Mortality prediction in chronic obstructive pulmonary disease comparing the GOLD 2015 and GOLD 2019 staging: a pooled analysis of individual patient data

Mo ali y p edic ion in ch onic obs uc i e pulmona y disease compa ing he GOLD 2015 and GOLD 2019 s aging: a pooled analysis o indi idual pa ien da a Elena Ga cía Cas illo 1,41 , Tama a Alonso Pé ez 1,41 , Julio Ancochea 1,17 , Ma ia Te esa Pas o Sanz 1 , Pe e Almag o 2 , Pablo Ma ínez-Camblo 3 , Ma c Mi a i lles 4 , Mónica Rod íguez-Ca ballei a 2 , Annie Na a o 5 , Be nd Lamp ech 6 , Ana S. Ramí ez-Ga cía Luna 7 , Be nha d Kaise 8 , Inmaculada Al ageme 9 , Ci o Casano a 11 , C is óbal Es eban 11 , Juan J. Sole -Ca aluña 12 ,JuanP.de-To es 13,14 , Ba olomé R. Celli 15 , Jose M. Ma ín 16,17 , Ge ben e Rie 18,19 ,Pa iciaSob adillo 20 , Pe e Lange 21 , Judi h Ga cia-Ayme ich 22,23,24 , Josep M. An o 22,23,24 ,AliceM.Tu ne 25 , MeiLan K. Han 26 , A nul Langhamme 27,26,28 ,Sig idAnnaAalbe gVikjo d 27 , Alice S e nbe g 29 ,LindaLei se h 30 ,Pe Bakke 31 , Ane Johannessen 32 ,To uOga 33 , Bo ja G. Cosío 34 , And és Echaza e a 35 , Nicolás Roche 36 ,Pie e-RégisBu gel 36 , Don D. Sin 37,38 , Milo A. Puhan 39 , Jose Luis López-Campos 17,40 , Lau a Ca asco 40 , Joan B. So iano 1,17 , o he 3CIA collabo a ion ABSTRACT In 2019, The Global Ini ia i e o Ch onic Obs uc i e Lung Disease (GOLD) modi ied he g ading sys em o pa ien s wi h COPD, c ea ing 16 subg oups (1A–4D). As pa o he COPD Coho s Collabo a i e In e na ional Assessmen (3CIA) ini ia i e, we aim o compa e he mo ali y p edic ion o he 2015 and 2019 COPD GOLD s aging sys ems. We s udied 17139 COPD pa ien s om he 3CIA s udy, selec ing hose wi h comple e da a. Pa ien s we e classi ied by he 2015 and 2019 GOLD ABCD sys ems, and we compa ed he p edic i e abili y o 5- yea mo ali y o bo h classi ica ions. In o al, 17139 pa ien s wi h COPD we e en olled in 22 coho s om 11 coun ies be ween 2003 and 2017; 8823 o hem had comple e da a and we e analysed. Mean±SD age was 63.9±9.8 yea s and 62.9% we e male. GOLD 2019 classi ied he pa ien s in milde deg ees o COPD. Fo bo h classi ica ions, g oup D had highe mo ali y. 5-yea mo ali y did no di e be ween g oups B and C in GOLD 2015; in GOLD 2019, mo ali y was g ea e o g oup B han C. Pa ien s classi ied as g oup A and B had be e sensi i i y and posi i e p edic i e alue wi h he GOLD 2019 classi ica ion han GOLD 2015. GOLD 2015 had be e sensi i i y o g oup C and D han GOLD 2019. The a ea unde he cu e alues o 5-yea mo ali y we e only 0.67 (95% CI 0.66–0.68) o GOLD 2015 and 0.65 (95% CI 0.63–0.66) o GOLD 2019. The new GOLD 2019 classi ica ion does no p edic mo ali y be e han he p e ious GOLD 2015 sys em. @ERSpublica ions GOLD 2019 s aging sys em c ea ed 16 subg oups. GOLD 2015 and GOLD 2019 a e no s ong p edic o s o mo ali y, and do no ha e su icien disc imina o y powe o be used as a ool o isk classi ica ion o mo ali y in pa ien s wi h COPD. h ps://bi .ly/3idBuaN Ci e his a icle as: Ga cía Cas illo E, Alonso Pé ez T, Ancochea J, e al. Mo ali y p edic ion in ch onic obs uc i e pulmona y disease compa ing he GOLD 2015 and GOLD 2019 s aging: a pooled analysis o indi idual pa ien da a. ERJ Open Res 2020; 6: 00253-2020 [h ps://doi.o g/ 10.1183/23120541.00253-2020]. Copy igh ©ERS 2020. This a icle is open access and dis ibu ed unde he e ms o he C ea i e Commons A ibu ion Non-Comme cial Licence 4.0. Recei ed: 7 May 2020 | Accep ed a e e ision: 31 July 2020 h ps://doi.o g/10.1183/23120541.00253-2020 ERJ Open Res 2020; 6: 00253-2020 ORIGINAL ARTICLE COPD In oduc ion COPD is a common cause o mo bidi y and mo ali y in he wo ld. COPD a ec s ∼328 million people wo ldwide, and COPD- ela ed dea hs amoun o 4 million e e y yea [1]. Assessmen o disease se e i y is essen ial o p edic p ognosis and o s anda dise ea men egimes. The Global Ini ia i e o Ch onic Obs uc i e Lung Disease (GOLD) documen is he mos widely used ea men guide o he s aging and managemen o COPD. The GOLD g ading sys em o COPD has signi ican ly e ol ed since i s publica ion in 2001 o he cu en e sion in 2019. Ini ially, in he GOLD 2007 classi ica ion, only pos -b onchodila o ai low limi a ion based on spi ome y o ced expi a o y olume in 1 s (FEV 1 )was used o g ade he se e i y o COPD [2]. La e on, some c i icism a ose on his g ading sco e because i elied only upon FEV 1 , which is no a good p edic o o dyspnoea, quali y o li e o exe cise ole ance. Fu he , o he impo an a iables o e alua e p ognosis in COPD, such as sub-pheno ypes, exace ba ions, dyspnoea se e i y o como bidi ies ha e been p oposed, among o he s. The e o e, GOLD 2011 p oposed a classi ica ion sys em o ou g oups, ABCD, combining FEV 1 and wo clinical pa ame e s: his o y o exace ba ions and espi a o y symp oms measu ed by he modi ied Medical Resea ch Council (mMRC) dyspnoea sco e, o he COPD Assessmen Tes sco e (CAT) [3]. The 2011 ABCD classi ica ion was conside ed an imp o emen in he managemen o pa ien s wi h COPD, p o iding an oppo uni y o u he guide he indi idualised ca e o hese pa ien s. GOLD 2011 p edic ed u u e exace ba ions be e han GOLD 2007, bu he e was no di e ence in mo ali y p edic ions o espi a o y ou comes [4–7]. In 2015 an upda ed epo was published wi h he same measu emen pa ame e s (FEV 1 , dyspnoea and exace ba ions) han he 2011 classi ica ion [8]. The la es GOLD upda e, published in 2019, uses a composi e o spi ome y, symp oms and exace ba ions, bu impo an ly sepa a ing he spi ome ic 1–4 s aging om he ABCD g oups [9]. This sepa a ion is ele an because i is known he e a e di e ences in he a e o exace ba ions o he mos se e e COPD pa ien s, depending on whe he he isk is based on pulmona y unc ion es s, on he his o y o exace ba ions o bo h [10]. All hese classi ica ions we e ini ially designed no o assess p ognosis, bu o aid A ilia ions: 1 Pneumology Dep , Hospi al Uni e si a io de la P incesa, Ins i u o de In es igación Hospi al Uni e si a io de la P incesa (IISP), Uni e sidad Au ónoma de Mad id, Mad id, Spain. 2 In e nal Medicine Depa men , Mú ua Te assa Uni e si y Hospi al, Ba celona, Spain. 3 Geisel School o Medicine a Da mou h, Hano e , NH, USA. 4 Pneumology Dep , Hospi al Uni e si a y Vall d’Heb on, CIBER de En e medades Respi a o ias (CIBERES), Ba celona, Spain. 5 Pneumology Se ice, Hospi al Uni e si a i Mú ua Te assa, Ba celona, Spain. 6 Dep o Pulmona y Medicine, Keple -Uni e si y-Hospi al, Facul y o Medicine, Johannes- Keple -Uni e si y Linz, Linz, Aus ia. 7 Facul ad de Medicina, Uni e sidad Au ónoma de San Luis Po osí, San Luis Po osí, Mexico. 8 Dep o Pulmona y Medicine, Pa acelsus Medical Uni e si y Hospi al, Salzbu g, Aus ia. 9 Depa amen o de Medicina, Uni e sidad de Se illa, HU Vi gen de Valme, Se ille, Spain. 10 Pulmonology Depa men , Hospi al Galdakao-Usansolo, Galdakao, Spain. 11 Pulmona y Depa men , Resea ch Uni , Hospi al Uni e si a io Nues a Seño a de La Candela ia, Uni e sidad de La Laguna, Tene i e, Spain. 12 Se icio de Neumología, Hospi al A nau de Vilano a, Valencia, Spain. 13 Clinica Uni e sidad de Na a a, Pamplona, Spain. 14 Respi ology and Sleep Medicine Di ision, Queen’s Uni e si y, Kings on, Canada. 15 Pulmona y and C i ical Ca e Medicine, Ha a d Uni e si y, B igham and Women’s Hospi al, Bos on, MA, USA. 16 Hospi al Uni e si a io Miguel Se e , Za agoza, Spain. 17 Cen o de In es igación Biomédica en Red de En e medades Respi a o ias (CIBERES), Ins i u o de Salud Ca los III, Mad id, Spain. 18 U ban Vi ali y –Cen e o Expe ise, Facul y o Heal h, Ams e dam Uni e si y o Applied Sciences, Ams e dam, The Ne he lands. 19 Dep o Ca diology, Ams e dam UMC, loca ion Academic Medical Cen e , Ams e dam, The Ne he lands. 20 Uni e si y Hospi al o C uces in Ba akaldo, Ba akaldo, Spain. 21 Sec ion o Social Medicine, Dep o Public Heal h, Copenhagen Uni e si y, Copenhagen Ci y Hea S udy, F ede iksbe g Hospi al, Copenhagen, Denma k. 22 ISGlobal, Ba celona, Spain. 23 Uni e si a Pompeu Fab a (UPF), Ba celona, Spain. 24 CIBER Epidemiología y Salud Pública (CIBERESP), Ba celona, Spain. 25 Ins i u e o Applied Heal h Resea ch, Uni e si y o Bi mingham, Edgbas on, UK. 26 In e nal Medicine, Di ision o Pulmona y and C i ical Ca e Medicine, Uni e si y o Michigan, Ann A bo , MI, USA. 27 Dep o Public Heal h and Nu sing, NTNU (No wegian Uni e si y o Science and Technology), T ondheim, No way. 28 Le ange Hospi al, No d-T øndelag Hospi al T us , Le ange , No way. 29 Dep o Epidemiology, Johns Hopkins Bloombe g School o Public Heal h, Bal imo e, MD, USA. 30 Cen e o Clinical Documen a ion and E alua ion, No he n No way Regional Heal h Au ho i y, T omso, No way. 31 Dep o Clinical Science, Uni e si y o Be gen, Be gen, No way. 32 Dep o Global Public Heal h and P ima y Ca e, Uni e si y o Be gen, Be gen, No way. 33 Dep o Respi a o y Ca e and Sleep Con ol Medicine, Kyo o Uni e si y, Kyo o, Japan. 34 Hospi al Uni e si a io Son Espases-IdISPa, Mallo ca, Spain. 35 Se icio de Neumonología, Hospi al San Juan de Dios de La Pla a, Buenos Ai es, A gen ina. 36 Respi a o y Medicine, Cochin Hospi al, APHP Cen e–Uni e si y o Pa is, Cochin Ins i u e (INSERM UMR1016), Pa is, F ance. 37 UBC Cen e o Hea Lung Inno a ion, Vancou e , BC, Canada. 38 Di ision o Respi a o y Medicine, Dep o Medicine, Uni e si y o B i ish Columbia, Vancou e , BC, Canada. 39 Epidemiology, Bios a is ics and P e en ion Ins i u e, Uni e si y o Zu ich, Zu ich, Swi ze land. 40 Unidad Médico Qui ú gica de En e medades Respi a o ias, Ins i u o de Biomedicina de Se illa (IBiS), Hospi al Uni e si a io Vi gen del Rocío/Uni e sidad de Se illa, Se ille, Spain. 41 These au ho s con ibu ed equally. Co espondence: Joan B. So iano, Hospi al Uni e si a io de la P incesa, Diego de León 62, Neumología 6ª plan a, 28006-Mad id, Spain. E-mail: jbso [email protected] h ps://doi.o g/10.1183/23120541.00253-2020 2 COPD | E. GARCÍA CASTILLO ET AL. clinicians in c ea ing op imal ea men egimens o pa ien s. Thus, he p ognos ic abili y o GOLD 2019 compa ed o p e ious classi ica ions is la gely unknown, wi h only a ew published s udies [11, 12]. To add ess his issue, we used pooled da a om 17139 pa ien s o 22 COPD coho s and 11 coun ies and compa ed he p ognos ic capaci y o he 2019 e sus 2015 GOLD s aging classi ica ions o p edic mo ali y. Me hods S udy popula ion In his in e na ional s udy, we assessed 17139 pa ien s om he COPD Coho s Collabo a i e In e na ional Assessmen (3CIA) ini ia i e. All we e p ospec i e coho s ha ec ui ed pa ien s wi hin he pe iod 1999 o 2017, excep o one which was a popula ion-based coho . All pa ien s had a de ini ion o COPD cha ac e ised by spi ome y, ha is pos -b onchodila o FEV 1 o o ced i al capaci y (FVC) a io <0.7 and a clinical diagnosis o COPD. Spi ome y was pe o med using he s anda ds p o ided by he Ame ican Tho acic Socie y and Eu opean Respi a o y Socie y [13]. The p ima y in es iga o s o each o he pa icipa ing 3CIA coho s p o ided indi idual pa ien ’s da a o pooled analysis. We ob ained a minimum indi idual da ase including he i al s a us (up o dea h, igh unca ion o 2017), age, sex, smoking s a us, p e-b onchodila o and pos -b onchodila o FEV 1 and FVC, and dyspnoea measu ed wi h he modi ied mMRC, among o he s [14]. Only in a numbe o 3CIA coho s we e da a o numbe o exace ba ions in he p e ious yea a ailable. Fo he cu en s udy, we selec ed exclusi ely hose coho s in which he numbe o exace ba ions in he p e ious yea we e a ailable in he da abase, since his a iable is equi ed o calcula e he GOLD 2015 and 2019 g ading sys ems. Fi een ou o a o al o 22 coho s con ained da a on his o y o exace ba ions, so ha he inal numbe o pa ien s a ailable o be classi ied acco ding o GOLD 2015 and GOLD 2019 was 8823. Symp oms we e assessed using he mMRC dyspnoea scale. To de e mine he isk desc ip o in he 2015 g ouping sys em, we used exace ba ions his o y and GOLD spi ome y s ages. The combina ion o symp oms (mMRC) and he wo se isk desc ip o (spi ome y o exace ba ion his o y) we e used o classi y pa ien s by he GOLD 2015 sys em. Pa icipan s we e classi ied using he GOLD 2019 sys em in o ou g ades (1–4) based on pos -b onchodila o FEV 1 pe cen age o p edic ion as s age 1 (FEV 1 ⩾80), s age 2 (FEV 1 79–50), s age 3 (FEV 1 30–49) and s age 4 (FEV 1 <30). G oups ABCD we e de ined by sel - epo ed se e i y o dyspnoea (mMRC) and numbe o exace ba ions in he p e ious yea . All pa icipan s p o ided in o med w i en consen , and each s udy was conduc ed wi h he o mal app o al o he local e hics ins i u ional commi ees ollowing he p inciples o he Decla a ion o Helsinki. Ou comes The p ima y ou come was he p edic ion abili y o all-cause mo ali y in he indi iduals by he wo GOLD sys ems. S a is ical analysis The 3CIA da abase manage quali y-con olled all da a cen ally and c ea ed a clean da abase wi h a da a dic iona y. All implausible o missing a iables we e que ied wi h he o iginal s udy in es iga o s, and da a we e emo ed om he cen al da abase i e o s could no be co ec ed. Because he coho s had di e en ollow-up imes, pa ien s we e igh -censo ed a 5 yea s o ollow-up. Desc ip i e s a is ics used mean and s anda d de ia ion o con inuous a iables and he numbe o cases and pe cen ages o ca ego ical a iables. Compa isons be ween g oups we e pe o med wi h he Chi-squa ed es o ca ego ical a iables and he - es o con inuous a iables. We es ima ed 5-yea all-cause mo ali y, acco ding o GOLD 2015 and 2019 s aging sys ems, using Kaplan–Meie su i al s a is ics. S a is ical compa isons we e pe o med using he log- ank es ecei e ope a ion cha ac e is ic (ROC) cu e analyses and a ea unde he cu e (AUC) and he 95% con idence in e als o he AUC we e calcula ed o measu e he p edic i e accu acy o mo ali y. Also, we compa ed he p edic ion abili y o mo ali y on bo h classi ica ions using sensi i i y, posi i e p edic i e alue and he Youden’s index wi h Epida 3.1 p og amme. Resul s We pooled da a om 22 COPD coho s wi h a o al o 17139 pa ien s, inally including 8823 pa ien s om 15 coho s ha had all comple e a iables o be classi ied as GOLD 2015 and GOLD 2019. A compa ison o baseline cha ac e is ics o included and no included pa ien s is p esen ed in able 1. The e we e s a is ically signi ican di e ences in many a iables gi en he la ge size, bu mos should be conside ed no clinically ele an ( able 1). The 8823 included pa ien s we e 62.9% male, wi h a mean±SD age o 63.9 ±9.8 yea s, body mass index 27.0±5.8 kg·m −2 and mMRC dyspnoea sco e o 1.8±1.4. Pos -b onchodila o FEV 1 was 54.8%±22.3 o he h ps://doi.o g/10.1183/23120541.00253-2020 3 COPD | E. GARCÍA CASTILLO ET AL. p edic i e alue, and 6-min walk dis ance was 376.9±129.1 m. Based on spi ome y s aging, 1153 (13%) had mild (s age 1), 3711 (42.1%) had mode a e (s age 2), 2654 (30.1%) had se e e (s age 3) and 1301 (14.8%) had e y se e e (s age 4) disease. TABLE 1 Compa ison o demog aphic and clinical cha ac e is ics in COPD Coho s Collabo a i e In e na ional Assessmen (3CIA) COPD pa ien s included/excluded in his analysis Excluded Included p- alue Subjec s n 8316 8823 Age yea s 64.2±10.7 63.9±9.8 0.08 Male sex 6232 (74.9%) 5552 (62.9%) <0.001 BMI kg·m -2 26.5±4.9 27.0±5.8 <0.001 Modi ied MRC dyspnoea scale 1.5±1.3 1.8±1.4 <0.001 0 1647 (23.4%) 1957 (22.2%) <0.001 1 2261(32.1%) 1886 (21.4%) 2 1641 (23.3%) 1772 (20.1%) 3 688 (9.8%) 1951 (22.1%) 4 805 (11.4%) 1257 (14.3%) Six-min walk dis ance m 415.4±108.8 376.9±129.1 <0.001 FEV 1 pos BD mL 1.7±0.7 1.6±0.8 <0.001 FEV 1 pos BD % 60.8±22.1 54.8±22.3 <0.001 Smoke <0.001 Fo me 3989 (49.1%) 5392 (61.4%) Cu en 3589 (44.2%) 3174 (36.1%) Ne e 542 (6.7%) 222 (2.5%) Pack-yea s 46.4±28.8 42.1±28.3 <0.001 Cough 1103 (54.2%) 342 (43.9%) <0.001 Spu um 1159 (42.1%) 341 (43.9%) 0.353 Diabe es 354 (6.7%) 303(16.6%) <0.001 Ca diac disease 1072 (30.8%) 467 (25.9%) <0.001 Ch onic b onchi is 166 (38.7%) 787 (69.5%) <0.001 Hype ension 454 (40.4%) 826 (44.8%) 0.028 As hma 1243 (26.1%) 209 (10.7%) <0.001 Spi ome y s aging <0.001 1 1567 (19%) 1153 (13.1%) 2 3892 (47.1%) 3711 (42.1%) 3 2126 (25.8%) 2654 (30.1%) 4 671 (8.1%) 1301 (14.8%) Long- e m oxygen he apy 119 (1.4%) 430 (4.8%) 0.259 Da a a e p esen ed as n (%), mean±SD o median (in e qua ile ange), unless o he wise s a ed. BMI: Body mass index; MRC: Medical Resea ch Council; FEV 1 : o ced expi a o y olume in 1 s; BD: b onchodila o . 45 40 35 30 25 20 15 10 5 0ABCD 5.8% 13.6% 17.8% 40% 38.6% 37.7% 29.9% 16.4% 2015 2019 FIGURE 1 Dis ibu ion o pa icipan s by classi ica ion in Global Ini ia i e o Ch onic Obs uc i e Lung Disease (GOLD) 2015 and GOLD 2019. h ps://doi.o g/10.1183/23120541.00253-2020 4 COPD | E. GARCÍA CASTILLO ET AL. The dis ibu ion o hese 8823 pa ien s acco ding o GOLD 2015 and GOLD 2019 is p esen ed in igu e 1. Wi h GOLD 2019 he e is a shi owa ds he less se e e s aging o disease (absolu e inc ease in s age A and B o 7.8% and 22.2% espec i ely; and absolu e dec ease in s age C and D o 7.8% and 22.2% espec i ely). The o e all 5-yea mo ali y a e was 18.3%. The all-cause 5-yea mo ali y a es acco ding o bo h classi ica ions a e shown in able 2. Figu e 2 shows Kaplan–Meie cu es o 5-yea mo ali y acco ding o GOLD 2015 ( igu e 2a) and GOLD 2019 ( igu e 2b). All-cause mo ali y a 5 yea s in he 2015 GOLD classi ica ion was highe in g ade D, ollowed by g ades B, C (wi h simila mo ali y), and inally g ade A, log- ank es p<0.001 ( able 2 and igu e 2a). G ade D di e ged om he beginning o ollow-up, while g ades B and C di e ged a e 1 yea o ollow-up. In GOLD 2019, he ou Kaplan–Meie cu es di e ge du ing he i s yea , bu in e es ingly, g ade B had highe mo ali y han g ade C, so mo ali y was highe in g oups B and D (mo e symp oms) han in g oups A and C ( ewe symp oms; igu e 2b). The deg ee o obs uc ion measu ed by FEV 1 % u he subclassi ied pa ien s in o 16 subg oups wi h di e en mo ali y a es, inc easing om 1A o 4D in he GOLD 2019 g ading sys em ( able 3). Figu e 3 shows Kaplan–Meie cu es o each o he spi ome y s a a. The highe mo ali y o g oup B o e g oup C pe sis ed in each o he s a a wi h he excep ion o spi ome y s a a 1, wi h a highe mo ali y in g oup C. Simila ly, o GOLD 2015, in GOLD 2019, g ades A and D had he lowes and highes mo ali y, espec i ely, wi h e y simila absolu e numbe s ( able 2, igu e 2). The p ima y ou come, he p edic ion capaci y as measu ed by he AUC o ROC cu e o mo ali y up o 5 yea s, was in e media e (<0.70) o bo h sys ems ( igu e 4). GOLD 2015 exhibi ed sligh ly be e disc imina ion in p edic ing mo ali y (AUC 0.67, 95% CI 0.66–0.68) han he GOLD 2019 classi ica ion (AUC 0.64, 95% CI 0.63–0.66). TABLE 2 Mo ali y isk among COPD pa ien s acco ding o Global Ini ia i e o Ch onic Obs uc i e Lung Disease (GOLD) 2019 and GOLD 2015 classi ica ions GOLD 2015 5-yea mo ali y % GOLD 2019 5-yea mo ali y % G oup A 5.8 7.5 G oup B 13.8 23.2 G oup C 14.1 14.8 G oup D 30.8 32.8 A GOLD 2015 GOLD 2019 B C D A B C D 1.0a) b) 0.8 0.6 0.4 0.2 0.0 Cumula i e su i al 1.0 0.8 0.6 0.4 0.2 0.0 Cumula i e su i al Pa ien s a isk A B C D 2388 1259 1129 3229 2285 1184 1070 2962 2129 1045 968 2609 1914 876 875 2128 1420 608 693 1525 937 365 563 977 0 1224364860 Follow-up mon hs Pa ien s a isk A B C D 3039 3053 478 1435 2909 2854 446 1292 2687 2539 410 1115 2419 2100 370 904 1806 1453 307 680 1238 873 262 469 0 1224364860 Follow-up mon hs Log ank es p<0.01 Log ank es p<0.01 FIGURE 2 Kaplan–Meie su i al cu es by a) Global Ini ia i e o Ch onic Obs uc i e Lung Disease (GOLD) 2015 and b) GOLD 2019. h ps://doi.o g/10.1183/23120541.00253-2020 5 COPD | E. GARCÍA CASTILLO ET AL. TABLE 3 Fi e-yea all-cause mo ali y (%) among spi ome y s a a wi hin Global Ini ia i e o Ch onic Obs uc i e Lung Disease (GOLD) 2019 ABCD classi ica ion GOLD 2019 A B C D Spi ome y I 3.4 9.8 16.2 5.5 Spi ome y II 7 14.9 8.3 22.1 Spi ome y III 13 23.7 17.7 32.6 Spi ome y IV 16.7 39.1 29.5 46 1.0a) b) c) d) 0.8 0.6 0.4 0.2 0.0 Pa ien s a isk A B C D A B C D A B C D A B C D A B C D 756 188 47 40 721 176 40 38 674 158 38 37 604 131 35 34 424 77 25 27 219 35 20 15 0 1224364860 Follow-up mon hs Pa ien s a isk A B C D 1659 1094 236 398 1591 1029 223 360 1481 907 205 317 1335 760 184 247 1020 542 152 180 738 338 132 129 0 1224364860 Follow-up mon hs Pa ien s a isk A B C D 539 1166 139 594 518 1090 130 535 459 985 121 469 413 814 111 400 309 567 98 308 233 351 80 217 01224364860 Follow-up mon hs Pa ien s a isk A B C D 85 605 55 399 79 559 52 358 73 489 45 291 67 395 40 223 53 267 32 165 48 149 30 108 01224364860 Follow-up mon hs Log ank es p<0.01 Log ank es p<0.01 Log ank es p<0.01 Log ank es p<0.01 Cumula i e su i al 1.0 0.8 0.6 0.4 0.2 0.0 Cumula i e su i al 1.0 0.8 0.6 0.4 0.2 0.0 Cumula i e su i al 1.0 0.8 0.6 0.4 0.2 0.0 Cumula i e su i al FIGURE 3 Kaplan–Meie su i al cu es by Global Ini ia i e o Ch onic Obs uc i e Lung Disease (GOLD) 2019 and spi ome y subg oups. a) Spi ome y subg oup I; b) spi ome y subg oup II; c) spi ome y subg oup III; d) spi ome y subg oup IV. h ps://doi.o g/10.1183/23120541.00253-2020 6 COPD | E. GARCÍA CASTILLO ET AL. Rega ding sensi i i y pa ame e s, bo h classi ica ions had qui e shallow alues. GOLD 2019 had a highe sensi i i y o p edic ing mo ali y on A and B g oups e sus he 2015 classi ica ion (19.1 e sus 11.1 and 43.6 e sus 11.6). On he o he hand, GOLD 2015 had a highe sensi i i y on g oups C and D e sus he 2019 classi ica ion (14.6 e sus 6.6 and 62.5 e sus 30.6) based on o e lapping 95% con idence in e als ( able 4). The posi i e p edic i e alues we e also highe in GOLD 2019 g oup A and B e sus he same g oups in GOLD 2015 (9.5 e sus 6.9 and 21.2 e sus 13.3); bu no di e en (o e lapping con idence in e als) in g oups C and D. The Youden indices we e qui e low o bo h classi ica ions, e en wi h nega i e alues, showing ha i is no an op imal classi ica ion sys em o assess mo ali y. Discussion Ou s udy e alua es mo ali y acco ding o he las wo GOLD classi ica ions, and i is one o he mos ex ensi e o da e. The mos impo an inding is ha he new GOLD 2019 classi ica ion (based on symp oms o exace ba ions along wi h s ages spi ome y analysed al oge he ) did no p edic 5-yea mo ali y be e han he GOLD 2015 classi ica ion (based on spi ome y, his o y o exace ba ion and symp oms). How o de ine and s age COPD exace ba ions is a ma e o in ense, long deba e [15]. Mos 3CIA indi idual coho s used subsequen i e a ions o GOLD-accep ed de ini ions in hei p o ocols, mos ly based on RODRIGUEZ-ROISIN e al.’s [16] seminal pape , including mild (symp om-only) exace ba ions. Howe e , when pooling o 3CIA, we only ocused on hose COPD exace ba ions ha equi ed heal h se ices use, eme gency oom admission o dea h. In he GOLD 2019 classi ica ion, subg oups B and D had he wo s mo ali y, highligh ing ha he highe bu den o symp oms con eys a highe mo ali y independen ly o spi ome y ( able 2 and igu e 2b). Finally, we show an impo an shi in he p opo ions o pa ien s be ween he 2015 and 2019 ABCD g ades, wi h milde se e i y in GOLD 2019. FIGURE 4 Recei e ope a ing cu es o all-cause mo ali y a 5 yea s ollow-up. 1.00 GOLD 2015 AUC 0.67 (95% CI 0.66–0.68) GOLD 2019 AUC 0.65 (95% CI 0.63–0.66) Re e ence line 0.75 0.50 0.25 0.00 0.00 0.25 0.50 0.75 1.00 1-speci ici y Sensi i i y TABLE 4 Accu acy o p edic ing mo ali y o he ABCD g oups classi ica ions by Global Ini ia i e o Ch onic Obs uc i e Lung Disease (GOLD) 2015 and GOLD 2019 schemes Classi ica ion Sensi i i y (95% CI) PPV (95% CI) Youden index (95% CI) GOLD 2015 g oup A 11.1 (9.6–12.7) 6.9 (5.9–7.9) −0.23 (−0.25–−0.22) GOLD 2015 g oup B 11.6 (10–13.2) 13.3 (11.5–15.1) −0.06 (−0.08–−0.04) GOLD 2015 g oup C 14.6 (12.8–16.4) 20.2 (17.9–22.6) +0.01 (−0.01–+0.03) GOLD 2015 g oup D 62.5 (60.1–64.9) 29.5 (28–31.1) +0.28 (+0.25–+0.31) GOLD 2019 g oup A 19.1 (17.2–21.1) 9.5 (8.4–10.5) −0.23 (−0.25–−0.21) GOLD 2019 g oup B 43.6 (41.1–46.1) 21.2 (19.8–22.6) +0.06 (+0.03–+0.09) GOLD 2019 g oup C 6.6 (5.3–7.8) 21.7 (18–25.5) +0.01 (0.0–+0.02) GOLD 2019 g oup D 30.6 (28.3–32.8) 32.8 (30.4–35.2) +0.16 (+0.14–+0.18) PPV: posi i e p edic i e alue. h ps://doi.o g/10.1183/23120541.00253-2020 7 COPD | E. GARCÍA CASTILLO ET AL. GOLD 2015 and GOLD 2019 classi ica ions had a low disc imina o y powe as pe he AUCs, anging om 0.67 o 0.65, and simila o o he s udies (by consensus, AUCs below 0.70 a e conside ed low o weak) [11]. Sensi i i y and posi i e p edic i e alues indica e ha he gene al pe o mance o he wo models is simila and e y low. Also, he Youden’s indices a e e y low wi h nega i e alues ha ha e no meaning ul in e p e a ion in p ac ice. These indings suppo he esul s o o he s udies, sugges ing ha GOLD classi ica ion is no a good p edic o o mo ali y [11, 12]. The e may be a ious easons o hese poo esul s. The main eason is ha hese classi ica ions we e concei ed o guide ea men , so i is no su p ising ha hei capaci y o p edic ing mo ali y is low. The e a e clinical pheno ypes such as he as hma-COPD, he equen exace ba o wi h emphysema o ch onic b onchi is como bidi ies and di e en indexes, ha a e signi ican p edic o s o mo ali y and a e no en i ely included in he GOLD s ages [17–21]. In ou s udy, sensi i i y and posi i e p edic i e alue a e highe in GOLD 2019 g oups A and B. On he con a y, sensi i i y is highe in GOLD 2015 in g oups C and D. These esul s sugges ha GOLD 2019 p edic s sligh ly be e mo ali y in low- isk g oups (A and B) and GOLD 2015 in high- isk g oups (C and D), equi ing mo e s udies o co obo a e i . The disc imina o y powe o GOLD 2019 was lowe han GOLD 2015 as shown in AUC alues. The pa i ion o FEV 1 as a di ec classi ie in GOLD 2019 educed abili y o disc imina e su i al, highligh ing he need o conside he se e i y o ai low obs uc ion in assessing mo ali y isk. When using GOLD 2019 wi h a composi e o spi ome y, exace ba ions and symp oms (16 subg oups 1A o 4D classi ica ion), we ound an inc ease in all-cause mo ali y be ween GOLD 2019 s age 1 and GOLD 2019 s age 4 ac oss g ades A, B and D, highligh ing he pe sis ing impo ance o FEV 1 as a p edic o o mo ali y ( able 3 and igu e 3). In g oup C, mo ali y was highe in spi ome y s age 1 han in s age 2, p obably due o a signi ican di e ence o he p opo ion in pa ien s be ween he wo s ages. Ou s udy con i ms ha pa ien s classi ied as GOLD A had he bes su i al, and pa ien s wi h GOLD D had he highe mo ali y in bo h classi ica ions [22–24]. Mo ali y o g oups B and C in GOLD 2015 lay in be ween A and D g oups and o en o e lapped. In GOLD 2019 mo ali y was signi ican ly highe in g oup B pa ien s han in g oup C. This inding is simila o o he p e ious epo s published [6, 23, 25], showing ha g oup B is an in e media e- o high- isk g oup associa ed wi h mo e exace ba ions and likely wi h o he como bidi ies ha may cause dyspnoea (such as hea ailu e). Fu he mo e, we show ha he bu den o symp oms ( ep esen ed by g oups B and D) ha e a p ognos ic alue addi i e bu independen o spi ome y. The cu en s udy demons a es ha he use o he GOLD 2019 classi ica ion scheme shi ed pa ien s wi h COPD o g oups o milde se e i y compa ed wi h GOLD 2015. This happened in keeping wi h p e ious epo s, bu in a smalle p opo ion o pa ien s (30% o pa ien s eassigned o g oup A o B compa ed o 53% in LEE e al. [25] s udy o 66% in TAN e al. [26] s udy). The u he dis ibu ion o spi ome ic pa ame e s om wo ca ego ies in GOLD 2015 (FEV 1 less o highe han 50%) o ou ca ego ies in GOLD 2019 was one o he possible easons o he pa ien s’shi om C and D in GOLD 2015 o A and B g oups in GOLD 2019. This phenomenon is opposi e o he one obse ed wi h he use o he e ised GOLD 2011 classi ica ion, which shi ed he pa ien s om he GOLD 2007 classi ica ion owa ds mo e ad anced s ages o he disease (D g oup inc eased almos h ee imes) [7]. Rema kably, in bo h classi ica ions, g oup C was consis en ly he smalles g oup ( igu e 1), as seen in o he epo s [23]. The implica ion o p og essi ely milde disease classi ica ions o he ea men choices clinicians make in eal li e p ac ice guided by GOLD is no ye clea gi en he ecen na u e o he la es GOLD i e a ion bu will be impo an o moni o . The s eng hs o ou s udy include a la ge sample size, he s udy design (a pooled analysis o indi idual pa ien da a om se e al coho s) and he di e en deg ees o se e i y o pa ien s om di e en coho s, maximising i s high ex e nal alidi y. P ospec i e da a collec ion o spi ome y wi h a pos -b onchodila o es , dyspnoea by mMRC scale, his o y o exace ba ion and mo ali y enabled di ec classi ica ion o pa ien s by he 2015 and 2019 GOLD s aging schemes. We also ha e a signi ican ep esen a ion o women, which o he COPD s udies migh no ha e achie ed [28]. Ou s udy has se e al limi a ions. Fi s , al hough we s a ed wi h 17139 pa ien s wi h COPD, a conside able numbe we e excluded because o missing in o ma ion on key a iables, mainly ega ding he his o y o exace ba ions. These missing da a a e unlikely o a ec he alidi y o ou esul s, as we can see in ou analysis compa ing included and non-included pa ien s. Second, he mo ali y analysis used all-cause dea h, and we ha e no da a ega ding speci ic causes o dea h ( his da a was no collec ed consis en ly in all coho s). Thi d, symp oms we e only e alua ed using he mMRC dyspnoea sco e, bu no wi h he COPD Assessmen Tes , o o he ins umen s [29]; howe e , ha is in line wi h o he epo ed coho s [30, 31]. Fou h, mos pa ien s came om hospi al-based coho s, so we likely ha e mo e pa ien s wi h mode a e o se e e disease and less pa ien s wi h mild and mode a e disease compa ed o an h ps://doi.o g/10.1183/23120541.00253-2020 8 COPD | E. GARCÍA CASTILLO ET AL. ou pa ien se ing, o a p ima y ca e popula ion. Indeed, he 22 coho s om 11 coun ies in 3CIA wi hin ou ini ia i e a e only a sample ep esen ing he es ima ed 300 million COPD pa ien s wo ldwide [32]. Finally, he ec ui men ime ame was long (10 yea s), so e ol ing ea men ecommenda ions migh in luence esul s. In conclusion, his s udy o COPD coho s, including 8823 pa ien s, showed ha nei he GOLD 2015 no GOLD 2019 a e s ong p edic o s o mo ali y. GOLD 2019 p edic ed mo ali y be e han GOLD 2015 in g oups A and B bu wo se in g oups C and D. Howe e , none o he GOLD classi ica ions has su icien disc imina o y powe o be used as a ool o isk classi ica ion o mo ali y in pa ien s wi h COPD. Ou s should be conside ed a cons uc i e exe cise and a c i ical app aisal o he las wo GOLD i e a ions de ining COPD. Wi hin 3CIA, we ha e al eady sugges ed se e al p oposals o u u e COPD s aging and g ading classi ica ions, by applying mo e e idence-based h esholds o e idence-based a iables [7, 21, 33–35]. Con lic o in e es : None decla ed. Re e ences 1GBD 2015 Ch onic Respi a o y Disease Collabo a o s. Global, egional, and na ional dea hs, p e alence, disabili y-adjus ed li e yea s, and yea s li ed wi h disabili y o ch onic obs uc i e pulmona y disease and as hma, 1990–2015: a sys ema ic analysis o he Global Bu den o Disease S udy 2015. Lance Respi Med 2017; 5: 691–706. 2Rabe KF, Hu d S, Anzue o A, e al. Global s a egy o he diagnosis, managemen , and p e en ion o ch onic obs uc i e pulmona y disease: GOLD execu i e summa y. Am J Respi C i Ca e Med 2007; 176: 532–555. 3Ves bo J, Hu d SS, Agus í AG, e al. Global s a egy o he diagnosis, managemen , and p e en ion o ch onic obs uc i e pulmona y disease: GOLD execu i e summa y. Am J Respi C i Ca e Med 2013; 187: 347–365. 4Han MK, Muelle o a H, Cu an-E e e D, e al. GOLD 2011 disease se e i y classi ica ion in COPDGene: a p ospec i e coho s udy. Lance Respi Med 2013; 1: 43–50. 5So iano JB, Al ageme I, Almag o P, e al. Dis ibu ion and p ognos ic alidi y o he new Global Ini ia i e o Ch onic Obs uc i e Lung Disease g ading classi ica ion. Ches 2013; 143: 694–702. 6Lange P, Ma o JL, Ves bo J, e al. P edic ion o he clinical cou se o ch onic obs uc i e pulmona y disease, using he new GOLD classi ica ion: a s udy o he gene al popula ion. Am J Respi C i Ca e Med 2012; 186: 975–981. 7So iano JB, Lamp ech B, Ramí ez AS, e al. Mo ali y p edic ion in ch onic obs uc i e pulmona y disease compa ing he GOLD 2007 and 2011 s aging sys ems: a pooled analysis o indi idual pa ien da a. Lance Respi Med 2015; 3: 443–450. 8Global Ini ia i e o Ch onic Obs uc i e Lung Disease (GOLD). Global S a egy o he diagnosis, managemen and p e en ion o COPD 2015. www.goldcopd.o g/ Da e las accessed: 21 June 2019, Da e las upda ed: 1 Oc obe 2020. 9Singh D, Agus i A, Anzue o A, e al. Global S a egy o he Diagnosis, Managemen , and P e en ion o Ch onic Obs uc i e Lung Disease: he GOLD science commi ee epo 2019. Eu Respi J 2019; 53: 1900164. 10 Goossens LM, Leime I, Me zdo N, e al. Does he 2013 GOLD classi ica ion imp o e he abili y o p edic lung unc ion decline, exace ba ions and mo ali y: a pos -hoc analysis o he 4-yea UPLIFT ial. BMC Pulm Med 2014; 14: 163. 11 Gedebje g A, Széplige i SK, Wacke hausen LH, e al. P edic ion o mo ali y in pa ien s wi h ch onic obs uc i e pulmona y disease wi h he new Global Ini ia i e o Ch onic Obs uc i e Lung Disease 2017 classi ica ion: a coho s udy. Lance Respi Med 2018; 6: 204–212. 12 A ambu u A, A os egui I, Mo aza J, e al. COPD classi ica ion models and mo ali y p edic ion capaci y. In J Ch on Obs uc Pulmon Dis 2019; 14: 605–613. 13 Mille MR, Hankinson J, B usasco V, e al. S anda disa ion o spi ome y. Eu Respi J 2005; 26: 319–338. 14 Bes all JC, Paul EA, Ga od R, e al. Use ulness o he Medical Resea ch Council (MRC) dyspnea scale as a measu e o disabili y in pa ien s wi h ch onic obs uc i e pulmona y disease. Tho ax 1999; 54: 581–586. 15 Ma hioudakis AG, Mobe g M, Janne J, e al. Ou comes epo ed on he managemen o COPD exace ba ions: a sys ema ic su ey o andomised con olled ials. ERJ Open Res 2019; 5: 00072-2019. 16 Rod iguez-Roisin R. Towa d a consensus de ini ion o COPD exace ba ions. Ches 2000; 117: 5 Suppl. 2, 398S–401S. 17 Golpe R, Suá ez-Valo M, Ma ín-Robles I, e al. Mo ali y in COPD pa ien s acco ding o clinical pheno ypes. In J Ch on Obs uc Pulmon Dis 2018; 13: 1433–1439. 18 Ma in JM, Al ageme I, Almag o P, e al. Mul icomponen indices o p edic su i al in COPD: he COCOMICS s udy. Eu Respi J 2013; 42: 323–332. 19 Izquie do JL, Mi a i lles M, Esquinas C, e al. Cha ac e is ics o COPD pa ien s managed in espi a o y medicine depa men s in Spain, acco ding o GOLD g oups and GesEPOC clinical pheno ypes. A ch B onconeumol 2018; 54: 559–567. 20 Rado ano ic D, Con oli M, Ma co FD, e al. Clinical and unc ional cha ac e is ics o COPD pa ien s ac oss GOLD classi ica ions: esul s o a mul icen e obse a ional s udy. COPD 2019; 16: 215–226. 21 Almag o P, Ma ínez-Camblo P, Mi a i lles M, e al. Ex e nal alida ion and ecalcula ion o he CODEX Index in COPD pa ien s. A 3CIAplus coho s udy. COPD 2019; 16: 8–17. 22 Han MZ, Hsiue TR, Tsai SH, e al. Valida ion o he GOLD 2017 and new 16 subg oups (IA-4D) classi ica ions in p edic ing exace ba ion and mo ali y in COPD pa ien s. In J Ch on Obs uc Pulmon Dis 2018; 13: 3425–3433. 23 Flynn RW, MacDonald TM, Chalme s JD, e al. The e ec o changes o GOLD se e i y s age on long e m mo bidi y and mo ali y in COPD. Respi Res 2018; 19: 249. h ps://doi.o g/10.1183/23120541.00253-2020 9 COPD | E. GARCÍA CASTILLO ET AL.