2-2 vs HP2-1 and HP1-1 (OR 1171 95 CI0735ndash1867 p = 0507) or other permutations of HP sub-groups (figure 3A and table 2)

Five studies (AndashD and G) with a total sample size of 308were included in the secondary outcome analysis for radio-logic infarction Age Fisher grade WFNS grade treatmenthypertension and aneurysm location were controlled for in allstudies except studies B and C Fisher grade was not con-trolled in studies B and C due to highly sparse cells Using the2-stage approach we identified no association between HPgenotype and radiologic infarction when comparing HP2-2 vsHP2-1 and HP1-1 (OR 1255 95 CI 0632ndash2490 p =0516) (figure 3B and table 2)

Five studies (BndashD F and I) with a total sample size of 431were included in the secondary outcome analysis for thepresence of angiographic evidence of vasospasm Age Fishergrade WFNS grade treatment and hypertension were con-trolled for in all studies Using the 2-stage approach weidentified no association between HP genotype and angio-graphic evidence of vasospasm when comparing HP2-2 vsHP2-1 and HP1-1 (OR 1130 95 CI 0498ndash2564 p =0771) (figure 3C and table 2)

Five studies (DndashG and I) with a total sample size of 465 wereincluded in the secondary outcome analysis for the presenceof TCD evidence of vasospasm Age Fisher grade WFNSgrade treatment and hypertension were controlled for in allstudies except study G Fisher grade and treatment were notcontrolled in study G due to highly sparse cells Using the2-stage approach we identified no association between HPgenotype and TCD evidence of vasospasm when comparingHP2-2 vs HP2-1 and HP1-1 (OR 0895 95 CI 0557ndash1439p = 0648) (figure 3D and table 2)

Sensitivity analysesThere were 2 main potential reasons to explain the discrep-ancy between the results here and the data from individual

e2156 Neurology | Volume 92 Number 18 | April 30 2019 NeurologyorgN

Copyright copy 2019 American Academy of Neurology Unauthorized reproduction of this article is prohibited

published studies The first was publication bias here weincluded all studies and unpublished studies are more likelyto be negative The second was controlling for covariatesbecause most published studies did not control well forcovariates We proceeded to investigate the relative

contribution of these 2 possibilities by repeating the primaryanalyses for primary (figure 2B) and secondary (figure 4)outcomes using data from published studies alone We also re-peated the primary (ie 2-stage IPLD) analyses withoutadjusting for covariates for both primary (figure 2C) and

Figure 2 Forest plots for 2-stage individual patientndashlevel data analysis for primary outcome (dichotomizedmodified RankinScale score in HP2-2 vs HP2-1 and HP1-1)

An odds ratio (OR) gt1 denotes a higherprobability of poor outcome (A) Adjustedfor covariates (B) same as A but includingpublished studies only and (C) un-adjusted for covariates CI = confidenceinterval HP = haptoglobin ID = identifier

NeurologyorgN Neurology | Volume 92 Number 18 | April 30 2019 e2157

Copyright copy 2019 American Academy of Neurology Unauthorized reproduction of this article is prohibited

secondary (figure 5) outcomes None of these analyses showeda significant difference or a trend suggesting an association

A number of other sensitivity analyses were conducted toevaluate the robustness of the results These were as fol-lows (1) exclusion of studies in which patients included inthe primary analysis were not in HWE (2) inclusion ofadditional covariates (3) inclusion of studies that did nothave all the essential covariates (4) sliding dichotomyanalysis (5) ordinal regression (6) 1-stage analysis (7)unpublished studies only and (8) including only covariatessignificant in univariate analysis For all the resultsremained consistent (figures e-6ndashe-11 and tables e-4ndashe-6available from Eprints eprintssotonacuk426525)

There were no trends suggesting an association betweenHP genotype and primary or secondary outcomes in any ofthese analyses

Heterogeneity and publication biasIn all analyses there was little evidence for heterogeneity(I2 range 0ndash358 Cochrane Q tests were not significantp gt 005) except for angiographic vasospasm (figure 3C andfigure e-10C available from Eprints eprintssotonacuk426525) There was no indication of publication bias fromfunnel plots (Egger regression test p gt 005 for all the tests)for any analyses (figures e-12ndashe-36 available from Eprintseprintssotonacuk426525)

Table 2 Summary of the 2-stage IPLD analysis results for all primary and secondary outcomes adjusted for covariates

Outcome Analysis OR (95 CI) p Value

Primary HP2-2 vs HP2-1 and HP1-1 0997 (0672ndash1421) 0905

HP2-2 vs HP1-1 0752 (0429ndash1321) 0322

HP2-1 vs HP1-1 0814 (0470ndash1410) 0462

HP2-2 vs HP2-1 1021 (0684ndash1524) 0921

HP2-2 and 2-1 vs HP1-1 0776 (0461ndash1305) 0339

Secondary

DCI HP2-2 vs HP2-1 and HP1-1 1171 (0735ndash1867) 0507

HP2-2 vs HP1-1 0878 (0437ndash1762) 0713

HP2-1 vs HP1-1 0735 (0393ndash1376) 0336

HP2-2 vs HP2-1 1187 (0707ndash1993) 0517

HP2-2 and 2-1 vs HP1-1 0851 (0476ndash1523) 0587

Radiologic infarction HP2-2 vs HP2-1 and HP1-1 1255 (0632ndash2490) 0516

HP2-2 vs HP1-1 0868 (0314ndash2402) 0785

HP2-1 vs HP1-1 0536 (0218ndash1319) 0174

HP2-2 vs HP2-1 1369 (0662ndash2832) 0397

HP2-2 and 2-1 vs HP1-1 0611 (0259ndash1441) 0260

Angiographic vasospasm HP2-2 vs HP2-1 and HP1-1 1130 (0498ndash2564) 0771

HP2-2 vs HP1-1 0942 (0445ndash1993) 0877

HP2-1 vs HP1-1 0862 (0285ndash2602) 0792

HP2-2 vs HP2-1 1184 (0422ndash3321) 0749

HP2-2 and 2-1 vs HP1-1 1015 (0484ndash2127) 0969

TCD evidence of vasospasm HP2-2 vs HP2-1 and HP1-1 0895 (0557ndash1439) 0648

HP2-2 vs HP1-1 0962 (0496ndash1867) 0909

HP2-1 vs HP1-1 1048 (0566ndash1940) 0881

HP2-2 vs HP2-1 0882 (0534ndash1456) 0662

HP2-2 and 2-1 vs HP1-1 0980 (0550ndash1746) 0945

Abbreviations CI = confidence interval HP = haptoglobin IPLD = individual patientndashlevel data OR = odds ratio TCD = transcranial DopplerAn OR gt1 denotes a higher probability of poor outcome

e2158 Neurology | Volume 92 Number 18 | April 30 2019 NeurologyorgN

Copyright copy 2019 American Academy of Neurology Unauthorized reproduction of this article is prohibited

Figure 3 Forest plots for 2-stage individual patientndashlevel data analysis for secondary outcomes adjusted for covariates(HP2-2 vs HP2-1 and HP1-1)

An odds ratio (OR) gt1 denotes a higher proba-bility of poor outcome (A) Delayed cerebral is-chemia (B) radiologic infarction (C)angiographic evidence of vasospasm and (D)transcranial Doppler evidence of vasospasm CI= confidence interval HP = haptoglobin ID =identifier

NeurologyorgN Neurology | Volume 92 Number 18 | April 30 2019 e2159

Copyright copy 2019 American Academy of Neurology Unauthorized reproduction of this article is prohibited

DiscussionAs is normal in IPLD methodology study size was driven bydata availability rather than by a predetermined sample sizePrestudy power by which we mean a power calculation usingestimates from prior studies was performed on data from the

largest published study in the IPLD (study I)11 which had aneffect size of an OR of 18 for comparing HP2-2 vs HP2-1 andHP1-1 on the primary outcome A logistic regression of thebinary response variable (mRS score) on the binary in-dependent variable (HP2-2 vs HP2-1 and HP1-1) witha sample size of 755 subjects (of whom 69 were HP2-1 and

Figure 4 Forest plots for secondary outcomes adjusted for covariates published studies only (HP2-2 vs HP2-1 and HP1-1)

An odds ratio (OR) gt1 denotes a higherprobability of poor outcome (A) Delayed ce-rebral ischemia (B) radiologic infarction (C)angiographic evidence of vasospasm and (D)transcranial Doppler evidence of vasospasmCI = confidence interval HP = haptoglobin ID= identifier

e2160 Neurology | Volume 92 Number 18 | April 30 2019 NeurologyorgN

Copyright copy 2019 American Academy of Neurology Unauthorized reproduction of this article is prohibited

Figure 5 Forest plots for secondary outcomes unadjusted for covariates (HP2-2 vs HP2-1 and HP1-1)

An odds ratio (OR) gt1 denotes a higher probability ofpoor outcome (A) Delayed cerebral ischemia (B) ra-diologic infarction (C) angiographic evidence of vaso-spasm and (D) transcranial Doppler evidence ofvasospasm CI = confidence interval HP = haptoglobinID = identifier

NeurologyorgN Neurology | Volume 92 Number 18 | April 30 2019 e2161

Copyright copy 2019 American Academy of Neurology Unauthorized reproduction of this article is prohibited

HP1-1 and 31 were HP2-2) achieved 94 power at a 0050significance level to detect a small to medium effect size of anOR of 18 with a 2-sidedWald test Even with the conservativeestimate of an OR of 16 (10 reduction of OR = 18) powerwas 79 at a 0050 significance level to detect the associationTherefore this study has conclusively proven that HP2-2 isnot associated with a poor long-term outcome as defined bythe mRS score down to a minimum OR of 16

Although the reason could simply be that there is no differ-ence in the relative protective effects of different HP geno-types there are several possible explanations of how a clinicaleffect could have been missed First the mRS and GOS maybe insufficiently sensitive outcome measures to detect subtleyet important outcome variation in patients after aSAH in-cluding cognitive impairment anxiety and return to work2526

Subarachnoid hemorrhage (SAH)ndashspecific outcome measurescovering these more subtle outcomes such as the SAH outcometool27may bemore sensitive in detecting an association betweenHP genotype and functional outcome Another possible reasonis that the early brain injury takes longer to settle and expose thefinal residual permanent deficit influenced by HP genotype Inthis study outcomeswere analyzed 2weeks to 1 year after aSAHhowever improvements have been demonstrated beyond thistime For example mRS score has been shown to improve in19of patients between 12 and 36months after aSAH28Hencefuture studies should consider longer follow-up periods

The negative result for all secondary outcomes is not con-sistent with the recent meta-analysis that provided evidencethat the HP2 allele was associated with worse short-termoutcome including DCI and vasospasm11 The previousmeta-analysis had a number of limitations that may underliethis discrepancy First the meta-analysis used a compositedefinition of short-term outcome grouping DCI or cerebralvasospasm by any definition into 1 binary outcome measureIn comparison the IPLD analysis here used specific defi-nitions of cerebral vasospasm and DCI which were analyzedseparately Second the meta-analysis did not control forcovariates known to affect outcome after aSAH the inclusionof which in the IPLD analysis may explain the different resultThe effect sizes observed for the secondary outcomes besidesnot achieving statistical significance or showing trends wereextremely small Taken together this demonstrates that thereis no meaningful clinically significant difference in theseoutcomes between HP genotypes

Although this IPLD analysis included a number of un-published studies which may have contributed significantly tothe negative result a sensitivity analysis of published studiesonly was still negative It has previously been noted that in-corporation of unpublished studies does not significantlychange the results of most meta-analyses29

The binding of Hp to Hb is thought to confer protection viaa number of mechanisms including limiting the oxidativedamage potential of Hb30 facilitating its clearance via the

CD163 membrane receptor on macrophagesmicroglia31

and generating an anti-inflammatory response32 The lack ofa clear effect of HP genotype on outcome after aSAH inhumans contrasts with observations in animal modelsTransgenic mice expressing a murine equivalent of humanHP2 experienced more vasospasm and functional deficit afterexperimentally induced SAH compared to wild-type mice33

However there are important biological differences betweenmice and humans The influence of Hp on the affinity ofCD163 to Hb is markedly different34 and CD163 sheddingoccurs in humans12 not mice35 These differences suggestthat the Hb scavenging system is sufficiently different betweenthe 2 species such that extrapolation of the detrimental effectof HP2 observed in this mouse model to humans should bedone with extreme caution

The basic unit of Hp protein is an Hp monomer consisting of1 α and 1 β subunit The HP1 allele codes for an α subunit(called α1) with 1 cysteine residue that enables dimerization ofthe Hp monomer by formation of a disulfide bond The HP2allele codes for an α2 subunit that contains an extra cysteineresidue compared to α1 and is therefore able to make multipledisulfide bonds resulting in several polymers of increasingsize in HP2-1 heterozygotes and HP2-2 homozygotes10

Whether there is a functional difference between the proteinsexpressed by different HP genotypes is controversial and verymuch depends on which characteristic of Hp one considers Itis well established that Hp expression is influenced by geno-type HP1-1 gt HP1-2 gt HP2-236 Some investigators havedemonstrated that the Hp1-1 dimer is more effective than theHp2-2 polymer in reducing the oxidative potential of Hb37ndash39

although other reports suggest that there is no difference40ndash42

Binding affinity to CD163 appears to be higher for Hb incomplex with Hp2-2 polymer compared to Hp1-1 dimer3143

However studies looking at the uptake of Hb-Hp complexesby CD163-expressing cells are less clear with some reportingno difference40 and others indicating an increased bindingaffinity of both Hp1-1 dimer43 and Hp2-2 polymer44

depending on the experimental conditions Differences mayextend to inflammatory effects because binding of Hp1-1-Hbcomplexes to CD163 results in secretion of the anti-inflammatory cytokine interleukin-103245 at levels several-fold higher compared to Hp2-2-Hb complexes45 It is alsoplausible that differences may be unrelated to Hb scavengingFor example the HP1-1 genotype appears to decrease en-dothelial progenitor cell cluster formation46 HP2 has alsobeen associated with poorer clinical outcome in people withdiabetes mellitus ischemic heart disease and infections36

suggesting that it may influence outcome after aSAH inindividuals with these comorbidities

HP genotype may not influence outcome after aSAH even ifthere are differences betweenHP genotypes in Hb scavengingefficiency Recently CD163 expression by neurons has beendemonstrated in animal models of cerebral hemorrhage47

Because Hb is normally predominantly taken up in the CNSbymicroglia it has been proposed that this increased neuronal

e2162 Neurology | Volume 92 Number 18 | April 30 2019 NeurologyorgN

Copyright copy 2019 American Academy of Neurology Unauthorized reproduction of this article is prohibited

CD163 expression may lead to increased neuronal toxicitythrough uptake of Hb48 It is therefore possible that any po-tential protective effect conferred by different HP genotypesmay be mitigated by increased neuronal toxicity To dateCD163 expression by human neurons in situ remains to bedemonstrated

This IPLD analysis provides the most robust evidence to dateexamining the relationship between HP genotype and out-come after aSAH and has a number of strengths First thisstudy has the largest sample size to date17 although stillsmaller than usual for genetic studies Second the studypopulation was in HWE excluding significant case missingnessor technical problems with genotypephenotype ascertainmentThird a number of covariates known to affect outcome afteraSAH that have not been consistently controlled for in previousstudies were included in the analysis age WFNS grade Fishergrade and treatment2349 Fourth this study includes a largenumber of unpublished studies identified through a network ofinvestigators worldwide Fifth it uses IPLD Sixth all analyseswere preplanned the protocol was published before the analysiswas started and the statisticians were blinded to the identities ofthe study andHP genotypes (deidentification of the studies wasperformed at the end table e-7 available from Eprints eprintssotonacuk426525) Finally a comprehensive array of statis-tical approaches was used

There are several limitations First there was minor but sig-nificant evidence of selection bias when patients in this studywere compared with both a hospital aSAH population23 anda typical aSAH randomized controlled study24 favoringpatients with a lower coiling rate compared to clipping anda higher incidence of DCI Second this study was retro-spective and despite the collection of IPLD the available datalimited the choice and number of covariates that could beused It does not control for other covariates known to beimportant in predicting outcome after aSAH including needfor CSF diversion and preoperative rebleeding23 because ofa lack of data availability In addition although we havecontrolled for follow-up time the duration varied significantlybetween studies Future studies could examine Hp subunitexpression because the Hp α1 chain band intensity may beprognostic in HP2-1 individuals50

Author contributionsDO Bulters and I Galea conceived the study B Gaastra DRen S Alexander DM Bielawski SL Blackburn MKBorsody S Dore J Galea K Iihara Y Kawamura PANyquist DO Bulters and I Galea contributed to the designof the study All authors contributed to different aspects ofdata acquisition B Gaastra D Ren T He PA Nyquist DOBulters and I Galea analyzed the data D Ren and T Heperformed statistical analyses B Gaastra D Ren and I Galeawrote the first draft of the manuscript B Gaastra D RenDO Bulters PA Nyquist and I Galea worked on sequentialdrafts of the manuscript All authors revised and approved thefinal version of the manuscript

AcknowledgmentThe authors thank the University of Southampton GlobalPartnership Award Scheme and NeuroSpring for theirsupport as well as Miss Emily Edwards (Mayo) ProfDaniel Laskowitz (Duke) Dr Satoshi Matsuo and DrRyota Kurogi (National Hospital Organization KyushuMedical Center) NeuroSpring is a nonprofit organizationwith the primary aim of supporting research in theneurosciences leading to new medical therapies (neuro-springorg) one of the authors (MB) is a senior scientistat NeuroSpring Samples and clinical data from the MayoClinic in Florida came from the Mayo Clinic FloridaFamilial Cerebrovascular Diseases Registry (JFM prin-cipal investigator)

Study fundingFunded by the University of Southampton (University ofSouthampton Global Partnership Award Scheme) and UKMedical Research Council

DisclosureThe authors report no disclosures relevant to the manuscriptGo to NeurologyorgN for full disclosures

Publication historyReceived by Neurology July 30 2018 Accepted in final form February 42019

References1 Rinkel GJ Algra A Long-term outcomes of patients with aneurysmal subarachnoid

haemorrhage Lancet Neurol 201110349ndash562 Teasdale GM Drake CG Hunt W Kassell K Pertuiset B De Villiers JC A universal

subarachnoid hemorrhage scale report of a Committee of the World Federation ofNeurosurgical Societies J Neurol Neurosurg Psychiatry 1988511457

3 Jennett B BondM Assessment of outcome after severe brain damage Lancet 19751480ndash484

4 Jaja BNR Saposnik G Lingsma HF et al Development and validation of outcomeprediction models for aneurysmal subarachnoid haemorrhage the SAHIT multina-tional cohort study BMJ 2018360j5745

5 Borsody M Burke A Coplin W Miller-Lotan R Levy A Haptoglobin and the de-velopment of cerebral artery vasospasm after subarachnoid hemorrhage Neurology200666634ndash640

6 Kantor E Bayır H Ren D et al Haptoglobin genotype and functional outcome afteraneurysmal subarachnoid hemorrhage J Neurosurg 2014120386ndash390

7 Leclerc JL Blackburn S Neal D Mendez JA Waters MF Dore S Haptoglobinphenotype predicts the development of focal and global cerebral vasospasm and mayinfluence outcomes after aneurysmal subarachnoid hemorrhage Proc Natl Acad SciUSA 20151121155ndash1160

8 Murthy SB Caplan J Levy AP et al Haptoglobin 2-2 genotype is associated withcerebral salt wasting syndrome in aneurysmal subarachnoid hemorrhage Neurosur-gery 20167871ndash76

9 Ohnishi H Iihara K Kaku Y et al Haptoglobin phenotype predicts cerebral vaso-spasm and clinical deterioration after aneurysmal subarachnoid hemorrhage J StrokeCerebrovasc Dis 201322520ndash526

10 Bulters D Gaastra B Zolnourian A et al Haemoglobin scavenging in in-tracranial bleeding biology and clinical implications Nat Rev Neurol 201814416ndash432

11 Gaastra B Glazier J Bulters D Galea I Corrigendum to haptoglobin genotype andoutcome after subarachnoid haemorrhage new insights from a meta-analysis OxidMed Cell Longev 201820188

12 Galea J Cruickshank G Teeling JL Boche P Perry VH Galea I The intrathecalCD163-haptoglobin-hemoglobin scavenging system in subarachnoid hemorrhageJ Neurochem 2012121785ndash792

13 Stewart LA Clarke M Rovers M et al Preferred reporting items for a systematicreview and meta-analysis of individual participant data the PRISMA-IPD statementJAMA 20153131657ndash1665

14 Gaastra BAS Bielawski D Blackburn S et al Haptoglobin after subarachnoid haemorrhageindividual patient level data (IPLD) analysis [online] Available at httpwwwcrdyorkacukPROSPEROdisplay_recordphpRecordID=70830ampVersionID=108574 AccessedMarch 25 2019

NeurologyorgN Neurology | Volume 92 Number 18 | April 30 2019 e2163

Copyright copy 2019 American Academy of Neurology Unauthorized reproduction of this article is prohibited

15 Farrell B Godwin J Richards SWarlow C The United KingdomTransient IschaemicAttack (UK-TIA) aspirin trial final results J Neurol Neurosurg Psychiatry 1991541044ndash1054

16 Jaja BN Cusimano MD Etminan N et al Clinical prediction models for aneurysmalsubarachnoid hemorrhage a systematic review Neurocrit Care 201318143ndash153

17 Riley RD Lambert PC Abo-Zaid G Meta-analysis of individual participant datarationale conduct and reporting BMJ 2010340c221

18 Murray GD Barer D Choi S et al Design and analysis of phase III trials with orderedoutcome scales the concept of the sliding dichotomy J Neurotrauma 200522511ndash517

19 Savitz SI Lew R Bluhmki E HackeW FisherM Shift analysis versus dichotomizationof the modified Rankin Scale outcome scores in the NINDS and ECASS-II trialsStroke 2007383205ndash3212

20 Roozenbeek B Lingsma HF Maas AI New considerations in the design of clinicaltrials for traumatic brain injury Clin Investig (Lond) 20122153ndash162

21 Ilodigwe DMurray GD Kassell NF Torner RSMolyneux AJ Macdonald RL Slidingdichotomy compared with fixed dichotomization of ordinal outcome scales in sub-arachnoid hemorrhage trials J Neurosurg 20131183ndash12

22 Vergouwen MD Vermeulen M van Gijn J et al Definition of delayed cerebralischemia after aneurysmal subarachnoid hemorrhage as an outcome event in clinicaltrials and observational studies proposal of a multidisciplinary research group Stroke2010412391ndash2395

23 Galea JP Dulhanty L Patel HC UK and Ireland Subarachnoid Hemorrhage DatabaseCollaborators Predictors of outcome in aneurysmal subarachnoid hemorrhagepatients observations from a multicenter data set Stroke 2017482958ndash2963

24 Kirkpatrick PJ Turner CL Smith C Hutchinson PJ Murray GD Simvastatin inAneurysmal Subarachnoid Haemorrhage (STASH) a multicentre randomised phase3 trial Lancet Neurol 201413666ndash675

25 Macdonald RL Jaja B Cusimano MD et al SAHIT Investigators on the outcome ofsome subarachnoid hemorrhage clinical trials Transl Stroke Res 20134286ndash296

26 Hutter BO Gilsbach JMWhich neuropsychological deficits are hidden behind a goodoutcome (Glasgow = I) after aneurysmal subarachnoid hemorrhage Neurosurgery199333999ndash1005

27 Pace A Mitchell S Casselden E et al A subarachnoid haemorrhage-specific outcometool Brain 20181411111ndash1121

28 Wilson DA Nakaji P Albuquerque FC McDougall CG Zabramski JM Spetzler RFTime course of recovery following poor-grade SAH the incidence of delayed im-provement and implications for SAH outcome study design J Neurosurg 2013119606ndash612

29 Schmucker CM Blumle A Schell LK et al Systematic review finds that study data notpublished in full text articles have unclear impact on meta-analyses results in medicalresearch PLoS One 201712e0176210

30 Cooper CE Schaer DJ Buehler PW et al Haptoglobin binding stabilizes hemoglobinferryl iron and the globin radical on tyrosine beta145 Antioxid Redox Signal 2013182264ndash2273

31 Kristiansen M Graversen JH Jacobsen C Sonne HJ Law SK Moestrup SK Iden-tification of the haemoglobin scavenger receptor Nature 2001409198ndash201

32 Philippidis P Mason JC Evans BJ Nadra KM Haskard DO Landis RC Hemoglobinscavenger receptor CD163 mediates interleukin-10 release and heme oxygenase-1synthesis antiinflammatory monocyte-macrophage responses in vitro in resolving

skin blisters in vivo and after cardiopulmonary bypass surgery Circ Res 200494119ndash126

33 Chaichana KL Levy AP Miller-Lotan R Shakur S Tamargo RJ Haptoglobin 2-2genotype determines chronic vasospasm after experimental subarachnoid hemor-rhage Stroke 2007383266ndash3271

34 Etzerodt A Kjolby M Nielsen MJ Maniecki M Svendsen P Moestrup SK Plasmaclearance of hemoglobin and haptoglobin in mice and effect of CD163 gene targetingdisruption Antioxid Redox Signal 2013182254ndash2263

35 Etzerodt A Rasmussen MR Svendsen P et al Structural basis for inflammation-driven shedding of CD163 ectodomain and tumor necrosis factor-alpha in macro-phages J Biol Chem 2014289778ndash788

36 Sadrzadeh SM Bozorgmehr J Haptoglobin phenotypes in health and disorders Am JClin Pathol 2004121(suppl)S97ndashS104

37 Asleh R Guetta J Kalet-Litman S Miller-Lotan R Levy AP Haptoglobin genotype-and diabetes-dependent differences in iron-mediated oxidative stress in vitro and invivo Circ Res 200596435ndash441

38 BammVVTsemakhovichVA ShaklaiM ShaklaiNHaptoglobin phenotypes differ in theirability to inhibit heme transfer from hemoglobin to LDL Biochemistry 2004433899ndash906

39 Melamed-Frank M Lache O Enav BI et al Structure-function analysis of the anti-oxidant properties of haptoglobin Blood 2001983693ndash3698

40 Lipiski M Deuel JW Baek JH Engelsberger WR Buehler PW Schaer DJ HumanHP1-1 and HP2-2 phenotype-specific haptoglobin therapeutics are both effective invitro and in guinea pigs to attenuate hemoglobin toxicity Antioxid Redox Signal 2013191619ndash1633

41 Mollan TL Jia Y Banerjee S et al Redox properties of human hemoglobin in complexwith fractionated dimeric and polymeric human haptoglobin Free Radic Biol Med201469265ndash277

42 Pimenova T Pereira CP Gehrig P Buehler PW Schaer DJ Zenobi R Quantitativemass spectrometry defines an oxidative hotspot in hemoglobin that is specificallyprotected by haptoglobin J Proteome Res 201094061ndash4070

43 Asleh R Marsh S Shilkrut M et al Genetically determined heterogeneity in hemo-globin scavenging and susceptibility to diabetic cardiovascular disease Circ Res 2003921193ndash1200

44 Na N Ouyang J Taes YE Delanghe JR Serum free hemoglobin concentrations inhealthy individuals are related to haptoglobin type Clin Chem 2005511754ndash1755

45 Guetta J Strauss M Levy NS Fahoum L Levy AP Haptoglobin genotype modulatesthe balance of Th1Th2 cytokines produced by macrophages exposed to free he-moglobin Atherosclerosis 200719148ndash53

46 Rouhl RP van Oostenbrugge RJ Damoiseaux JG et al Haptoglobin phenotype mayalter endothelial progenitor cell cluster formation in cerebral small vessel disease CurrNeurovasc Res 2009632ndash41

47 Liu R Cao S Hua Y Keep RF Huang Y Xi G CD163 expression in neurons afterexperimental intracerebral hemorrhage Stroke 2017481369ndash1375

48 Chen-Roetling J Regan RF Haptoglobin increases the vulnerability of CD163-expressing neurons to hemoglobin J Neurochem 2016139586ndash595

49 Pegoli M Mandrekar J Rabinstein AA Lanzino G Predictors of excellent functionaloutcome in aneurysmal subarachnoid hemorrhage J Neurosurg 2015122414ndash418

50 Kim BJ Kim Y Kim S-E Jeon JP Study of correlation between Hp α1 expression ofhaptoglobin 2-1 and clinical course in aneurysmal subarachnoid hemorrhage WorldNeurosurg 2018117e221ndashe227

e2164 Neurology | Volume 92 Number 18 | April 30 2019 NeurologyorgN

Copyright copy 2019 American Academy of Neurology Unauthorized reproduction of this article is prohibited

DOI 101212WNL0000000000007397201992e2150-e2164 Published Online before print April 5 2019Neurology

Ben Gaastra Dianxu Ren Sheila Alexander et al data analysis

Haptoglobin genotype and aneurysmal subarachnoid hemorrhage Individual patient

This information is current as of April 5 2019

ServicesUpdated Information amp

httpnneurologyorgcontent9218e2150fullincluding high resolution figures can be found at

References httpnneurologyorgcontent9218e2150fullref-list-1

This article cites 49 articles 17 of which you can access for free at

Citations httpnneurologyorgcontent9218e2150fullotherarticles

This article has been cited by 1 HighWire-hosted articles

Subspecialty Collections

httpnneurologyorgcgicollectionsubarachnoid_hemorrhageSubarachnoid hemorrhage

httpnneurologyorgcgicollectionassociation_studies_in_geneticsAssociation studies in genetics

httpnneurologyorgcgicollectionall_cbmrt_null_hypothesisAll CBMRTNull Hypothesisfollowing collection(s) This article along with others on similar topics appears in the

Permissions amp Licensing

httpwwwneurologyorgaboutabout_the_journalpermissionsits entirety can be found online atInformation about reproducing this article in parts (figurestables) or in

Reprints

httpnneurologyorgsubscribersadvertiseInformation about ordering reprints can be found online

rights reserved Print ISSN 0028-3878 Online ISSN 1526-632X1951 it is now a weekly with 48 issues per year Copyright copy 2019 American Academy of Neurology All

reg is the official journal of the American Academy of Neurology Published continuously sinceNeurology