John Q. Su

Linear Combinations of Multiple Diagnostic Markers

Abstract The receiver operating characteristic (ROC) curve is a simple and meaningful measure to assess the usefulness of diagnostic markers. To use the information carried by multiple markers, we note that Fishers linear discriminant function provides a linear combination of markers to maximize the sensitivity over the entire specificity range uniformly under the multivariate normal distribution model with proportional covariance matrices. With no restriction on covariance matrices, we also provide a solution of the best linear combination of markers in the sense that the area under the ROC curve of this combination is maximized among all possible linear combinations. We illustrate both situations discussed in the article with a cancer clinical trial data.

Prognostic significance of p53 immunostaining in epithelial ovarian cancer.

PURPOSE To evaluate the prognostic significance of p53 expression in epithelial ovarian cancer, including a subset of stage I patients, and to look for correlations between p53 expression and other disease parameters, including stage, grade, age, histologic subtype, second-look results, ploidy, and percent S phase. PATIENTS AND METHODS We analyzed p53 expression in 284 patients with epithelial ovarian cancer using immunohistochemical techniques in paraffin-embedded specimens. There were 36 patients with stage I disease, 20 with stage II disease, 186 with stage III disease, and 42 with stage IV disease. RESULTS p53 immunoreactivity was present in 177 cases (62%). p53 expression was associated with grade 3 to 4 disease (P = .003). The following factors were associated with a decrease in overall survival in a univarate analysis: stage III or IV disease (P = .0001), grade 3 or 4 disease (P = .0001), age above the median (P = .0002), and p53 reactivity (P = .04). In a multivariate analysis, stage, grade, and age retained independent prognostic significance. In the subset of 36 stage I patients, p53 positively approached statistical significance (P = .10) as a negative prognostic factor in a univariate analysis. CONCLUSION Abnormalities of p53 expression occur commonly in epithelial ovarian cancer. Although associated with decreased survival in a univariate analysis, this biologic marker did not retain independent prognostic significance in a multivariate analysis. p53 expression should be studied in a larger cohort of early-stage patients, where accurate prognostic information is needed to direct therapy.

A Lack-of-Fit Test for the Mean Function in a Generalized Linear Model

Abstract A supremum-type statistic, based on partial sums of residuals, is proposed to test the validity of the mean function of the response variable in a generalized linear model. The proposed test does not need a partition of the space of covariates to handle the case of nonreplication. The new test is consistent against the alternative, under which the deterministic part of the model is not the one specified in the null hypothesis. The asymptotic null distribution of the test statistic can be approximated through simulations. This approximation is valid even if the variance of the response variable is misspecified in the model. For practical sample sizes, the adequacy of this large-sample approximation to the null distribution of the test statistic is carefully examined. Power comparisons with other lack-of-fit tests are also performed to show the advantage of the test. In particular, we find that the new procedure is rather sensitive to detect a misspecified mean function, if the assumed mean functio...

Limited-stage small-cell lung cancer: patterns of intrathoracic recurrence and the implications for thoracic radiotherapy.

PURPOSE This analysis was performed to determine the most appropriate volume that should be encompassed by thoracic radiation treatments (TRTs) for patients with limited-stage small-cell lung cancer (LSSCLC) who have responded to initial chemotherapy. PATIENTS AND METHODS A retrospective review of all patients (N = 67) with LSSCLC who were not entered onto a research protocol and were treated at our institution between the years of 1982 and 1990 was performed. Fifty-nine of 67 patients had adequate information regarding the size of the tumor before the start of chemotherapy (computed tomographic [CT] scan of chest or chest x-ray), the size of the tumor before TRT, and the TRT field size based on a simulation radiography. All 59 patients were treated with cyclophosphamide-based chemotherapy, and TRT was generally delivered concomitantly with chemotherapy following two to three cycles of chemotherapy alone. RESULTS Of 59 patients, 28 were treated with TRT field sizes that encompassed postchemotherapy tumor volumes, and 31 patients were treated with TRT field sizes that encompassed prechemotherapy tumor volumes (defined as a volume that included at least a 1.5-cm margin on the prechemotherapy tumor volume). Nineteen patients had an intrathoracic recurrence of disease as the first site of recurrent small-cell carcinoma: 10 of 31 patients treated with TRT fields that encompassed prechemotherapy tumor volumes and nine of 28 patients treated with TRT fields that encompassed postchemotherapy tumor volumes. For the 28 patients treated with TRT fields that encompassed postchemotherapy tumor volumes, the greatest distance that the prechemotherapy tumor volume (without margins) extended beyond the edge of the TRT field was 0.5 to 5.0 cm, with a median of 2.5 cm. All 19 of the intrathoracic recurrences were in-field failures, although two patients (one prechemotherapy volume and one postchemotherapy volume) did have concurrent pleural effusions. CONCLUSION These results indicate that the use of TRT fields that encompass postchemotherapy tumor volumes does not increase the risk of marginal failures or intrathoracic failures outside the TRT field.

Prophylactic cranial irradiation in complete responders with small-cell lung cancer: analysis of the Mayo Clinic and North Central Cancer Treatment Group data bases.

PURPOSE To determine whether prophylactic cranial irradiation (PCI) has an impact on brain failure and survival in patients with small-cell lung cancer (SCLC) who have achieved a complete response to chemotherapy with or without thoracic radiation therapy (TRT). METHODS Between 1975 and 1990, the Mayo Clinic and North Central Cancer Treatment Group entered 1,617 patients on 15 phase II and III SCLC protocols of chemotherapy with or without TRT and PCI. RESULTS Of 772 patients with limited disease, 457 (59%) achieved a complete response, compared with 200 of 845 patients (24%) with extensive disease. With follow-up durations of 2 to 17 years (median, 4), the median survival time and 2-, 5-, and 10-year survival rates for the 457 completely responding limited-disease (LD-CR) patients were 19.6 months, 41%, 17%, and 5%, compared with 13.9 months, 26%, 8%, and 5%, respectively, for the 200 completely responding extensive disease (ED-CR) patients (P = .0001). Multiple prognostic factors, including whether the patient did or did not receive PCI (30 to 38 Gy in 2- to 3.6-Gy fractions) were analyzed. In both univariate and multivariate analyses, PCI was not associated with improved (or worsened) survival. The brain relapse rate was 37% for LD-CR patients who did not receive PCI versus 9% for those who did (P = .0001). In ED-CR patients, the brain relapse rate was 31% without PCI and 8% with (P = .009). Essentially all patients who developed brain relapse died within 2 years, with a median survival time of 3.7 months following relapse. Severe, life-threatening, or fatal CNS toxicity occurred in approximately 3% of patients who received PCI. CONCLUSION The use of PCI remains controversial outside the setting of a clinical trial.

Malignant tracheoesophageal fistula in patients with esophageal cancer

Background. Patients with esophageal cancer and a malignant tracheoesophageal fistula (TEF) have an extremely poor prognosis. Additionally, these patients often are denied treatment with radiation therapy because there is concern that these treatments may increase the size and associated problems of the TEF.

Radiation therapy for diabetes insipidus caused by Langerhans cell histiocytosis

Hypothalamic-pituitary radiation therapy has been the standard treatment for the diabetes insipidus of Langerhans cell histiocytosis. The goal of this study was to assess the role of radiation therapy in Langerhans cell histiocytosis-associated diabetes insipidus and to compare the results with nonirradiated controls. Forty-seven patients with pathologically confirmed Langerhans cell histiocytosis were diagnosed with diabetes insipidus between 1950 and 1989 and were treated at the Mayo Clinic. These patients were divided into two groups on the basis of treatment for the diabetes insipidus: The first group (radiation group) included 30 patients (28 of whom were evaluable for response) who received hypothalamic-pituitary radiation therapy, and the second group (control group) included 17 patients who did not. A partial response to treatment was defined as a reduction in vasopressin dosage or improvement in computed tomography (CT) or magnetic resonance imaging (MRI). A complete response was defined as no further need for vasopressin therapy or normalization of CT or MRI. End points analyzed included treatment response, patient characteristics, morbidity, dose-response relationship, and survival. Patient characteristics of the two groups were similar except for age and lung involvement, both of which were significantly less in the radiation group. Thirty-six percent of patients (10 of 28) in the radiation group responded to hypothalamic-pituitary radiation therapy (22% complete response and 14% partial response), whereas none in the control group responded. Five of the six complete responders were irradiated within 14 days of the diagnosis of diabetes insipidus. The mean dose used in the responding and nonresponding patients was 11.2 and 10 Gy, respectively. Three of five patients (60%) treated with more than 15 Gy responded compared to seven of 23 (30%) treated with less than 15 Gy. Eight of the 10 responders (80%), compared to 16 of 35 nonresponders (46%), were female. Only one in 20 patients with concomitant lung histiocytosis responded. Complications of therapy may include insufficiency in other hypothalamic-pituitary axes in the treated patients. Actuarial survivals at 5, 10, 20, and 40 years for the entire group were 80%, 78%, 75%, and 65%, respectively, with a median follow-up in living patients of 14.7 years.

Splenectomy in advanced chronic lymphocytic leukemia: A single institution experience with 50 patients

PURPOSE To assess the efficacy of splenectomy in the treatment of refractory cytopenias associated with advanced chronic lymphocytic leukemia (CLL). PATIENTS AND METHODS The histories of 57 patients with CLL who underwent splenectomy at the Mayo Clinic between 1975 and 1991 were retrospectively reviewed. Of the 57 patients, 50 underwent splenectomy for reasons directly related to their disease process such as cytopenias or symptomatic splenomegaly. The histories from these 50 patients were studied to assess the response to splenectomy and the operative morbidity and mortality. RESULTS Ninety-four percent of patients were in Rai stage III or IV with extensive marrow infiltration, massive splenomegaly, and cytopenias refractory to chemotherapy. A positive response to splenectomy was defined at 3 months of follow-up as: (1) a hemoglobin level of 11 g/dL or greater in a patient with a preoperative value less than 11 g/dL; or (2) a platelet count of 100 x 10(3)/mm3 or greater in a patient with a preoperative value less than 100 x 10(3)/mm3. A positive response was achieved in 77% of patients with anemia, 70% of patients with thrombocytopenia, and 64% of patients with both anemia and thrombocytopenia. The response was sustained at 1 year of follow-up in 86%, 84%, and 85% of the patients, respectively. Postoperative transfusion requirements decreased correspondingly. The operative morbidity was 26%, and the operative mortality was 4%. The mean duration of hospitalization was 9.8 days (median: 9 days; range: 5 to 24 days). The actuarial median survival after splenectomy was 41 months in responders and 14 months in nonresponders. We found no preoperative parameters that were clearly predictive of a poor hematologic response. In particular, outcome was not affected by preoperative spleen size or the degree of marrow infiltration by CLL. All patients with symptomatic splenomegaly had an improved sense of well-being. CONCLUSION In this, the largest single institution study to date, we found splenectomy to be efficacious in providing durable remissions of refractory cytopenias and in relieving symptomatic splenomegaly in the majority of patients with CLL. The procedure is associated with a low perioperative mortality. Although the impact on survival is uncertain, the improved peripheral blood counts may allow the administration of adequate doses of myelosuppressive chemotherapy.

A phase II study of recombinant human alpha-interferon in advanced hormone-refractory prostate cancer

To determine the efficacy of recombinant human leukocyte alpha‐interferon (IFL‐RA) in advanced hormone‐refractory prostate cancer, the authors treated 40 patients with IFL‐RA administered intramuscularly at a dose of 10 × 108 U/m2 three times weekly. Toxicity was substantial and necessitated at least a 50% dose reduction in all but five patients during the first 1‐2 months of therapy. No responses were observed in patients with bone metastases, but complete and partial regression of nodal disease were observed in two patients with extraosseous disease (overall response rate, 5%; 95% confidence interval, 0.64‐17.75%). The authors conclude that IFL‐RA cannot be recommended at this dose and schedule in patients with advanced prostate cancer, but additional study of its use in patients with nodal disease may be warranted. Cancer 1992; 702310‐2312.

Measurable or assessable disease in lung cancer trials: does it matter?

PURPOSE The goals of this study were to analyze and compare the major clinical response rates and survival of patients with either measurable or assessable disease status to treatment with systemic chemotherapy. PATIENTS AND METHODS All patients had stage IIIB or IV non-small-cell lung cancer (NSCLC) and were enrolled onto three consecutive phase III clinical trials. Patients were stratified by disease status (measurable or assessable) before randomization to systemic chemotherapy. The three trials were conducted in the setting of a multicenter cooperative oncology group. Composite data were obtained for the three trials. Major clinical responses, time to progression, and survival were analyzed and compared in patients with measurable or assessable disease using standard statistical methods. RESULTS Four hundred twenty-six patients were enrolled onto the three trials from June 1981 through August 1990. Measurable disease was present in 236 patients (55%) and assessable disease in 190 (45%). Each study was well balanced for the number of patients with measurable or assessable disease on either treatment regimen. A major clinical response was observed in 71 patients with measurable disease (30%; 95% confidence interval [CI], 24 to 36). Forty patients with assessable disease responded to treatment (21%; 95% CI, 16 to 28) (P = .04). The time to progression for all patients (P = .23) and for responders only (P = .10) was not significantly different based on disease status. Overall survival and survival of responders only was not significantly different, but patients with assessable disease tended to do better. Using multivariate analysis, sex and disease status had a borderline influence on the major response rate (P = .05). Performance score (PS) was the only factor that was significantly correlated with survival. CONCLUSION NSCLC patients with assessable disease have major clinical response rates, time to progression, and survival that are similar to those of NSCLC patients with measurable disease. This study supports the inclusion of patients with assessable-disease lung cancer in both phase II and III trials conducted in the cooperative group setting.