ARTICLE OPEN ACCESS

Phenome-wide examination of comorbidityburden and multiple sclerosis disease severityTingting Zhang PhD Matthew Goodman PhD Feng Zhu MSc Brian Healy PhD Robert Carruthers MD

Tanuja Chitnis MD HowardWeiner MD Tianxi Cai ScD Philip De JagerMD PhD Helen Tremlett PhD and

Zongqi Xia MD PhD

Neurol Neuroimmunol Neuroinflamm 20207e864 doi101212NXI0000000000000864

Correspondence

Dr Xia

zxia1pittedu

AbstractObjectiveWe assessed the comorbidity burden associated with multiple sclerosis (MS) severity byperforming a phenome-wide association study (PheWAS)

MethodsWe conducted a PheWAS in 2 independent cohorts a discovery (Boston United States1993ndash2014) and extension cohort (British Columbia Canada 1991ndash2008) We includedadults with MS ge1 Expanded Disability Status Scale (EDSS) score and ge1 InternationalClassification of Diseases (ICD) code other than MS We calculated the Multiple SclerosisSeverity Score (MSSS) using the EDSS We mapped ICD codes into PheCodes (phenotypes)using a published system with each PheCode representing a unique medical condition Asso-ciation between the MSSS and the presence of each condition was assessed using logisticregression adjusted for covariates

ResultsThe discovery and extension cohorts included 3439 and 4876 participants respectively AfterBonferroni correction and covariate adjustments a higher MSSS was associated with 37coexisting conditions in the discovery cohort Subsequently 16 conditions including genito-urinary infectious metabolic epilepsy and movement disorders met the reporting criteriareaching the Bonferroni threshold of significance with the same direction of effect in thediscovery and extension cohort Notably benign neoplasm of the skin was inversely associatedwith the MSSS

ConclusionThe phenome-wide approach enabled a systematic interrogation of the comorbidity burdenand highlighted clinically relevant medical conditions associated with MS severity (beyondMS-specific consequences) and defines a roadmap for comprehensive investigation ofcomorbidities in chronic neurologic diseases Further prospective investigation of thebidirectional relationship between disability and comorbidities could inform the individualizedpatient management

These authors contributed equally as first authors

These authors contributed equally as senior authors

From the Department of Health Services (TZ) Policy and Practice Brown University Providence RI Department of Biostatistics (MG T Cai) Harvard T H Chan School of PublicHealth Boston MA Department of Medicine (Neurology) (FZ RC HT) and the Djavad Mowafaghian Centre for Brain Health University of British Columbia Vancouver CanadaDepartment of Neurology (BH T Chitnis HW PDJ ZX) Brigham and Womenrsquos Hospital Boston MA Department of Neurology (PDJ) Columbia University New York NY andDepartment of Neurology (ZX) University of Pittsburgh PA

Go to NeurologyorgNN for full disclosures Funding information is provided at the end of the article

The Article Processing Charge was funded by the authors

This is an open access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License 40 (CC BY-NC-ND) which permits downloadingand sharing the work provided it is properly cited The work cannot be changed in any way or used commercially without permission from the journal

Copyright copy 2020 The Author(s) Published by Wolters Kluwer Health Inc on behalf of the American Academy of Neurology 1

People with chronic neurologic diseases such as MS exhibitvariable disease trajectories People with MS often developcoexisting conditions Prior efforts relying on candidate phe-notype approaches have linked comorbidities to diagnosticdelay treatment delay increased hospitalization decreasedquality of life and increased mortality12 Thus comorbiditieslikely constitute an important facet of the individual variationin MS

Phenome-wide association studies (PheWASs) were origi-nally designed to interrogate the multiple phenotypes asso-ciated with a given genetic variant3 A central contribution ofthe PheWAS approach is the creation of a clinically in-formative disease classification system that consolidates re-lated International Classification of Diseases (ICD) codes foreach medical condition into a unique phenotype code (Phe-Code)4 This approach enables a more comprehensive andorganized examination of the comorbidity burden than can-didate approaches

We conducted a PheWAS to systematically interrogate therelationship between MS severity and comorbidity burden inthis proof-of-concept study We hypothesized that this novelapplication of the PheWAS approach to comprehensivelyassess comorbidities could offer insights into the associationbetween disability and the occurrence of comorbidities andguide future prospective studies and individualized manage-ment strategies

MethodsData sourcesFollowing a common protocol we conducted independentanalyses using data from 2 large MS cohorts (PartnersHealthCare Boston United States British Columbia [BC]Canada) We preassigned the Boston cohort as discovery andBC as extension for analyses

The discovery cohort comprised patients with MS seen at thePartners MS Center (Boston) between 1993 and 2014 in-cluding patients enrolled in the Comprehensive LongitudinalInvestigations in MS (CLIMB) study a well-characterizedclinic-based prospective study5 We previously merged elec-tronic health records (EHRs) and CLIMB research data forthese patients6 We obtained all available inpatient and out-patient ICD-9 codes from the Partners HealthCare EHRs

The extension cohort comprised patients with MS registeredat MS clinics in BC between 1991 and 2004 with follow-up

until 2008 We accessed linked clinical and health adminis-trative data The British Columbia MS database capturedclinical data We determined residential status in BC usingprovincial health care plan registration dates and captureddemographic information from the BC Registration andPremium Billing Files7 and all available ICD codes from thepopulation-based physician and hospital data (the MedicalService Plan and Discharge Abstract Databases)89 ICD-10codes were converted to ICD-9 codes10

Standard protocol approvals registrationsand patient consentsThe relevant Institutional Review Boards at PartnersHealthCare and University of British Columbia approved thedeidentified data usage for this study

Eligibility criteriaWe included all individuals aged ge18 years with a neurologist-confirmed MS diagnosis1112 and ge1 recorded ICD codeother than MS and ge1 recorded Expanded Disability StatusScale (EDSS) measure (assessed gt3 months from a relapseand within 3ndash40 years after MS symptom onset) (figure 1)

MS disease severityWe extracted relevant clinical data including date of symptomonset relapsing or progressive onset and disability (EDSSscores from clinic visits) for all patients We calculated theMultiple Sclerosis Severity Score (MSSS)13 independently ineach cohort using the EDSS scores observed at each diseaseduration in the respective local cohort and assigned a singlescore to each individual using the median of their 3 mostrecent MSSS measures if available The MSSS provides apercentile-based cross-sectional assessment of severity thataccounts for disease duration Ten percent of each cohortbelonged to each severity decile at each disease duration Weincluded EDSS scores measured gt3 months from a relapsedate (to capture measures taken during periods of stability)and within 3ndash40 years after symptom onset because (1) theMSSS is unstable for disease durations lt3 years and (2)individuals with EDSS scores gt40 years after symptom onsetwere rare and differential survival of older individuals may biasthe analyses

Coexisting conditionsWe determined the occurrence of coexisting medical condi-tions in patients with MS Using a published disease classifi-cation system that consolidates multiple related ICD-9 codesof each unique medical condition14 we mapped each ICD-9code to a single clinically informative medical condition rep-resented by a unique phenotype code (PheCode) To reduce

GlossaryBC = British Columbia CLIMB = Comprehensive Longitudinal Investigations in MS EDSS = Expanded Disability StatusScale EHR = electronic health record ICD = International Classification of DiseasesMSSS = Multiple Sclerosis Severity ScorePheWAS = phenome-wide association study

2 Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 6 | November 2020 NeurologyorgNN

the potential bias from rule-out codes (eg billing codes en-tered to indicate testing for rather than confirming a medicalcondition) we excluded patients with lt3 of a given ICD-9code in the discovery analysis1516 whereas we assigned suchpatients to control status in the extension analysis due tocomplete capture of codified data in BC made possible by itsuniversal health care system To mitigate double countingmultiple counts of a given ICD code on the same day and fromthe same clinic location were counted once

Statistical analysisIn each cohort we independently assessed associations be-tween the MSSS and the presence (or absence) of eachPheCode with the MSSS modeled linearly as a continuousindependent variable The MSSS was modeled as a linearexposure because the alternative categorical division faced atleast 2 challenges the ideal threshold could differ for eachcomorbidity and the effect on comorbidity could be non-monotonic neither increasing nor decreasing We adjustedthe model for confounders including sex age at symptom onsetdisease course at symptom onset ICD code follow-up duration(defined as the time from the first to the last available ICD code)and disease duration (defined as the time from MS symptomonset to the last available EDSS) We adjusted for disease du-ration to account for comorbidities that are predicted to occurwith increasing disease duration In the discovery analysis we

additionally adjusted for race and ethnicity (self-reported andgrouped as non-Hispanic European ancestry Hispanic or non-European ancestry and unknown) smoking status and having aprimary care provider who practices within the same hospitalsystem as the EHR data These additional covariates (raceethnicity smoking status and presence of primary care provider)were not available in the extension cohort

To reduce finite sample and rare event bias and prevent type Ierror we excluded medical conditions for which the observedprevalence in each cohort was lt317 In both cohorts wedefined the significance threshold as p lt 005 after Bonferronicorrection Our predefined reporting criteria for each findingrequired that (1) the association achieved the Bonferronithreshold of significance in both cohorts and (2) displayed thesame direction of effect

We used a previously published PheWAS table andR package tomap ICD-9 codes into PheCodes14 Analyses were performedusing the Statistical Package for the Social Sciences (Version200 IBM Corp Armonk NY) and R (Version 302)18

Data availabilityCode for analysis and figure generation is available at tinyurlcomy9lu7t3t Summary data that support the findings of thisstudy as well as anonymous data from the Boston cohort that

Figure 1 Eligibility criteria for the discovery and extension study cohorts

EDSS = Expanded Disability Status Scale ICD = International Classification of Diseases

NeurologyorgNN Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 6 | November 2020 3

were used for this study are available on request to the cor-responding author The data from the BC cohort that wereused for this study were accessed through PopulationData BC(popdatabcca) and reside on a limited access secure re-search environment For legal reasons (Canada) the datacannot leave this secure research environment

Results

Comparison of the study cohortsPatient characteristics of both cohorts were generally similarexcept that the extension cohort had a longer median follow-up duration from the first to the last EDSS value (table 1) Inthe discovery cohort 3439 patients with MS met the eligi-bility criteria for analysis (figure 1) whereas 109 medicalconditions met the 3 prevalence threshold In the extensioncohort 4876 patients with MS met the eligibility criteria foranalysis (figure 1) whereas 365 medical conditions met the3 prevalence threshold In total 78 PheCodes or medicalconditions met the 3 prevalence threshold in both cohorts

The extension analysis (based on population-linked healthadministrative data in a universal health care system) sys-tematically capture all available ICD codes whereas the dis-covery analysis (based on EHRs in an open health caresystem) had fewer analyzable medical conditions that met theminimum case threshold As such we included the presenceof primary care provider in the same EHR system as a keycovariate in the discovery analysis Furthermore the longerfollow-up duration in the extension cohort contributed tohigher measured prevalence of medical conditions which af-fected the observability of medical conditions that reached thethreshold for inclusion in the analysis Specifically 287PheCodes reached the 3 case threshold in the extensioncohort but were excluded because they failed to reach the 3case threshold in the discovery cohort For comparison 31PheCodes reached the 3 case threshold in the discoverycohort but were excluded due to failure to reach the 3 casethreshold in the extension cohort (supplementary materiallinkslwwcomNXIA294)

Comorbidities meeting the reporting criteriaIn the discovery cohort a higher MSSS was significantly as-sociated with 33 (of the 78 42) coexisting medical condi-tions after Bonferroni correction and covariate adjustment Atotal of 16 (of the 33 48) conditions met the predefinedcriteria for reporting having reached the Bonferroni thresholdof significance (on 78 tests) and displaying a consistent di-rection of effect in both cohorts Table 2 provides the prev-alence of the coexisting conditions that met the reportingcriteria as well as the effect size and p values from the dis-covery and extension analysis

Among the reported coexisting medical conditions associatedwith a higher MSSS the 6 genitourinary conditions (otherdisorders of bladder [discovery OR = 145 95 CI =

139ndash152 p = 477 times 10minus58 extension OR = 125 95 CI =122ndash128 p = 182 times 10minus78] functional disorders of bladderother symptomsdisorders of the urinary system hematuriaretention of urine and urinary incontinence) had the stron-gest associations with MS severity as a category These geni-tourinary conditions likely indicated MS-related morbiditiesrather than comorbidities (table 2) Two coexisting infectionslikely result from MS immune-modifying treatments pneu-monia (discovery OR = 133 95 CI = 123ndash144 p = 284 times10minus12 extension OR = 116 95 CI = 114ndash119 p = 104 times10minus32) and fever of unknown origin (discovery OR = 12395 CI = 116ndash130 p = 218 times 10minus11 extension OR = 12895 CI = 122ndash135 p = 189 times 10minus20) Other infections arelikely consequences of worsening physical and genitourinaryfunction in more disabled patients superficial cellulitis andabscess (discovery OR = 117 95CI = 109ndash125 p = 426 times10minus6 extension OR = 107 95 CI = 105ndash109 p = 533 times10minus10) and urinary tract infection (discovery OR = 132 95CI = 126ndash138 p = 741 times 10minus33 extension OR = 136 95CI = 133ndash140 p = 590 times 10minus11)

Beyond the genitourinary and infectious conditions othercomorbidities meeting the predefined reporting criteria (table2) included epilepsy (epilepsy recurrent seizures convulsions[discovery OR = 119 95 CI = 111ndash127 p = 331 times 10minus7extension OR = 113 95 CI = 109ndash116 p = 265 times 10minus14]convulsions [discovery OR = 121 95 CI = 112ndash130 p =260 times 10minus7 extension OR = 123 95 CI = 117ndash130 p =249 times 10minus15]) metabolic disturbance (disorders of fluidelectrolyte and acid-base balance [discovery OR = 126 95CI = 116ndash136 p = 108 times 10minus8 extension OR = 127 95CI = 122ndash132 p = 141 times 10minus35]) abnormal movement(discovery OR = 124 95 CI = 119ndash130 p = 139 times 10minus19extension OR = 111 95 CI = 108ndash114 p = 997 times 10minus14)and gastrointestinal discomfort (nausea and vomiting [dis-covery OR = 121 95 CI = 112ndash130 p = 136 times 10minus6extension OR = 107 95 CI = 104ndash112 p = 173 times 10minus4])Of interest benign neoplasm of the skin was inversely asso-ciated with the MSSS in both cohorts (discovery OR = 08995 CI = 085ndash094 p = 493 times 10minus6 extension OR = 09295 CI = 090ndash094 p = 167 times 10minus14) The observed cor-relations among the reported comorbidities suggest that thepredetermined Bonferroni threshold of significance was likelyconservative (figure 2) but it was important for the mainfindings to adhere to the predefined reporting criteria

Comorbidities not meeting thereporting criteriaA number of conditions that did not meet the stringent pre-defined reporting criteria are noteworthy First we founddifferent coexisting psychiatric conditions with greater MSseverity in the 2 cohorts Mood disorders such as depression(OR = 117 95 CI = 112ndash122 p = 283 times 10minus14) andadjustment reaction (OR= 115 95CI = 110ndash120 p = 186times 10minus9) reached significance threshold only in the discoveryanalysis (table e-1 linkslwwcomNXIA294) whereasschizophrenia (OR = 112 95 CI = 107ndash116 p = 994 times

4 Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 6 | November 2020 NeurologyorgNN

10minus8) and psychosis (OR = 114 95CI = 108ndash119 p = 169times 10minus7) did so only in the extension analysis (table e-2)Second although osteoporosis (OR = 140 95 CI =130ndash151 p = 208 times 10minus19) and osteoporosis osteopeniaand pathologic fracture (OR = 121 95 CI = 116ndash127 p =147 times 10minus15) reached the significance threshold only in thediscovery analysis (table e-1) related conditions such asseveral types of fractures (eg fracture of lower limb [OR =111 95 CI = 107ndash115 p = 270 times 10minus9]) were also sig-nificant in the extension analysis (table e-2) Third severalnotable vascular conditions did not meet the stringent pre-defined reporting criteria in our study (tables e-1 and e-2)Diabetes (discovery OR = 117 95 CI = 108ndash125 p = 420times 10minus5) cerebrovascular (extension OR = 108 95 CI =104ndash111 p = 843 times 10minus6) and peripheral vascular disease(extension OR = 115 95 CI = 110ndash120 p = 128 times 10minus8)reached significance in only 1 cohort Hyperlipidemia did notreach significance in either cohort Hypertension had oppo-site direction of effect (discovery OR = 109 95 CI =104ndash114 p = 550 times 10minus4 extension OR = 093 95 CI =091ndash095 p = 552 times 10minus10) Finally some comorbiditiesexpected in severe MS such as physical impairment (egother paralytic syndromes [OR = 142 95 CI = 137ndash146 p= 185 times 10minus10]) and chronic ulcers (eg decubitus ulcers[OR = 180 95 CI = 166ndash194 p = 773 times 10minus49]) reachedsignificance only in the extension analysis but not in the dis-covery analysis (tables e-1 and e-2)

Independent discovery analysisAnalyzing the discovery cohort data independently weidentified 42 (of the 109 that met the 3 prevalence thresh-old 39) coexisting medical conditions that were signifi-cantly associated with a higher MSSS after Bonferroni

correction and covariate adjustment (table e-1 linkslwwcomNXIA294 figures 3A and 4A) Of the 31 PheCodesthat reached the 3 case threshold in the discovery cohortdata (but were excluded from the discovery analysis due tofailure to reach the 3 case threshold in the extension co-hort) 9 medical conditions are significantly associated withthe MSSS after Bonferroni adjustment in an independentanalysis of discovery data Of these 9 conditions 2 hadprevalence above 10 in the discovery cohort

Independent extension analysisAnalyzing the extension cohort data independently weidentified 130 (of the 365 that met the 3 prevalencethreshold 36) coexisting medical conditions that were sig-nificantly associated with a higher MSSS after Bonferronicorrection and covariate adjustment (table e-2 linkslwwcomNXIA294 figures 3B and 4B) Of the 287 PheCodesthat reached the 3 case threshold in the extension cohort(but were excluded from the extension analysis due to failureto reach the 3 case threshold in the discovery cohort) wefound that 90 of these medical conditions are significantlyassociated with the MSSS after Bonferroni adjustment in theindependent analysis of the BC extension cohort data Ofthese 90 medical conditions 39 had prevalence above 10 inthe extension cohort

DiscussionIn this study of 2 large independent cohorts we deployed aphenome-wide approach to comprehensively assess thecomorbidity burden associated with higher disease severity inMS Notwithstanding the differences between the 2 cohorts

Table 1 Characteristics of the MS cohorts

Patient characteristicsDiscovery cohort (Boston UnitedStates) N = 3439

Extension cohort (British ColumbiaCanada) N = 4876

Age at symptom onset y median (Q1ndashQ3) 330 (271ndash409) 319 (258ndash395)

Female N () 2524 (73) 3535 (72)

Non-Hispanic European N () 2947 (86) Not available

Disease duration (MS symptom onset to the last EDSS) ymedian (Q1ndashQ3)

134 (83ndash209) 145 (87ndash227)

Unique ICD-9 code count N median (Q1ndashQ3) 7 (3ndash18) 17 (3ndash45)

Duration from the first to the last ICD-9 code y median(Q1ndashQ3)

113 (65ndash154) 129 (60ndash155)

Duration from the first to the last EDSS value y median(Q1ndashQ3)

60 (30ndash98) 105 (42ndash145)

Disease course n ()

Progressive onset 378 (11) 433 (89)

Relapsing onset 3061 (89) 4443 (91)

Abbreviations EDSS = Expanded Disability Status Scale ICD = International Classification of Diseases

NeurologyorgNN Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 6 | November 2020 5

(health care systems billing practices and data integration)we reported 16 coexisting medical conditions that met thepredefined stringent reporting criteria for reaching the Bon-ferroni threshold of significance and displaying the same di-rection of effect in both cohorts Our application of thePheWAS approach sets the road map for comprehensivecomorbidity investigation in other chronic neurologicaldiseases

Consistent with prior targeted approaches investigating pre-determined candidate conditions19 we found an over-representation of genitourinary conditions among the reportedcomorbidities associated with MS severity after accounting forconditions that are predicted to occur with increasing diseaseduration It is important to note that some of the reportedcoexisting conditions particularly the genitourinary disordersare known MS-related morbidities rather than independentcomorbidities Importantly the disproportionally large numberof genitourinary conditions that met the stringent reporting

criteria not only confirms the clinical experience of the largeimpact of disease severity on the genitourinary system inMS butalso reaffirms the validity of the PheWAS approach in high-lighting this category of conditions

Another comorbidity category associated with a higher MSSSwas infectious diseases including urinary tract infection pneu-monia superficial cellulitis and abscess and fever of unknownorigin (likely secondary to an underlying infection) Priorstudies reported higher prevalence of infections amongMS thanthe general population2021 attributable to older age or the useof disease-modifying therapies22 Infections such as abscess andcellulitis can also result from physical debilitation Infectionscontribute to MS mortality23 and increase MS relapse risk24

Furthermore relapses associated with an infection might resultin greater neurologic damage (relative to relapses withoutcomorbid infections) and greater long-term severity25 Ourstudy again highlights infections associated with MS severityindependent of age while reaffirming the PheWAS approach

Table 2 Coexisting medical conditions associated higher MS severity scores

PheCodea Reported medical conditionbc

Discovery Extension

Prevd

() ORe OR 95 CI p ValuePrev() OR OR 95 CI p Value

216 Benign neoplasm of skin 94 089 085 094 493Eminus06 302 092 090 094 167Eminus14

276 Disorders of fluid electrolyte and acid-basebalance

38 126 116 136 108Eminus08 89 127 122 132 141Eminus35

345 Epilepsy recurrent seizures convulsions 40 119 111 127 331Eminus07 114 113 109 116 265Eminus14

3453 Convulsions 33 121 112 130 260Eminus07 40 123 117 130 249Eminus15

350 Abnormal movement 108 124 119 130 139Eminus19 149 111 108 114 997Eminus14

480 Pneumonia 36 133 123 144 284Eminus12 210 116 114 119 104Eminus32

591 Urinary tract infection 145 132 126 138 741Eminus33 258 136 133 140 590Eminus11

593 Hematuria 54 129 121 138 548Eminus15 53 112 107 117 111Eminus06

596 Other disorders of bladder 159 145 139 152 477Eminus58 309 125 122 128 182Eminus78

5965 Functional disorders of bladder 156 144 138 151 140Eminus56 156 131 127 135 616Eminus69

599 Other symptomsdisorders of the urinarysystem

188 133 128 138 181Eminus42 595 113 111 115 657Eminus30

5992 Retention of urine 40 135 125 145 528Eminus14 68 132 126 138 761Eminus33

5994 Urinary incontinence 79 140 132 148 383Eminus28 93 116 112 120 389Eminus17

681 Superficial cellulitis and abscess 49 117 109 125 426Eminus06 368 107 105 109 533Eminus10

783 Fever of unknown origin 61 123 116 130 218Eminus11 44 128 122 135 189Eminus20

789 Nausea and vomiting 33 121 112 130 136Eminus06 72 107 104 112 173Eminus04

a Related International Classification of Diseases Ninth Revision codes of a single medical condition are mapped to a single phenotype code or PheCode(column 1)b Predefined reporting criteria are meeting the Bonferroni threshold of statistical significance and having the same direction of effect in both the discoverycohort and the extension cohortc Description of the PheCode or the medical condition (column 2) is directly extracted from a publicly available mapping tool without further editing by theauthors14d Prevalence (Prev) is defined as the percentage of the patients in the cohort having a medical condition as indicated by the PheCodee OR represents the estimated factor by which the odds of the presence of the givenmedical condition changes per unit increase of theMS Severity Score (onthe decile scale) holding controlled covariates constant

6 Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 6 | November 2020 NeurologyorgNN

Among the reported findings epilepsy and disorders of fluidelectrolyte and acid-base balance are underappreciated fortheir association with MS severity though consistent withanecdotal clinical observation Building on the known prev-alence of epilepsy among MS (2)26 our finding of epilepsyin association with MS severity is consistent with a priorstudy27 and may be due to irritation from MS lesions Dis-orders of fluid electrolyte and acid-base as a category ofcomorbid metabolic disturbance also commonly occur inpatients with heart diseases or renal failure and may also re-flect poor diet fluid intake or swallowing difficulties in severeMS28 Indeed another reported finding is nausea and vomitingwhich could contribute to metabolic disturbance Finally

abnormal movement was a reported finding that warrantstargeted investigation particularly given report of increasedfrequency of movement disorder even early in MS29

Although we set out to investigate the comorbidity burdenassociated with MS severity we found an interesting inverseassociation between MS severity and certain cancers in-cluding benign neoplasm of the skin Prior studies reported alower incidence of all-cause cancers among people with MSrelative to matched general population3031 Diagnostic ne-glect particularly in those with more severe disability couldbe contributory30 As an alternative explanation less sun ex-posure and lower serum vitamin D level early in the disease

Figure 2 Correlation structure of the reported coexisting medical conditions (as indicated by PheCodes) associated withhigher MS severity scores

As shown in a heat map (A) and in a hierarchicalclustering (B) format The color saturation and thecircle size in the heatmap both indicate the strength of(positive or negative) correlation between 2 medicalconditions Reporting criteria were defined as meet-ing the Bonferroni threshold of statistical significanceand having the same direction of effect in both thediscovery cohort and the extension cohort Strengthof association minuslog(p value) from the extension cohortis shown Refer to table 2 for explanation of eachphenotype code (PheCode)

NeurologyorgNN Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 6 | November 2020 7

Figure 3 Manhattan plots of the mapped coexisting medical conditions associated with higher MS severity scores

Conditions that met the requirement of minimum cases and respective ndashlog (p values) findings from independent analysis in the discovery cohort (A) andextension cohort (B) Refer to table e-1 and table e-2 (linkslwwcomNXIA294) for details of each condition The red line indicates the threshold of significanceafter Bonferroni correction and covariate adjustment the blue line represents p = 005

8 Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 6 | November 2020 NeurologyorgNN

course may contribute to greater MS relapse risk and greaterdisease worsening in some patients with MS32

We made several other interesting observations with respect tothe comorbidity burden in MS First we found comorbidpsychiatric conditions with MS severity in the 2 cohorts al-though the relationships are inconsistent Of interest a priorstudy reported that the presence of psychiatric conditions in-creased the severity of subsequent neurologic disability33 Wecould not rule out the possibility that (1) population

differences in the prevalence andmanagement of mental healthconditions within each cohort contributed to these differences(2) validation could be achieved with less stringent criteria or(3) disease severity as measured by the MSSS may not con-sistently correlate with psychiatric comorbidities Second os-teoporosis reached the significance threshold in the discoveryanalysis only whereas related conditions (eg several types offractures) were significant in the extension analysis Togetherthese findings suggest a higher risk of osteoporosis and relatedcomplications in those with a greater MS severity which may

Figure 4 Plot of estimated OR and cohort prevalence formapped coexistingmedical conditions associated with higherMSseverity scores (MSSS)

Findings from independent analysisin the discovery cohort (A) and ex-tension cohort (B) Results belowthe dotted lines are inverse associ-ations ORs represent per unit in-crease of the MSSS (on the decilescale)

NeurologyorgNN Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 6 | November 2020 9

be associated with vitamin D deficiency early in the disease andreduced mobility in more disabled individuals and is consistentwith the report of higher prevalence of osteoporosis inMS thancontrols34 Third several notable vascular conditions such ascerebrovascular and peripheral vascular disease diabetes hy-perlipidemia and hypertension did not meet the stringentpredefined reporting criteria in our study Vascular diseaseswere previously reported to be associated with MSseverity35ndash38 Finally some comorbidities expected in severeMS such as physical impairment (eg hemiplegia and paralyticsyndromes) and chronic ulcers (eg decubitus ulcers) werefound only in the extension analysis but not in the discoveryanalysis Some of the inconsistencies with prior reports couldreflect differences in the study design study population and themethod with which comorbidities were identified whereasdifferences in reaching the stringent reporting criteria betweenthe 2 cohorts could be attributable to differences in billingpractices between the 2 distinct health care systems the greaterability to systematically capture all ICD codes in the extensioncohort (Canada) than the discovery cohort (United States)and longer follow-up duration in the extension cohort than thediscovery cohort Although the application of stringentreporting criteria is a strength of our study those medicalconditions identified only in a single cohort warrant furtherinvestigation

Our study had several limitations First the ICD systemcaptures a broad spectrum of diagnoses but studies based onICD data inevitably encounter potential incompleteness andmisclassification We took steps to mitigate these issues bysystematically capturing all available ICD codes consolidatingICD codes related to a single medical condition and estab-lishing minimum case requirement for each medical condi-tion Second a related limitation pertinent to the discoverycohort was that not all patients had primary care providers inthe same health care system as the source EHRs data Primarycare providers are more likely to record all available ICD codesin the EHRs than subspecialty physicians In a multipayerhealth care system such as the United States incompletecapture of ICD codes is more common for patients whoseprimary care providers are not in the same hospital systembecause the EHRs of one hospital system may not capturecodes for care received at outside facilities To mitigate thisissue in the discovery analysis we included the presence ofprimary care providers within the same hospital system as theEHRs data as a covariate Despite these differences betweenthe 2 cohorts and the use of stringent reporting criteria ourstudy found 16 coexisting comorbidities which point to therobustness of the PheWAS approach Finally our cross-sectional association study does not draw causal conclusionsand the PheWAS approach may not sufficiently distinguishthe direction of effect or the direct consequence between MSseverity and comorbidities Indeed we calculated theMSSS toobtain a stable cross-sectional measure of severity based on upto the 3 most recent EDSS scores and associated diseasedurations If we were to require the EDSS to precede (orfollow) the occurrence of an ICD code there would be

insufficient number of events (ICD-9 codes) to permit ade-quately powered analyses and the analyses may have mis-measurement of the timing of the first occurrence of an ICD-9code due to incomplete records Our present study paves theway for future prospective investigations which will be criticalto establish temporal relationships between disability assess-ment and coexisting condition occurrence and account for allpotentially important confounders

In summary our novel application of the phenome-wide ap-proach systematically identified the comorbidity burden andinformed clinically relevant and lesser appreciated comorbiditiesassociated with disease severity in a chronic neurologic diseaseMS Future prospective investigation of the comorbidity burdenin relation to worsening disabilitymay help realize individualizedmonitoring and management of patients with severe MS

AcknowledgmentThe authors thankMariann Polgar-Turcsanyi AndrewCaganand Vivian Gainer for providing technical assistance for theBoston cohort data extraction They gratefully acknowledge theBritish Columbia MS Clinic neurologists who contributed to thestudy through patient examination and data collection (currentmembers at the time of data extraction listed here by primaryclinic) UBC MS Clinic A Traboulsee MD FRCPC (UBCHospital MS Clinic Director and Head of the UBC MSPrograms) A-L Sayao MD FRCPC V Devonshire MDFRCPC S Hashimoto MD FRCPC (UBC and Victoria MSClinics) J Hooge MD FRCPC (UBC and Prince George MSClinic) L Kastrukoff MD FRCPC (UBC and Prince GeorgeMS Clinic) J Oger MD FRCPC Kelowna MS Clinic DAdams MD FRCPC D Craig MD FRCPC S Meckling MDFRCPC PrinceGeorgeMSClinic L DalyMD FRCPCVictoriaMS Clinic O Hrebicek MD FRCPC D Parton MD FRCPCKAtwell-PopeMD FRCPCThe authors gratefully acknowledgePartners MS Clinic neurologists who contributed to the studythrough patient examination and data collection The viewsexpressed in this article do not necessarily reflect the views of eachindividual acknowledged The British Columbia Ministry ofHealth approved access to and use of data from British ColumbiaCanada facilitated by Population Data British Columbia Allinferences opinions and conclusions drawn in this study are thoseof the authors and do not reflect the opinions or policies of theData Stewards

Study fundingT Zhang M Goodman F Zhu R Carruthers T Chitnis HWeiner and T Cai report no disclosures relevant to the manu-script B Healy has received research funding fromMerck SeronoSA Verily Life Sciences Genzyme and Novartis and has servedon an advisory board for Biogen Idec (none relevant to this study)P De Jager was a Harry Weaver Neuroscience Scholar of theNational Multiple Sclerosis Society (JF2138A1) and has researchfunding from Biogen and SanofiGenzyme (none relevant to thisstudy) H Tremlett is the Canada Research Chair for Neuro-epidemiology and Multiple Sclerosis Current research supportreceived from the National Multiple Sclerosis Society the

10 Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 6 | November 2020 NeurologyorgNN

Canadian Institutes of Health Research the Multiple SclerosisSociety of Canada and the Multiple Sclerosis Scientific ResearchFoundation In addition in the last 5 years has received researchsupport from the UKMS Trust and travel expenses to present atCME conferences from the Consortium of MS Centres (2018)the National MS Society (2016 and 2018) ECTRIMSACTRIMS (2015 2016 2017 2018 and 2019) and AmericanAcademy of Neurology (2015 2016 and 2019) Speaker hono-raria are either declined or donated to an MS charity or to anunrestricted grant for use by HTrsquos research group Z Xia was arecipient of the Clinician Scientist Development Award from theNationalMultiple Sclerosis Society and the American Academy ofNeurology and is supported by NIH R01-NS098023

DisclosureNone relevant to the study Go to NeurologyorgNN for fulldisclosures

Publication historyReceived by Neurology Neuroimmunology amp NeuroinflammationMarch 2 2020 Accepted in final form June 17 2020

References1 Marrie RA Comorbidity in multiple sclerosis implications for patient care Nat Rev

Neurol 201713375ndash382 doi 101038nrneurol2017332 Tettey P Siejka D Simpson S et al Frequency of comorbidities and their association

with clinical disability and relapse in multiple sclerosis Neuroepidemiology 201646106ndash113 doi 101159000442203

3 Denny JC Bastarache L Roden DM Phenome-wide association studies as a tool toadvance precision medicine Annu Rev Genom Hum Genet 201617353ndash373 doi101146annurev-genom-090314-024956

4 Denny JC Ritchie MD Basford MA et al PheWAS demonstrating the feasibility of aphenome-wide scan to discover gene-disease associations Bioinformatics 2010261205ndash1210 doi 101093bioinformaticsbtq126

5 Gauthier SA Glanz BI Mandel M Weiner HL A model for the comprehensiveinvestigation of a chronic autoimmune disease the multiple sclerosis CLIMB studyAutoimmun Rev 20065532ndash536 doi 101016jautrev200602012

6 Xia Z Secor E Chibnik LB et al Modeling disease severity in multiple sclerosis usingelectronic health records PLoS One 20138e78927 doi 101371journalpone0078927

7 BC Ministry of Health [creator] Consolidation File (MSP Registration amp PremiumBilling) BC Ministry of Health 2011 Data Extract MOH (2009) Available atwww2govbccagovcontenthealthconducting-health-research- evaluationdata-access-health-data-central Accessed June 8 2020

8 Canadian Institute of Health Information [creator] Discharge Abstract Database(Hospital Separations) British Columbia Ministry of Health 2011 Data ExtractMOH (2011) Available at www2govbccagovcontenthealthconducting-health- research-evaluationdata-access-health-data-central Accessed June 82020

9 BC Ministry of Health [creator] Medical Services Plan (MSP) Payment InformationFile Population Data BC 2011 Data Extract MOH (2010) Available at www2govbccagovcontenthealthconducting-health-research-evaluationdata-access-health-data-central Accessed June 8 2020

10 Canadian Institute for Health Information (CIHI) Canadian Coding Standards forVersion 2018 ICD-10-CA and CCI Available at httpssecurecihicafree_prod-uctsCodingStandards_v2018_ENpdf Accessed August 5 2020

11 Poser CM Paty DW Scheinberg L et al New diagnostic criteria for multiple sclerosisguidelines for research protocols Ann Neurol 198313227ndash231 doi 101002ana410130302

12 McDonald WI Compston A Edan G et al Recommended diagnostic criteria formultiple sclerosis guidelines from the International Panel on the diagnosis of multiplesclerosis Ann Neurol 200150121ndash127 doi 101002ana1032

13 Roxburgh RHSR Seaman SR Masterman T et al Multiple Sclerosis Severity Scoreusing disability and disease duration to rate disease severity Neurology 2005641144ndash1151 doi 10121201WNL000015615519270F8

14 Carroll RJ Bastarache L Denny JC R PheWAS data analysis and plotting tools forphenome-wide association studies in the R environment Bioinformatics 2014302375ndash2376 doi 101093bioinformaticsbtu197

15 Denny JC Bastarache L Ritchie MD et al Systematic comparison of phenome-wideassociation study of electronic medical record data and genome-wide associationstudy data Nat Biotechnol 2013311102ndash1110 doi 101038nbt2749

16 Hebbring SJ Schrodi SJ Ye Z Zhou Z Page D Brilliant MH A PheWAS approach instudying HLA-DRB11501 Genes Immun 201314187ndash191 doi 101038gene20132

17 Liao KP Sparks JA Hejblum BP et al Phenome-wide association study of autoan-tibodies to citrullinated and noncitrullinated epitopes in rheumatoid arthritis ArthritisRheumatol 201769742ndash749 doi 101002art39974

Appendix Authors

Name Location Contribution

TingtingZhang PhD

Brown University Designed and conceptualizedthe study analyzed the dataand drafted and revised themanuscript for intellectualcontent

MatthewGoodmanPhD

Harvard TH ChanSchool of PublicHealth

Designed and conceptualizedthe study analyzed the dataand drafted and revised themanuscript for intellectualcontent

Feng ZhuMSc

University ofBritish Columbia

Major role in the acquisition ofthe data and analyzed the data

BrianHealyPhD

Brigham andWomenrsquos Hospital

Major role in the acquisition ofthe data interpreted data andrevised the manuscript forintellectual content

RobertCarruthersMD

University ofBritish Columbia

Major role in the acquisition ofthe data

TanujaChitnis MD

Brigham andWomenrsquos Hospital

Major role in the acquisition ofthe data and revised themanuscript for intellectualcontent

HowardWeiner MD

Brigham andWomenrsquos Hospital

Major role in the acquisition ofthe data

Tianxi CaiScD

Harvard TH ChanSchool of PublicHealth

Interpreted data and revised themanuscript for intellectualcontent

Philip DeJager MDPhD

ColumbiaUniversity

Designed and conceptualizedthe study major role in theacquisition of the datainterpreted data and revisedthe manuscript for intellectualcontent

Appendix (continued)

Name Location Contribution

HelenTremlettPhD

University ofBritish Columbia

Designed and conceptualized thestudy major role in theacquisition of the datainterpreted data and revised themanuscript for intellectualcontent

Zongqi XiaMD PhD

University ofPittsburgh

Designed and conceptualized thestudy major role in theacquisition of the data analyzeddata interpreted data anddrafted and revised themanuscript for intellectualcontent

T Zhang M Goodman H Tremlett and Z Xia had full access to all the datain the study and take responsibility for the integrity of the data the accuracyof the data analysis and the decision to submit for publication All authorshave seen and approved the submitted version of the manuscript

NeurologyorgNN Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 6 | November 2020 11

18 R Core Team R A Language and Environment for Statistical Computing 2016Available at R-projectorg Accesse April 1 2020

19 Ineichen BV Schneider MP Hlavica M Hagenbuch N Linnebank M Kessler TMHigh EDSS can predict risk for upper urinary tract damage in patients with multiplesclerosis Mult Scler J 201724529ndash534 doi 1011771352458517703801

20 Wijnands JM Kingwell E Zhu F et al Infection-related health care utilization amongpeople with and without multiple sclerosis Mult Scler J 2016651352458516681198doi 1011771352458516681198

21 Montgomery S Hillert J Bahmanyar S Hospital admission due to infections inmultiple sclerosis patients Eur J Neurol 2013201153ndash1160 doi 101111ene12130

22 Wijnands JMA Zhu F Kingwell E et al Disease-modifying drugs for multiple scle-rosis and infection risk a cohort study J Neurol Neurosurg Psychiatry 2018891050ndash1056 doi 101136jnnp-2017-317493

23 Manouchehrinia A Tanasescu R Tench CR Constantinescu CS Mortality in mul-tiple sclerosis meta-analysis of standardised mortality ratios J Neurol NeurosurgPsychiatry 201687324ndash331 doi 101136jnnp-2015-310361

24 Buljevac D Flach HZ Hop WCJ et al Prospective study on the relationship betweeninfections and multiple sclerosis exacerbations Brain 2002125952ndash960

25 Perry VH Newman TA Cunningham C The impact of systemic infection on theprogression of neurodegenerative disease Nat RevNeurosci 20034103ndash112 doi 101038nrn1032

26 Catenoix H Marignier R Ritleng C et al Multiple sclerosis and epileptic seizuresMult Scler J 20111796ndash102 doi 1011771352458510382246

27 Zhang T Tremlett H Zhu F et al Effects of physical comorbidities on disabilityprogression inmultiple sclerosis Neurology 201890e419ndashe427 doi 101212WNL0000000000004885

28 Guan X-L Wang H Huang H-S Meng L Prevalence of dysphagia in multiple scle-rosis a systematic review and meta-analysis Neurol Sci 201536671ndash681 doi 101007s10072-015-2067-7

29 Abboud H Yu XX Knusel K Fernandez HH Cohen JA Movement disorders in earlyMS and related diseases a prospective observational study Neurol Clin Pract 2019924ndash31 doi 101212CPJ0000000000000560

30 Kingwell E Bajdik C Phillips N et al Cancer risk in multiple sclerosis findingsfrom British Columbia Canada Brain 20121352973ndash2979 doi 101093brainaws148

31 Thormann A Koch-Henriksen N Laursen B Soslashrensen PS Magyari M Inversecomorbidity in multiple sclerosis findings in a complete nationwide cohort MultScler Relat Disord 201610181ndash186 doi 101016jmsard201610008

32 Simpson S van der Mei I Lucas RM et al Sun exposure across the life coursesignificantly modulates early multiple sclerosis clinical course Front Neurol 2018916 doi 103389fneur201800016

33 McKay KA Tremlett H Fisk JD et al Psychiatric comorbidity is associated withdisability progression in multiple sclerosis Neurology 201890e1316ndashe1323 doi 101212WNL0000000000005302

34 Bisson EJ Finlayson ML Ekuma O Leslie WD Marrie RA Multiple sclerosis isassociated with low bone mineral density and osteoporosis Neurol Clin Pract 20199391ndash399 doi 101212CPJ0000000000000669

35 Roshanisefat H Bahmanyar S Hillert J Olsson T Montgomery S Multiple sclerosisclinical course and cardiovascular disease riskmdashSwedish cohort study Eur J Neurol2014211353ndashe88 doi 101111ene12518

36 Marrie RA Rudick R Horwitz R et al Vascular comorbidity is associated with morerapid disability progression in multiple sclerosis Neurology 2010741041ndash1047 doi101212WNL0b013e3181d6b125

37 Tettey P Simpson S Taylor BV van der Mei IAF Vascular comorbidities in the onsetand progression of multiple sclerosis J Neurol Sci 201434723ndash33 doi 101016jjns201410020

38 Kowalec K McKay KA Patten SB et al Comorbidity increases the risk of relapse inmultiple sclerosis a prospective study Neurology 2017892455ndash2461 doi 101212WNL0000000000004716

12 Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 6 | November 2020 NeurologyorgNN

DOI 101212NXI000000000000086420207 Neurol Neuroimmunol Neuroinflamm

Tingting Zhang Matthew Goodman Feng Zhu et al severity

Phenome-wide examination of comorbidity burden and multiple sclerosis disease

This information is current as of August 19 2020

ServicesUpdated Information amp

httpnnneurologyorgcontent76e864fullhtmlincluding high resolution figures can be found at

References httpnnneurologyorgcontent76e864fullhtmlref-list-1

This article cites 33 articles 4 of which you can access for free at

Subspecialty Collections

httpnnneurologyorgcgicollectionmultiple_sclerosisMultiple sclerosis

httpnnneurologyorgcgicollectionelectronic_medical_recordsElectronic medical recordsfollowing collection(s) This article along with others on similar topics appears in the

Permissions amp Licensing

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

Reprints

httpnnneurologyorgmiscaddirxhtmlreprintsusInformation about ordering reprints can be found online

Academy of Neurology All rights reserved Online ISSN 2332-7812Copyright copy 2020 The Author(s) Published by Wolters Kluwer Health Inc on behalf of the AmericanPublished since April 2014 it is an open-access online-only continuous publication journal Copyright

is an official journal of the American Academy of NeurologyNeurol Neuroimmunol Neuroinflamm