- Email: [email protected]
Prognostic Value of the Tissue Polypeptide Antigen in Lung Cancer
Prognostic Value of the Tissue Polypeptide Antigen in Lung Cancer* Gianfranco Buccheri, M.D., F.C.C.R; antI-Domenico Ferrigno, M.D.
The prognostic impact of TPA was evaluated by assaying the marker in the serum of 563 patients with a newly diagnosed bronchogenic carcinoma. The group included patients with squamous cell cancers and others with tumors of diverse or unde6ned histologies. Raised levels of TPA were clearly associated with a shortened survival, even adjusting for the stage of disease. A Cox's proportional hazards regression analysis, incorporating all major prognostic factors, selected TPA as an independent survival predictor. In the regression model, however, TPA was less important than disease extent, KPS and weight loss. Another
issue polypeptide antigen is a single-chain polyT peptide which has been isolated from cell mem-
branes and smooth endoplasmic reticulin of malignant
eells.P Increased concentrations of the antigen are
found in a variety oftumors, 3-5 including lung cancer.6-10 Serum determinations of TPA appear to be useful in all lung cancer histotypes, probably more useful than the conventional CEA assays," In these tumors, TPA can be of help in assisting in the diagnosis, 9 in evaluating the extent of disease.s" in monitoring the clinical evolution.s" and in predicting the survival probability of affected patients.6-8,10 However, the prognostic significance of the marker has been studied by univariate methods alone,6-8 not taking into account the other prognostically important variables. 6 •7 •1o This latter evaluation, in particular, would be required by the clear correlation existing between TPA and the tumor burden.s" Another interesting issue, not yet addressed, concerns the relative merit ofTPA in the different histotypes oflung cancer: prior studies, in fact, included tumors ofany cell type, considered as a unique pathologic entity6"8 or evaluated only the subgroup with small cell caneer.!" We designed this study to assess TPA with reference to other major prognostic factors, both in the general population with lung cancer and within subgroups with different histotypes of the tumor. PATIENTS AND METHODS ltJtients
Between 1984 and 1990, 563 new patients with cytologically or ·From the Divisions of Pneumology, A. Carle Hospital of Chest Diseases, Cuneo, Italy. Presented at the 57th Annual Scientific Assembly, American College of Chest Physicians, San Francisco, November 4-8, 1991. Manuscript received June 14; revision_8CCeRted September 11. &print requests: Dr. Buccheri, Via liepUbblic4 101C, RoccavWne
(CN), Italy 112018
multivariate analysis was made in a subgroup of 121 patients who, because of poor KPS or advanced age, had undergone a limited staging worlcup; TPA came out as the Grst most important factor. This study shows that .TPA is an important prognostic factor and that it should be included among laboratory data evaluated in lung cancer studies. (Chat 1992; 101:1187-92)
= =
= =
BMDP biomedical package; CEA carcinoembryonic antigen; CI con8dence intervalJ US Kamo£sky performance status; TPA = tissue polypeptide antigen
I
histopathologically proved bronchogenic carcinoma were seen at the Antonio Carle Hospital and assayed for TPA before starting treatment. Males constituted 89 percent (504 of 563) of the cohort studied. The median age was 64 years (range: 31 to 88 years). Histopathologic diagnosis was carried out in accordance with the revised WHO classificationof lung tumors" by the same pathologist. Pathologic diagnoses comprised the following: 319 squamous cell cancers, 67 small and 36 large cell anaplastic tumors, 55 adenocarcinomas. Eighty-six patients bad tumors with unde6ned or mixed bistol~ Routine pretreabnent staging procedures included: physical examination, blood chemistry and hematologic counts, chest radiographs and tomograms, bronchoscopy Computed tomography of the thorax was performed in all patients evaluated for surgery or irradiation and in many others with good performance status, not too advanced age, and in general with a fair physical and psychic compliance (in all, 429 patients). The same subjects were also tested with a total body 87Ga scan or with a combination of thorax, abdomen and brain (total body) CT scanning. During the last three years, an investigational procedure of total body scintigraphy with anti-eEA monoclonalantibody fragments was added to the baseline evaluation of the consenting patients who were candidates for surgery or irradiation. Ultrasound and radioisotope scans, along with other imaging tests and biopsies, were performed as clinically indicated. Only selected surgical candidates underwent mediastinoscopy Patients were classified using the new International Union Against Cancer (UICC) staging system. u Accordingly; 57 patients were ~1assified into stage I disease, 53 into stage II, 138 into stage IlIa and 109 into stage IIIb, 203 into stage N Three patients were not staged because of insufficient data. Follow-up consisted of clinical and radiographic examinations performed monthly (or more frequently) during the treatment, every three months during the following two years, and subsequently every six months. Very few patients abandoned their outpatient visit program; when they did, the physical status was obtained through a telephone interview with either the patient or the referring physician. In order to assess the relative value ofTPA, multiple factors were analyzed in conjunction with the marker. Included in multivariate analyses were: sex, age, the presence of weight loss in the six months preceding the diagnosis, the performance status according to the Kpsl3 and various parameters of disease extent, such as the stage of disease, the T and N factors, and number and site, if any, of metastases. Table 1 summarizes the variables studied, the number of observations and the median value and range (or the frequency CHEST I 101 I 5 I MA'f, 1992
1287
Table 1- Patient ChartlCferiBtica Variable·
No. Observations
Median
563 563
64
31-88
80
20-100
Sex (malelfemale) Age (yr) Weight loss (y/n)t KPS Stage of disease (I/II/IIIalIllblIV) T factor (I/II/IIIIIV) N factor (OII/II/III) M factor (OII/II/III/Iv)t Lung metastases (yIn) Brain metastases (yIn) Liver metastases (yIn) Adrenal metastases (yIn) Bone metastases (yIn) Histology (E/SIAIUU)§ TPA(UIL)
539
557 560 560 560 560 560 560 560 560 560 563 563
Range
Frequency 504159
2141325 57~13811091203
30-5203
130
43/229/1:n'155 126197/249/88 357/1551361913 65/495 471513 361524 481512 551505 319/67155136186
·Numeric codes for the subsequent multivariate analyses as in the Table, except for: male = I, female=O; y= I, n=O; stage 3a=3, stage 3b=4, stage 4=5; E,S,A,L,U histology (y/ri) = 1/0. t During the six months preceding diagnosis. tExpressed as number of metastatic sites. §E: epidermoid carcinoma; S: small cell carcinoma; A: adenocarcinoma; L: large cell carcinoma; U: unclassified. of distribution) of each variable. Bio-marker Assays
The TPA measurements were performed by radioimmunoassay on serum samples stored at - 2
on the survival duration was estimated using Cox's proportional hazards regression model." Factors having significant influence on the model were included in a stepwise fashion. The significance of each factor was calculated by the maximum likelihood ratio. A p value of 0.1 was set to enter, while a value of 0.15 was set to remove a variable from the model. The Cox's proportional hazards model is given by the formula:
Statistical Analysis
h(t;~)
Survival duration was measured for all patients from the time of pathologic diagnosis. Categorization of TPA for univariate analyses was accomplished using, as a cutoff point, the median of its distribution (130 U/L). Univariate survival analyses were based on the Kaplan-Meier product-limit estimates of the survival distribution. No deaths were censored in the analysis. Differences between survival curves were tested statistically using the Mantel-Cox logrank test and the Breslow-Gehan test, offered by the BMDP program." The relative importance of multiple prognostic factors
= h, (t)exp(~'~)
where h(t;z) is the hazard rate of an individual patient with covariate vector ~; ~ is a vector of the unknown regression coefficients, and h, (t) is the unknown hazard function for an individual with covariate vector z equal to o. A p ';alue of less than 0.05 was regarded as statistically significant. All tests were two-sided. Data were processed with the BMDP statistical software produced by Health Science Computing Facility, University of California at Los Angeles."
1 DEAD TPAsl30 Uti TPA>l30 Uti
P
~.75
Mantel-Cox Test BreslowTest
B A B .5 I L I
208
CENSORED 84
211
60
STATISTIC
P-VALUE
27.944 33.645
0.0(0) 0.0(0)
T.25
Y
o ... o 1288
...._ _....
..._ _......, .-_ _...
12
36
24
48
60
SURVIVAL (months)
..._ _..... FIGURE 1. Actuarial survival curves of the entire population (heavy solid Une: patients with TPA serum levels up to 130 UIL; dashed line: patients with TPA levels greater than 130 U/L).
72
84
Tissue Polypeptide Antigen in LungCancer (Buccheri,Ferrigno)
1PASl30~
1PA>130 /I
.75
ManIeI-Cac 'felt
hIIowfelt
DEAD
CENSORED
15 18
15 7
STA11S1lC
P-VAUE
1Q.212 9A03
0.0014 0.0022
75
R .25 0
25
0
A
0
24
12
36
A~NOCARCINOMA
B I
I
y
ManIeI-Cac felt
hIIowTeit
72
60 OEAD
CENSORED
36
9
STA11S1IC
P-VAUE
0.114
0.7359 0.1_
lPAsl30U/I lPA>l30U/I
L .75 T
48
15
1.732
7
37 36
stA11S11C
P-VALUE
12.229 16.210
cum1
hIIow'felt
.5
B
CENSOREO
121 126
MarII-...eacTeIt
.5
P
OEAD
lPAsl30U/I '1PA>l30U/I
0
84
0
·10
20 30 40 50 60 70 SQUAMOUS CELL CARCINOMA 1PASl30~
1PA>130 ill
75
.5
.5
.25
25
.. '=....
0
8
16 24 32 40 48
56 SMALL CELL CARCINOMA
64
72
0
~
iii••••••:
................ ··.....: ~
0
3
CENSOREO
14 13
7 2
1.190 1.226
8IlIIIcw 1eIt
80
DEAD
STA11S1lC
MarII-...eac'felt
~
0
o.cxD
~
Q.27U Q.26I3
.....•... ····5.
6 9 12 15 18 21 LARGE CELL CARCINOMA
24 27
SURVIVAL (months)
FIGURE 2. Actuarial survival curves of histologically homogeneous subgroups (heavy solid line: patients with TPA serum levels up to 130 UIL; dashedline: patients with TPA levels greater than 130 UIL). RESULTS
Univariate Analyses
As shown in Table 1, for the whole population the median value ofTPA was 130 UIL (range: 30 to 5,203 UIL). Patients with TPA up to 130 UIL survived much longer (median survival time: 10.41 months; CI, 8.47 Table I-Facton l ~
to 12.35) than patients with higher TPA values (median: 5.57; CI, 4.84 to 1). As shown in Figure 1, this difference was highly significant (p = 0.ססoo). We split the series in subgroups homogeneous as far as the extension of disease is concerned (that is, patients with stages I and II, and stage IlIa, IIIb and IV were
Inftuencing SuroiwJl: BaultI oftlte Multifactorial AntJlrIaia 1· Standard Error
Improvement ofchisquare*
p Value§
0.2162
0.0549
75.807
0.000
2 3
-0.0273 0.2570
0.0040 0.0946
56.759 11.106
0.000 0.001
4
0.2689
0.1072
7.702
0.006
5
0.0003
0.0001
6.045
0.014
Variables
Stept
Stage of disease (I1IIIIIIalIIIblIV) KPS M factor (MIIIIIIIIIV) W,ight loss (yIn) TPA(UIL)
1
Coefficient
*Five hundred thirty-one patients had no missing values and were included in the model. tStep at which the variable was included in the model. *Global chi square = 185.456. IGlobal p=O.()()()(). CHEST 1101 1 5 1 MA~
1992
1288
Table 3-Facton Independently Influencing Suroiool: Besulta ofthe MultifGCtorial Analysia 2· Variables
Stept
Coefficient
Standard Error
Improvement of chi squarer
p Value§
TPA(UIL) Stage of disease (I1IIIIIIalIIIblIV) Age (yr)
1 2
-0.0013 0.3886
0.0003 0.0937
17.956 11.666
O.()OO 0.001
3
-0.0315
0.0109
8.735
0.003
*Limited to 121 patients with a poor staging workup and no missing data. tStep at which the variable was included in the model. iGlobaI chi square =43.116. §Global p = 0.ססOO.
considered separately). In any subdivision, the same trend, observed in the entire population, was recognized. The tendency was statistically significant in the more advanced stages (p<.005 for stage IlIa and I~ by the Breslow test). In Figure 2, the TPA-survival relationship is analyzed as a function of the histotypes. This analysis is rather hindered by the severe imbalance among cell types, but again an identical trend is observed in all subgroups. The marker was best correlated with survival of adenocarcinoma patients (median survivals: 28 and 4 months for low and high TPA values, respectively).
Multivariate Analyses A first multivariate analysis, incorporating all the potentially significant prognostic factors described in Table 1, confirmed not only the independence ofTPA from tumor burden and histology (which could be guessed from the results of the univariate analyses), but also its independence from age, gender, weight loss and the performance status. However, TPA was selected as the less important independent factor (Table 2). Staging a patient is a time- and resource-demanding
process, which is not ever essential for therapeutic decisions and sometimes impossible to perform with sufficient confidence. In our series, for example, 121 patients were staged on the basis of thoracic radiographs, bronchoscopy and clinical examination, but without thoracic cr and a total body scintigraphy In this subgroup of patients, given the already known correlation existing between TPA and the stage of disease.s" one might anticipate an increased weight of the marker in proportion to the minor importance of the unreliable staging evaluation. This is exactly what we found to happen in a second multivariate analysis limited to the aforementioned patients (Table 3). Finally, multivariate analyses were run on histologically homogeneous subgroups, using the same covariates of the previous analyses (Table 4). Again, TPA emerged as a significant independent factor, and was the most important variable in both adenocarcinomas and anaplastic tumors. DISCUSSION
As a marker of neoplastic diseases, TPA has never been as well recognized as CEA.16.17 Nevertheless, it was discovered and characterized in the late 195081.2 -
Table 4-Facton Independently Influencing Survival in the DijJenmt Hiatoypea ofwng Cancer Variables
Step*
Coefficient
Squamous cell cancer (cases: 300; global chi square: 136; p=O.OOOO): -0.0325 KPS 1 Stage of disease 2 0.2543 (IIII/I IIalII IbIIV) TPA(UIL) 0.0007 3 Weight loss (yIn) 4 0.3564 Bone metastases (yIn) 0.6726 5 N factor (OII/IIIIII) 0.1720 .6 Small cell cancer (cases: 62; global chi square: 13.83; p = 0.(010): TPA (UIL) 1 0.0026 Weight loss (yIn) 0.5352 2 Large cell cancer (cases: 34; global chi square: 10.25; p = 0.0059): TPA(UIL) 0.0017 1 Weight loss (yIn) 0.9391 2 Adenocarcinoma (cases: 52; global chi square: 30.33; p=O.OOOO): TPA (UIL) 1 0.0018 Stage of disease 2 0.3008 (OIIIII/IIIalIIIbIIV)
Standard Error
Improvement of chi square
p Value
0.0052 0.0643
51.661 41.000
0.000 0.000
0.0003 0.1416 0.2527 0.0834
11.461 5.767 5.760 4.357
0.001 0.016 0.016 0.037
0.0009 0.2993
7.051 3.159
0.008 0.076
0.0007 0.5211
3.626 3.077
0.057 0.079
0.0006 0.1356
12.837 5.221
0.000 0.022
*Step at which the variable was included in the model.
1290
TISSue Polypeptide Antigen in Lung cancer (Buccheri, Ferrigno)
before the original CEA description's-e-and proposed as a nonspecific marker for various human tumors." likewise CEA.?Jl In many clinical applications, TPA has been reported to be superior to several other tumor markers, including CEA. 7 In lung cancer patients, the pretreatment serum levels ofTPA have been shown to correlate positively with either the primary tumor characteristics (defined as Tl, T2 and T3), or the nodal involvement (as NO, N 1 and N2), or the metastatic spread (as MO and Ml) or the classification into four stages of disease.v" In general, the greater the tumor bulk at diagnosis, the more elevated is the serum concentration of the marker," The same studies have indicated that posttreatment TPA may vary in accordance with the obvious changes in disease statusv? and may sometimes precede them." TIssue polypeptide antigen has been proposed also as a tool for predicting the course of disease. 6-8,10 Actually, the prognostic capability ofTPA may be anticipated by the knowledge of the molecule metabolism. The substance is synthesized during the S- to M-phase of the cell cycle and released upon proliferation into the blood stream." Thus, the concentration of the antigen is an indicator of the rate of cell division and tumor aggressiveness, and, therefore, of the host survival. This study was primarily conducted to address a not yet fully clarified issue, rather than simply to confirm the prognostic capability of TPA, which we believe to be already well established. In a previous study, reported in 1988 in this journal," we found that cohorts of patients having the same stage of disease tended to be divided into poor- and good-risk subgroups on the basis of their pretreatment TPA value, but the tendency did not reach the statistical level of significance. Emphasizing the negativity of the statistic and the confirmed correlation of the marker with the stage of disease, we supposed that TPA could directly reflect tumor load-rather than the cell proliferation rate - and only indirectly could indicate the prognosis. 8 This is not the case, however, since our current data prove by both univariate and multivariate methods the independence of TPA from the stage of disease. Indeed, TPA surpassed the prognostic value of the same stage when this was obtained by a limited staging workup. A second aim of the study was to evaluate in greater depth the prognostic significance of the marker, particularly as far as the relationship between TPA and the prognosis of the diverse cell types is concerned. In this respect, we have elucidated that TPA is a greater predictor of the outcome in adenocarcinomas and in anaplastic cancers. This is as one might expect on the basis of the marker metabolism" and of the knowledge of the proliferation rate of each cell type of lung cancer. 22 The identification of prognostic factors is useful in
designing and comparing clinical trials and in advising individual patients. The efforts to define a set of prognostic determinants important for lung cancer have continued during the last decade and have included anthropometric and clinical characteristics, pathologic aspects, in addition to a large number of laboratory parameters.P On the basis of the current findings, we would suggest including TPA in any future multifactorial analysis of survival, especially in the non-squamous cell histologic variants of lung cancer. ACKNOWLEDGMENTS: The authors wish to thank Mr. James Beauchamp for linguistic editing and Mr. Francesco Bontadi for his graphic aid. REFERENCES
1 Bjorklund B, Bjorklund ~ Antigenicity of pooled human malignant and normal tissue by cyto-immunologica1 technique: presence of an insoluble heat-labile tumor antigen. Int Arch Allergy 1957; 10:153-83 2 Bjorklund B, Lundblad G, Bjorklund ~ Antigenicity of pooled human malignant and normal tissues by cyto-immunologica1 technique: II. nature of tumor antigen. Int Arch Allergy 1958; 12:241-61 3 Nemoto T, Constantine R, Chu TM. Human tissue polypeptide antigen in breast cancer. J Natl Cancer Inst 1979;63:1247-1350 4 Adolphs HD, Oher ~ Significance of plasma tissue polypeptide antigen determination for the diagnosis and follow-upof uretelial bladder cancer. Urol Res 1984; 12:125-28 5 Kew MC, Berger EL. The value of serum concentrations of tissue polyptide antigen in the diagnosis of hepatocellular carcinoma. Cancer 1986;58:868-72 6 Buccheri GF, Violante B, Sartoris AM, Ferrigno D, Curcio A, Vola F. Clinical value of a multiple biomarker assay in patients with bronchogenic carcinoma. Cancer 1986; 57:2389-96 7 Buccheri GF, Ferrigno D, Sartoris AM, Violante B, Vola F, Curcio A. Tumor markers in bronchogenic carcinoma. Superiority of tissue polypeptide antigen to carcinoembryonic antigen and carbohydrate antigenic detenninant 19-9. Cancer 1987; 60:42-50 8 Buccheri GF, Ferrigno D. Usefulness of tissue polypetide antigen in staging, monitoring, and prognosis of lung cancer. Chest 1988;93:565-70 9 Mizushima Y, Hirata H, Izumi S, Hoshino K, Konishi K, et ale Clinical significance of the number of positive tumor markers in assisting the diagnosis of lung cancer with multiple tumor marker assay. Oncology 1990; 47:43-48 10 Granowitz JS, Bergstrom R, Nou E, Pahlman S, Brodin 0, Nilsson S, et ale Clinical and serologic markers of stage and prognosis in small cell lung cancer: a multivariate analysis. Cancer 1990;66:722-32 11 World Health Organization. Histologic typing of lung tumours. Am J Clin Pathol 1982; 77:123-36 12 De Angelis G, Cipri A, Flore F, Munno R. UICC: TNM classification of malignant tumours. ed 4. Geneva, Switzerland: International Union Against Cancer, 1987 13 Karnofsky DA, Burchenal JH. The clinical evaluation of combination chemotherapy in cancer. In: Mac Leod CM, ed. Evaluation of chemotherapeutic agents. New York: Columbia University Press, 1949:199-205 14 BMDI! Statistical software manual. Berkeley, Cal: University of California Press, 1988 15 Cox DR. Regression models and life tables (with discussion). J R Stat (series B) 1972;34:187-220 16 Schein PS. Tumor markers. In: Wyngaarden JB, Smith LH, eds. Text book of medicine. 16th ed. Philadelphia: WB Saunders ce., 1982:1020-22 CHEST I 101 I 5 I MA~
1992
1281
17 Broder LE, Primack A. Marker substances in bronchogenic carcinoma. In: Straus MJ, ed. Lung cancer: clinical diagnosis and treatment. 2nd ed. New York: Grune &: Stratton Inc, 1983:37-61 antigens 18 Gold ~ Freedman SO. Demonstration of tumo~sPecific in human colonic carcinomata by immunological tolerance and absorption techniques. J Exp Med 1965; 121:439-62 19 Bjorklund B. 00 the nature and clinical use of tissue polypetide antigen. Tumor Diagnostik 1980; 1:9-20 20 NHI. Carcinoembryonic antigen: its role as a marker in the
management of cancer. Br Moo J 1981; 282:373-75 21 Bjorklund B, Bjorklund ~ Specificity and basis of the tissue polypetide antigen. Cancer Detect Prevent 1983; 6:41-50 22 Straus MJ, Moran RE, Shackney SE. Growth characteristics or lung cancer. In: Straus MJ, ed. Lung cancer: clinical diagnosis and treatment. 2nd ed. New York: Grune &: Stratton Inc, 1983:21-35 23 George SL. Identification and assessment of prognostic factors. Semin Oocoll988; 15:462-71
Plan to Attend ACCP's
58th Annual SCientific Assembly Chicago --~ OCtober 25-29, 1992
cA ~.---.......~=
1282
TIuue Polypeptide AntIgenin Lung Cancer(BuccherI, FerrIgno)
The prognostic impact of TPA was evaluated by assaying the marker in the serum of 563 patients with a newly diagnosed bronchogenic carcinoma. The group included patients with squamous cell cancers and others with tumors of diverse or unde6ned histologies. Raised levels of TPA were clearly associated with a shortened survival, even adjusting for the stage of disease. A Cox's proportional hazards regression analysis, incorporating all major prognostic factors, selected TPA as an independent survival predictor. In the regression model, however, TPA was less important than disease extent, KPS and weight loss. Another
issue polypeptide antigen is a single-chain polyT peptide which has been isolated from cell mem-
branes and smooth endoplasmic reticulin of malignant
eells.P Increased concentrations of the antigen are
found in a variety oftumors, 3-5 including lung cancer.6-10 Serum determinations of TPA appear to be useful in all lung cancer histotypes, probably more useful than the conventional CEA assays," In these tumors, TPA can be of help in assisting in the diagnosis, 9 in evaluating the extent of disease.s" in monitoring the clinical evolution.s" and in predicting the survival probability of affected patients.6-8,10 However, the prognostic significance of the marker has been studied by univariate methods alone,6-8 not taking into account the other prognostically important variables. 6 •7 •1o This latter evaluation, in particular, would be required by the clear correlation existing between TPA and the tumor burden.s" Another interesting issue, not yet addressed, concerns the relative merit ofTPA in the different histotypes oflung cancer: prior studies, in fact, included tumors ofany cell type, considered as a unique pathologic entity6"8 or evaluated only the subgroup with small cell caneer.!" We designed this study to assess TPA with reference to other major prognostic factors, both in the general population with lung cancer and within subgroups with different histotypes of the tumor. PATIENTS AND METHODS ltJtients
Between 1984 and 1990, 563 new patients with cytologically or ·From the Divisions of Pneumology, A. Carle Hospital of Chest Diseases, Cuneo, Italy. Presented at the 57th Annual Scientific Assembly, American College of Chest Physicians, San Francisco, November 4-8, 1991. Manuscript received June 14; revision_8CCeRted September 11. &print requests: Dr. Buccheri, Via liepUbblic4 101C, RoccavWne
(CN), Italy 112018
multivariate analysis was made in a subgroup of 121 patients who, because of poor KPS or advanced age, had undergone a limited staging worlcup; TPA came out as the Grst most important factor. This study shows that .TPA is an important prognostic factor and that it should be included among laboratory data evaluated in lung cancer studies. (Chat 1992; 101:1187-92)
= =
= =
BMDP biomedical package; CEA carcinoembryonic antigen; CI con8dence intervalJ US Kamo£sky performance status; TPA = tissue polypeptide antigen
I
histopathologically proved bronchogenic carcinoma were seen at the Antonio Carle Hospital and assayed for TPA before starting treatment. Males constituted 89 percent (504 of 563) of the cohort studied. The median age was 64 years (range: 31 to 88 years). Histopathologic diagnosis was carried out in accordance with the revised WHO classificationof lung tumors" by the same pathologist. Pathologic diagnoses comprised the following: 319 squamous cell cancers, 67 small and 36 large cell anaplastic tumors, 55 adenocarcinomas. Eighty-six patients bad tumors with unde6ned or mixed bistol~ Routine pretreabnent staging procedures included: physical examination, blood chemistry and hematologic counts, chest radiographs and tomograms, bronchoscopy Computed tomography of the thorax was performed in all patients evaluated for surgery or irradiation and in many others with good performance status, not too advanced age, and in general with a fair physical and psychic compliance (in all, 429 patients). The same subjects were also tested with a total body 87Ga scan or with a combination of thorax, abdomen and brain (total body) CT scanning. During the last three years, an investigational procedure of total body scintigraphy with anti-eEA monoclonalantibody fragments was added to the baseline evaluation of the consenting patients who were candidates for surgery or irradiation. Ultrasound and radioisotope scans, along with other imaging tests and biopsies, were performed as clinically indicated. Only selected surgical candidates underwent mediastinoscopy Patients were classified using the new International Union Against Cancer (UICC) staging system. u Accordingly; 57 patients were ~1assified into stage I disease, 53 into stage II, 138 into stage IlIa and 109 into stage IIIb, 203 into stage N Three patients were not staged because of insufficient data. Follow-up consisted of clinical and radiographic examinations performed monthly (or more frequently) during the treatment, every three months during the following two years, and subsequently every six months. Very few patients abandoned their outpatient visit program; when they did, the physical status was obtained through a telephone interview with either the patient or the referring physician. In order to assess the relative value ofTPA, multiple factors were analyzed in conjunction with the marker. Included in multivariate analyses were: sex, age, the presence of weight loss in the six months preceding the diagnosis, the performance status according to the Kpsl3 and various parameters of disease extent, such as the stage of disease, the T and N factors, and number and site, if any, of metastases. Table 1 summarizes the variables studied, the number of observations and the median value and range (or the frequency CHEST I 101 I 5 I MA'f, 1992
1287
Table 1- Patient ChartlCferiBtica Variable·
No. Observations
Median
563 563
64
31-88
80
20-100
Sex (malelfemale) Age (yr) Weight loss (y/n)t KPS Stage of disease (I/II/IIIalIllblIV) T factor (I/II/IIIIIV) N factor (OII/II/III) M factor (OII/II/III/Iv)t Lung metastases (yIn) Brain metastases (yIn) Liver metastases (yIn) Adrenal metastases (yIn) Bone metastases (yIn) Histology (E/SIAIUU)§ TPA(UIL)
539
557 560 560 560 560 560 560 560 560 560 563 563
Range
Frequency 504159
2141325 57~13811091203
30-5203
130
43/229/1:n'155 126197/249/88 357/1551361913 65/495 471513 361524 481512 551505 319/67155136186
·Numeric codes for the subsequent multivariate analyses as in the Table, except for: male = I, female=O; y= I, n=O; stage 3a=3, stage 3b=4, stage 4=5; E,S,A,L,U histology (y/ri) = 1/0. t During the six months preceding diagnosis. tExpressed as number of metastatic sites. §E: epidermoid carcinoma; S: small cell carcinoma; A: adenocarcinoma; L: large cell carcinoma; U: unclassified. of distribution) of each variable. Bio-marker Assays
The TPA measurements were performed by radioimmunoassay on serum samples stored at - 2
on the survival duration was estimated using Cox's proportional hazards regression model." Factors having significant influence on the model were included in a stepwise fashion. The significance of each factor was calculated by the maximum likelihood ratio. A p value of 0.1 was set to enter, while a value of 0.15 was set to remove a variable from the model. The Cox's proportional hazards model is given by the formula:
Statistical Analysis
h(t;~)
Survival duration was measured for all patients from the time of pathologic diagnosis. Categorization of TPA for univariate analyses was accomplished using, as a cutoff point, the median of its distribution (130 U/L). Univariate survival analyses were based on the Kaplan-Meier product-limit estimates of the survival distribution. No deaths were censored in the analysis. Differences between survival curves were tested statistically using the Mantel-Cox logrank test and the Breslow-Gehan test, offered by the BMDP program." The relative importance of multiple prognostic factors
= h, (t)exp(~'~)
where h(t;z) is the hazard rate of an individual patient with covariate vector ~; ~ is a vector of the unknown regression coefficients, and h, (t) is the unknown hazard function for an individual with covariate vector z equal to o. A p ';alue of less than 0.05 was regarded as statistically significant. All tests were two-sided. Data were processed with the BMDP statistical software produced by Health Science Computing Facility, University of California at Los Angeles."
1 DEAD TPAsl30 Uti TPA>l30 Uti
P
~.75
Mantel-Cox Test BreslowTest
B A B .5 I L I
208
CENSORED 84
211
60
STATISTIC
P-VALUE
27.944 33.645
0.0(0) 0.0(0)
T.25
Y
o ... o 1288
...._ _....
..._ _......, .-_ _...
12
36
24
48
60
SURVIVAL (months)
..._ _..... FIGURE 1. Actuarial survival curves of the entire population (heavy solid Une: patients with TPA serum levels up to 130 UIL; dashed line: patients with TPA levels greater than 130 U/L).
72
84
Tissue Polypeptide Antigen in LungCancer (Buccheri,Ferrigno)
1PASl30~
1PA>130 /I
.75
ManIeI-Cac 'felt
hIIowfelt
DEAD
CENSORED
15 18
15 7
STA11S1lC
P-VAUE
1Q.212 9A03
0.0014 0.0022
75
R .25 0
25
0
A
0
24
12
36
A~NOCARCINOMA
B I
I
y
ManIeI-Cac felt
hIIowTeit
72
60 OEAD
CENSORED
36
9
STA11S1IC
P-VAUE
0.114
0.7359 0.1_
lPAsl30U/I lPA>l30U/I
L .75 T
48
15
1.732
7
37 36
stA11S11C
P-VALUE
12.229 16.210
cum1
hIIow'felt
.5
B
CENSOREO
121 126
MarII-...eacTeIt
.5
P
OEAD
lPAsl30U/I '1PA>l30U/I
0
84
0
·10
20 30 40 50 60 70 SQUAMOUS CELL CARCINOMA 1PASl30~
1PA>130 ill
75
.5
.5
.25
25
.. '=....
0
8
16 24 32 40 48
56 SMALL CELL CARCINOMA
64
72
0
~
iii••••••:
................ ··.....: ~
0
3
CENSOREO
14 13
7 2
1.190 1.226
8IlIIIcw 1eIt
80
DEAD
STA11S1lC
MarII-...eac'felt
~
0
o.cxD
~
Q.27U Q.26I3
.....•... ····5.
6 9 12 15 18 21 LARGE CELL CARCINOMA
24 27
SURVIVAL (months)
FIGURE 2. Actuarial survival curves of histologically homogeneous subgroups (heavy solid line: patients with TPA serum levels up to 130 UIL; dashedline: patients with TPA levels greater than 130 UIL). RESULTS
Univariate Analyses
As shown in Table 1, for the whole population the median value ofTPA was 130 UIL (range: 30 to 5,203 UIL). Patients with TPA up to 130 UIL survived much longer (median survival time: 10.41 months; CI, 8.47 Table I-Facton l ~
to 12.35) than patients with higher TPA values (median: 5.57; CI, 4.84 to 1). As shown in Figure 1, this difference was highly significant (p = 0.ססoo). We split the series in subgroups homogeneous as far as the extension of disease is concerned (that is, patients with stages I and II, and stage IlIa, IIIb and IV were
Inftuencing SuroiwJl: BaultI oftlte Multifactorial AntJlrIaia 1· Standard Error
Improvement ofchisquare*
p Value§
0.2162
0.0549
75.807
0.000
2 3
-0.0273 0.2570
0.0040 0.0946
56.759 11.106
0.000 0.001
4
0.2689
0.1072
7.702
0.006
5
0.0003
0.0001
6.045
0.014
Variables
Stept
Stage of disease (I1IIIIIIalIIIblIV) KPS M factor (MIIIIIIIIIV) W,ight loss (yIn) TPA(UIL)
1
Coefficient
*Five hundred thirty-one patients had no missing values and were included in the model. tStep at which the variable was included in the model. *Global chi square = 185.456. IGlobal p=O.()()()(). CHEST 1101 1 5 1 MA~
1992
1288
Table 3-Facton Independently Influencing Suroiool: Besulta ofthe MultifGCtorial Analysia 2· Variables
Stept
Coefficient
Standard Error
Improvement of chi squarer
p Value§
TPA(UIL) Stage of disease (I1IIIIIIalIIIblIV) Age (yr)
1 2
-0.0013 0.3886
0.0003 0.0937
17.956 11.666
O.()OO 0.001
3
-0.0315
0.0109
8.735
0.003
*Limited to 121 patients with a poor staging workup and no missing data. tStep at which the variable was included in the model. iGlobaI chi square =43.116. §Global p = 0.ססOO.
considered separately). In any subdivision, the same trend, observed in the entire population, was recognized. The tendency was statistically significant in the more advanced stages (p<.005 for stage IlIa and I~ by the Breslow test). In Figure 2, the TPA-survival relationship is analyzed as a function of the histotypes. This analysis is rather hindered by the severe imbalance among cell types, but again an identical trend is observed in all subgroups. The marker was best correlated with survival of adenocarcinoma patients (median survivals: 28 and 4 months for low and high TPA values, respectively).
Multivariate Analyses A first multivariate analysis, incorporating all the potentially significant prognostic factors described in Table 1, confirmed not only the independence ofTPA from tumor burden and histology (which could be guessed from the results of the univariate analyses), but also its independence from age, gender, weight loss and the performance status. However, TPA was selected as the less important independent factor (Table 2). Staging a patient is a time- and resource-demanding
process, which is not ever essential for therapeutic decisions and sometimes impossible to perform with sufficient confidence. In our series, for example, 121 patients were staged on the basis of thoracic radiographs, bronchoscopy and clinical examination, but without thoracic cr and a total body scintigraphy In this subgroup of patients, given the already known correlation existing between TPA and the stage of disease.s" one might anticipate an increased weight of the marker in proportion to the minor importance of the unreliable staging evaluation. This is exactly what we found to happen in a second multivariate analysis limited to the aforementioned patients (Table 3). Finally, multivariate analyses were run on histologically homogeneous subgroups, using the same covariates of the previous analyses (Table 4). Again, TPA emerged as a significant independent factor, and was the most important variable in both adenocarcinomas and anaplastic tumors. DISCUSSION
As a marker of neoplastic diseases, TPA has never been as well recognized as CEA.16.17 Nevertheless, it was discovered and characterized in the late 195081.2 -
Table 4-Facton Independently Influencing Survival in the DijJenmt Hiatoypea ofwng Cancer Variables
Step*
Coefficient
Squamous cell cancer (cases: 300; global chi square: 136; p=O.OOOO): -0.0325 KPS 1 Stage of disease 2 0.2543 (IIII/I IIalII IbIIV) TPA(UIL) 0.0007 3 Weight loss (yIn) 4 0.3564 Bone metastases (yIn) 0.6726 5 N factor (OII/IIIIII) 0.1720 .6 Small cell cancer (cases: 62; global chi square: 13.83; p = 0.(010): TPA (UIL) 1 0.0026 Weight loss (yIn) 0.5352 2 Large cell cancer (cases: 34; global chi square: 10.25; p = 0.0059): TPA(UIL) 0.0017 1 Weight loss (yIn) 0.9391 2 Adenocarcinoma (cases: 52; global chi square: 30.33; p=O.OOOO): TPA (UIL) 1 0.0018 Stage of disease 2 0.3008 (OIIIII/IIIalIIIbIIV)
Standard Error
Improvement of chi square
p Value
0.0052 0.0643
51.661 41.000
0.000 0.000
0.0003 0.1416 0.2527 0.0834
11.461 5.767 5.760 4.357
0.001 0.016 0.016 0.037
0.0009 0.2993
7.051 3.159
0.008 0.076
0.0007 0.5211
3.626 3.077
0.057 0.079
0.0006 0.1356
12.837 5.221
0.000 0.022
*Step at which the variable was included in the model.
1290
TISSue Polypeptide Antigen in Lung cancer (Buccheri, Ferrigno)
before the original CEA description's-e-and proposed as a nonspecific marker for various human tumors." likewise CEA.?Jl In many clinical applications, TPA has been reported to be superior to several other tumor markers, including CEA. 7 In lung cancer patients, the pretreatment serum levels ofTPA have been shown to correlate positively with either the primary tumor characteristics (defined as Tl, T2 and T3), or the nodal involvement (as NO, N 1 and N2), or the metastatic spread (as MO and Ml) or the classification into four stages of disease.v" In general, the greater the tumor bulk at diagnosis, the more elevated is the serum concentration of the marker," The same studies have indicated that posttreatment TPA may vary in accordance with the obvious changes in disease statusv? and may sometimes precede them." TIssue polypeptide antigen has been proposed also as a tool for predicting the course of disease. 6-8,10 Actually, the prognostic capability ofTPA may be anticipated by the knowledge of the molecule metabolism. The substance is synthesized during the S- to M-phase of the cell cycle and released upon proliferation into the blood stream." Thus, the concentration of the antigen is an indicator of the rate of cell division and tumor aggressiveness, and, therefore, of the host survival. This study was primarily conducted to address a not yet fully clarified issue, rather than simply to confirm the prognostic capability of TPA, which we believe to be already well established. In a previous study, reported in 1988 in this journal," we found that cohorts of patients having the same stage of disease tended to be divided into poor- and good-risk subgroups on the basis of their pretreatment TPA value, but the tendency did not reach the statistical level of significance. Emphasizing the negativity of the statistic and the confirmed correlation of the marker with the stage of disease, we supposed that TPA could directly reflect tumor load-rather than the cell proliferation rate - and only indirectly could indicate the prognosis. 8 This is not the case, however, since our current data prove by both univariate and multivariate methods the independence of TPA from the stage of disease. Indeed, TPA surpassed the prognostic value of the same stage when this was obtained by a limited staging workup. A second aim of the study was to evaluate in greater depth the prognostic significance of the marker, particularly as far as the relationship between TPA and the prognosis of the diverse cell types is concerned. In this respect, we have elucidated that TPA is a greater predictor of the outcome in adenocarcinomas and in anaplastic cancers. This is as one might expect on the basis of the marker metabolism" and of the knowledge of the proliferation rate of each cell type of lung cancer. 22 The identification of prognostic factors is useful in
designing and comparing clinical trials and in advising individual patients. The efforts to define a set of prognostic determinants important for lung cancer have continued during the last decade and have included anthropometric and clinical characteristics, pathologic aspects, in addition to a large number of laboratory parameters.P On the basis of the current findings, we would suggest including TPA in any future multifactorial analysis of survival, especially in the non-squamous cell histologic variants of lung cancer. ACKNOWLEDGMENTS: The authors wish to thank Mr. James Beauchamp for linguistic editing and Mr. Francesco Bontadi for his graphic aid. REFERENCES
1 Bjorklund B, Bjorklund ~ Antigenicity of pooled human malignant and normal tissue by cyto-immunologica1 technique: presence of an insoluble heat-labile tumor antigen. Int Arch Allergy 1957; 10:153-83 2 Bjorklund B, Lundblad G, Bjorklund ~ Antigenicity of pooled human malignant and normal tissues by cyto-immunologica1 technique: II. nature of tumor antigen. Int Arch Allergy 1958; 12:241-61 3 Nemoto T, Constantine R, Chu TM. Human tissue polypeptide antigen in breast cancer. J Natl Cancer Inst 1979;63:1247-1350 4 Adolphs HD, Oher ~ Significance of plasma tissue polypeptide antigen determination for the diagnosis and follow-upof uretelial bladder cancer. Urol Res 1984; 12:125-28 5 Kew MC, Berger EL. The value of serum concentrations of tissue polyptide antigen in the diagnosis of hepatocellular carcinoma. Cancer 1986;58:868-72 6 Buccheri GF, Violante B, Sartoris AM, Ferrigno D, Curcio A, Vola F. Clinical value of a multiple biomarker assay in patients with bronchogenic carcinoma. Cancer 1986; 57:2389-96 7 Buccheri GF, Ferrigno D, Sartoris AM, Violante B, Vola F, Curcio A. Tumor markers in bronchogenic carcinoma. Superiority of tissue polypeptide antigen to carcinoembryonic antigen and carbohydrate antigenic detenninant 19-9. Cancer 1987; 60:42-50 8 Buccheri GF, Ferrigno D. Usefulness of tissue polypetide antigen in staging, monitoring, and prognosis of lung cancer. Chest 1988;93:565-70 9 Mizushima Y, Hirata H, Izumi S, Hoshino K, Konishi K, et ale Clinical significance of the number of positive tumor markers in assisting the diagnosis of lung cancer with multiple tumor marker assay. Oncology 1990; 47:43-48 10 Granowitz JS, Bergstrom R, Nou E, Pahlman S, Brodin 0, Nilsson S, et ale Clinical and serologic markers of stage and prognosis in small cell lung cancer: a multivariate analysis. Cancer 1990;66:722-32 11 World Health Organization. Histologic typing of lung tumours. Am J Clin Pathol 1982; 77:123-36 12 De Angelis G, Cipri A, Flore F, Munno R. UICC: TNM classification of malignant tumours. ed 4. Geneva, Switzerland: International Union Against Cancer, 1987 13 Karnofsky DA, Burchenal JH. The clinical evaluation of combination chemotherapy in cancer. In: Mac Leod CM, ed. Evaluation of chemotherapeutic agents. New York: Columbia University Press, 1949:199-205 14 BMDI! Statistical software manual. Berkeley, Cal: University of California Press, 1988 15 Cox DR. Regression models and life tables (with discussion). J R Stat (series B) 1972;34:187-220 16 Schein PS. Tumor markers. In: Wyngaarden JB, Smith LH, eds. Text book of medicine. 16th ed. Philadelphia: WB Saunders ce., 1982:1020-22 CHEST I 101 I 5 I MA~
1992
1281
17 Broder LE, Primack A. Marker substances in bronchogenic carcinoma. In: Straus MJ, ed. Lung cancer: clinical diagnosis and treatment. 2nd ed. New York: Grune &: Stratton Inc, 1983:37-61 antigens 18 Gold ~ Freedman SO. Demonstration of tumo~sPecific in human colonic carcinomata by immunological tolerance and absorption techniques. J Exp Med 1965; 121:439-62 19 Bjorklund B. 00 the nature and clinical use of tissue polypetide antigen. Tumor Diagnostik 1980; 1:9-20 20 NHI. Carcinoembryonic antigen: its role as a marker in the
management of cancer. Br Moo J 1981; 282:373-75 21 Bjorklund B, Bjorklund ~ Specificity and basis of the tissue polypetide antigen. Cancer Detect Prevent 1983; 6:41-50 22 Straus MJ, Moran RE, Shackney SE. Growth characteristics or lung cancer. In: Straus MJ, ed. Lung cancer: clinical diagnosis and treatment. 2nd ed. New York: Grune &: Stratton Inc, 1983:21-35 23 George SL. Identification and assessment of prognostic factors. Semin Oocoll988; 15:462-71
Plan to Attend ACCP's
58th Annual SCientific Assembly Chicago --~ OCtober 25-29, 1992
cA ~.---.......~=
1282
TIuue Polypeptide AntIgenin Lung Cancer(BuccherI, FerrIgno)