Recei ed: Augus 8, 2014; Re ised: Decembe 17, 2014; Accep ed: Ma ch 11, 2015
JNCI J Na l Cance Ins (2015) 107(7): dj 095
doi:10.1093/jnci/dj 095
Fi s published online Ap il 11, 2015
A icle
1 o 6
a icle
© The Au ho 2015. Published by Ox o d Uni e si y P ess.
a icle
P edic ing High-G ade Cance a Ten-Co e P os a e
Biopsy Using Fou Kallik ein Ma ke s Measu ed in
Blood in he P o ecT S udy
Richa d J.B yan *, Daniel D.Sjobe g*, And ew J.Vicke s, Ma y C.Robinson,
Rajee Kuma , LukeMa sden, MichaelDa is, Pe e T.Sca dino,
JennyDono an, Da id E.Neal, HansLilja, F eddie C.Hamdy
A ilia ions o au ho s: Nu ield Depa men o Su gical Sciences, Uni e si y o Ox o d, UK (RJB, RK, LM, HL, FCH); Depa men o Epidemiology & Bios a is ics, Memo ial
Sloan Ke e ing Cance Cen e , New Yo k, NY (DDS, AJV); Depa men o Cellula Pa hology, Royal Vic o ia In i ma y, Newcas le upon Tyne, UK (MCR); School o Social
and Communi y Medicine, Uni e si y o B is ol, UK (MD); Depa men o Su ge y, U ology Se ice, Memo ial Sloan Ke e ing Cance Cen e (PTS, HL); Depa men
o Oncology, Uni e si y o Camb idge, UK (DEN); Depa men s o Labo a o y Medicine (Clinical Chemis y Se ice) and Medicine (Geni ou ina y Oncology Se ice),
Memo ial Sloan Ke e ing Cance Cen e , New Yo k, NY (HL); Depa men o Labo a o y Medicine and Clinical Sciences in Malmö, Lund Uni e si y, Skåne Uni e si y
Hospi al, Malmö, Sweden; and Ins i u e o Biomedical Technology, Uni e si y o Tampe e, Finland (HL).
* Au ho s con ibu ed equally o his wo k.
Co espondence o: Hans Lilja, MD, PhD, Memo ial Sloan Ke e ing Cance Cen e , 1275 Yo k A enue, New Yo k, NY 10065 (e-mail: [email p o ec ed]g).
Abs ac
Backg ound: Many men wi h ele a ed p os a e-speci ic an igen (PSA) le els in se um do no ha e agg essi e p os a e cance
and unde go unnecessa y biopsy. Re ospec i e s udies using c yop ese ed se um sugges ha ou kallik ein ma ke s can
p edic biopsy ou come.
Me hods: F ee, in ac and o al PSA, and kallik ein- ela ed pep idase 2 we e measu ed in c yop ese ed blood om 6129 men
wi h ele a ed PSA (≥3.0 ng/mL) pa icipa ing in he p ospec i e, andomized ial P os a e Tes ing o Cance and T ea men .
Ma ke le els om 4765 men p o iding an icoagula ed plasma we e inco po a ed in o s a is ical models o p edic any-g ade
and high-g ade (Gleason sco e ≥7) p os a e cance a 10-co e biopsy. The models we e co ec ed o op imism by 10- old c oss
alida ion and independen ly alida ed using ma ke s measu ed in se um om 1364 men. All s a is ical es s we e wo-sided.
Resul s: The ou kallik eins enhanced p os a e cance de ec ion compa ed wi h PSA and age alone. A ea unde he cu e
(AUC) o he ou kallik eins was 0.719 (95% con idence in e al [CI]=0.704 o 0.734) s 0.634 (95% CI=0.617 o 0.651, P <
.001) o PSA and age alone o any-g ade cance , and 0.820 (95% CI=0.802 o 0.838) s 0.738 (95% CI=0.716 o 0.761) o
high-g ade cance . Using a 6% isk o high-g ade cance as an illus a i e cu o , o 1000 biopsied men wi h PSA le els
o 3.0 ng/mL o highe , he model would educe he need o biopsy in 428 men, de ec 119 high-g ade cance s, and delay
diagnosis o 14 o 133 high-g ade cance s. Models exhibi ed excellen disc imina ion on independen alida ion among men
wi h only se um samples a ailable o analysis.
Conclusions: A s a is ical model based on kallik ein ma ke s was alida ed in a la ge p ospec i e s udy and educes
unnecessa y biopsies while delaying diagnosis o high-g ade cance s in ew men.
Risk o dea h om p os a e cance is s ongly associa ed wi h
le els o p os a e-speci ic an igen (PSA) in blood measu ed in
middle-aged men (1). E idence om andomized sc eening
ials in Eu ope shows ha PSA-based sc eening can educe
dea hs om p os a e cance (2–4), bu also leads o o e diag-
nosis and he isk o o e ea men among elde ly men wi h
This is an Open Access a icle dis ibu ed unde he e ms o he C ea i e Commons A ibu ion License
(h p://c ea i ecommons.o g/licenses/by/3.0/), which pe mi s un es ic ed euse, dis ibu ion, and ep oduc ion in
any medium, p o ided he o iginal wo k is p ope ly ci ed.
a Tampe e Uni e si y Lib a y. Depa men o Heal h Sciences on No embe 17, 2016h p://jnci.ox o djou nals.o g/Downloaded om
B yan e al. | 2 o 6
a icle
a icle
a limi ed li e expec ancy (5,6). Al hough he PSA es de ec s
an inc eased isk o p os a e cance a an ea ly s age o he
disease, i has low speci ici y (7) such ha mos men wi h an
ele a ed PSA ei he do no ha e p os a e cance o ha e low-
isk disease ha is unlikely o a ec quali y o leng h o li e
i le un ea ed. An ele a ed PSA is he main indica ion o
he app oxima ely one million p os a e biopsies pe o med
pe annum in he Uni ed S a es (8). Annual p os a e cance
incidence in he Uni ed S a es is close o 250 000, illus a ing
he unme need o ma ke s ha con ibu e speci ici y beyond
ha o o al PSA in o de o disc imina e be ween men wi h
cance s likely o in luence he leng h o quali y o li e and
hose wi h indolen disease o benign condi ions associa ed
wi h PSA ele a ion inblood.
P e ious esea ch sugges ed ha a panel o ee PSA
( PSA), “in ac ” PSA (iPSA—de ec ing only nonca aly ic single-
chain ee PSA bu no mul ichain- ee PSA in e nally clea ed
be ween Lys145 o Lys146 [9]), and o al PSA ( PSA), as well as
human kallik ein– ela ed pep idase 2 (hK2) measu ed in blood,
is mo e accu a e in p edic ing he ou come o p os a e biopsy
han o al PSA alone among p e iously unsc eened (10–12)
and p e iously sc eened (11,12) men and men wi h a p e ious
nega i e biopsy (13). Decision analyses showed ha a s a is i-
cal model based on he ou kallik ein ma ke s in blood can
imp o e clinical decision-making abou biopsy o men wi h
a PSA abo e 3 ng/mL (10,11,14). These da a sugges ha he
numbe o men unde going biopsy could be educed o hal
using 20% o g ea e cance isk as a en a i e h eshold o
biopsy, wi h app oxima ely 20% o cance s emaining unde-
ec ed among p e iously unsc eened men. Howe e , mos o
hese cance s would be low-g ade and low-s age cance s ypi-
cally associa ed wi h o e diagnosis, while ew high-g ade can-
ce s would be missed.
The P os a e Tes ing o Cance and T ea men (P o ecT)
s udy in he Uni ed Kingdom is a p ospec i e andomized con-
olled ial e alua ing he cos -e ec i eness o con en ional
ea men modali ies in PSA-de ec ed clinically localized p os-
a e cance . A o al o 8565 o 82 428 (10.4%) men ec ui ed o
P o ecT had a PSA o 3.0 ng/mL o g ea e and we e o e ed
a s anda d 10-co e p os a e biopsy. O he 7413 pa icipan s
ecei ing biopsies, 2894 we e ound o ha e e idence o p os-
a e cance (15,16). In p e ious s udies, se um samples had
been used o measu e he ou kallik ein ma ke s. Howe e , i
is well ecognized ha e hylenediamine e aace ic acid (EDTA)
an icoagula ed plasma has ad an ages o e se um wi h ee
and in ac PSA, being less p one o deg ada ion in plasma, ena-
bling mo e accu a e bioma ke analyses wi h samples shipped
o labo a o ies dis an om he poin o ca e (9,17). The majo -
i y o pa icipan s en olled in P o ecT had EDTA an icoagu-
la ed blood collec ed, which p o ided unique oppo uni ies
o compa e he kallik ein ma ke s measu ed in plasma s
se um. P e ious e alua ions o he ou kallik ein ma ke s
we e limi ed o men unde going sex an p os a e biopsies
(10–13), while he s udy epo ed he ein has he added alue
o assessing he ma ke s in he mo e con empo a y ex ended
10-co e biopsy p o ocol used in P o ecT (15), as his leads o
highe a es o cance de ec ion (18,19). We pe o med e o-
spec i e measu emen s o ou kallik ein ma ke s ( PSA, PSA,
iPSA, and hK2) in c yop ese ed EDTA an icoagula ed blood o
se um in he con ex o a la ge andomized p ospec i e clini-
cal ial in ol ing con empo a y ex ended 10-co e biopsies o
de e mine whe he his panel o ma ke s imp o es p edic ion
o biopsy ou comes compa ed wi h o al PSA andage.
Me hods
Pa ien Coho
In he P o ecT s udy, 228 926 men aged 50 o 69yea s we e in i ed
be ween 2001 and 2008 o ecei e PSA es ing. O hose, 82 429
(36%) pa icipan s we e es ed, 8565 men wi h a se um PSA
measu emen o 3.0 ng/mL o g ea e we e in i ed o unde go a
10-co e p os a e biopsy, 7471 (87%) men we e biopsied, and p os-
a e cance was de ec ed in 2637 pa icipan s. Two men wi h
missing age in o ma ion a biopsy and h ee men wi h missing
Gleason g ading we e omi ed om all analyses. C yop ese ed
blood was e ie ed o 82% o biopsied P o ecT pa icipan s:
EDTA an icoagula ed plasma om 4765 men, se um om 1860
men, and bo h plasma and se um om 496 men. Pa icipan s
had consen ed o sample collec ion and analysis by en oll-
men in he P os a e Cance Mechanisms o P og ession and
T ea men (P oMPT) s udy (e hics app o al NRES 01/04/061).
Pa hological Analysis
Assessmen o biopsies was unde aken blind o ma ke esul s
by u ological pa hologis s using s anda dized p o ocols and
ag eed-upon epo ing p o o mas (20), cance s we e g aded
using he s anda d Gleason sys em (21), as de ailed in a ecen
publica ion p o iding in o ma ion on he p ocessing and epo -
ing o p os a e co es and changes in Gleason sco es o e ime
(22).
Labo a o y Me hods
Fo consen ed indi iduals unde going PSA es ing wi hin he
P o ecT s udy, blood was collec ed, cen i uged, and ozen a
-80° C.An icoagula ed plasma o se um was ob ained a e cen-
i uga ion a 3000 g o 10 minu es wi hin 30 o 60 minu es o
enipunc u e. Immunoassays o PSA, PSA (17,23), iPSA, and
hK2 (24,25) we e pe o med as p e iously epo ed (11) using
c yop ese ed samples shipped on d y ice o analysis in Hans
Lilja’s labo a o y a he Wallenbe g Resea ch Labo a o ies,
Depa men o Labo a o y Medicine, Lund Uni e si y, Skåne
Uni e si y Hospi al, Malmö, Sweden. Sample aliquo s we e sub-
jec o wo eeze- haw cycles, and analyses we e pe o med
wi h esea che s blinded o p os a e biopsy ou come.
S a is ical Me hods
P e ious s a is ical models we e based on kallik ein le els meas-
u ed in se um o p e iously unsc eened men unde going sex-
an p os a e biopsy as pa o he Eu opean Randomized S udy
o Sc eening o P os a e Cance (ERSPC) s udy in Ro e dam
(10,11). As le els o some o he kallik ein ma ke s di e in
an icoagula ed plasma s se um and he biopsied men in he
P o ecT coho we e subjec ed o ex ended 10-co e biopsy, we
gene a ed new p edic ion models. We used mul i a iable logis-
ic eg ession o build models p edic ing p esence o any-g ade
o high-g ade disease on biopsy based on a man’s age, PSA,
PSA, iPSA, and hK2. Res ic ed cubic splines wi h kno s a he
e iles we e included in he model o PSA and PSA bu no
iPSA and hK2. P edic ions o men wi h a PSA le el o g ea e
han 25 ng/mL we e based on PSA le els alone. The use o
splines o PSA and PSA, bu no iPSA o hK2, and he use o
a 25 ng/mL PSA cu o we e p ede e mined based on ou p io
a Tampe e Uni e si y Lib a y. Depa men o Heal h Sciences on No embe 17, 2016h p://jnci.ox o djou nals.o g/Downloaded om
3 o 6 | JNCI J Na l Cance Ins , 2015, Vol. 107, No. 7
a icle
a icle
esea ch, and he only model i ing was o es ima e an in e cep
and coe icien s o each ma ke and nonlinea e m. We epo
he a ea unde he cu e (AUC), o disc imina ion, o he newly
de eloped models. AUCs we e compa ed wi h models based on
s anda d clinically a ailable p edic o s using he DeLong es
(26). We also in es iga ed whe he he model based on he ou -
kallik ein panel could educe he numbe o men unde going
biopsy wi hou delaying he diagnosis o high-g ade disease in
many men. We used decision cu e analysis (27) o in es iga e
he po en ial clinical e ec s o ou models. All epo ed s a is-
ics based on modelling we e co ec ed o o e i using 10- old
c oss- alida ion. As se um samples we e a ailable o some
men in he coho , we conduc ed explo a o y analyses applying
ou p e iously de eloped models o his subg oup. All P alues
epo ed a e wo-sided. Analyses we e conduc ed using S a a
12.0 (S a aCo p., College S a ion, TX).
Resul s
Cha ac e is ics o he 6129 biopsied men wi h blood samples a e
shown in Table1. An icoagula ed plasma (p ima y analysis) was
a ailable om 4765 men (3032 wi h no cance , 1733 wi h can-
ce ), and se um (seconda y analysis) was a ailable om ano he
1860 men (1221 wi h no cance and 639 wi h cance ). Finally,
we measu ed he ou kallik ein ma ke s in aliquo s om 496
men who had bo h plasma and se um samples (Supplemen a y
Table 1, a ailable online). O e all, p io PSA es ing was low
among biopsied men (18% o men p o iding plasma and 16%
o men p o iding se um).
The disc imina o y accu acy o each combina ion o kal-
lik ein ma ke s, as measu ed in EDTA an icoagula ed plasma,
is ou lined in Table2. The disc imina o y accu acy o he base
model (age plus PSA), as measu ed by he AUC, was 0.634 (95%
con idence in e al [CI]=0.617 o 0.651) o any-g ade p os a e
cance . Adding PSA, iPSA, and hK2 o his model imp o ed he
p edic i e accu acy (AUC 0.719 [95% CI=0.704 o 0.734], inc e-
men 0.085, P < .001). Use o he base model o p edic e idence
o Gleason sco e 7 o highe (high-g ade) p os a e cance a
10-co e biopsy ga e an AUC o 0.738 (95% CI=0.716 o 0.761),
while he addi ional kallik ein ma ke s s a is ically signi ican ly
enhanced he AUC o 0.820 (95% CI=0.802 o 0.838; inc emen
0.082, P < .001) (Table2).
To explo e he implica ions o using he models in clinical
p ac ice, we conduc ed a decision analysis o simula e ou -
comes i biopsy decisions had been based on a ious cu poin s
om he model. The esul s a e shown in Table3 and Figu e1.
Fo an illus a i e h eshold ep esen ing a 6% isk o Gleason
sco e 7 o highe (high-g ade) disease, use o his model would
educe he numbe o biopsies by 428 pe 1000 biopsied men
(43%), de ec 119 high-g ade cance s, and delay he diagnosis
o 14 o 133 high-g ade cance s, ou o which would ha e had
p ima y Gleason g ade 4. A decision-cu e analysis demon-
s a es ha use o he model con ibu es added clinical alue
(Supplemen a y Figu e3, a ailable online).
In a seconda y analysis, we examined he p ope ies
o he ou kallik ein ma ke s measu ed in se um om
ano he 1860 P o ecT pa icipan s (comple e da a shown in
he Supplemen a y Ma e ials, a ailable online). Applying he
p e iously epo ed “ERSPC-Ro e dam” (11) model imp o ed
he AUC compa ed wi h he base model (PSA plus age) om
0.665 o 0.709 (inc emen o 0.043, P=.010) o e idence o any-
g ade cance a biopsy, and om 0.785 o 0.836 (inc emen o
0.052, P=.010) o high-g ade cance a biopsy. Howe e , as
he “ERSPC-Ro e dam” model unde es ima ed isk o bo h
endpoin s (Supplemen a y Figu e 1B, a ailable online), and
ERSPC used sex an biopsies in con as o he minimum o
a 10-co e p os a e biopsy used in P o ecT, we ees ima ed he
coe icien s o he kallik ein ma ke s among men p o iding a
Table1. Cha ac e is ics o men in he P o ecT s udy coho *
EDTA an icoagula ed blood plasma Se um
Cha ac e is ics
No cance de ec ed
n=3032, 64%
No. (%)
Diagnosed wi h cance
n=1733, 36%
No. (%) P
No cance de ec ed
n=1221, 66%
No. (%)
Diagnosed wi h cance
n=639, 34%
No. (%) P
Clinical cha ac e is ics
Age (IQR), y 62 (58 o 66) 63 (59 o 67) <.001 62 (58 o 66) 63 (59 o 67) .005
P io PSA sc een 629 (21) 232 (13) <.001 207 (17) 83 (13) .024
Unknown 63 (2.1) 35 (2.0) 27 (2.2) 13 (2.0)
To al PSA (IQR), ng/mL 4.3 (3.6 o 5.7) 5.4 (3.9 o 8.6) <.001 4.5 (3.6 o 5.8) 5.6 (4.2 o 9.7) <.001
F ee PSA (IQR), ng/mL 1.00 (0.76 o 1.37) 0.97 (0.70 o 1.46) .3 0.97 (0.71 o 1.35) 0.92 (0.68 o 1.46) .7
In ac PSA (IQR), ng/mL 0.40 (0.27 o 0.58) 0.41 (0.26 o 0.66) .095 0.38 (0.25 o 0.57) 0.43 (0.27 o 0.70) <.001
hK2, ng/mL 0.043 (0.030 o 0.062) 0.049 (0.035 o 0.073) <.001 0.041 (0.029 o 0.061) 0.053 (0.036 o 0.081) <.001
Tumo cha ac e is ics
Gleason sum sco e
≤ 6 1099 (63) 464 (73)
7 542 (31) 143 (22)
≥ 8 92 (5.3) 32 (5.0)
S age
T1 1016 (59) 330 (52)
T2 331 (19) 88 (14)
T3 127 (7.3) 63 (10)
T4 4 (0.2) 2 (0.3)
Unknown 255 (15) 156 (24)
* The coho unde wen a 10-co e p os a e biopsy and had EDTA-an icoagula ed plasma and/o se um a ailable o e ospec i e measu emen s o o al, ee, and
in ac p os a e-speci ic an igens (PSAs) and hK2, in c yop ese ed sample aliquo s. Da a a e median (in e qua ile ange) o equency (pe cen age). IQR=in e qua -
ile ange; hK2=human kallik ein- ela ed pep idase 2; PSA=p os a e-speci ic an igen.
a Tampe e Uni e si y Lib a y. Depa men o Heal h Sciences on No embe 17, 2016h p://jnci.ox o djou nals.o g/Downloaded om
B yan e al. | 4 o 6
a icle
a icle
se um sample in he P o ecT s udy and pe o med an in e -
nal alida ion o he upda ed model again u ilizing 10- old
c oss- alida ion. The AUC now imp o ed om 0.665 o 0.757
(inc emen 0.092, P < .001) o e idence o any-g ade cance ,
and om 0.785 o 0.859 (inc emen 0.075, P < .001) o high-
g ade cance as shown in Supplemen a y Table 2 (a ailable
online). Nex , we e alua ed whe he he disc imina o y p op-
e ies o he kallik ein ma ke measu emen s we e di e en
in c yop ese ed se um s an icoagula ed plasma using pai ed
se um and plasma samples om 496 P o ecT pa icipan s
(Supplemen a y Table3, a ailable online). We ound no di e -
ences in he p edic i e accu acy o he ma ke s measu ed in
se um compa ed wi h plasma.
Finally, as an ex e nal alida ion o he model de eloped o
p edic high-g ade cance based on kallik ein ma ke s meas-
u ed in an icoagula ed plasma om 4765 biopsied P o ecT
pa icipan s, we assessed he disc imina o y accu acy o his
model using 1364 biopsied P o ecT pa icipan s wi h only se um
samples a ailable o measu e he ou kallik ein ma ke s and
excluding 496 P o ecT pa icipan s wi h bo h an icoagula ed
plasma and se um samples a ailable. Using his independen
g oup o 1364 biopsied P o ecT pa icipan s, ou ex e nal ali-
da ion showed ha he model exhibi s excellen disc imina o y
accu acy (AUC=0.849, 95% CI=0.814 o 0.883) (Supplemen a y
Da a Sec ion 5, a ailable online).
Discussion
In his s udy we demons a e ha a panel o ou kallik ein ma k-
e s— o al PSA, ee PSA, in ac PSA, and hK2—can p edic he
esul o p os a e biopsy. A s a is ical model based on he ou
ma ke s imp o ed his p edic ion abo e and beyond bo h PSA
and age, as well as beyond a combina ion o o al and ee PSA
(Table2). Adecision analysis indica ed ha use o he s a is ical
model o guide biopsy decisions would educe he numbe o men
ecei ing unnecessa y biopsies, wi hou subs an ially a ec ing
he diagnosis o Gleason sco e 7 o highe (high-g ade) cance s.
Two ini ial s udies on he Gö ebo g a m o he ERSPC demon-
s a ed ha he panel o ma ke s imp o ed p edic ion o biopsy
ou come o bo h unsc eened men (10) and hose wi h a p e ious
Table2. Disc imina o y accu acy o each kallik ein model*
Model
Plasma
AUC (95% CI) Inc emen o e “Age + o al PSA” (P alue)
Any-g ade p os a e cance
Age + o al PSA 0.634 (0.617 o 0.651)
Age + o al PSA and ee- o- o al PSA a io 0.710 (0.695 o 0.725) 0.076 (P < .001)
Age + panel o ou kallik ein ma ke s 0.719 (0.704 o 0.734) 0.085 (P < .001)
High-g ade p os a e cance
Age + o al PSA 0.738 (0.716 o 0.761)
Age + o al PSA and ee- o- o al PSA a io 0.799 (0.779 o 0.819) 0.060 (P < .001)
Age + panel o ou kallik ein ma ke s 0.820 (0.802 o 0.838) 0.082 (P < .001)
* This able ou lines he h ee combina ions o ma ke s o p edic ing any-g ade o Gleason sco e 7 o highe (high-g ade) cance a 10-co e p os a e biopsy based on
kallik ein ma ke measu emen s in an icoagula ed plasma p o ided by 4765 biopsied P o ecT pa icipan s. A eas unde he cu e we e compa ed wi h models based
on s anda d clinically a ailable p edic o s using he DeLong es (26). All s a is ical es s we e wo-sided. CI=con idence in e al; PSA=p os a e-speci ic an igen.
Table3. Resul s o di e ing biopsy s a egies pe 1000 men sc eened a a ying h esholds o isk o any-g ade cance o high-g ade cance
among men wi h an icoagula ed plasma
Th eshold
Biopsies
Any-g ade p os a e
cance
Gleason sco e 7 o
highe (high-g ade)
P ima y Gleason
sco e 4 o highe *
Pe o med A oided (%) Found Delayed Found Delayed Found Delayed
Risk o any-g ade cance
Biopsy all men 1000 0 (0) 364 0 133 0 47 0
Risk by age and o al PSA
≥20% 997 3 (0.3) 363 0 133 0 47 0
≥30% 642 358 (36) 264 99 111 22 41 6
Risk by age and panel o ou kallik ein ma ke s
≥20% 834 166 (17) 334 30 129 4 46 1
≥30% 545 455 (46) 264 100 116 17 43 4
Risk o high-g ade cance
Biopsy all men 1000 0 (0) 364 0 133 0 47 0
Risk by age and o al PSA
≥4% 974 26 (2.6) 357 7 132 1 47 0
≥6% 876 124 (12) 332 32 127 6 44 3
≥8% 715 285 (28) 284 79 117 16 42 5
≥10% 533 467 (47) 235 129 105 28 39 8
Risk by age and panel o ou kallik ein ma ke s
≥4% 734 266 (27) 313 51 127 6 46 1
≥6% 572 428 (43) 270 94 119 14 43 4
≥8% 442 558 (56) 229 135 110 23 42 5
≥10% 361 639 (64) 203 161 103 30 39 8
* Includes cases wi h any Gleason G ade 5 componen .
a Tampe e Uni e si y Lib a y. Depa men o Heal h Sciences on No embe 17, 2016h p://jnci.ox o djou nals.o g/Downloaded om
5 o 6 | JNCI J Na l Cance Ins , 2015, Vol. 107, No. 7
a icle
a icle
PSA in he no mal ange (28). Ou esul s con i m hese indings.
The assays we e subsequen ly modi ied, and a new s a is ical
model was buil based on a g oup o unsc eened men in he
Ro e dam a m o ERSPC. This was e med he “Ro e dam” model
and was shown o imp o e he p edic ion o biopsy ou come
o e and abo e PSA in se e al independen alida ion coho s,
including unsc eened men (11,14), men wi h p io sc eening (29),
hose wi h p io nega i e biopsy (13), and hose subjec o clini-
cal wo k-up be o e biopsy (12). The AUC esul s in hese p e i-
ous s udies a e e y simila o hose ound he e, including an
AUC o 0.820 o high-g ade cance , ep esen ing an inc emen o
0.082 o e age and PSA alone. Fo ins ance, in unsc eened men in
ERSPC, we epo ed an AUC o 0.825, 0.049 highe han he base
model (11), and o p e iously sc eened men he AUC was 0.793,
an inc emen o 0.094 beyond he base model (29).
The Ro e dam model was also shown o p edic clinical cance
endpoin s in men who we e ne e sc eened om he Malmö Die
and Cance s udy, a popula ion-based coho o 11 063 Swedish
men aged 45 o 73yea s who p o ided blood be ween 1991 and
1996. Subsequen diagnosis o p os a e cance was assessed by
linking wi h he Swedish Cance Regis y, upda ed o he end o
2006, and his occu ed in 943 men. Ra es o PSA sc eening we e
e y low in Sweden du ing he s udy. The Ro e dam model was
applied o a subse o men wi h a o al PSA le el 3.0 ng/mL o
mo e a baseline. The conco dance index o clinical diagnosis o
cance was highe o he Ro e dam model compa ed wi h PSA
alone (0.65 s 0.75, P < .001). Fo e e y 1000 men wi h a o al PSA
le el 3 ng/mL o mo e a baseline, he model would classi y as
high- isk 131 o 152 (86%) o he cance case pa ien s diagnosed
clinically wi hin i e yea s, and 421 men would be classi ied as
low- isk by he panel and ecommended o o go biopsy. O hese,
only wo men would be expec ed o be diagnosed wi h ad anced
p os a e cance (clinical T3 o T4 o me as ases) wi hin i e yea s.
Hence, while some men wi h a low sco e on he Ro e dam model
including he ou kallik ein ma ke s do indeed ha e biopsy-
de ec able cance , only a e y small p opo ion ha e agg essi e
disease ha would become appa en o e  ime.
The cu en s udy goes abo e and beyond he p e ious esea ch
no only in e ms o size—wi h a sample size app oaching all p e-
ious pape s on he kallik ein ma ke s combined—bu also in he
use o plasma samples, which appea o be mo e app op ia e o
analysis in clinical ca e, as he bioma ke s a e less p one o decay
in plasma compa ed wi h se um samples. Mo eo e , pa ien s in
he s udy unde wen a minimum o 10 co e biopsies, a he han
he ou da ed sex an biopsy used in p e ious s udies. Ou s udy
also p o ides clea e idence ha he s a is ical model based on
he ou kallik ein ma ke s could be used in e e yday clinical
p ac ice in o de o aid decisions abou p os a e biopsy.
The s udy has a numbe o limi a ions ha would need o be
add essed in subsequen esea ch. Fi s , esh samples would
need o be assayed in a ou ine clinical labo a o y a he han
in a esea ch se ing using c yop ese ed samples. Second, he
P o ecT ial p o ocol o e ed biopsies o all men wi h PSA le -
els o 3 ng/mL o g ea e , and no ably 87% o men accep ed he
o e o a biopsy. In ou ine clinical p ac ice, a man p esen ing o
a u ologis wi h an ele a ed PSA would be subjec ed o clinical
wo k-up, including assessmen o benign disease and equen ly
a epea PSA (30). Biopsy migh no be indica ed i he PSA is lowe
on epea es ing o i he PSA ele a ion is a ibu ed o a benign
cause. I may be ha some o he in o ma ion p o ided by he
kallik ein model is al eady cap u ed in such a clinical wo k-up. I
so, i is plausible ha while he kallik ein model may be o alue
whe e all men wi h ele a ed PSA a e biopsied, i is no o alue o
men selec ed o biopsy acco ding o a u ologis ’s clinical judg-
men . Thi d, he P o ecT coho does no e lec clinical p ac ice
in he Uni ed Kingdom, whe e li le oppo unis ic PSA es ing
akes place in he communi y (31) in con as wi h many o he
coun ies whe e PSA es ing and e es ing p e ails.
A de ini i e ou -kallik ein panel e alua ion will equi e he
p ospec i e analysis o samples aken om men wi h ele a ed
PSA le els in whom u ologis s ha e made he clinical judgmen
o pe o m a biopsy. The coho should include all-come s,
i espec i e o p io sc eening, PSA cu o used, o biopsy his-
o y. Kallik ein ma ke s should be measu ed om plasma in a
clinical labo a o y wi hin one o wo days o he blood d aw.
The biopsy would in ol e he 12 o 14 co e app oaches com-
mon in US p ac ice. Ma ke alues should be inco po a ed in o
a p especi ied s a is ical model including da a on digi al ec al
examina ion and his o y o p io biopsy. Such s udies a e cu -
en ly unde way.
In conclusion, we p o ide e idence ha a panel o ou kal-
lik ein ma ke s, inco po a ing o al PSA, ee PSA, in ac PSA,
Figu e1. Clinical implica ions o a ious biopsy s a egies using a model de eloped o p edic he isk o Gleason sco e 7 o highe (high-g ade) p os a e cance based
on ou kallik ein ma ke s measu ed in an icoagula ed plasma collec ed om 4765 biopsied P o ecT pa icipan s. The g aph illus a es he esul s o di e ing biopsy
s a egies pe 1000 biopsied P o ecT-pa icipan s, wi h he x-axis deno ing he isk o high-g ade cance and he y-axis indica ing he numbe o men biopsied (black
line) o de ec ed wi h e idence o high-g ade cance (g een line) using di e en biopsy s a egies. The do ed e ical blue line illus a es a en a i e cu poin (6% isk
o high-g ade cance ) a which only 572 o 1000 o he men would be biopsied, which would esul in he de ec ion o 119 o 133 high-g ade cance s.
a Tampe e Uni e si y Lib a y. Depa men o Heal h Sciences on No embe 17, 2016h p://jnci.ox o djou nals.o g/Downloaded om

B yan e al. | 6 o 6
a icle
a icle
and hK2, is supe io o o al PSA alone in p edic ing he esul o
p os a e biopsy. The ma ke s di e en ially de ec ed high-g ade
disease. In a decision analysis we ound ha implemen a ion
o a s a is ical model based on he ma ke s would educe by
close o hal he numbe o unnecessa y biopsies unde aken,
while delaying diagnosis o only a small numbe o high-g ade
cance s. These indings need o be con i med in p ospec i e
esea ch using clinical coho s.
No es
Au ho Con ibu ions: D . Lilja had ull access o all o he da a
in he s udy and akes esponsibili y o he in eg i y o he
da a and he accu acy o he da a analysis. S udy concep and
design: HL and AJV. Acquisi ion, analysis, o in e p e a ion o
da a: HL, RB, LM, RK, MD, FCH, DDS, MCR, and AJV. D a ing o
he manusc ip : RB, DDS, AJV, and HL. S a is ical analysis: DDS
and AJV. S udy supe ision: AJV, FCH, and HL. FCH, JD, and DEN
a e P incipal In es iga o s o he NIHR HTA P o ecT s udy.
Hans Lilja holds pa en s o ee PSA, hK2, and in ac PSA
assays, and is named, along wi h And ew J.Vicke s, on a pa en
applica ion o a s a is ical me hod o de ec p os a e cance . The
ma ke assay pa en s and he pa en applica ion o he s a is i-
cal model has been licensed and comme cialized as he 4K sco e
by OPKO Diagnos ics. D . Vicke s, Sca dino, and Lilja may ecei e
oyal ies om sales o his es . Addi ionally, D s. Sca dino and
Lilja own s ock and D . Vicke s owns s ock op ions in OPKO.
We a e g a e ul o Gemma Ma sden and A hene Lane o
hei help wi h his p ojec and o he P o ecT pa hology g oup:
Jon Oxley (Chai ), John Goepel, Mu ali Va ma, Da id G i i hs, Ken
G igo , Nick Maye , Ann Wa en, Naynee a Deshmukh, Selina
Ba ha ai, John Do me , and L. A. Adamczyk. We hank Gun-
B i E iksson, Mona Hassan Al-Ba a , and AnnaPe i E landsson
o expe assis ance wi h immunoassay measu emen s o he
blood samples a he Wallenbe g Resea ch Labo a o ies a Lund
Uni e si y in Malmö, Sweden.
Funding
This wo k is suppo ed by he Na ional Cance Ins i u e a he
Na ional Ins i u es o Heal h (R01 CA160816 o HL and AV, P50
CA092629 o C aig Thompson, and R01 CA175491 o Robe Klein),
he Sidney Kimmel Cen e o P os a e and U ologic Cance s,
Da id H. Koch h ough he P os a e Cance Founda ion, he
Na ional Ins i u e o Heal h Resea ch (NIHR) Ox o d Biomedical
Resea ch Cen e P og am, he Cance Resea ch UK Ox o d Cen e,
he Swedish Cance Socie y (p ojec no.11–0624 o HL), a FiDIP o-
p og am awa d o HL om TEKES, and Fundacion Fede ico SA. The
P o ecT s udy is unded by he UK Na ional Ins i u e o Heal h
Resea ch Heal h Technology P og amme (96/20/99).
Re e ences
1. Vicke s AJ, Ulme D, Sjobe g DD, e al. S a egy o de ec ion o p os a e can-
ce based on ela ion be ween p os a e speci ic an igen a age 40–55 and
long e m isk o me as asis: case-con ol s udy. BMJ. 2013;346: 2023.
2. Hugosson J, Ca lsson S, Aus G, e  al. Mo ali y esul s om he Go ebo g
andomized popula ion-based p os a e-cance sc eening ial. Lance Oncol.
2010;11(8):725–732.
3. Sch ode FH, Hugosson J, Roobol MJ, e  al. P os a e-cance mo ali y a
11yea s o ollow-up. N Engl J Med. 2012;366(11):981–990.
4. Sch ode FH, Hugosson J, Roobol MJ, e al. Sc eening and p os a e cance mo -
ali y: esul s o he Eu opean Randomized S udy o Sc eening o P os a e
Cance (ERSPC) a 13yea s o ollow-up. Lance . 2014;384(9959):2027–2035.
5. Welch HG, Albe sen PC. P os a e cance diagnosis and ea men a e he
in oduc ion o p os a e-speci ic an igen sc eening: 1986–2005. J Na l Cance
Ins . 2009;101(19):1325–1329.
6. Klo z L. P os a e cance o e diagnosis and o e ea men . Cu Opin Endo-
c inol Diabe es Obes. 2013;20(3):204–209.
7. Thompson IM, Anke s DP, Chi C, e al. Ope a ing cha ac e is ics o p os a e-
speci ic an igen in men wi h an ini ial PSA le el o 3.0 ng/ml o lowe . JAMA.
2005;294(1):66–70.
8. Fada e O, Wang S, Ma iappan MR. P ac ice pa e ns o clinicians ollowing
isola ed diagnoses o a ypical small acina p oli e a ion on p os a e biopsy
specimens. A ch Pa hol Lab Med. 2004;128(5):557–560.
9. Nu mikko P, Pe e sson K, Pii onen T, Hugosson J, Lilja H. Disc imina ion o
p os a e cance om benign disease by plasma measu emen o in ac , ee
p os a e-speci ic an igen lacking an in e nal clea age si e a Lys145-Lys146.
Clin Chem. 2001;47(8):1415–1423.
10. Vicke s AJ, C onin AM, Aus G, e al. A panel o kallik ein ma ke s can educe
unnecessa y biopsy o p os a e cance : da a om he Eu opean Randomized
S udy o P os a e Cance Sc eening in Go ebo g, Sweden. BMC Med. 2008;6:19.
11. Vicke s A, C onin A, Roobol M, e al. Reducing unnecessa y biopsy du ing
p os a e cance sc eening using a ou -kallik ein panel: an independen ep-
lica ion. J Clin Oncol. 2010;28(15):2493–2498.
12. Benchikh A, Sa age C, C onin A, e al. A panel o kallik ein ma ke s can p e-
dic ou come o p os a e biopsy ollowing clinical wo k-up: an independen
alida ion s udy om he Eu opean Randomized S udy o P os a e Cance
sc eening, F ance. BMC Cance . 2010;10:635.
13. Gup a A, Roobol MJ, Sa age CJ, e  al. A ou -kallik ein panel o he p e-
dic ion o epea p os a e biopsy: da a om he Eu opean Randomized
S udy o P os a e Cance sc eening in Ro e dam, Ne he lands. B J Cance .
2010;103(5):708–714.
14. Vicke s AJ, Gup a A, Sa age CJ, e al. A panel o kallik ein ma ke p edic s
p os a e cance in a la ge, popula ion-based coho ollowed o 15yea s
wi hou sc eening. Cance Epidemiol Bioma ke s P e . 2011;20(2):255–261.
15. Dono an J, Hamdy F, Neal D, e al. P os a e Tes ing o Cance and T ea men
(P o ecT) easibili y s udy. Heal h Technol Assess. 2003;7(14):1–88.
16. Lane JA, Dono an JL, Da is M, e al. Ac i e moni o ing, adical p os a ec omy,
o adio he apy o localized p os a e cance : s udy design and diagnos ic
and baseline esul s o he P o ecT andomized phase 3 ial. Lance Oncol.
2014;15(10):1109–1118.
17. Ulme D, Becke C, Nilsson JA, e al. Rep oducibili y and accu acy o meas-
u emen s o ee and o al p os a e-speci ic an igen in se um s plasma
a e long- e m s o age a -20 deg ees C. Clin Chem. 2006;52(2):235–239.
18. Eichle K, Hempel S, Wilby J, Mye s L, Bachmann LM, Kleijnen J. Diagnos ic
alue o sys ema ic biopsy me hods in he in es iga ion o p os a e cance :
a sys ema ic e iew. J U ol. 2006;175(5):1605–1612.
19. Heiden eich A, Bas ian PJ, Bellmun J, e al. EAU guidelines on p os a e can-
ce . pa 1: sc eening, diagnosis, and local ea men wi h cu a i e in en -
upda e 2013. Eu U ol. 2014;65(1):124–137.
20. Ca o JW, Robinson MC, Albe sen PC, e al. Sui abili y o PSA-de ec ed local-
ized p os a e cance s o ocal he apy: expe ience om he P o ecT s udy.
B J Cance . 2011;105(7):931–937.
21. Gleason DF. Ve e ans Adminis a ion Coope a i e U ological Resea ch
G oup. His ological g ading and s aging o p os a ic ca cinoma. In: Tan-
nenbaumM, edi o . U ologic pa hology. Philadelphia, PA: Lea and Febige ;
1977, 171–198.
22. Oxley J, Simpkin A, Goepel J, e al. Gleason d i in he NIHR P o ecT S udy.
His opa hology. 2015;66(3):438–446.
23. Pii onen T, Pe e sson K, Suonpaa M, e al. In i o s abili y o ee p os a e-
speci ic an igen (PSA) and p os a e-speci ic an igen (PSA) complexed o alpha
1-an ichymo ypsin in blood samples. U ology. 1996;48(6A Suppl):81–87.
24. Vaisanen V, Pel ola MT, Lilja H, Nu mi M, Pe e sson K. In ac ee p os a e-
speci ic an igen and ee and o al human glandula kallik ein 2.Elimina-
ion o assay in e e ence by enzyma ic diges ion o an ibodies o F(ab’)2
agmen s. Anal Chem. 2006;78(22):7809–7815.
25. Mi unen K, Pe e sson K, Pii onen T, Bjo k T, Lilja H, Lo g en T. Dual-label
one-s ep immunoassay o simul aneous measu emen o ee and o al
p os a e-speci ic an igen concen a ions and a ios in se um. Clin Chem.
1995;41(8 P 1):1115–1120.
26. DeLong ER, DeLong DM, Cla ke-Pea son DL. Compa ing he a eas unde wo
o mo e co ela ed ecei e ope a ing cha ac e is ic cu es: a nonpa ame ic
app oach. Biome ics. 1988;44(3):837–845.
27. Vicke s AJ, Elkin EB. Decision cu e analysis: a no el me hod o e alua ing
p edic ion models. Med Decis Making. 2006;26(6):565–574.
28. Vicke s AJ, C onin AM, Aus G, e al. Impac o ecen sc eening on p edic ing
he ou come o p os a e cance biopsy in men wi h ele a ed p os a e-spe-
ci ic an igen: da a om he Eu opean Randomized S udy o P os a e Cance
Sc eening in Go henbu g, Sweden. Cance . 2010;116(11):2612–2620.
29. Vicke s AJ, C onin AM, Roobol MJ, e al. A ou -kallik ein panel p edic s p os-
a e cance in men wi h ecen sc eening: da a om he Eu opean Rand-
omized S udy o Sc eening o P os a e Cance , Ro e dam. Clin Cance Res.
2010;16(12):3232–3239.
30. Kawachi MH, Bahnson RR, Ba y M, e  al. NCCN clinical p ac ice guide-
lines in oncology: p os a e cance ea ly de ec ion. J Na l Comp Canc Ne w.
2010;8(2):240–262.
31. Williams N, Hughes LJ, Tu ne EL, e al. P os a e-speci ic an igen es ing a es
emain low in UK gene al p ac ice: a c oss-sec ional s udy in six English ci -
ies. BJU In . 2011;108(9):1402–1408.
a Tampe e Uni e si y Lib a y. Depa men o Heal h Sciences on No embe 17, 2016h p://jnci.ox o djou nals.o g/Downloaded om