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.