Variable selection under multiple imputation using the bootstrap in a prognostic study

Background  

Missing data is a challenging problem in many prognostic studies. Multiple imputation (MI) accounts for imputation uncertainty that allows for adequate statistical testing. We developed and tested a methodology combining MI with bootstrapping techniques for studying prognostic variable selection.     

Use of E mu-PIM-1 transgenic mice short-term in vivo carcinogenicity testing: lymphoma induction by benzo[a]pyrene, but not by TPA

E mu-pim-1 transgenic mice are predisposed to develop lymphomas. Due to their low spontaneous tumour incidence and their increased sensitivity towards the lymphomagen ethylnitrosourea these mice may present an interesting model for short-term carcinogenicity testing. Here, we report on the further exploration of this transgenic mouse model with two additional carcinogens known to have, among others, the lymphohaematopoietic system as target, i.e. benzo[a]pyrene (B[a]P) and 12-O-tetradecanoylphorbol-13-acetate (TPA). B[a]P, given three times a week (by gavage) for 13 weeks at 4.3, 13 or 39 mg/kg body weight, resulted in a dose-related increase in lymphomas up to a 90% incidence in E(mu)-pim-1 mice during the observation period of 40 weeks. B[a]P also induced tumours of the forestomach within this observation period, though at a lower incidence and apparently equally effective in wildtype and transgenic mice. TPA, on the other hand, was unable to induce lymphomas (or tumours in any other organ) in either transgenic or wildtype animals within the observation period of 44 weeks, when applied dermally at the maximum tolerated dose of 3 microg/mouse, twice a week for 35 weeks. Molecular analysis showed that B[a]P-induced lymphomas in transgenic mice were of T-cell origin, 80% of which had elevated levels of c-myc expression. None of the lymphomas had increased N-myc expression and mutation analysis of the ras-gene family revealed a K-ras mutation in only one out of eight tumours investigated. Also, none of the lymphomas showed aberrant expression of p53 as determined by immunohistochemistry. It is concluded that the E mu-pim-1 mouse model will not be very suitable for short-term carcinogenicity testing in general: only genotoxic chemicals that have the lymphohaematopoietic system as target for carcinogenesis in wild- type mice, appear to be efficiently identified.      

Treatment of refractory adult lymphoblastic leukemia (ALL) with 4'(9- acridinylamino) methanesulfon-M-anisidide (AMSA)

Twenty-four adults with ALL were treated with AMSA alone or in combination. Twenty-two were treated at time of relapse and two patients after failing primary induction therapy. All had been treated with anthracyclines prior to receiving AMSA. Of the 22 patients with ALL in relapse, 4 achieved a complete remission. Two of these patients have relapsed while receiving maintenance chemotherapy; one died 1 mo after achieving remission due to the occurrence of cholycystitis in the setting of pancytopenia and one patient underwent bone marrow transplantation and is in remission at 8 mo after the second remission. Both patients who failed primary induction therapy remain in remission at 11 and 36 mo, respectively. The use of AMSA should be considered for patients with ALL who fail primary induction as well as those whose leukemia becomes resistant to conventional agents.     

Effectiveness of methyl-GAG (methylglyoxal-bis[guanylhydrazone]) in patients with advanced malignant lymphoma

We treated 51 patients with advanced malignant lymphoma refractory to conventional therapy with methyl-glyoxal-bis(guanylhydrazone) (methyl- GAG) at doses ranging from 400 to 800 mg/sq m. Therapy was started on a weekly schedule and was switched to every other week in responding patients at the onset of toxicity. Partial responses were observed in 6 of 13 evaluable patients with Hodgkin's disease (46%), 5 of 10 patients with diffuse poorly differentiated lymphocytic lymphoma (50%), 2 of 4 patients with nodular poorly differentiated lymphocytic lymphoma (50%), and 3 of 13 patients with diffuse histiocytic lymphoma (23%). Two of six patients with mycosis fungoides showed objective improvement in cutaneous disease. Toxicity was generally mild and included muscular weakness, myalgia, mucositis, and diarrhea; two patients developed bronchospasm following drug infusions. We conclude that methyl-GAG has major antitumor activity when administered on this schedule to patients with advanced malignant lymphoma. The low degree of toxicity, unique mechanism of action, and minimal myelosuppressive effects suggest that methyl-GAG will prove useful in future trials of combination chemotherapy regimens for the treatment of lymphoma.     

Percutaneous removal of residual intrahepatic stones

In the Orient there is a high frequency of residual intrahepatic stones after biliary tract surgery. Percutaneous removal of residual intrahepatic stones was attempted in a group of 74 patients. Stones were exclusively intrahepatic in 57 patients, whereas 17 patients also had stones in the common bile duct. Biliary strictures were present in 60 cases (81%). A combination of techniques was used, including preshaped angulated catheters, irrigation-suction, balloon dilation of strictures, and crushing of large stones. In 36 cases all stones were removed and in 14 cases most stones were removed, for a success rate of 67.6%. Biliary stricture was the factor most often responsible for failure.     

Breast and axillary tissue MR imaging: correlation of signal intensities and relaxation times with pathologic findings

We tested a variety of inversion-recovery (IR) and spin-echo (SE) sequences by imaging the breast masses of 22 patients before surgery and 23 tissue specimens with magnetic resonance (MR) imaging at 0.6 T to determine the most effective pulse sequences to evaluate breast disease. An SE pulse sequence using a long repetition time (TR) of 1,600 msec and a long echo time (TE) of 90 msec was found to be the most sensitive in depicting carcinoma in the excised tissue specimens, with all of the carcinomas (n = 15) demonstrating irregular areas of higher signal intensity (SI) than that of the adjacent fat. However, only five of 11 breast carcinomas present in the preoperative patients produced a higher SI than that produced by fat on the same T2-weighted sequence. Five of the remaining six carcinomas in the preoperative patients appeared as localized distortions of fibroductular architecture on both T2-weighted SE and IR sequences. In axillary tissue specimens, both metastatic carcinoma and hyperplastic lymph nodes produced a high SI on T2-weighted SE sequences. However, metastatic carcinoma had a significantly longer T2 relaxation time than did hyperplastic lymph nodes.