This material may be protected by Copyright law (Title 17 U.S. Code)
`
`Prostate-Specific Antigen as a Measure of Disease Outcome
`in Metastatic Hormone-Refractory Prostate Cancer
`
`By Wm. Kevin Kelly, Howard I. Scher, Modhu Mazumdor, Vaio Vlornis, Morton Schwartz, and Sophie D. Fosso
`
`Purpose: To evaluate the prognostic significance of
`-Y-FB"I'eU1'fI'IGlIt parameters and posttherapy declines in
`prostate-specific antigen ll‘-‘SA: in relation to the survival
`lit patients with hormone-refractory prostate cancer.
`Patients and Methods: One hundred ten assessable
`
`gatients treated on seven sequential protocols at Me-
`morial Sloan-Kettering Concer Center [MSKCCJ for hor—
`.none-refractory prostate cancer were evaluated for 29
`different pretherapy and posttheropy parameters, in-
`cluding a posttherapy decline in PSA of 50% and 30%
`from baseline.
`Results:
`In the univariate analysis, initial Karnotsky
`performance status {KPSJ 2 30% was associated with a
`favorable outcome [P = .0051, while age, extent,ot dis-
`ease on bone scan, and individual sites of metastatic
`disease were not significant. No difference in survival
`was observed between patients with measurable or as-
`sessable [bone only] disease. Initial hemoglobin {HGB_: P
`= .0012], alkaline phosphatase {ALl(; P = .0015], and
`lactate dehydrogenase {LDH,- P = .0002] levets were
`
`N 1992. PROSTATE CANCER resulted in the death
`of more than 34.000 men in the United States. The
`
`majority of these deaths result from cells that proliferate
`despite androgcrrablative therapies. Indeed. oncc h0r—
`monc-independent disease is documented. median sur-
`vivals are measured in months."3 and no standard therapy
`exists. Clearly new agents and strategies are required. This
`effort has been hindered in part by the clinical manifes-
`
`tations of horrnonorcfractory tumors in that upwards of
`90% of patients have sclerotic bone metastasis. which are
`not assessable by classic phase [1 response criteria.” As a
`result, many groups have restricted investigational studies
`ofncw agents to the |0% ofpatients with bidirhcnsionally
`measurable lesions. Others have criticized this approach
`with the vicw that these patients may not be representative
`of the more typical cases with osseous disease only, and
`that the prognosis for this group is inherently poor. To
`circumvent this problem. a variety of response criteria
`have been proposed.‘ Unfortunately, these criteria have
`not been universally accepted and do not correlate well
`with outcome. As a result. the reported survival benefit
`ofrcspondcrs versus nonrespondcrs in many trials was an
`imbalance of prognostic factors between treatment groups
`and not a true beneficial e-.fl'cct frorn therapy.’ Further-
`more, the majority of the trials were performed before
`the availability of serum prostatc—s]JccifiC antigen IPSA)
`determinations.
`
`PSA is a 34,000-d glycoprotcin found in prostatic tissue
`and seminal plasmas‘ Plasina levels correlate well with
`extent ol'diseasc.“"l although absolute values do not de-
`
`significant discriminators, while the initial PSA was not.
`Using a landmark analysis, a significantly longer median
`survival rate was observed for patients with a 2 50%
`decline in PSA [median not reached} versus patients with
`a less than 50% decline in PSA (median, 8.6 months; P
`= .0001]. Multivariate analysis using the Cox propor-
`tional hazards model showed that a 2 50% decline in
`PEA [P = .0004] and the natural log of LDH {P = .0001]
`were the two most significant variables predicting sur-
`vival. The model was confirmed on on independent data
`set from the Norwegian Radium Hospital [NRH] in Oslo,
`Norway.
`Conclusion: The results suggest that posttherapy PSA
`declines can be used as a surrogate end pointto evaluate
`new agents in hormone-retractory prostate cancer. The
`criteria for response need prospective validation in
`phase Ill trials.
`J Clin Oncol l H607-615. <'_C'3 I993 by American Society
`of Clinical Oncology.
`
`tcrminc stage. Changes in PSA have been shown to cor-
`relate with clinical outcome in patients treated by radical
`surgery, radiation therapy. or androgen ablation. For ex-
`ample. following successful removal of the prostate by
`radical surgery. levels should decline by l1a1f—|iFe and re-
`main undetectable. Similarly. following radiation therapy
`for localized disease. serial declines in PSA were indepen-
`dently prcdictivc oi‘outcome." while increasing levels were
`indicative of active di5cu.sc.'”"' In patients with metastatic
`disease treated with androgen ablation, posttherapy
`changes in PSA were found to be more sensitive than
`changes in acid phosphatase (ACP) with rcspcct to mea-
`surable discasc response.” while posttherupy declines in
`PSA of80".-‘ta or more within 1 month after treatment were
`
`associated with long progression-free intervals.”
`Use of posttherapy declines in PSA as an outcome
`measure for clinical trials in patients with hormone-re-
`
`From Hit’ Gr'nilr)m'i'nu:'_1' 0nt'o.’og1' .5'£‘I"l‘t}."(’, D:‘i'r'.\'t'i:n t.rf'Sir!iri Tumor‘
`0:ii'r;L-J,e_1'.
`.Dc,Ht:i'liJttvtl nl
`rI;’t'(.r‘r't'r'm'_' Dt'w'.rirm if!" B.r'i1rli1li'.rrit'.t, Dc-
`pttl‘ftil't‘tii' t§l'Epi'tic.=niol'n,§{_l' and Bit).'c'f(iii.Tii£’.5‘.' mid Dclporrmcm i._=,I"C’."t'u—
`icof C‘r’tt'mi.\'t:'_t'_ ,.l-ltvnuritir’ .S'.lrruit-!(t'rtci'i;:,r: C‘niict=v' Ccmcr, New I’or.lc.'
`Dtgnumircnl ry".-1i:.vl'i<':'ric. Clnr.-ici"l'
`t’.'.=i'vcr_\‘i'l_1'
`.'li'tJr.|'ictn’ C'trlt’t{g*t'. New
`l":Jrk_ N l". tmri Di-'pummvt.r 9;".-l-.lm’i't'tt:’ 0Jit'ra’r'3g_1' tum’ R::r."i':.liitc:'rip}'.
`Nnrit'cgimr Rtidimtr Ho.\‘pittn’. Om). Non:-n_1'.
`SirH'm:t'.rrt't'i' ()t‘tiIhct' 5 , I ‘)'§'_'’,' :rt't'c,ra'c*t! Nm't-rmhcr 25. I 993.
`.Ei1tp,r;.-trim’ h_|‘_t:i'um‘.i no. C‘.-'l—Ll5t'i'.?i‘= and C.<l—n'_i9'.7I‘) .7’. Hit’ {}m‘u'.ii:1im/
`Ft‘J‘t‘_l’ Pm.\'tt.=n" C'um'tv' Rc.tcrn't'.l'r Firm)’. and’ the .'Ut'!ti'iJi Hr'.-rmrcr‘ {‘I'Hilt'l'.
`.-lt7'd:'c.r.\' rcprint rct,rtit~'.n‘.s‘ in .l‘.-'.=1n'nn'i' I .5't'J't(-r. MD, :1.’ :=:Itni'iul .S‘lt:uli-
`}\'ct:t=i':'::,1{ Cnrrtcr t;'cnh:*r.
`l_7?5 i’r.u'k .-lrc, Non’ i'}1rlt'. N 1’ ltidfi.
`’-Q 1'99} by .-lim.'l‘it'tm ._\‘r.1r.‘ict_1' n,v"C'a’t'm't'ci)‘ 0:tt'r:!r:.r:_1'_
`t.l?_i3—/o'.t.\'/93/a’ ii!-'I—(llltl4.5i:i.iJt')g"U
`
`Joumol of Clinical Oncology, Vol
`
`I I, No 4 {April}, 1993: pp 60?—ol 5
`
`WCK1025
`Page 1
`
`WCK1025
`Page 1
`
`

`
`608
`
`KELLY ET AL
`
`fractory disease has been reported by several groups. “"7
`However. the criteria for response with respect to degree
`or duration ofdecline have not been standardized or pro-
`spectively validated. '3 Changes in PSA values after treat-
`ment must be interpreted cautiously for several reasons.
`First. the therapy itsell‘ may reduce PSA secretion without
`tumor-cell kill.” In a clinical trial oftrimetrexate, serum
`levels seemed to fluctuate in parallel with drug adminis-
`tration?" Second. androgen withdrawal has been shown
`to reduce serum PSA levels without a parallel reduction
`in tumor volume in androgen-stimulated lymph node
`cancer of the prostate {LNCaP) cells in vivo.3' It is also
`recognized that PSA staining of tumors is not homoge-
`neous. raising the possibility ofindependent therapeutic
`etfects on PSA secreting and nonsecreting ee|ls.”'4' Rec-
`ognizing these caveats, we investigated the relationship
`between posttherapy declines in PSA and survival in pa-
`tients with hormone-refractory prostatic cancer treated
`with systemic therapy. A univariate and multivariate
`analysis was performed and a multi variable model derived.
`
`For validation, an independent data set ofpatients treated
`at the Norwegian Radium Hospital {NR}-I}. Oslo. Norway,
`was studied.
`
`PATIENTS AND METHODS
`
`Patielir PiJpiilnti'mi
`
`One hundred ten assessable patients with histologically confirmed
`metastatic hormone—rel'ractory prostatic cancer treated on seven
`consecutive treatment programs (suramin. rheniurn 186. e5tramus-
`tincxvinblastine. ttimetreitate. or gemcitabine) between [987 and l99l
`at Memorial Sloan-Kettering Cancer Center {MSKCCL New York.
`NY. were considered. All had documented progressive disease. despite
`castrate levels oi’ testosterone. For this analysis. disease progression
`was defined as the presence of at least one ofthe following events:
`(I) an increase of PSA of greater than 50% from baseline on three
`successive occasions. (2) new metastatic lesions on bone scan. (3) a
`greater than 25% increase in a bidimensionally measurable tumor
`mass. or (4) new symptoms related to tumor growth. While the in-
`dividual treatmcnt programs dilfered with respect to the requirement
`For measurable or assessable disease. all patients underwent a complete
`history and physical examination. Laboratory evaluation included
`an automated blood and platelet count. I2-channel screening profile
`(alkaline phosphatase [ALK], lactic dehydrogenase [LDH], aspartate
`transglutaminase, blood urea nitrogen, creatinine. calcium. phos-
`phorus. uric acid. total protein. albumin. and total bilirubinl, serum
`ACP. and PSA. Imaging studies included an abdominal and pelvic
`computed tomographic {CT} scan. bone scan. and chest radiograph.
`Patients were treated according to the individual program on which
`they were enrolled. To assess response. all were reevaluated at 8-week
`intervals (12 weeks for bone scans} unless earlier reevaluation was
`clinically indicated. Posttherapy FSA determinations were performed
`at a minimum of biweekly intervals. All patients were monitored
`until death or censored at last follow-up. Patients who were en rolled
`in more than one study were evaluated only on their first investiga-
`tional treatment.
`
`.S'lttu’_1r Pam:-iterei'.r
`Twenty-nine variables were considered. Demographics included
`age as a dichotomous (> 65 v 5 65 years] or continuous variable.
`initial Karnofsky perfonnance status {[KPS] _>_ 8t'l‘lli or «c 80%). and
`pathologic grade at the time ofinitial diagnosis. Prior treatment was
`divided into the extent ofprior radiation therapy, prior surgery. and
`the type oi‘. and the duration ofrcsponse to. primary hormone therapy
`(5 6 1’) 6 months. and 5 I2 1' > I2 months). The latter was inves-
`tigated on the basis ofreports showing the significance ofthe duration
`of response to primary hormonal therapy and survival.2"'35
`
`Extent of Disease
`
`Extent of disease in bone was assessed by a review of‘ the pretreat-
`ment scans by two independent reviewers using the method reported
`by Soloway et al.” This scaling system involves countirlgthe number
`oflesions and classifying a patient as 0 (normal).
`I (< six lesions}. 3
`(six to 30 lesions), 3 l:> 20 lesions). or 4 la sttpersctin or > 75%
`involvement of the vertebrae. pelvis. or ribs)?“ individual sites of
`soft tissue spread. including liver. lung. rctroperitoneal lymph nodes
`{RFLNL epidural, or other sites (peripheral lymph nodes. pelvic side
`wall. bladder, adrenal. prostate. and rectal masses) were recorded.
`Osseous sprelsd nlone versus osseous plus solt tissue or soft tissue
`only was alsti considered.
`
`Baseline Bioclientfcri.-’ty"rti'riiiielei‘.s'
`Baseline biochemical parameters were evaluated as discrete t normal
`v abnormal) and continuous variables. AC1‘. ALK. creatinine (CR).
`hemoglobin (HUB). AST. and LDH (5 230 U,tL ‘I’ > 230 U/Ll were
`considered. Baseline PSA values were analyzed in discrete ranges (5
`4 v> 4-. 5 201-> 30'. § 50 t-:> 50: 5 I00 1-‘) I00: and 5130»:
`150} and as a continuous variable. All PS/t values were performed
`in the Department ol'Cli1tical Chemistry at MSKCC using the Tan-
`dern—E FSA immtinoenzymatic assay by the Hybritech Corporation
`{San Diego. CA}.
`
`l--iiriistb.-’c.5‘
`P0d'h’il1'E'l'&'p_l-'
`Posttherapy PSA values were obtained following treatrnent at a
`minimum of biweekly intervals. Recognizing that fluctuations in PSA
`can occur?“ and that maintaining a decline would be of importance,
`we used the following criteria for response based on a preclclincd
`posttherapy decline: (l) minimum of four PSA values. I week apart.
`and (3) a 2 50% or 2 lfiO% reduction in PSA from baseline by 60
`days and maintained for two additional consecutive determinations
`I week (ir more apart.
`
`Struts! ice! A lirt.’_t-*si's
`
`Recognizing the potential for positive bias when comparing re-
`sponders versus nonresponders. the landmark method was used. This
`technique requires establishing a landmark time and includes only
`those patients who survive to this time point. Patients are then fol-
`lowed forward ch ronologically to ascertain whether survival from the
`landmark time depends on the patients response status at the land-
`mark point.” Using this definition. patients who achieved a 2 50%
`decline after the 60-day period were not considered in a response
`category. The landmark time of 60 days was chosen, since this time
`point corresponded to the usual interval at which a restaging eval-
`uation to assess the elTecl of treatment would occur. The survival
`distributions were estimated by the Kaplan-Meier method,“ and
`median survivals were reported from the initiation oltreatment. All
`patients were monitored until death and patients alive at last Follow-
`
`WCK1025
`Page 2
`
`WCK1025
`Page 2
`
`

`
`PSA AND PROSTATE CANCER
`
`up were considered to be censored. Median follow-u p duration was
`only reported for the survivors.
`The survival distributions ol‘ the pretherapy variables that were
`analyzed as categoric variables were compared by the log-rank test.“
`with level of significance 5 .05. To avoid the subjectivity involved
`in the choice of the break point. continuous variables were also eval-
`uated using the score test. Whenever data were skewed. natural log-
`transformations were used to make the distribution symmetric.
`'1‘ he posttherupy PSA decline was used to define responders (2
`5(lE*tn or 2 30‘-rt. decline in PSA from baseline) versus non responders
`[no 50% or 30% decline) at the landmark time. and survival distri-
`butions were compared using the log—rartlt test. To investigate further
`the balance with respect to other variables in the group of responders
`and non responders. eategoric variables in each group were evaluated
`by Fishcr‘s exact test and contin uous variables by the Mann-Whitney
`test.
`Based on the significant univariate factors excluding KPS, a mul-
`tivariate Cox model was developed based on the risk ratios. Cate-
`gorization of high— and low-risk groups was rtcrlorrned usiing Cox
`scores.
`
`Indepenr.’en! Data Set’
`An independent data set consisting of‘ 85 patients treated at the
`Norvvcgiari Radium Hospital (NR1-I} was evaluated. All were patients
`with progressive hormone-refractory prostatic cancer with castrate
`levels of testosterone entered on seven in vestigationul programs be-
`tween l98'i and 1991. The patiet1t’s age. Eastern Cooperative On-
`cology Grotlp performance status. and baseline biochemical param-
`eters (ACP. ALK. CR. HUB. LDl-|_ AST. and PSA} were evaluated.
`The patient demographics were similar to those for the MSKCC series.
`Eighty patients (94%) had disease limited to_‘osscous sites with the
`extent of disease on the bone scan graded by the method reported
`by Solotvay el all” as previously described. Serial PSA values were
`again evaluated and posttherapy PSA declines of 2 50% and 2 30%
`|‘rom baseline at the landmark were used to classify a patient in a
`rcsponse or no response category. All patients were monitored until
`February I992. Some ofthe biochemical "parameters in the indepen-
`dent data set were measured in different units. In these cases. a cott-
`version factor was estimated on the basis of the upper limit of the
`normal range for both data sets. This conversion factor was Found
`to have minimal variability when comparing quaniles and. hence,
`minimal noise in the data from the conversions is expected
`
`RESULTS
`
`Prtricnl Den-togr'rtpi'rt'cs'
`
`One hundred nineteen patients were treated on seven
`
`invcstigational studies at MSKCC during the time period.
`Seven patients died before the landmark time of60 days
`and two were lost to follow-up, which left 110 assessable
`
`patients (30 deceased. 30 censored) for analysis. A sum-
`mary of the demographic and initial baseline parameters
`is listed in Table 1. Prior radiation therapy included none
`
`(36 cases). prostate (1-1). iodine I25-implantation (six).
`palliative to a metastatic focus (34). or a combination of
`local and palliative (20); prior surgery included no previous
`surgery (38 cases) versus previous surgery, which included
`radical prostatcctomy with pelvic lymphadenectomy (l0),
`transurethral resection (38). multiple urologic procedures
`
`Table I. Patient Charctcierisfies: MSKCC
`
`Demographics
`No. of pettiemts
`No. of tzissessctbls patients
`Median age, years
`Median KFS l'3t’:tl
`Median survival lrnonlhsl
`
`E-iocherniectl parameters
`Serum ACP lLllLl
`Alli: lUll.l
`Serum CR lntglclll
`HG8 lgielll
`LDH lU.t'Ll
`AST lUtlI.l
`PSA lnglml; TctndI.=.rn—E PEA,
`San Diego, CA}
`
`(21). eystectomy (one). or nonurologic surgery (two).
`Hormonal
`therapy included a luteinizing horrnone-re-
`leasing hormone (LI-I-RH) agonist i tlutamide (21 cases),
`diethylstilbestrol (10). or orchiectotny 1- llutamide (78),
`and other (one). The median time to progression following
`primary hormonal therapy was 14 months and overall
`median survival was ll.l months.
`
`Uitivorlrtte Atrrtl,t*.rl'_v
`
`The individual parameters were evaluated for possible
`correlations with survival by univariate analysis (Table
`2).
`
`DcJmogt‘rt;.tiiit's. As shown, an initial KPS of 2 80%
`was associated with a longer median survival (|3.2
`months) than a lower baseline KPS (median. 8.5 months;
`P = .005; Fig l). Age as a eategorie or continuous variable.
`treatment program, and initial pathology did not influence
`survival.
`
`Poor rretrrtmenr. The type and extent of prior radiation
`(P = .10). prior surgery (P = .61). and type of hormonal
`therapy (P = .58) did not have an impact on survival.
`The criteria for progression on primary hormonal therapy
`are not standardized and could only be assessed retro-
`
`spectively in 53 cases. In these patients, progression of
`disease was confirmed at MSKCCT. Nevertheless. the du-
`
`ration of response to primary treatment analyzed as a
`discrete (5 6 months 1’ > 6 months. P = .18; or 5 12
`months it > 12 months. P = .39) or as a continuous vari-
`able (P = .3) was not predictive.
`Ewen.’ o_,t"dt'sense. Extent of disease on bone scan as-
`sessed using the method reported by Soloway et at“ did
`not distinguish difl"erent prognostic categories (P = .08).
`However. by combining groups 0. l. and 2 versus groups
`3 and 4. a significant difference in survival (12.1 v 6.4
`
`WCK1025
`Page 3
`
`WCK1025
`Page 3
`
`

`
`610
`
`KELLY ET AL
`
`months, P = .02] was observed. Sixty percent of the pa-
`tienis had bidimensionally measurable disease. which was
`used for tumor assessment in four ofseven investigational
`studies. No specific site of metastasis was prognostic.
`Noteworthy was the observation that the survival distri-
`butions of patients with measurable and nonrneasurable
`disease were identical (P = .47: Fig 2).
`A Ll(_. HGB, and
`Baseliiie bt'0ciienri'cal' parnnretei's.
`LDH were significant in the univariate analysis when
`evaluated as categoric and continuous variables (Fig 3).
`but ACP was only significant as a continuous variable.
`The baseline PSA evaluated both as a categoric and
`a continuous variable had no impact on survival.
`The initial AST and CR were also not predictive of
`outcome.
`
`P(J.S'l’l‘l'l't"i‘ttp_l-’ vririables. There were 108 patients who
`had four or more serial PSA determinations 1 week or
`
`more apart. The survival distribution of the 93 patients
`who did not demonstrate a 2 50% decline in PSA at the
`
`landmark was compared with that ofthe I5 patients who
`did meet this criteria. Comparing the two groups. a sig-
`nificant difference. median 8.6 months versus not reached
`
`(P = .001), was observed (Fig 4). The median follow-up
`duration for the 15 patients with a 2 50% decline in PSA
`was 15.2 months (range. ll.0 to 24.6). The median num-
`ber of days required to achieve the 50% decline in PSA
`from baseline was 2? days (mean 1 SD. 28.7 4; 18.7’).
`Patients whose PSA declined by 2 50% maintained the
`reduction (calculated from the time patient met the criteria
`for decline to the time when the PSA has risen above the
`established decline level for two consecutive determina-
`
`tions”) for a median of 16? days (mean i SD, 212 at
`l6l.[).
`
`The demographics between the two groups were com-
`pared to assess for balances in known prognostic factors.
`The median age. KPS. baseline biochemical parameters,
`and time to progression on primary hormonal therapy
`were not significantly different between the 50% decline
`versus no 50% decline groups. Extent of disease on bone
`scan and the proportion of patients with soil tissue me-
`tastases were similar in the two groups. Only four patients
`met the criteria ofa 2 80% decline. and therefore no sur-
`vival comparisons were undertaken.
`An additional eight patients achieved a 2 50% decline
`in PSA from baseline after the landmark time of60 days.
`The mean time to a 2 50% decline in PSA was I 15 days
`(SD. 39.? days) with a median survival of 20.8 months
`
`(median follow-up, 18 months") for this cotton ofpatients
`with delayed response.
`
`A-2"iilrivei'i'azc A nah-=.t'i'.s
`
`The univariate analysis identified six covariates as pre-
`dictive ofsurvival. In order ofsignificance these included:
`
`Table 2. Results of Univariute Analysis
`Median
`Survival
`lmonlhsl
`
`'-lb
`
`Age. years
`:> 65
`s 65
`KP5
`2 30%
`-: 80%
`Extent of disease
`Siles of rnelostosis
`Bone only
`Bone and soft tissue
`Liver
`lung
`RPLN
`
`Epidural
`Other nonbone
`Bone scans
`
`Solowcly scale
`C‘ [norrnull
`I {< L‘: hot spotsl
`2 ts—2o hot spotsl
`312 20 hot spolsl
`4 [superscon]
`Pretherapy biochemical
`parameters
`ACP S 0.6 UH
`ACP 2 0.6 Uil
`ALK 5 l l5 UEL
`ALK2 ll5UilL
`CR -1. 1.3 rngfdl
`CR2 'l.3 rngtdl.
`A51 2 3?
`
`Ag-T 5 37
`HGB .2 I3 gicll.
`HGB < I3 gildl
`{DH 5 230 UIL
`LDH 2 230 UIL
`initial PSA 2 4 ngi'n1l.
`Initial PSA 5 4 ngil-11L
`Initial PSA 2 l00 ng!mL
`Initial PSA 5 l0{} ngilrnt
`Postlherolay PSA decline
`-3 50% decline
`No 2 50% decline
`E 50% decline
`No 2 80% decline
`
`46
`54
`
`55
`45
`
`8.6
`I33
`
`l3.1'
`8.5
`
`l6
`94
`35
`.75
`98
`l2
`2i
`
`39
`2?
`S3
`65
`45
`
`3
`64
`46
`I03
`l5
`93
`4
`I04
`
`l5
`35
`32
`63
`3?
`ll
`l9
`
`8|
`25
`.75
`59
`41
`9‘?
`3
`58
`42
`
`l4
`86
`4
`‘Q6
`
`NR
`B.t5
`NR
`9.9
`
`'13
`
`Abbreviations: RPLN. retroperitoneul lymph node; NR, median survival
`not reached.
`°Leg—ronls lest.
`I'Uni\roriate Cox analysis of the associated continuous variables.
`
`|n(ACP), KPS.
`ln{'LDH),
`a 2 50% decline in PSA.
`ln(ALI()_. and ln(HGB). As performance status is a sub-
`jective parameter that may not be reproducible across
`multiple investigators. we chose to develop a prognostic
`model excluding this factor. The remaining variables were
`used in a multivariate Cox regression. and identified 2
`
`WCK1025
`Page 4
`
`WCK1025
`Page 4
`
`

`
`PSA AND PROSTATE CANCER
`
`(ill
`
`.50
`
`I
`
`.40
`
`
`
`proportionsurviving
`
`l32.00
`
`30.00
`
`24.00
`15.00
`time (months)
`
`Fig 1. Overall survival with initial KP5 < 30% [El] v KPS 2 30%
`[Ol-
`
`50% decline in PSA and the natural log ofthe initial LDH
`
`as the two most significant factors based on relative risk
`ratios {Table 3). A patient‘s relative risk of death compaied
`with that of an average patient may be described by the
`following multivariate model:
`
`In [RR] = 2.1 Deel 50 + 1.09 [ln(LDH) — 5.45]
`
`where RR is the ratio of death rates. and Decl 50 = 0 if
`
`the patient obtains a .2 50% reduction in the PSA from
`the baseline as defined previously, or 1 if no 2 50% decline
`was obtained. The model indicates that the death rate is
`eight times higher For the group who did not achieve a
`posttherapy decline of 50% or more in PSAI In addition,
`the death rate increases with the higher value of‘ initial
`LDH.
`'
`The Cox score calculated from the above muitivarlate
`
`proportional hazards model was used to form low- and
`high-risk categories. The low- (Cox score -5. 2.0) and high-
`{Cox score 2 2.0) risk groups had median survival rates
`of l'r'.l months and 3.1 months, respectively. which were
`
`significantly different (P = .0001: Fig 5).
`
`l’nn’epeira'cnI Dara Set
`
`Of the 92 patients treated at the. NRH. six died before
`the initial 60-day landmark time and one was lost to fol-
`low-up, which left 85 patients assessable for analysis. The
`median age was 70 years (range, 46 to 81). and the median
`Eastern Cooperative Oncology Group performance status
`was 1 (range, 0 to 3). The demographic, pretherapy. and
`biochemical parameters were similar to those of the
`MSKCC series (Table 1). Overall, 53 patients have died,
`with a median survival of 10.12 months (range, 2.0 to
`
`46.0}. which is similar to the MSKCC series (Fig 6}. The
`median follow-up was 7.4 months.
`
`Fitting the MSKCC model to the independent data set
`demonstrated a threefold increase in the death rate (95%
`
`confidence interval [CI]. 1.2 to 7.5) for those patients who
`did not achieve a posttherapy PSA decline of 50% or
`greater. Similarly.
`the death rate increased with higher
`values of initial LDH (95% CI. 1.3 to 6.4). Applying the
`MSKCC multivariate proportional hazards model
`for
`survival time to the independent data set, the model was
`able to classify the patients into two distinct risk categories.
`The median survivals for the low (10.8 months) and high-
`
`(8.5 months) risk groups were found to be significantly
`different (P = .01).
`
`DISCUSSION
`
`This study extends the evaluation of prognostic factors
`for patients with hormone-refractory prostatic cancer to
`include posttherapy declines in a measured biochemical
`parameter. PSA. In the univariate and multivariate anal-
`ysis. a posttherapy decline in PSA of 2 50%. which was
`documented on multiple determinations and maintained
`over time, was the most significant variable. A landmark
`method“ was used to avoid difiieulties in comparing re-
`
`sponders versus nonresponders,3“‘“ since such a compar-
`ison has inherent systemic bias. Ofequal importance was
`the observation that the survival distributions for patients
`with measurable and assessable (bone only) disease were
`
`comparable. Previously, the ibrrner group was believed
`to have an inherently worse prognosis.‘''‘3 The high pro-
`portion of patients with measurable disease in this series
`reflected the eligibility requirements of the individual
`
`
`
`proportionsurviving
`
`.
`
`24.00
`13.00
`12.00
`time (months)
`
`30.00
`
`. Ova roll survival {or nonrneu surcrble [O1 v rneusurn ble disease
`
`WCK1025
`Page 5
`
`WCK1025
`Page 5
`
`

`
`.50
`
`
`
`proportionsurviving .40
`
`KELLY ET AL
`
`B
`
`
`
`proportionsurviving
`
`.60
`
`.2a,-to
`
`21.50
`1E.UU
`12.00
`time (rnon ths)
`
`$0.00
`
`56.00
`
`24.00
`15.00
`12.00
`time {months}
`
`30.00
`
`30.00
`
`C
`
`Fig 3. Overall survival for {A} initial Alli’. 5 I15 U_!L ID} V ALK >
`I15 U,r‘L IO}; [E] iniliol HGB 2 I3 g,-‘ell. {O} V HGB < 13 g;‘clL {El}: [Cl
`initial LDH 5 230 Ufl {DI v LDH '> 230 Ufl. [0]. I593 ‘Fable I for
`significance levels}
`
`median interval between the documentation of relapse
`based on PSA and clinical symptoms. Thus. once eleva-
`tions in PSA are documented. the natural history ofthe
`disease ‘has been determined, and alternate treatment(s)
`should be considered.
`
`Other investigators have reported the prognostic sig-
`nificance of a pretreatment performance status
`:2
`80%. "”“‘35--‘3 However, as performance status is subjective
`and diflicult to assess reproducibly among multiple in-
`vestigators.35'“ we opted to delete this factor from our
`prognostic model. Baseline biochemical parameters have
`been previously evaluated.
`In univariate analyses. pre-
`treatment ALK.“3'-“ and LDH have shown prognostic
`importance? These observations were confirmed in the
`present study. It was of interest that PSA. considered as
`a eategoric or continuous variable. did not show prognostic
`significance. In fact, three oftlie responding patients had
`initial PSA values of 2.245, 3.740. and 4,000 ng/mL. re-
`
`WCK1025
`Page 6
`
`.60
`
`.40
`
`.20
`
`U‘C
`>
`'>L
`Jm
`ED
`‘ZL
`L0L
`D.
`
`0C
`
`13.00
`t irns
`
`24.00
`13.00
`(months)
`
`30.00
`
`35.00
`
`protocols. The comparable median survival times ob-
`served in the MSKCC and the independent data set from
`Norway, with less than [(1% of patients with measurable
`disease. shows the similarities between the hormone-re-
`fractory populations.
`The median survival rate was higher than in our pre-
`vious report on a hormone-refractory group. 6.? months.”
`This is due to multiple factors. First. the analysis excluded
`the poor-risk patients who died before 60 days. If these
`deaths were included. median survival rates of 9.’! months
`
`For the MSKCC‘ and 9.1 months for the Norwegian series
`are obtained. In addition, patient‘s participation in clinical
`trials are usually dependent on their ability to commute
`to treatment centers and to fit minimum performance
`criteria. Last. there is earlier documentation and initiation
`
`oftreatment ofhormone-relapsed patients based on PSA
`elevations as opposed to waiting for clinical or radiologic
`progression. Recently, Newling et al"' reported a 3~month
`
`WCK1025
`Page 6
`
`

`
`PSA AND PROSTATE CANCER
`
`.69
`
`I
`
`
`
`proportionsurviving.%O.40
`
`
`
`proportionsurviving
`
`.60
`
`.20.40
`
`24.ou
`1a.oe
`I2.:m
`t irne (months)
`
`amen
`
`"
`
`amen
`
`24.00
`ifi.00
`12.00
`time (men the}
`
`30.00
`
`30.00
`
`Fig 4. Overall survival for patients with a 2 50% decline in PSA
`{O} v no 2 50% decline in PSA {D}.
`
`Fig .5. Overall survival tar patients in the low—rislt 1
`group (OI.
`
`} v high-risk
`
`spectively. Similar results were observed by Fossa et at.“
`in contrast, Myers et al” used a cutoff of less than 100
`ng,-‘mL versus greater than 100 ng,a'mL. and found a sig-
`nificant correlation with survival.
`The correlation between the pretreatment ALK. and extent
`ofdisease on bone scan is not direct. as a variety ofchanges
`
`in bone may occur. This is probably clue to the Fact that
`serum ALK may increase during disease regression, or as a
`consequence of bone healing during effective therapy. Fur-
`thermore, increased uptake on bone scan may represent a
`nonnec-plastic or bone-repairing process. Therefore. serum
`ALK and bone scan. at best, are crude discriminators for
`prognostic groups. However. we did observe a difference in
`survival between patients with less versus more extensive
`disease on bone scan using grouped categories '0. 1. and 2
`( 12.1 months) versus 3 and 4 (6.4 months] as noted by So-
`loway et at.“ The latter may therefore be considered a strat-
`ification variable in phase III trials.
`Previously. others have reported that postthcrapy
`changes in ACP can be of prognostic importance." How-
`ever. this parameter is subject to large intrapatient vari-
`ability independent of treatment, and has been largely
`replaced with PSA. We, and others. are investigating
`
`posttherapy declines in PSA as the outcome measure.‘°'”‘33
`However, the criteria for response needs to be standardized
`and validated prospectively. While the criteria for response
`are evolving, we initially proposed a 2 30% decrease from
`the baseline value for a minimum of three biweekly de-
`terminations to classify a patient in a response category.
`This was based on the observed correlation between ob-
`
`jective tumor regression and PSA decline in patients with
`hornt one-naive disease.” A similar outcome measure was
`
`reported by Myers et at“; however, no comment was made
`regarding the duration of the decline. or the minimum
`number of determinations documenting the decline that
`
`
`
`proportionsurviving
`
`.60
`
`.20.40
`
`Table 3. Multivariate Proparlioruai Hazards Model
`for Survival Time
`Estimated
`Reg: ass i on
`Caeltlcierir
`
`Variable
`
`-50% decline in PSA'
`In llDHl
`
`2. l 0
`L09
`
`Role Ratio (95% Eli
`8.? (2.6 to 2£i.0i
`3.0ll.7to 5.ll
`
`F‘
`.0004
`.0001
`
`I
`32.00
`24.00
`18.00
`time (months)
`
`r—'
`40.00
`
`I
`43.00
`
`‘Dad 50 = 0 if the patient achieves a 2 50% reduction in the PSA from
`the baseline, or 1
`it no 2 50% decline was observed.
`
`Fig 6. Overall survival {or patients treated at MSKCC [Cl] v NRH
`[Q].
`
`WCK1025
`Page 7
`
`WCK1025
`Page 7
`
`

`
`614
`
`KELLY ET AL
`
`were required to classify a patient in a response category.
`Furthermore. potential imbalances in prognostic param-
`eters between the responding and nonresponding groups
`were not considered. nor was an independent data set
`evaluated. The requirement For a number of determina-
`tions is essential. as transient decreases or increases in
`
`PSA can be observed. It is unlikely that the PSA changes
`observed in the present study were transient. as patients
`had to demonstrate multiple sequential PSA elevations
`before initiating therapy. and posttherapy declines in PSA
`had to be maintained for prolonged periods and docu-
`mented by multiple determinations.
`The 2-month interval used for the landmark analysis
`was preselected on the basis ofthe time usually chosen to
`reevaluate the effect oftreatment. However. only 65% ( l 5
`of 23) of the patients who had a at 50% decline in PSA
`achieved the decline by 60 days. Eight patients (35%) had
`a delayed response and would not have been classified as
`responders if the analysis were restricted to the 60-day
`landmark. The median time to a 2 50% decline for the
`
`23 patients who achieved it was 49 days