Analysing longitudinal continuous quality of life data with dropout.

Quality of Life (QL) is becoming an increasingly popular endpoint in phase III cancer clinical trials. However, there is still no agreement as to what is the optimal approach to analysis. In this paper we review some concepts which should be considered during a QL analysis. We present two modelling approaches that have been substantively developed in other research fields: selection models and pattern-mixture models. These models are compared using data from an EORTC clinical trial in poor-prognosis prostate cancer patients. It is illustrated that, although selection models and pattern mixture are probabilistically equivalent, they may shed completely different light on data from a modeller's point of view.

Why phase III trials of maximal androgen blockade versus castration in M1 prostate cancer rarely show statistically significant differences.

BACKGROUND: The meta-analysis of maximal androgen blockade (MAB) concluded that there is no survival advantage of MAB over castration alone. However, the results from the largest trials yield conflicting results. METHODS: The design and results of three trials were examined. RESULTS: Most studies were planned to detect an over-optimistic difference in survival and immature data were published. The survival curves show that statistical assumptions are not fulfilled. Excluding from the meta-analysis all trials where a negative impact of disease flare on survival could not be excluded resulted in no difference in survival between MAB and castration. CONCLUSIONS: Trials of MAB should be planned to detect differences of no more than 5-10% in median survival. The analyses should only be carried out on mature data and should take into account the possibility of a negative impact on survival due to disease flare if no anti-androgen has been given initially with an LH-RH agonist.

Differences in risk factors for local and distant recurrence after breast-conserving therapy or mastectomy for stage I and II breast cancer: pooled results of two large European randomized trials.

PURPOSE: Risk factors for local and distant recurrence after breast-conserving therapy and mastectomy were compared to define guidelines for the decision making between both treatments. PATIENTS AND METHODS: The data of two randomized clinical trials for stage I and II breast cancer patients were pooled. The total number of patients in the study was 1,772, of whom 879 underwent breast conservation, and 893, modified radical mastectomy. Representative slides of the primary tumor were available for histopathologic review in 1,610 cases (91%). RESULTS: There were 79 patients with local recurrence after breast-conservation and 80 after mastectomy, the 10-year rates being 10% (95% confidence interval [CI], 8% to 13%) and 9% (95% CI, 7% to 12%), respectively. Age no more than 35 years (compared with age >60: hazard ratio [HR], 9.24; 95% CI, 3.74 to 22.81) and an extensive intraductal component (HR, 2.52; 95% CI, 1.26 to 5.00) were significantly associated with an increased risk of local recurrence after breast-conserving therapy. Vascular invasion was predictive of the risk of local recurrence, irrespective of the type of primary treatment (P <.01). Tumor size, nodal status, high histologic grade, and vascular invasion were all highly significant predictors of distant disease after breast-conserving therapy and mastectomy (P <.01). Age no more than 35 years and microscopic involvement of the excision margin were additional independent predictors of distant disease after breast-conserving therapy (P <.01). CONCLUSION: Age no more than 35 years and the presence of an extensive intraductal component are associated with an increased risk of local recurrence after breast-conserving therapy. Vascular invasion causes a higher risk of local recurrence after mastectomy as well as after breast-conserving therapy and should therefore not be used for deciding between the two treatments.

Long-term results of a randomized trial comparing breast-conserving therapy with mastectomy: European Organization for Research and Treatment of Cancer 10801 trial.

BACKGROUND: Breast-conserving therapy (BCT) has been shown to be as effective as mastectomy in the treatment of tumors 2 cm or smaller. However, evidence of its efficacy, over the long term, in patients with tumors larger than 2 cm is limited. From May 1980 to May 1986, the European Organization for Research and Treatment of Cancer carried out a randomized, multicenter trial comparing BCT with modified radical mastectomy for patients with tumors up to 5 cm. In this analysis, we investigated whether the treatments resulted in different overall survival, time to distant metastasis, or time to locoregional recurrence. METHODS: Of 868 eligible breast cancer patients randomly assigned to the BCT arm or to the modified radical mastectomy arm, 80% had a tumor of 2.1-5 cm. BCT comprised lumpectomy with an attempted margin of 1 cm of healthy tissue and complete axillary clearance, followed by radiotherapy to the breast and a supplementary dose to the tumor bed. The median follow-up was 13.4 years. All P values are two-sided. RESULTS: At 10 years, there was no difference between the two groups in overall survival (66% for the mastectomy patients and 65% for the BCT patients; P =.11) or in their distant metastasis-free rates (66% for the mastectomy patients and 61% for the BCT patients; P =.24). The rate of locoregional recurrence (occurring before or at the same time as distant metastasis) at 10 years did show a statistically significant difference (12% of the mastectomy and 20% of the BCT patients; P =. 01). CONCLUSIONS: BCT and mastectomy demonstrate similar survival rates in a trial in which the great majority of the patients had stage II breast cancer.

Orchiectomy and orchiectomy plus mitomycin C for metastatic prostate cancer in patients with poor prognosis: the final results of a European Organization for Research in Cancer Therapy Genitourinary Group Trial.

PURPOSE: The outcome of patients with symptomatic metastatic prostate cancer is poor and improved treatment regimens are urgently needed. Theoretically, the combination of orchiectomy and chemotherapy could reduce androgen sensitive and insensitive cells in the prostate. This European Organization for Research in Cancer Therapy Genitourinary Group randomized, multicenter phase III trial demonstrates the outcome of orchiectomy alone versus orchiectomy followed by intravenous mitomycin C. MATERIALS AND METHODS: A total of 189 patients with metastatic prostate cancer and poor prognostic factors were randomized in this trial by 42 institutions. Of these patients 184 (97%) were eligible for study, including 90 treated with orchiectomy alone (orchiectomy only arm) and 94 treated with orchiectomy followed by 15 mg./m.2 mitomycin C in 1 week (combined treatment arm). Mitomycin C was administered every 6 weeks and treatment was continued as long as tolerance and patient compliance allowed, and no progression was observed. Objective and subjective criteria for progression were clearly defined in the protocol. RESULTS: Patient and tumor characteristics were well balanced between the 2 treatment arms. At a median followup of 4.2 years 144 patients had died, including 112 of prostate cancer. No significant differences for time to overall (p = 0.17), subjective (p = 0.25) and objective (p = 0.08) progression were found between the 2 treatment groups. For progression-free survival no difference was noted (p = 0.67) between the 2 treatment groups but a trend in favor of orchiectomy alone was observed for overall survival (p = 0.04). Mitomycin C induced considerable hematological, gastrointestinal, renal and pulmonary toxicity leading to discontinuation in 31% of patients with pulmonary toxicity and 7% with renal deterioration. In addition, the quality of life evaluation revealed significant reduction in the combined treatment arm. CONCLUSIONS: Based on the results of this randomized phase III study orchiectomy plus mitomycin C for metastatic prostate cancer in patients with poor prognostic factors cannot be recommended due to failure of improvement in survival and reduced quality of life parameters.

Impact of the treating institution on survival of patients with "poor-prognosis" metastatic nonseminoma. European Organization for Research and Treatment of Cancer Genito-Urinary Tract Cancer Collaborative Group and the Medical Research Council Testicular Cancer Working Party.

BACKGROUND: Because metastatic nonseminomatous germ cell cancer is a rare but treatable cancer, we have explored whether there is an association between the experience of the treating institution with this disease and the long-term clinical outcome of the patients, particularly patients with a poor prognosis. METHODS: We analyzed data on 380 patients treated in one of 49 institutions participating in the European Organization for Research and Treatment of Cancer/ Medical Research Council randomized trial of four cycles of bleomycin-etoposide-cisplatin followed by two cycles of etoposide-cisplatin versus three cycles of bleomycin-vincristine-cisplatin followed by three cycles of etoposide-ifosfamide-cisplatin-bleomycin, both treatment regimens given with or without filgrastim (granulocyte colony-stimulating factor). Institutions were divided into four groups based on the total number of patients entered in the trial. The groups were compared by use of the Cox proportional hazards model stratified for treatment with filgrastim and for patient prognosis as defined by the International Germ Cell Consensus Classification Group. With the use of this classification, only 65 % of the patients had a poor prognosis. RESULTS: Patients treated in the 26 institutions that entered fewer than five patients into the trial had an overall survival that was statistically significantly worse (two-sided P = .010; hazard ratio = 1.85; 95% confidence interval = 1.16-3.03) than that of patients treated in the 23 institutions that entered five patients or more. Overall survival and failure-free survival were similar among institutions that entered at least five patients. The observed effect may be related to differences in adherence to the chemotherapy protocol and in the frequency and extent of surgery for residual masses, although only the differences in dose intensity achieved statistical significance. CONCLUSIONS: Patients treated in institutions that entered fewer than five patients into the trial appeared to have poorer survival than those treated in institutions that entered a larger number of patients with "poor-prognosis" nonseminoma.

Sample size estimation in phase III cancer clinical trials.

This paper deals with the basic principles involved in sample size calculation of phase III cancer clinical trials. It illustrates the concepts and factors determining the sample size. Various examples of phase III cancer clinical trials are provided and the sample size is calculated taking into account the assumptions made. The examples provided include sample sizes for comparing proportions and sample sizes for comparing survival times. Several special topics are also discussed including choice of endpoint, number of treatment groups, factorial designs and equivalence trials.

Identifying the types of missingness in quality of life data from clinical trials.

This paper discusses methods of identifying the types of missingness in quality of life (QOL) data in cancer clinical trials. The first approach involves collecting information on why the QOL questionnaires were not completed. Based on the reasons provided one may be able to distinguish the mechanisms causing missing data. The second approach is to model the missing data mechanism and perform hypothesis testing to determine the missing data processes. Two methods of testing if missing data are missing completely at random (MCAR) are presented and applied to incomplete longitudinal QOL data obtained from international multi-centre cancer clinical trials. The first method (Ridout, 1991) is based on a logistic regression and the second method (Park and Davis, 1993) is based on an adaptation of weighted least squares. In one application (advanced breast cancer) missing data was not likely to be MCAR. In the second application (adjuvant breast cancer) the missing mechanism was dependent on the QOL scale under study. MCAR and missing at random (MAR) have distinct consequences for data analysis. Therefore it is relevant to distinguish between them. However, if either MCAR or MAR hold, likelihood or Bayesian inferences can be based solely on the observed data, although for MAR, depending on the research question, modelling the dropout mechanism may still be necessary. Distinguishing between MAR and missing not at random (MNAR) is not trivial and relies on fundamentally untestable assumptions.

The importance of prognostic factors in the interpretation of two EORTC metastatic prostate cancer trials. European Organization for Research and Treatment of Cancer (EORTC) Genito-Urinary Tract Cancer Cooperative Group.

INTRODUCTION AND OBJECTIVES: The EORTC conducted two randomized phase III trials of maximal androgen blockade (MAB) in 695 patients with metastatic prostate cancer. Trial 30,843 compared orchidectomy or buserelin to buserelin plus cyproterone acetate and showed no significant difference in survival while trial 30,853 showed that Zoladex plus flutamide had a significantly longer survival than orchidectomy. Reasons for this discrepancy were sought. METHODS: In order to determine whether differences in patient characteristics could explain these possibly contradictory results, a Cox proportional hazards regression model was used to identify prognostic factors for survival in each study. Patients were divided into risk groups (good or poor prognosis with 3.5 and 1.75 years' median survival, respectively) based on their alkaline phosphatase, hemoglobin, performance status, pain score, T category and G grade at entry on study. RESULTS: The survival advantage of MAB in 30,853 was limited to patients with a good prognosis (164/302 (54%) of the patients). In 30,843, only 93/337 patients (28%) had a good prognosis so there were insufficient data to draw separate conclusions in these patients. Despite the limitations of subgroup analyses, these results show that patients in 30,843 had on the average a worse prognosis than patients in 30,853. Hence there were fewer good prognosis patients who could potentially benefit from MAB, thus providing one possible explanation for the overall negative conclusion. CONCLUSIONS: These studies once again underline the importance of taking into account patient characteristics when designing and interpreting metastatic prostate cancer trials. They also provide criteria which may be used to define risk groups as part of a protocol's patient eligibility criteria. In the design of future trials assessing MAB, a sufficient number of good prognosis patients should be entered to reliably assess treatment efficacy in this subgroup.

Overview of randomized perioperative polychemotherapy trials in women with early-stage breast cancer.

PURPOSE: To determine whether perioperative polychemotherapy (PeCT) can significantly prolong the overall survival of women with early-stage breast cancer. METHODS: A meta-analysis that used updated individual patient data from all available randomized trials of PeCT, both published and unpublished, was conducted. Data on 6,093 patients (1,124 deaths and 1,912 recurrences) from five clinical trials were available (median follow-up duration, 5.3 years; maximum, 11.3 years). RESULTS: No significant effect of PeCT on overall survival was observed. However, patients who received PeCT had a significantly longer disease-free survival (hazards ratio [HR], 0.89; 95% confidence interval [CI], 0.82 to 0.98; P = .02). Time to local recurrence was significantly prolonged in the PeCT arm (HR, 0.68; 95% CI, 0.58 to 0.80; P < .0001). Likewise, there was a borderline significant difference in favor of PeCT in terms of time to distant metastases (HR, 0.90; 95% CI, 0.81 to 1.00; P = .05). Subgroup analyses suggest that node-negative women benefited the most from treatment. CONCLUSION: At present, there is no evidence that PeCT is able to prolong overall survival in patients with early-stage breast cancer; however, further follow-up evaluation is required. PeCT significantly prolongs disease-free survival, especially in node-negative women, which emphasizes once more the need for clinical trials in this subgroup.

Randomized trial of bleomycin, etoposide, and cisplatin compared with bleomycin, etoposide, and carboplatin in good-prognosis metastatic nonseminomatous germ cell cancer: a Multiinstitutional Medical Research Council/European Organization for Research and Treatment of Cancer Trial.

PURPOSE: This prospective randomized multicenter trial was designed to evaluate the efficacy of carboplatin plus etoposide and bleomycin (CEB) versus cisplatin plus etoposide and bleomycin (BEP) in first-line chemotherapy of patients with good-risk nonseminomatous germ cell tumors. PATIENTS AND METHODS: Between September 1989 and May 1993, a total of 598 patients with good-risk nonseminomatous germ cell tumors were randomized to receive four cycles of either BEP or CEB. In each cycle, the etoposide dose was 120 mg/m2 on days 1, 2, and 3, and the bleomycin dose was 30 U on day 2. BEP patients received cisplatin at 20 mg/m2/d on days 1 to 5 or 50 mg/m2 on days 1 and 2. For CEB patients, the carboplatin dose was calculated from the glomerular filtration rate to achieve a serum concentration x time of 5 mg/mL x minutes. Chemotherapy was recycled at 21-day intervals to a total of four cycles. RESULTS: Of patients assessable for response, 253 of 268 (94.4%) of those allocated to receive BEP achieved a complete response, compared with 227 of 260 (87.3%) allocated to receive CEB (P = .009). There were 30 treatment failures in the 300 patients allocated to BEP and 79 in the 298 allocated to CEB (log-rank chi 2 = 26.9; P < .001), which led to failure-free rates at 1 year of 91% (95% confidence interval [CI], 88% to 94%) and 77% (95% CI, 72% to 82%), respectively. There were 10 deaths in patients allocated to BEP and 27 in patients allocated to CEB (log-rank chi 2 = 8.77; P = .003), which led to 3-year survival rates of 97% (95% CI, 95% to 99%) and 90% (95% CI, 86% to 94%), respectively. CONCLUSION: With these drug doses and schedules, combination chemotherapy based on carboplatin was inferior to that based on cisplatin. This BEP regimen that contains moderate doses of etoposide and bleomycin is effective in the treatment of patients with good-prognosis metastatic nonseminoma.

Ten-year results of a randomized trial evaluating prolonged low-dose adjuvant chemotherapy in node-positive breast cancer: a joint European Organization for Research and Treatment of Cancer-Dutch Breast Cancer Working Party Study. Cooperating Investigators.

PURPOSE: To investigate whether treatment with prolonged low-dose adjuvant chemotherapy could improve survival of patients with axillary node-positive breast cancer. PATIENTS AND METHODS: Four hundred fifty-two patients with axillary node-positive breast cancer who received postoperative irradiation were prospectively randomized in a trial (European Organization for Research and Treatment of Cancer [EORTC] 09771) that compared surgery followed by prolonged low-dose chemotherapy versus surgery alone. Chemotherapy was given for a period of 2 years and consisted of monthly courses of cyclophosphamide 50 mg/m2 orally on days 1 to 14, methotrexate 15 mg/m2 intravenously on days 1 and 8, and fluorouracil 350 mg/m2 intravenously on days 1 and 8 (CMF). RESULTS: At a median follow-up time of 10 years, the overall survival duration was significantly prolonged in the chemotherapy arm (hazards ratio, 0.75; 95% confidence interval, 0.56 to 0.99; P = .04). Ten-year overall survival rates (+/- SE) were 59% (+/- 3.6%) for the chemotherapy arm and 50% (+/- 3.7%) for the control arm. Time to local relapse was significantly prolonged in the chemotherapy arm (hazards ratio, 0.63; 95% confidence interval, 0.42 to 0.94; P = .02). Patients with one to three positive axillary nodes and patients with estrogen receptor-negative tumors especially benefited from chemotherapy. Toxicity was observed in 93% of patients. CONCLUSION: We conclude that prolonged low-dose adjuvant CMF can significantly prolong overall survival in patients with node-positive breast cancer. However, considering the fact that toxicity was still considerable despite reducing the dose of chemotherapy by 50%, we believe that conventionally dosed short-term regimens are preferable in the treatment of node-positive breast cancer.

Assessment of estrogenic recruitment before chemotherapy in advanced breast cancer: preliminary results of a double-blind randomized study of the EORTC Breast Cancer Cooperative Group.

We investigated whether estrogenic recruitment could enhance the antitumor effect of chemotherapy in 165 patients with advanced breast cancer, presumably sensitive to hormonal treatments (ER + and/or PgR + lesions). The therapeutic regimen consisted of: (a) estrogenic suppression by aminoglutethimide 1 g/day + hydrocortisone 40 mg/day; surgical castration in premenopausal patients only; (b) FAC (5FU 500 mg/m2; ADM 50 mg/m2; CPA 500 mg/m2) for 3 weeks; (c) following randomization, exactly 24 h prior to chemotherapy, patients had to take 1 tablet of either placebo (PL) or 50 microgram ethinylestradiol (EE2). Tolerance, responses, time to progression and median survival were identical in both groups. Thus, EE2 before chemotherapy did not contribute to the efficacy of this particular therapeutic regimen, which yielded an overall response rate of 64%. We conclude that the validity of the hormonal recruitment concept has not yet been established in clinical practice, so that this approach remains experimental.

A bayesian approach to the design of phase II clinical trials.

A new strategy for the design of Phase II clinical trials is presented which utilizes the information provided by the prior distribution of the response rate, the costs of treating a patient, and the losses or gains resulting from the decisions taken at the completion of the study. A risk function is derived from which one may determine the optimal Bayes sampling plan. The decision theoretic/Bayesian approach is shown to provide a formal justification for the sample sizes often used in practice and shows the conditions under which such sample sizes are clearly inappropriate.