Investigating the accuracy, risk impact, and cost-effectiveness of component-resolved diagnostic test for food allergy: a systematic review protocol

PROTOCOL OPEN In es iga ing he accu acy, isk impac , and cos -e ec i eness o componen - esol ed diagnos ic es o ood alle gy: a sys ema ic e iew p o ocol Ja ie Flo es Kim 1 , B igh I. Nwa u 1,2 , Nicola McClea y 1 , And ew S odda 1,3 and Aziz Sheikh 1 npj P ima y Ca e Respi a o y Medicine (2017) 27:10 ; doi:10.1038/s41533-017-0015-0 BACKGROUND Food alle gy—pa icula ly immunoglobulin E (IgE)-media ed ood alle gy—is now a global public heal h p oblem. I esul s in conside able mo bidi y, heal hca e u ilisa ion, and impai men in quali y o li e. 1–6 Food alle gy eac ions ange in se e i y om ela i ely mild ea u es o li e- h ea ening anaphylaxis. 2,3 Es i- ma es o he p e alence o ood alle gy a y, bu o e all li e ime p e alence has been es ima ed as a ec ing up o 7% o child en and 6% o adul s. 4,5 Ca e ul a oidance o o ending oods is he mains ay p e en i e s a egy o ood alle gy. 2,3 Accu a e diagnosis is essen ial in o de o p o ide app op ia e, po en ially li e-sa ing ad ice on how o p e en and manage eac ions. This will also help o p e en unwa an ed die a y es ic ions. 2,3 The diagnosis o ood alle gy is ypically based on a combina ion o clinical his o y and an objec i e ma ke o alle gic sensi isa ion, namely, skin p ick es s (SPT) and/o immunoassays o se um-specific IgE le els. While hese app oaches ha e easonable sensi i i y, hey ha e sub- op imal specifici y and a e in gene al poo ly p edic i e o he se e i y o eac ions. Fu he diagnos ic confi ma ion wi h a double-blinded placebo-con olled ood challenge (DBPCFC)— he gold s anda d diagnos ic es —is he e o e o en equi ed. 2,3,7 DBPCFC is howe e cos ly, echnically challenging, ime-consum- ing, labou -in ensi e, and associa ed wi h impo an sa e y isks as he p ocedu e can igge anaphylaxis. 2,3,8 Ne e heless, al hough isky, DBPCFCs o e a eliable diagnosis, which in u n inc eases pa ien s’quali y o li e. 9 The limi a ions o con en ional me hods o diagnosing ood alle gy ha e s imula ed p omising de elopmen s in molecula - based diagnos ic echniques, collec i ely e e ed o as componen - esol ed diagnosis (CRD). 2,8,10 Molecula diagnosis o ood alle gy in ol es examining he specific p o eins in oods, p ima ily a he molecula le el. 8 In his case, a he han p o iding a summa y o all he alle gen p o eins in a pa icula ood (e.g., peanu s), molecula -based me hods quan i y he indi idual p o eins wi hin a ood ha may be esponsible o alle gic eac ions. 8 In CRD, specific IgE esponses a e e alua ed agains indi idual alle genic molecules o he epi opes o hose alle gens. The e o e, CRD echniques ha e he po en ial o enhance he specifici y and sensi i i y o se um-specific IgE esponses o oods. They may also: acili a e ou abili y o de e mine specific ood alle gy pheno ypes; help imp o e compila ion o a pa ien - ailo ed isk p ofile o specific ood alle gens, gi en ha IgE an ibodies o ood molecules a y om pa ien o pa ien and also geog aphically; and boos he abili y o dis inguish be ween p ima y and seconda y sensi ise s. 2,8,10 Impo an ly, in CRD app oaches, diagnosis can be unde aken ei he in single es o ma s o in a mic oa ay by simul aneously e alua ing a ange o alle gens. 2,8,10 The e is now a subs an ial body o e idence on using CRD o diagnose ood alle gy, which needs o be syn hesised in o de o gene a e obus es ima es o diagnos ic accu acy and assess cos - e ec i eness in compa ison wi h con en ional app oaches. Fu he mo e, a syn hesis o he unde lying e idence will help o in o m delibe a ions on whe he and how CRD should be deployed o (i) diagnosing ood alle gy, and (ii) assessing pa ien isk h ough he p edic ion o ood alle gy se e i y (i.e., he likelihood o li e- h ea ening anaphylaxis as opposed o ela i ely mild ea u es) ac oss di e ing heal hca e se ice con ex s. AIMS We seek o iden i y, c i ically app aise, and unde ake me a- analyses o s udies using CRD o he diagnosis o ood alle gy. Specifically, we aim o: a. De e mine he accu acy (i.e., sensi i i y, specifici y, posi i e and nega i e p edic i e alues) o CRD o diagnosis o cow’s milk, whea , egg, peanu , soy, ee nu s, fish, and shellfish alle gy. b. Summa ise he e idence on he cos and cos -e ec i eness o CRD in compa ison wi h con en ional echniques o he diagnosis o cow’s milk, whea , egg, peanu , soy, ee nu s, fish, and shellfish alle gy. c. Summa ise he e idence on he abili y o CRD o p edic he se e i y o cow’s milk, whea , egg, peanu , soy, ee nu s, fish, and shellfish alle gy. METHODS The P e e ed Repo ing I ems o Sys ema ic Re iew and Me a-Analysis P o ocols (PRISMA-P) checklis has been used o guide he epo ing o his p o ocol. 11 Recei ed: 8 Oc obe 2016 Re ised: 13 Decembe 2016 Accep ed: 20 Decembe 2016 1 As hma UK Cen e o Applied Resea ch, Cen e o Medical In o ma ics, Ushe Ins i u e o Popula ion Heal h Sciences and In o ma ics, The Uni e si y o Edinbu gh, Edinbu gh, UK; 2 School o Heal h Sciences, Uni e si y o Tampe e, Tampe e, Finland and 3 Edinbu gh Clinical T ials Uni , Cen e o Medical In o ma ics, Ushe Ins i u e o Popula ion Heal h Sciences and In o ma ics, The Uni e si y o Edinbu gh, Edinbu gh EH8 9AG, UK Co espondence: Aziz Sheikh (a[email p o ec ed]) www.na u e.com/npjpc m Published in pa ne ship wi h P ima y Ca e Respi a o y Socie y UK S udy eligibili y c i e ia Types o s udies. We will include p ospec i e, e ospec i e, and c oss- sec ional s udies. S udies mus ha e su ficien da a o calcula e he ele an diagnos ic measu es, including sensi i i y, specifici y, and posi i e and nega i e p edic i e alues. All s udies mus ha e a defined s udy popula ion, wi h ei he consecu i e o andom sampling o pa icipan s. I he ec ui men echnique used o selec pa icipan s is no indica ed in a s udy, we will con ac he au ho s o eques u he in o ma ion abou hei s udy. In he case we do no ecei e a esponse om au ho s, we will include such s udy and unde ake ele an sensi i i y analyses o assess hei impac on he o e all es pe o mance. Types o pa icipan s. We will include s udies ha ha e examined pa ien s wi h suspec ed ood alle gy o any age. Ta ge condi ions. We will include s udies examining he CRD diagnos ic accu acy o alle genic molecules o he eigh mos common IgE-media ed ood alle gies, namely, cow’s milk, egg, whea , soy, peanu , ee nu s (including, bu no limi ed o hazelnu , walnu , cashew, b azil nu , pis achio, almond), fish (including, bu no limi ed o cod and ca p), and shellfish (sh imp). Index es . We will include s udies ha ha e u ilised molecula diagnos ic echniques ei he based on alle genic molecules (using CRD) o he epi opes o hose alle gens (using epi ope mapping o p ofiling). Re e ence s anda d. We will use he DBPCFC as he gold s anda d o e alua ing es accu acy. We will include s udies in which DBPCFC has been unde aken in a leas 50% o subjec s in o de o p o ide su ficien sample o s eng hen he in e nal alidi y o each s udy. Seconda ily, wi h ega d o e alua ing impac on decision-making, we will also compa e CRD pe o mance wi h cu en s anda d p ac ice (i.e., clinical assessmen plus SPT and/o specific IgE). Exclusion c i e ia. We will exclude he ollowing epo s: e iews, discussion pape s, non- esea ch le e s and edi o ials; quali a i e s udies; case s udies, case se ies; animal s udies; and s udies ha include pa icipan s on he basis o a posi i e ood alle gy es esul . Sea ch s a egy We will sea ch he ollowing da abases om 1 Janua y 2000 o 31 July 2016 o s udies o diagnos ic es s o ood alle gy: AMED (O id), CAB Abs ac s (O id), he Coch ane Lib a y, CINAHL (EBSCO), Embase (O id), Global Heal h (O id), MEDLINE (O id), PsycINFO (O id), Web o Science Co e Collec ion (Thomson Reu e s), WHO’s Global Heal h Lib a y, and he Heal h Economic E alua ions Da abase. Al hough CRD me hods o ood alle gy we e al eady desc ibed in he 1990s, 12 hei applica ion o ood alle gy diagnosis came in o o ce ully in he 2000s, hence we ha e chosen he beginning o 2000 as a easonable s a ing ime o he li e a u e sea ch. We will ob ain addi ional e e ences by sea ching he e e ences ci ed in iden ified s udies, by con ac ing in e na ional expe s and au ho s who ha e published in he field (we ha e compiled a p elimina y lis o expe s o con ac ), and by sea ching he ISI Con e ence P oceedings Ci a ion Index. We will sea ch ial egis ies, such as Cu en Con olled T ials (h p://www.con olled- ials.com), ClinicalT ials.go (h p://www. clinical ials.go ), he Aus alian and New Zealand Clinical T ials Regis y (h p://www.anzc .o g.au), and WHO’s In e na ional Clinical T ials Regis y Pla o m (h p://www.who.in /ic p/en/) o iden i y ongoing s udies. Using he O id in e ace o MEDLINE, we ha e de eloped a sea ch s a egy (Appendix 1) o iden i y and e ie e ele an s udies o he e iew and his will be adap ed in sea ching he o he da abases. The e a e no language es ic ions o included s udies: li e a u e in languages o he han English will be ansla ed whe e possible, and any li e a u e ha we a e unable o ansla e will be epo ed. Da a managemen and selec ion p ocess We will expo e ie ed eco ds om all da abases o Endno e Lib a y o s udy sc eening, de-duplica ion, and managemen o he e ie ed eco ds. Ti les and abs ac s o e ie ed eco ds will be sc eened and ull ex copies o po en ially eligible s udies will be assessed by wo independen e iewe s; a hi d e iewe will a bi a e any disc epancies. Da a ex ac ion We will de elop and pilo a da a ex ac ion o m ha will be used o collec ele an da a om eligible s udies. Two e iewe s will independen ly ex ac ele an s udy da a om included s udies on o he o m; a hi d e iewe will a bi a e any disc epancies. Da a i ems We will p oduce desc ip i e ables o abula e and summa ise ele an da a om included s udies. We will cap u e he ollowing minimum se o da a i ems om included s udies: s udy au ho , coun y o s udy, yea o publica ion, ype o s udy design, s udy size, sou ce o s udy popula ion, ype o ood alle gy, ype o CRD used o diagnosed ood alle gy, key summa y esul s o diagnos ic accu acy, any epo ed isk assessmen measu es. Fo cos -e ec i eness analyses, he ollowing minimum se o da a i ems will be ex ac ed: compe ing es s analysed, economic s udy design ype, economic pe spec i e, es cos , measu ed ou come and i s alue, and u u e cos s o he ou come and he es . We will use he PRISMA checklis o guide he epo ing o he sys ema ic e iew. 13 Quali y assessmen o s udies Risk o bias in eligible s udies will be assessed by wo e iewe s; a hi d e iewe will a bi a e any disc epancies. We will assess he quali y o included s udies using he Quali y Assessmen o Diagnos ic Accu acy S udies (QUADAS-2) ool. 14 We will quali y assess he economic s udies using he D ummond c i e ia. 15 Fo all s udies, we will quali y assess he ollowing componen s o each s udy: pa ien popula ion; index es used and in e p e a ion; ype o e e ence s anda d used; and da a analysis. The quali y o each o hese componen s will be g aded as high, mode a e, o low. Da a syn heses As a fi s s ep o syn hesising he e idence om included s udies, we will g aph he diagnos ic es accu acy o he indi idual s udies. The pai ed esul s o sensi i i y and specifici y will be plo ed as ecei e ope a ing cha ac e is ic (ROC) cu es, using hese o highligh he co a ia ion be ween he sensi i i y and specifici y in he indi idual s udies. These ini ial g aphical ep esen a ions will also enable isual assessmen o he a ia ion be ween s udies and acili a e subg oup analyses o explo ing he e ec o ce ain cha ac e is ics (e.g., sex, age, coun y/geog aphical egion) on es pe o mance. Expec edly, he cu -o s used o define es posi i i y may a y ac oss s udies; he e o e, o summa ise and compa e he accu acy o he es ac oss s udies o each endpoin , we will employ he hie a chical summa y ROC (HSROC) model. 16 Th ough he inclusion o andom e ec s, he HSROC model accoun s o he a iabili y ac oss s udies. The model uses s udy-specific es ima es o he ue posi i e a e (sensi i i y) and he alse posi i e a e (1—specifici y) o es ima e a summa y ROC cu e. Whe e s udies ha e used a common o simila cu -o and ha e analysed popula ions wi h simila cha ac e is ics, we will use pa ame e es ima es om he models o compu e summa y sensi i i ies and specifici ies wi h 95% confidence in e als. We will in es iga e be ween-s udy he e ogenei y wi h he HSROC model using he ollowing co a ia es o e alua e he associa ion be ween s udy ( ype o s udy design, sampling me hods employed, coun y o o igin, cu -o s o es posi i i y) and pa ien cha ac e is ics (age, sex, disease se e i y) and es pe o - mance. The esul s om he quali y assessmen will be used o unde ake sensi i i y analyses by e alua ing whe he he quali y o s udies (low, medium, high quali y) has any impac on he o e all conclusion. All analyses will be pe o med wi h S a a e sion 14.0 (S a aCo p LP, College S a ion, TX, USA). A na a i e syn hesis o he da a will also be unde aken and will be he app oach pa icula ly employed in summa ising he e idence pe aining o isk p edic ion and cos -e ec i eness. Publica ion bias We will e alua e publica ion bias by using unnel plo s and Begg and Egge es s. 17,18 E alua ing confidence in he o e all e idence We will e alua e he s eng h and quali y o he o e all e idence by using he GRADE (G ading o Recommenda ions Assessmen , De elopmen and E alua ion) app oach. 19 The GRADE sys em is use ul o e alua ing he quali y o e idence and s eng h o ecommenda ions. 19 A e iew p o ocol o ood alle gy JF Kim e al 2 npj P ima y Ca e Respi a o y Medicine (2017) 10 Published in pa ne ship wi h P ima y Ca e Respi a o y Socie y UK P o ocol egis a ion A de ailed p o ocol o he e iew will be egis e ed wi h he In e na ional P ospec i e Regis e o Sys ema ic Re iews (PROSPERO): h p://www.c d. yo k.ac.uk/p ospe o/ p io o commencing he e iew. DISCUSSION The majo alle gens in di e en ood alle gies ha e now been measu ed using he CRD echnique ac oss di e en se ings. 8,10,20–23 Fo example, A a h 2 o peanu alle gy, omega-5-gliadin o whea alle gy, Gly m 4 o soy milk alle gy, and a numbe o ee nu s, ui s, ege ables, and fish ha e now been es ablished. 8,10,20–23 Despi e he po en ial o CRD in imp o ing diagnosis o ood alle gy, a numbe o impo an ques ions ega ding i s clinical applica ion emain. These include a clea de e mina ion o whe he hese es s o e benefi o e -and-abo e exis ing p ocedu es and, i so, whe e in he clinical pa hway CRD fi s—whe he as a sc eening ool o confi ma o y me hod; an assessmen o he impac o CRD diagnosis on he isk p ofile o pa ien s; and a clea app ecia ion o he cos implica ions o i s use in clinical p ac ice in compa ison wi h cu en app oaches. Add essing hese ques ions will p o ide a s ong basis o deciding on he significance o CRD o he diagnosis o ood alle gy in he clinical se ing. This is he fi s sys ema ic e iew o he diagnos ic accu acy o CRD o ood alle gy: in addi ion o syn hesising he e idence on he diagnos ic measu es o he eigh mos common ood alle gies (cow’s milk, egg, whea , soy, peanu , ee nu s, fish, and shellfish), we will, whe e possible, unde ake isk assessmen o CRD on pa ien s, and summa ise he es ima es o he cos -e ec i eness o CRD o diagnosing hese ood alle gies. ACKNOWLEDGEMENTS We a e hank ul o Ma shall Dozie , Senio Liaison Lib a ian and Liaison Di ec o o he College o Medicine and Ve e ina y Medicine a he Uni e si y o Edinbu gh, o assis ance in de eloping he sea ch s a egy. This wo k is unded by he Chie Scien is O fice, Edinbu gh (HIPS/16/42). I is pa o J.F.K. disse a ion o he awa d o MPH deg ee a he Uni e si y o Edinbu gh. B.N. is suppo ed by a ellowship om he Uni e si y o Tampe e, Finland, wi h addi ional suppo om he Fa Ins i u e and As hma UK Cen e o Applied Resea ch. AUTHOR CONTRIBUTIONS A.S. concei ed he idea o his wo k. I was d a ed by J.F.K. and B.N. and was hen e ised a e se e al ounds o c i ical commen s om N.M., A.S ., and A.S. COMPETING INTERESTS A.S. is Join Edi o -in-Chie o npj P ima y Ca e Respi a o y Medicine, bu was no in ol ed in he edi o ial e iew o , no he decision o publish, his a icle. The emaining au ho s decla e no compe ing in e es s. REFERENCES 1. Allen, J. K. & Koplin, J. J. The epidemiology o IgE-media ed ood alle gy and anaphylaxis. Immunol. Alle gy Clin. N. Am. 32,35–50 (2012). 2. Mu a o, A. e al. EAACI ood alle gy and anaphylaxis guidelines: diagnosis and managemen o ood alle gy. Alle gy 69, 1008–1025 (2014). 3. Siche e , S. H. & Sampson, H. A. Food alle gy: epidemiology, pa hogenesis, diagnosis, and ea men . J. Alle gy Clin. Immunol. 133, 291–307 (2014). 4. Nwa u, B. I. e al. The epidemiology o ood alle gy in Eu ope: a sys ema ic e iew and me a-analysis. Alle gy 69,62–75 (2014). 5. Rona, R. J. e al. The p e alence o ood alle gy: a me a-analysis. J. Alle gy Clin. Immunol. 120, 638–646 (2007). 6. Mu a o A., Robe s G. (ed). Food Alle gy and Anaphylaxis Guidelines. (The Eu opean Academy o Alle gy and Clinical Immunology (EAACI), 2014). 7. Soa es-Weise , K. e al. The diagnosis o ood alle gy: a sys ema ic e iew and me a-analysis. Alle gy 69,76–86 (2014). 8. Ho mann-Somme g ube , K., P ei e , S. & Bublin, M. Applica ions o molecula diagnos ic es ing in ood alle gy. Cu . Alle gy As hma Rep. 15, 56 (2015). 9. an de Velde, J. L. e al. Food alle gy- ela ed quali y o li e a e double-blind, placebo-con olled ood challenges in adul s, adolescen s, and child en. J. Alle gy Clin. Immunol. 130, 1136–1143 (2012). 10. Elle , E. & Bindsle -Jensen, C. Clinical alue o componen - esol ed diagnos ics in peanu -alle gic pa ien s. Alle gy 68, 190–194 (2013). 11. Shamsee , L. e al. P e e ed epo ing i ems o sys ema ic e iew and me a- analysis p o ocols (PRISMA-P) 2015: elabo a ion and explana ion. B . Med. J. 349, g7647 (2015). 12. Valen a, R., Lidholm, J., Niede be ge , V., Hayek, B., K a , D. & G önlund, H. The ecombinan alle gen-based concep o componen - esol ed diagnos ics and immuno he apy. Clin. Exp. Alle gy 29, 896–904 (1999). 13. Mohe , D., Libe a i, A., Te zla , J. & Al man, D. G. P e e ed epo ing i ems o sys ema ic e iews and me a-analyses: he PRISMA s a emen . PLoS Med. 6, e1000097 (2009). The PRISMA G oup. 14. Whi ing, P. F. e al. QUADAS-2: a e ised ool o he quali y assessmen o diagnos ic accu acy s udies. Ann. In e n. Med. 155, 529–536 (2011). 15. D ummond, M. F., Sculphe , M. J., To ance, G. W., O’B ien, B. J. & S odda , G. L. Me hods o he Economic E alua ion o Heal h Ca e P og ammes. 3 d edn. (Ox o d Uni e si y P ess, 2005). 16. Wang, F. & Ga sonis, C. A. Hie a chical models o ROC cu e summa y measu es: design and analysis o mul i- eade , mul i-modali y s udies o medical es s. S a . Med. 27, 243–256 (2008). 17. Begg, C. B. & Mazumda , M. Ope a ing cha ac e is ics o a ank co ela ion es o publica ion bias. Biome ics 50, 1088–1101 (1994). 18. Egge , M., Smi h, G. D., Schneide , M. & Minde , C. Bias in me a-analysis de ec ed by a simple, g aphical es . B . Med. J. 315, 629–634 (1997). 19. Schunemann, H. J. e al. G ading quali y o e idence and s eng h o ecommenda ion o diagnos ic es s and s a egies. B . Med. J. 336, 1106 (2008). 20. Ballme -Webe , B. K. & Ho mann-Somme g ube , K. Molecula diagnosis o ui and ege able alle gy. Cu . Opin. Alle gy Clin. Immunol. 11, 229–235 (2011). 21. Caube , J. C., Kondo, Y., U isu, A. & Nowak-Weg zyn, A. Molecula diagnosis o egg alle gy. Cu . Opin. Alle gy Clin. Immunol. 11, 210–215 (2011). 22. Fiocchi, A., Bouygue, G. R., Alba ini, M. & Res ani, P. Molecula diagnosis o cow’s milk alle gy. Cu . Opin. Alle gy Clin. Immunol. 11, 216–221 (2011). 23. Nicolaou, N. & Cus o ic, A. Molecula diagnosis o peanu and legume alle gy. Cu . Opin. Alle gy Clin. Immunol. 11, 222–228 (2011). This wo k is licensed unde a C ea i e Commons A ibu ion 4.0 In e na ional License. The images o o he hi d pa y ma e ial in his a icle a e included in he a icle’s C ea i e Commons license, unless indica ed o he wise in he c edi line; i he ma e ial is no included unde he C ea i e Commons license, use s will need o ob ain pe mission om he license holde o ep oduce he ma e ial. To iew a copy o his license, isi h p://c ea i ecommons.o g/licenses/by/ 4.0/ © The Au ho (s) 2017 Supplemen a y In o ma ion accompanies he pape on he npj P ima y Ca e Respi a o y Medicine websi e (doi:10.1038/s41533-017-0015-0). A e iew p o ocol o ood alle gy JF Kim e al 3 Published in pa ne ship wi h P ima y Ca e Respi a o y Socie y UK npj P ima y Ca e Respi a o y Medicine (2017) 10