Clonogenic Growth In Vitro: An Independent Biologic Prognostic Factor in Ovarian Carcinoma By Christian Dittrich, Evelyn Dittrich, Paul Sevelda, Marcus Hudec, Heinrich Salzer, Thomas Grunt, and James Eliason A retrospective analysis was performed to investigate the prognostic value of growth in a human tumor clonogenic assay system for 84 ovarian cancer patients. A significant difference in survival probability (determined by the method of Kaplan-Meier) was found by univariate analysis between patients with ovarian carcinoma whose tumors manifested clonogenic growth (defined as growth of Ž five colonies per plate) and patients whose tumors did not grow. Clonogenic growth in vitro was associated with worse prognosis (P = .007, log-rank test). A number of generally accepted prognostic factors, International Federation of Gynecology and Obstetrics (FIGO) stage (P = .003),

THE

EFFECTIVENESS of therapy in ovarian

carcinoma has improved substantially in recent years resulting in a 60% to 80% complete remission rate in advanced-stage disease under cisplatin-containing regimens. However, the overall survival of these patients is still very poor1 ; 30% of the patients with International Federation of Gynecology and Obstetrics (FIGO) stage I tumors and 50% to 60% with FIGO stage II tumors will not be alive 5 years from diagnosis.2 Less than half of the optimally debulked FIGO stage III patients and only 40% to 70% of the patients reaching pathologic complete remission will be long-term

survivors.1-3 So far, prognostic factors seem to determine patient's outcome in ovarian carcinoma more than therapy. 4 6 On the basis of a number of

retrospective analyses, the size of the postoperative residual tumor mass, FIGO stage, grade, histologic type, and age have been the factors most often identified as prognostic determinators for the natural course of ovarian carcinoma. 4'7'4 These factors represent a static classification rather than dynamic processes. Patients will be allocated to a

few subgroups only according to these conventional risk factors, although the prognosis of the patients within these subgroups varies considerably.'1 Attempts to develop biologic tests that would better determine an individual tumor's aggressiveness have identified a variety of more specific tumor-cell-related parameters such as morphometric features,' 2 ploidy, 13"14 DNA content,'3 "5 and

residual tumor mass (P < .001), and grade (P = .011), were also of prognostic importance in our patient population. Multivariatp analysis, based on the Cox regression model, identified clonogenic growth as a significant independent prognostic parameter in ovarian carcinoma (P = .031), in addition to the conventional risk factors. Estimation of survival of individual patients was best accomplished by combining the factors of residual tumor mass (P < .05), age (P < .01), and clonogenic growth (P < .05) (in sequence of decreasing potential of risk). J Clin Oncol 9:381-388. o 1991 by American Society of Clinical Oncology. especially more dynamic factors such as S-phase 5 16 fraction,1 ' mitotic index,12,17 thymidine labeling index,18'19 in vitro chemosensitivity,20'21 oncogene product expression,22 growth factor production, and growth factor receptor expression.2 3 Univariate and multivariate analyses indicate that the independence and prognostic significance of these parameters is not yet clear. Clonogenic growth in semisolid media is thought 24 27 to be related to the malignancy of cancer cells. Thus, assessing the presence or absence of clonogenic growth in vitro might provide a means to estimate the aggressiveness of tumor cells. We have examined this hypothesis in ovarian carcinoma comparing the clonogenicity of tumors from patients with ovarian carcinoma with their clinical

From the Laboratoryfor Applied and Experimental Tumor Cell Biology, Department of Chemotherapy, Department of Obstetrics and Gynecology I, and Institute of Statistics and Computer Sciences, Universityof Vienna, Vienna, Austria;and Departmentof ExperimentalDermatologyand Oncology, Hoffmann-La Roche Ltd, Basle, Switzerland. Submitted March 21, 1990; acceptedAugust 21, 1990. Supported by grantsfrom the Kamillo Eisner-Stiftung, Jubiliiumsfonds der Osterreichischen Nationalbank, and Medizinisch-WCssenschaftlicherFonds des Burgermeisters der BundeshauptstadtWien. Address reprint requests to ChristianDittrich, MD, Department of Chemotherapy, Division of Oncology, University of Vienna, Lazarettgasse 14, A-1090 Vienna, Austria. © 1991 by American Society of Clinical Oncology. 0732-183X/91/0903-0003$3.00/0

Journalof Clinical Oncology, Vol 9, No 3 (March), 1991: pp 381-388

Downloaded from ascopubs.org by 37.44.207.18 on January 27, 2017 from 037.044.207.018 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.

381

382

DITTRICH ET AL

outcome. The independent prognostic significance of clonogenic growth in vitro was then evaluated in relation to conventional prognostic parameters in ovarian cancer patients.

ovarian cancer patients.

MATERIALS AND METHODS Between November 1981 and March 1985, a total of 111 tumor samples from patients with histologically verified epithelial ovarian carcinoma-ranging from one to four samples per patient-was cultivated using a clonogenic assay system. Only the result of the first assay for each patient was used for evaluating the correlation between in vitro growth and corresponding clinical outcome, thus resulting in 84 correlations. Patients Patients included in this investigation were staged, treated, and evaluated by the staff of the Department of Obstetrics and Gynecology I, University of Vienna, Austria (P.S., H.S.), according to guidelines of treatment protocols of the Austrian Collaborative Ovarian Cancer Study Group for first-line therapy and according to treatment plans of the department for second-line therapy. Table 1 details various characteristics of the patients at the time of diagnosis. The majority (87%) had advancedstage disease (according to FIGO'). One third belonged to the favorable prognostic group with no residual disease (ND) or optimal disease ([OD]; tumor masses of < 2 cm in diameter remaining after primary surgery), whereas 64% of the patients presented with nonoptimal disease ([NOD]; tumor masses of > 2 cm in diameter). Most of the tumors were of serous histology (World Health Organization criteria 29). A three-stage grading system combining the architec3 tural pattern and cytologic features was used. Grade 1 describes a tumor with well-developed papillary structures consisting of cells with a low degree of pleomorphism, grade 3 tumors show predominantly solid formations with pleomorphic tumor cells, and grade 2 is a combination of these patterns. Mean age of the population studied was 59 years. Sixty-one of the patients from whom tumor samples were derived had not been treated, whereas seven patients were on therapy (polychemotherapy, n = 5; polychemo-hormone therapy, n = 2) at the time. For 10 patients, the last therapy (chemo-/radio-/hormone therapy) had been administered within 6 months before cultivation, and for six others, the last treatment had been more than half a year before. Tumor Materialand Culture Technique Sample collection, tumor processing, and cultivation were and performed according to the method of Hamburger 3 233 3 Salmon" with modifications described previously. Solid tumor tissue obtained by surgery was mechanically disaggregated into small pieces, and the remaining tissue fragments were incubated in an enzyme cocktail (0.15% collagenase, type IA; 0.015% deoxyribonuclease, type I) at 37 0 C for 1 to 2 hours. Effusions obtained by paracentesis were mixed with preservative-free heparin (50 U/mL) to prevent clotting and centrifuged at 600g for 20 minutes. The resulting cell suspensions were washed twice, resuspended

Table 1. Distribution of Risk Factors of Ovarian Cancer Patients According to the InVitro Growth Potential of the Corresponding Tumor Samples in the Clonogenic Assay Risk Factor Distribution Among the Patient Population Studied

Growth*

No Growth

N

In Vitro N (%)

In Vitro N (%)

9 2 57 16

4 (5) 0 (0) 23 (27) 8 (10)

5 2 34 8

16 14 54

5 (6) 5 (6) 25(30)

11 (13) 9 (11) 29 (34)

NS

12 24 48

6 (7) 8 (10) 21 (25)

6 (7) 16 (19) 27 (32)

NS

57 7 4 1 10 5

29 (35) 2 (2) 1 (1) 0 (0) 3 (4) 0 (0)

28 (33) 5 (6) 3 (4) 1 (1) 7 (8) 5 (6)

58 + 13

60 + 10

Pt

FIGO stage*

I II III IV Residual tumor mass (diameter in cm) 0 < 2 > 2 Grade 1 2 3 Histology Serous Mucinous Endometrioid Clear cell Undifferentiated Mixed Age (years) Mean - SD

(6) (2) (40) (10)

NS

NA

NS

Abbreviations: NS, not significant; NA, nonassessable. *Colony formation of > five colonies per 5 x 10' viable nucleated cells plated. tDifferences in frequency distributions were calculated using X' tests. Comparison of age groups was performed using the t-test. :FIGO stages I and IIwere combined for statistical evaluation. in medium, and passed through a 25 pm mesh filter to obtain a single-cell suspension. Total and viable cell numbers were determined by trypan blue dye exclusion. The cell concentration was adjusted to 3 x 106 viable nucleated cells per milliliter. The agar double-layer system was used for cultivation. Briefly, 5 x 10S cells per 35 mm petri dish were seeded in 1 mL containing enriched Connaught Medical Research Laboratories 1066 medium and 0.3% agar above a 1 mL underlayer consisting of 0.5% agar in enriched McCoy's 5A medium. On the day of cultivation, one plate was fixed with 2.5% buffered glutaraldehyde and another plate was treated with sodium azide (100 vig/mL) to maintain the existing state immediately after plating and before any putative growth. The plates were checked for the presence of aggregates before incubation in a humidified 5% carbon dioxide atmosphere at 37"C for 2 to 3 weeks. Samples with aggregates of _ 30 l.m in diameter were discarded. Cultures were examined periodically to assess the optimal time for study evaluation. Evaluationof Clonogenic Growth Colonies were defined as aggregates of > 60 ptm in diameter. Plates with five or more colonies were considered

Downloaded from ascopubs.org by 37.44.207.18 on January 27, 2017 from 037.044.207.018 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.

PROGNOSTIC FACTOR IN OVARIAN CARCINOMA

383

"positive" for growth. Plates treated with sodium azide and glutaraldehyde were used as criteria of "no growth." Cloning efficiency was defined as the number of colonies scored, divided by the number of viable nucleated cells plated, and is expressed as a percentage.

ing clinically to therapy (according to the criteria of the Gynecologic Oncology Group38 ) gave rise to at least five colonies per 5 x 10' cells plated, the tumors of nonresponding patients exhibited clonogenic growth in 15 of 28 cases. This difference was not statistically significant.

Characterizationof PlatingSuspensions and Colonies Cytocentrifuge slides prepared from each plating suspension were stained with May-Griinwald-Giemsa and Papanicolaou for morphologic evaluation, with nonspecificesterase for identification of macrophages and monocytes, and with peroxidase for granulocytes. For morphologic evaluation of cells in individual in vitro colonies, permanent slides were prepared according to the method of Salmon and Buick" and stained with Papanicolaou. Chromosomal analyses of the plating suspensions as well as of colonies were performed regularly to confirm malignancy. In a few cases, transmission and scanning electron microscopy were used.

Criteriafor Exclusion Samples having insufficient material for cultivation, cell viability < 10%, negative cytology, presence of aggregates, or microbial contamination, as well as samples of patients lost to follow-up, were excluded from the final evaluation.

StatisticalMethods Association of clonogenic growth with other risk factors was evaluated using the classical X' test. The difference in the mean ages of patients with and without clonogenic growth was assessed by the conventional two-sample t-test. Estimation of survival probabilities was performed accord35 ing to the method of Kaplan and Meier. The significance of differences between various groups of patients was assessed using the log-rank test. A simultaneous analysis of risk factors evaluating their relative importance was performed by means of the multivariate regression model of Cox." All reported P values derived from this multivariate analysis are based on one-sided tests. Computations were performed by 37 the P1L and P2L programs of BMDP-81.

Clonogenic Growth In Vitro and Survival Survival analyses showed that patients whose tumors gave rise to colonies in the clonogenic assay had a significantly worse prognosis than the patients with tumors with no detectable clonogenic cells (P = .007) (Fig 1). Association Between ConventionalRisk Factorsand Prognosis As illustrated in Fig 2, the importance of previously identified risk factors could be confirmed in our analysis. The median survival of the entire patient population was 20.5 months (Fig 2A) and the expected overall 5-year survival was 17.0%. For the survival analysis, patients with FIGO stage I (11%) and II (2%) were combined. The median survival of these early-stage patients has not been reached yet, but for patients with FIGO stages III and IV, median survival estimates were 19.0 and 5.0 months, respectively (Fig 2B). These differences were significant.

(U

0) RESULTS Association of Clonogenic Growth In Vitro With Pretreatmentand ClinicalResponsiveness

4-

0

Clonogenic growth in vitro was exhibited by 35 of the 84 (42%) ovarian cancer samples tested (Table 1). One fourth of the patients had been pretreated when their tumors were processed for cloning in vitro. Nine of the 23 (39%) tumors of pretreated patients exhibited clonogenic growth in vitro compared with 26 of 61 (43%) of those without prior treatment. There was no growth advantage in favor of tumor samples from patients responding to therapy. Whereas 12 of 37 tumors of patients respond-

0

10

20

30

40

Patient Survival (in Months) Fig 1. Survival of patients whose tumors gave rise to colonies in the clonogenic assay is compared with that of patients with nonclonogenic tumors. (0-0) Growth: median in months, 5.0; total patients, 35; patients who died, 22. (e--) No growth: median in months, 32.0; total patients, 49; patients who died, 18; P = .007.

Downloaded from ascopubs.org by 37.44.207.18 on January 27, 2017 from 037.044.207.018 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.

384

DITTRICH ET AL

SURVIVAL FROM PRIMARY SURGERY

STAGE

SFIGO

1.11 S FIGO III SFIGO IV

MEDIAN (months) 910 19.0 50

PATIENTS TOTAL DEAD 11 3 57 26 16 11

RESIDUAL DISEASE (max diameter in cm) NONE (ND ) : 2 cm(NOD)

100-

100-

80

80-

60-

60-

40-

40-

S20.

>

PATIENTS TOTALDEAD 84 40

MEDIAN (months) 205

20A

MEDIAN (months) 910 320 165

PATIEINTS TOTAL DEAD 16 2 14 5 54 33

p60

MEDIAN (months) 230 170

PATIENTS TOTALDEAD 43 14 41 26

8060-

40-

40-

D

PATIENTS TOTALDEAD 57 30

100-

60-

20

MEDIAN (months) 190

SEROUS

20-

000 S10 20 30 40 50 60 70 80 90 100

E•

0 -,

,

,

,

,

fl ln 20n0Ln5n 6n7n 8n

Patient

Survival

0n

(in

Months)

Fig 2. Survival curves of 84 patients with epithelial ovarian carcinoma whose tumor growth was evaluated in the clonogenic assay classified according to stage, residual tumor mass, grade, histology, and age at diagnosis. (A) Overall patient survival. (B) 2 Stage: classification according to the FIGO criteria." Stage I (N = 9) and II (N = 2) patients were combined for evaluation. (C) Residual disease: postoperative tumor mass classified as ND, OD, and NOD. (D) Grading: performed according to classification criteria from Day et al" discerning well-differentiated tumors (GI), moderately well-differentiated tumors (G2), and undifferentiated tumors (G3). (E) Histology: determined according to classification criteria from Serov et al. 2 ' The majority of the patient tumors (N = 57) was of serous histology. Because of the low number of patients with other histologies, separate survival curves could not be calculated. (F) Age: 60 years was arbitrarily chosen as discriminant.

Fourteen of the 16 patients (88%) with ND are still alive after 5 years, whereas none of the patients with tumor remaining after primary surgery reached the 5-year follow-up (Fig 2C). Median survival estimates for patients with ND, OD, and NOD were 91.0, 32.0, and 16.5 months, respectively. Figure 2D illustrates the differences in survival according to histologic tumor grade. More than 50% of the patients with grade 1 tumors are still alive after 5 years. Their median survival was 91.0 months. The estimated median survival for grade 2 and grade 3 tumor patients was only 29.5 and 17.0 months, respectively. Two thirds of the tumors had serous histology. Median survival of these patients was estimated to be 19.0 months, and less than 20% of the patients were alive after 5 years (Fig 2E). No other histologic subgroup was large enough for statistical analysis. The age of 60 years was arbitrarily chosen to discriminate between two age groups of patients.

As shown in Fig 2F, no significant difference in survival between these age groups was found in our series. The median survival times calculated for the patients aged up to 60 years and for the older patients' group were 23.0 and 17.0 months, respectively. Independent PrognosticSignificanceof Colony FormationEvaluatedby MultivariateAnalysis No association between any of the conventional risk factors examined and colony growth in vitro was found (Table 1). Because of the low number of tumors of other than serous histologies, no association analysis could be performed for this prognostic factor. A proportional hazards regression analysis36 was performed including the following factors: stage of disease, residual tumor mass, tumor grade, patient's age, and colony formation in vitro. The resulting relevant factors are summarized in Table 2. As demonstrated, colony formation in

Downloaded from ascopubs.org by 37.44.207.18 on January 27, 2017 from 037.044.207.018 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.

385

PROGNOSTIC FACTOR IN OVARIAN CARCINOMA Table 2. Multivariate Analysis of Prognostic Factors in Ovarian Carcinoma Covariate

Covariate Value

Parameter Value

Relative Risk*

Residual tumor mass (diameter in cm)

0 < 2 > 2 Each year < 60 60 Each year > 60 No growth Growth

0 2.065 2.994 0.044 0 0.044 0 0.627

1 7.88 19.97 0.91 1 1.05 1 1.87

Age

Clonogenic growth in vitrot

P .031 .002 .005 .005 .031

NOTE. Parameter estimates have been assessed expressing the relative risk for a given covariate value with respect to a basic condition for each covariate. These basic conditions are reflected by parameter values 0 and relative risk 1. *The relative risk is derived from the exponential of the parameter value. tColony formation of Ž 5 colonies per 5 x 105 viable nucleated cells plated.

vitro was of significant value as a new independent prognostic parameter in patients with ovarian carcinoma (P = .031). By means of the Cox regression model, survival estimates for individual patients based on the information of the respective age category, the residual tumor mass and the clonogenic growth were calculated. As illustrated in Fig 3, for a chosen age category, exemplified for age < 60 years or greater than 60 years, patients with ND whose tumors did not exhibit clonogenic growth in vitro showed the highest probability of survival, whereas patients with NOD and clonogenic growth had the worst prognosis.

.o te

0

a61

wi

0

10

20

30

40

50

Patient Survival

60 (in

70

80

90

100

Months)

Fig 3. Survival estimates from multivariate analyses based on the Cox regression model. The following risk-factor constellations are considered: (A-A) no clonogenic growth, no residual disease; (0-0) clonogenic growth, no residual disease; (A-A) no clonogenic growth, optimal disease; (0--0) clonogenic growth, optimal disease; (i--) no clonogenic growth, nonoptimal disease; and (e--) clonogenic growth, nanoptimal disease.

DISCUSSION

Clonogenic growth of ovarian tumor samples in our system is related to the proportion of tumor cells in the specimens.33 The fact that not all of the samples that contained tumor cells actually gave rise to colonies may depend on the presence of an intrinsically more aggressive tumor-cell population or subpopulation in the specimens that grew. It is clear from the low cloning efficiencies of solid tumor cells in semisolid media that the requirement for attachment-independent growth in the absence of exogenous growth factors, other than those present in animal sera, represents a severe selection pressure for cells with a potentially "aggressive" phenotype. In theory, this clonogenic fraction should be composed of transitional cells that subsequently give rise to nondividing, differentiated end cells and of so-called stem cells that maintain their own numbers by self-renewal. 39" From a conceptual point of view, this cell subpopulation should represent the primary target of cancer therapy. It still has to be determined whether the stem-cell population represents a specific cell entity41,42 or rather a distinct functional state of cells. Experiments with mouse mammary tumors show that the in vitro colony-forming efficiencies are closely associated with the ability of the cells to form lung colonies after intravenous inoculation, confirming the suggestion that the clonogenic cell content may be the most important determinant of metastatic cancer.43 Based on conceptual considerations and experimental results, clonogenic growth in vitro has been claimed to reflect the malignant potential of tumors. Colony formation has been reported to be

Downloaded from ascopubs.org by 37.44.207.18 on January 27, 2017 from 037.044.207.018 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.

386

DITTRICH ET AL

related to patient prognosis in squamous cell carcinoma of the head and neck,""' multiple myeloma,46 neuroblastoma,"7 breast cancer,4849 and 5 miscellaneous solid tumors." Limited data from Bertoncello et al,51 demonstrating a significant inverse correlation between clonogenicity and patient survival for a small group of patients with ovarian carcinoma, were the first to indicate the possibility of a prognostic significance of clonogenic growth with this tumor. The present study was designed to analyze in more detail if clonogenic growth in vitro represents an independent prognostic factor for patients with ovarian carcinoma. As illustrated in Fig 2, the factors generally accepted to be predictive for the survival of ovarian cancer patients were similarly shown to be of importance in our patient population. A high comparability with respect to the patient distribution according to residual tumor mass,7 histologic subtypes,4' 7' and grade8 is evident between our data and those published by others,4 7' '8 although the average age was higher and more advanced FIGO stages were preponderant in our study. The possibility that only tumors from patients responding to therapy grow in our system can be excluded, supporting the findings of Hug et al.52 There are some reports of interrelationships between certain risk factors such as stage and grade 9 or histologic subtype and grade.7 No such interdependence was found for any of the accepted risk factors and clonogenic growth in our study (Table 1). In a consecutively performed multivariate analysis using the Cox regression model, clonogenic growth was shown to be a significant (P = .031) independent prognostic discriminator for patients with ovarian carcinoma. According to the regression model illustrated in Fig 3, the survival probability of an individual patient can be determined in advance. Clonogenic growth in vitro was found in 42% of 51 5 the specimens, comparable to other reports. ' 3-55 Whereas a significant association between overall clonogenic growth in vitro and survival was observed in our study, no quantitative relationship was apparent between cloning efficiency and survival. There is, in fact, no reason a priori to assume that such a quantitative relationship should exist between a parameter, such as cloning efficiency, that measures the frequency of clonogenic cells in

a tumor and survival that may depend more on the total number of cells. Furthermore, if the ability of solid tumor cells to grow in an attachmentindependent manner was dependent on a mutational event and if these cells were responsible for the eventual death of the patients, then only a quantitative relationship between global survival and presence of clonogenic cells should be expected. It might be reasonable in this case to find a quantitative relationship between time to death and total numbers of clonogenic cells (cloning efficiency times tumor mass) when comparing only those patients having clonogenic cells. A much larger series of patients would be needed to test this hypothesis, however. Our findings are supported by those of Rey et al15 who have reported a statistically significant difference in survival between ovarian cancer patients whose tumors displayed a high cloning efficiency, and those with a low cloning efficiency but without a linear relationship between cloning efficiency and survival. Our results are also in accordance with the observations of others who found no quantitative relationship between clonogenic growth and survival in breast cancer patients.449,56

Our results indicate that the presence of cells capable of attachment-independent growth under our selective conditions represents an independent prognostic factor which, in combination with conventional risk factors, could be useful for identifying patients who would benefit from more aggressive therapy. Its discriminating power should be evaluated prospectively with FIGO stage I and II patients, with optimally debulked stage III patients, and with patients having pathologically complete remissions. On the basis of clonogenic assay results, patients with otherwise favorable prognostic features could be submitted to more aggressive therapy than conventionally used for the respective subgroup of patients. According to the regression model illustrated in Fig 3, the individual patient survival could possibly be determined in advance. This approach would finally allow a more individualized choice of therapy. However, before this goal can be reached, the value of clonogenic growth as a prognostic parameter should be independently assessed as a stratification criterion in randomized clinical trials in

Downloaded from ascopubs.org by 37.44.207.18 on January 27, 2017 from 037.044.207.018 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.

PROGNOSTIC FACTOR IN OVARIAN CARCINOMA

387

ovarian carcinoma. Only the confirmation of the discriminating power of presence or absence of clonogenic growth, found in our retrospective evaluation, by univariate and multivariate analyses of a prospectively randomized study in which

clonogenic growth and nongrowth have been used as stratification criteria in a larger patient base will be proof that clonogenic growth represents an independent prognostic factor for ovarian carcinoma.

REFERENCES 1. Ozols RF, Young RC: Chemotherapy of ovarian cancer. Semin Oncol 11:251-263, 1984 2. Young RC: Initial therapy for early ovarian carcinoma. Cancer 60:2042-2049, 1987 3. Wharton JT, Edwards CL, Rutledge FN: Long-term survival after chemotherapy for advanced epithelial ovarian carcinoma. Am J Obstet Gynecol 148:997-1005, 1984 4. Swenerton KD, Hislop TG, Spinelli J, et al: Ovarian carcinoma: A multivariate analysis of prognostic factors. Obstet Gynecol 65:264-270, 1985 5. van Houwelingen JC, ten Bokkel Huinink WW, van der Burg MEL, et al: Predictability of the survival of patients with advanced ovarian cancer. J Clin Oncol 7:769773, 1989 6. Stein G, Schubert-Frischle G: Einflup von prognostischen Faktoren und Therapie auf die Prognose von Patientinnen mit epithelialem Ovarialkarzinom. Tumor Diagnostik Therapie 10:191-197, 1989 7. Dembo AJ, Bush RS: Choice of postoperative therapy based on prognostic factors. Int J Radiat Oncol Biol Phys 8:893-897, 1982 8. Schray M, Martinez A, Cox R, et al: Radiotherapy in epithelial ovarian cancer: Analysis of prognostic factors based on long-term experience. Obstet Gynecol 62:373-382, 1983 9. Malkasian GD, Melton LJ, O'Brien PC, et al: Prognostic significance of histologic classification and grading of epithelial malignancies of the ovary. Am J Obstet Gynecol 149:274-282, 1984 10. Sjbvall K, Silfverswlird C, Einhorn N: Ovarian cancer: The challenge of local tumor control and its impact on survival. Int J Radiat Oncol Biol Phys 12:567-571, 1986 11. Feinstein AR: A clinical method for estimating the rate of growth of a cancer. Yale J Biol Med 41:422-433, 1969 12. Wils J, van Geuns H, Baak J: Proposal for therapeutic approach based on prognostic factors including morphometric and flow-cytometric features in stage III-IV ovarian cancer. Cancer 61:1920-1925, 1988 13. Klemi PJ, Joensuu H, Mienpiiii J, et al: Influence of cellular DNA content on survival in ovarian carcinoma. Obstet Gynecol 74:200-204, 1989 14. Friedlander ML, Hedley DW, Swanson C, et al: Prediction of long-term survival by flow cytometric analysis of cellular DNA content in patients with advanced ovarian cancer. J Clin Oncol 6:282-290, 1988 15. Erba E, Ubezio P, Pepe S, et al: Flow cytometric analysis of DNA content in human ovarian cancers. Br J Cancer 60:45-50, 1989 16. Kallioniemi OP, Punnonen R, Mattila J, et al: Prognostic significance of DNA index, multiploidy, and S-phase fraction in ovarian cancer. Cancer 61:334-339, 1988 17. Haapasalo H, Collan Y, Atkin NB, et al: Prognosis of

ovarian carcinomas: Prediction by histoquantitative methods. Histopathology 15:167-178, 1989 18. Conte PF, Alama A, Rubagotti A, et al: Cell kinetics in ovarian cancer. Cancer 64:1188-1191, 1989 19. Khoo SK, Hurst T, Webb MJ, et al: Measurement of tumor cell activity in short-term primary culture. Cancer 61:1579-1586, 1988 20. Silvestrini R, Sanfilippo O, Daidone MG, et al: Inhibition of DNA synthesis in selection of anticancer therapy, in Hall TC (ed): Prediction of Response to Cancer Therapy, vol 276. New York, NY, Liss, 1988, pp 93-103 21. Volm M, Briiggemann A, Giinther M, et al: Prognostic relevance of ploidy, proliferation, and resistancepredictive tests in ovarian carcinoma. Cancer Res 45:51805185, 1985 22. Rodenburg CJ, Koelma JA, Nap M, et al: Immunohistochemical detection of the ras oncogene product p21 in advanced ovarian cancer. Arch Pathol Lab Med 112:151154, 1988 23. Bauknecht Th, Janz I, Kohler M, et al: Human ovarian carcinomas: Correlation of malignancy and survival with the expression of epidermal growth factor receptors (EGF-R) and EGF-like factors (EGF-F). Med Oncol Tumor Pharmacother 6:121-127, 1989 24. MacPherson I, Montagnier L: Agar suspension culture for the selective assay of cells transformed by polyoma virus. Virology 23:291-294, 1964 25. Salmon SE: Human tumor clonogenic assays: Growth conditions and applications. Cancer Genet Cytogenet 19:2128, 1986 26. Von Hoff DD: In vitro predictive testing: The sulfonamide era. Int J Cell Cloning 5:179-190, 1987 27. Dittrich Ch: Klonieren von soliden Tumoren. Therapiesimulation, Therapieoptimierung und Prognoseerstellung am Beispiel des Ovarialkarzinoms. New York, NY, Springer, 1987 28. Kottmeier H, Kolstad P: Annual Report on the Results of Treatment in Gynecological Cancer, vol 17. Stockholm, Sweden, Radiumhemmet, 1979 29. Serov SF, Scully RE, Sobin LH: Histological typing of ovarian tumors, in International Histological Classification of Tumors, vol 9. Geneva, World Health Organization, 1973 30. Day TG, Gallager HS, Rutledge FN: Epithelial carcinoma of the ovary: Prognostic importance of histologic grade. Natl Cancer Inst Monogr 42:15-18, 1975 31. Hamburger AW, Salmon SE: Primary bioassay of human tumor stem cells. Science 197:461-463, 1977 32. Dittrich Ch, Jakesz R, Wrba F, et al: The human tumor cloning assay in the management of breast cancer patients. Br J Cancer 52:197-203, 1985 33. Dittrich Ch, Dittrich E, Wrba F, et al: Analysis of

Downloaded from ascopubs.org by 37.44.207.18 on January 27, 2017 from 037.044.207.018 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.

388

DITTRICH ET AL

factors influencing clonogenic growth in vitro of cells from ovarian carcinoma patients. Cancer Lett 50:183-189, 1990 34. Salmon SE, Buick RN: Preparation of permanent slides of intact soft-agar colony cultures of hematopoietic and tumor stem cells. Cancer Res 39:1133-1136, 1979 35. Kaplan EL, Meier P: Nonparametric estimation from incomplete observations. J Am Stat Assoc 53:457-481, 1958 36. Cox DR: Regression models and life-tables. J Roy Stat Soc (B) 34:187-220, 1972 37. Dixon WJ, Brown MB, Engelman L, et al: BMDP Statistical Software. Berkeley, CA, University of California Press, 1981 38. Blessing JA: Design, analysis and interpretation of chemotherapy trials in gynecologic cancer, in Deppe G (ed): Chemotherapy of Gynecologic Cancer. New York, NY, Liss, 1984, pp 49-83 39. Buick RN: Cell heterogeneity in human ovarian carcinoma. J Cell Physiol 3:117-122, 1984 (suppl) 40. Mackillop WJ, Ciampi A, Till JE, et al: A stem cell model of human tumor growth: Implications for tumor cell clonogenic assays. J Natl Cancer Inst 70:9-16, 1983 41. Resnicoff M, Medrano EE, Podhajcer OL, et al: Subpopulations of MCF 7 cells separated by Percoll gradient centrifugation: A model to analyze the heterogeneity of human breast cancer. Proc Natl Acad Sci USA 84:72957299, 1987 42. Grunt ThW, Dittrich E, Somay C, et al: Clonogenic and differentiated cell phenotypes separated by density gradient centrifugation of ovarian cancer cells (HOC-7). (letter) (submitted) 43. Price JE: Clonogenicity and experimental metastatic potential of spontaneous mouse mammary neoplasms. J Natl Cancer Inst 77:529-535, 1986 44. Mattox DE, Von Hoff DD: In vitro stem cell assay in head and neck squamous carcinoma. Am J Surg 140:527530, 1980 45. Johns ME, Mills SE: Cloning efficiency. A possible prognostic indicator in squamous cell carcinoma of the head and neck. Cancer 52:1401-1404, 1983 46. Durie BGM, Young LA, Salmon SE: Human myeloma in vitro colony growth: Interrelationships between

drug sensitivity, cell kinetics, and patient survival duration. Blood 61:929-934, 1983 47. Von Hoff DD, Casper J, Bradley E, et al: Direct cloning of human neuroblastoma cells in soft agar culture. Cancer Res 40:3591-3597, 1980 48. Aapro MS, Eliason JF, Krauer F, et al: Colony formation in vitro as a prognostic indicator for primary breast cancer. J Clin Oncol 5:890-896, 1987 49. Nomura Y, Tashiro H, Hisamatsu K: In vitro clonogenic growth and metastatic potential of human operable breast cancer. Cancer Res 49:5288-5293, 1989 50. Rey AA, Cupissol D, Ursule E, et al: The clonogenic assay in prognosis and as an indicator of drug activity in human cancer patients, in Dittrich Ch, Salmon SE (eds): Chemotherapy Sensitivity Testing, Proceedings of the 13th International Congress of Chemotherapy, part 224. Vienna, Austria, Egermann, 1983, pp 1-4 51. Bertoncello I, Bradley TR, Campbell JJ, et al: Limitations of the clonal agar assay for the assessment of primary human ovarian tumor biopsies. Br J Cancer 45:803-811, 1982 52. Hug V, Thames H, Johnston D, et al: The true predictive value of the human tumor stem cell assay: Does a workable assay select for treatment responders? J Clin Oncol 2:42-45, 1984 53. Ozols RF, Hogan WM, Young RC: Direct cloning of human ovarian cancer in soft agar: Clinical limitations and pharmacologic applications, in Hofmann V, Berens ME, Martz G (eds): Predictive Drug Testing on Human Tumor Cells, Recent Results in Cancer Research, vol 94. New York, NY, Springer, 1984, pp 41-50 54. Salmon SE, Hamburger AW, Soehnlen B, et al: Quantitation of differential sensitivity of human tumor stem cells to anticancer drugs. N Engl J Med 298:1321-1327, 1978 55. Sridhar KS, Ohnuma T, Nieburgs HE, etal: Relationship between the in vitro colony growth of human tumors and cytologic diagnosis. Cancer 64:2275-2278, 1989 56. Polyzos A, Hug V, Johnston D, et al: The colony forming ability of tumors can predict the survival duration of patients with breast carcinoma. Proc Am Soc Clin Oncol 5:68, 1986 (abstr 264)

Downloaded from ascopubs.org by 37.44.207.18 on January 27, 2017 from 037.044.207.018 Copyright © 2017 American Society of Clinical Oncology. All rights reserved.