Transcript PSA Volatility Research
PSA VOLATILITY INDEX (PVI) & UPDATE ON PSA KINETICS John Phillips, MD 19 November, 2010 New York Medical College Hackensack University Medical Center
PSA VOLATILITY INDEX (PVI) & UPDATE ON PSA KINETICS John Phillips, MD 19 November, 2010 New York Medical College Hackensack University Medical Center
PSA VOLATILITY INDEX (PVI) & UPDATE ON PSA KINETICS 203,415 men developed prostate cancer 28,372 men died from prostate cancer John Phillips, MD 19 November, 2010 New York Medical College Hackensack University Medical Center
•34 kD •Serine protease •19q13 •Kallikrein-related peptidase III •Half-life total PSA 2-3 days •Free PSA <8 hours J Mol Biol (2008), 376, 1021-33
Serum Proteins Alpha-2-macroglobulin Alpha-Chymotrypsin
Free isoforms: B-PSA, I-PSA
, and
pro-PSA
SCR-270 Radar
Balancing true signal with interference
Receiver Operating Characteristic (ROC)
Receiver Operating Characteristic (ROC)
Area Under Curve* 0.500= worthless 0.650=lukewarm 0.700=very good 0.800=outstanding *trapezoidal method
PSA cut-offs
•1986 “99% of healthy men < 4.0” •1992 20-30% men > 4.0
•1994 24,000 screened q 6 months # cancers 2.6-4 identical to 4-10 ng/ml •1995 6,691 men PSA 4.1 missed 82% of cancers in men <50 •2001 PSA 2.5 ng/ml for men <50
PSA cut-offs •1986 “99% of healthy men < 4.0” •1992 20-30% men > 4.0
•1994 24,000 screened q 6 months # cancers 2.6-4 identical to 4-10 ng/ml •1995 6,691 men PSA 4.1 missed 82% of cancers in men <50 •2001 PSA 2.5 ng/ml for men <50 •Placebo arm of the 2004 PCPT trial •<2.1 ng/ml 15 % of men had cancer •Significant number had high grade disease •Stacey Loeb 14,000 BLAS •Median PSA 0.7 age 40-49: 0.7-2.5= 14.6 fold increased risk of prostate ca •Median PSA 0.9 age 50-59: 0.9-2.5=7.6 fold increased risk of prostate ca
•AUA Guidelines •PSA testing begin at 40 years of age + DRE if > 10 year life expectancy •Repeat PSA annually especially if above median for age •Prostate Cancer exists at all PSA levels •No absolute cut-off but PSA represents a “continuum of risk” •PSA presents an odds to a patient, then they can determine comfort zone
Other Scalar Tests
• [-2] pro PSA AUC=.76 Sokoll et al. CEBM 2010 • TMPRSS2-ERG • PSAD >0.155
• TZ-PSAD • PCA3
Dynamic Tests
• PSA doubling time (PSAdt) • Classically described after radiation therapy • Originally identified as a surrogate for failure • PSA dt < 3 months associated with decreased CSS • PSA dt < 6 months associated with increased BCF • Difficult to calculate
n n
i
1 log 2 log( * (
time n
1
PSA n
1
time PSA n
)
n
)
i
Dynamic Tests
• PSA Velocity • Carter B.L.S.A.
• PSA < 4 • PSAV was associated with CSS 25 years later • 92 % survival <.35 ng/ml/yr • 54% survival >.35 ng/ml/yr • Loeb: 0.75 cut off PSA 4-10 • 0.35 cut off PSA 2.5-4.0
Dynamic Tests
• Can PSAV be used to help with screening?
• Vickers et al. J U 184: 907-912 2010 • European Randomized Screening Study for Prostate Cancer ERSPC • N=2,742 • Median PSA 4.47 vs 4.69 (No Ca versus Ca) • Median volume 47 vs 36 (No Ca versus Ca) • PSAV 0.26 vs 0.28 ng/ml/yr (No Ca versus Ca) • AUC PSA 0.522 PSAV 0.551 PSA+PSAV 0.621
• “Little support for any clinically useful role for PSA velocity to help determine initial biopsy” • “May be useful after an initial negative biopsy”
Dynamic Tests
• • Can PSAV be used as a trigger in active surveillance?
• Andrew Loblaw et al. (Laurence Klotz’ group) JU 184: 1942, 2010 • N=305 untreated men 6.1 years (range 0.5-13.3) • Zero prostate cancer deaths • Trigger for Treatment (TT) • PSA of 10 ng/ml TT 38% of the time • PSAV associated with TT 42-84% of the time www.ASURE.com
PSA
• Incredibly useful • Incredibly misused • Multiple forms • PSA, free PSA, proPSA • PSAV, PSAdt, Ln(PSA) • TMPRSS2ERG • PCA3
Volatility: A measure of bidirectional variation over time
Volatility: A measure of bidirectional variation over time .
time, t
Volatility: A measure of bidirectional variation over time .
time, t
Volatility: A measure of bidirectional variation over time .
time, t
Volatile (Bidirectional) Variables Options Pricing Treasuries Sunspot Activity Global Warming PSA Source: Stockcharts.com
Volatile (Bidirectional) Variables Options Pricing Treasuries Sunspot Activity Global Warming PSA Source: Stockcharts.com
6-month-ahead Eurodollar rate (Rudebusch et al.)
Volatile (Bidirectional) Variables Options Pricing Treasuries Sunspot Activity Global Warming PSA Daily GOES readings: Nat’l Geophysical Data Canter
Volatile (Bidirectional) Variables Options Pricing Treasuries Sunspot Activity Global Warming PSA Source:
Global Temperature Land-Ocean Index: Goddard Institute, NASA
Volatile (Bidirectional) Variables Options Pricing Treasuries Sunspot Activity Global Warming PSA
PSA kinetics
• PSAV: 1 ng/dl/mo 14 12 10 8 6 4 2 0 1 2 3 4 5 6 7
Year
8 9 10 11 12
PSA velocity
• PSAV: How accurate with volatile PSAs?
14 12 10 8 6 4 2 0 1 2 3 4 5 6 7
Year
8 9 10 11 12 • Same average PSA, s , t-test
14 12 10 8 6 4 2 0 1 2 3 4 5 6 7
Year
8 9 10 11 12
14 12 10 8 6 4 2 0 1 2 3 4 5 6 7
Year
8 9 10 11 12
14 12 10 8 6 4 2 0 1 2 3 4 5 6 7
Year
8 9 10 11 12
• Hypothesis : Can a PVI discriminate benign from malignant states?
14 12 10 8 6 4 2 0 1 2 3 4 5 6 7
Year
8 9 10 11 12
PVI: Retrospective Study Population INCLUSION At least 3 PSAs At least 1 (one) prostate biopsy (TRUS or TP) EXCLUSION No Urothelial Ca No h/o pelvic RT No h/o hormone deprivation
Quantitating PVI
Volatility
s log
t Q t
1
Q
Financial Models
C
SN
ln(
s
/
K
) (
r
s
t
s 2 / 2 )
t
Ke
rt N
s
t
(
Black & Sholes, 1973
) s 2 2
T
i
K i K i
2
e RT Q
(
K i
) 1
T
F K
0 1 2 (
Rattray & Shah, 2003)
PSA Volatility Index (PVI)
SE e slope
HYPOTHESIS An e-based normalization for PSA scatter and a PVI may predict Ca from BPH
Retrospective Study Population N 1 =932 N 2 =279 107 Cancer 172 BPH (e.g. BPH, atrophy, prostatitis, PIN) Independent variables Age, race, IPSS, ASA, Meds, TRUS data, Grade, volume/core, # cores, surgical pathology aPSA, aPSAV, aPSAdt, PSAD, Ln(PSA), PVI
Descriptives
30 25 20 15 10 5 0 Jun-94 Oct-95 Mar-97 Jul-98 Dec-99 Apr-01 Sep-02 Jan-04 May-05 Oct-06 Feb-08 Jul-09 Prostate Ca (n=107) 30 20 10 0 Jan-93 Oct-95 Jul-98 Apr-01 Jan-04 Oct-06 Jul-09 BPH (n=179) Total PSA ng/dl 1993-2010
Age Caucasian (%) ASA Class I-II (%) ASA Class III-IV (%) AUA-SS (0-40) Nocturia (voids/night) Study Interval (yrs) Treated for LUTS (%) Number of PSAs Abnormal DRE (%) TRUS vol (cc) Number of Biopsies Demographics BENIGN n=172 64.75 ± 8.67
82.94
66.48
33.52
CANCER n=107 65.14 ± 9.22
80.95
64.58
35.42
P-Value 8.15 ± 5.67
1.40 ± 1.16
5.10 ± 3.33
58.82
6.90 ± 3.82
13.69
44.08 ± 25.01
1.75 ± 1.03
6.42 ± 5.58
1.38 ± 0.89
4.04 ± 2.96
48.82
5.05 ± 2.78
20.41
40.24 ± 20.59
1.50 ± 0.88
0.89
0.85
0.88
<.01
0.49
<.01
0.07
<.01
0.15
0.21
0.06
Descriptives
30 25 20 15 10 5 0 Jun-94 Oct-95 Mar-97 Jul-98 Dec-99 Apr-01 Sep-02 Jan-04 May-05 Oct-06 Feb-08 Jul-09 30 20 10 0 Jan-93 Oct-95 Jul-98 Apr-01 Jan-04 Oct-06 Jul-09 PSA ng/ml free PSA (%) PSAV ng/ml/yr PSA dt (mo) slope ln(PSA)*1000 PVI PSAD (ng/ml/cc) BENIGN n=172 5.60 ± 3.05
19.78 ± 9.33
-0.39 ± 6.28
12.81 ± 336.95
-0.06 ± .00
1.56 ± 1.29
0.15 ± 0.09 CANCER n=107 6.36 ± 5.53
16.17 ± 8.51
0.71 ± 2.86
241.28 ± 240.18
0.38 ± 1.1
0.83 ± 0.83
0.19 ± 0.20
p-value 0.7
0.04
0.09
0.06
<.0001
<.0001
.04
PVI Correlations
30 25 20 15 10 5 0 Jun-94 Oct-95 Mar-97 Jul-98 Dec-99 Apr-01 Sep-02 Jan-04 May-05 Oct-06 Feb-08 Jul-09 30 20 10 0 Jan-93 Oct-95 Jul-98 Apr-01 Jan-04 Oct-06 Jul-09 X PVI PVI PVI PVI PVI PVI Y PSAV PSAD Proscar P value 0.11
# biopsies 0.20
IPSS BPH 0.29
<0.00
<0.00
<0.00
R 2 .06
0.05
0.06
PVI Correlations
30 25 20 15 10 5 0 Jun-94 Oct-95 Mar-97 Jul-98 Dec-99 Apr-01 Sep-02 Jan-04 May-05 Oct-06 Feb-08 Jul-09 30 20 10 0 Jan-93 Oct-95 Jul-98 Apr-01 Jan-04 Oct-06 Jul-09 X PVI PVI PVI PVI PVI PVI Y PSAV PSAD Proscar P value 0.11
# biopsies 0.20
IPSS BPH 0.29
<0.00
<0.00
<0.00
R 2 .06
0.05
0.06
Linear Regression y=mx+b
Linear Regression y=mx+b
For each increase of 1 unit of x, y increases by 4.88 %
PSA volatility and % cancer Or, as cancer volume decreases, PSA volatility increases
Logistic Regression = yes or no
f
(
z
) 1 1
e
z
For each change in x, the probability of the outcome goes up by y
Logistic Regression = yes or no
UNIVARIATE
free PSA PSAD AUA SS Ln(PSA)*1000 † PVI
MULTIVARIATE
Coeff -0.05
2.07
-0.06
1.43
-0.77
p-value 0.05
0.06
0.02
<.001
<.001
Ln(PSA) + AUA SS PVI + AUA SS Coeff 1.23
-0.05
-0.78
p-value <.001
<.001
-0.06
0.05
0.03
ROCs AUC=.800
Nomograms=Prediction Tools
Nomograms=Prediction Tools Formula for the oxygen consumption of rainbow trout as a function of weight and water temperature. Ron Doerfler Dead Reckonings
Nomograms=Prediction Tools •Too Many Variables can Compromise the Model •Increasing Complexity may Reduce Clinical Usefulness •“Precision versus Parsimony”-Sengupta and Blute 2006
PVI Nomogram
PVI
•PVI > 2 benign, esp. in total PSA range 4-20 •PVI < 1 malignant •PVI quantifies the PSA instability over time •www.psastatistics.com
•Multivariable analysis
SE e slope
WWW.PSASTATISTICS.COM/PVI
PVI
•PVI > 2 benign, esp. in total PSA range 4-20 •PVI < 1 malignant •PVI quantifies the PSA instability over time •www.psastatistics.com
•Multivariable analysis
SE e slope
Thanks
Muhammad Choudhury, MD (NYMC) Andrew Fishman, MD (NYMC) Basir Tareen, MD (BIMC) Harris Nagler, MD (BIMC) Devesh Shah, PhD (Goldman Sachs)