Survival Analysis in the Presence of Competing Risks: The Example of Waitlisted Kidney Transplant Candidates

Competing events (or risks) preclude the observation of an event of interest or alter the probability of the event's occurrence and are commonly encountered in transplant outcomes research. Transplantation, for example, is a competing event for death on the waiting list because receiving a transplant may significantly decrease the risk of long-term mortality. In a typical analysis of time-to-event data, competing events may be censored or incorporated into composite end points; however, the presence of competing events violates the assumption of "independent censoring," which is the basis of standard survival analysis techniques. The use of composite end points disregards the possibility that competing events may be related to the exposure in a way that is different from the other components of the composite. Using data from the Scientific Registry of Transplant Recipients, this paper reviews the principles of competing risks analysis; outlines approaches for analyzing data with competing events (cause-specific and subdistribution hazards models); compares the estimates obtained from standard survival analysis, which handle competing events as censoring events; discusses the appropriate settings in which each of the two approaches could be used; and contrasts their interpretation.

Hypoxia Marker GLUT-1 (Glucose Transporter 1) is an Independent Prognostic Factor for Survival in Bladder Cancer Patients Treated with Radical Cystectomy

BACKGROUND

Tumour hypoxia, which is frequent in many cancer types, is associated with treatment resistance and poor prognosis. The role of hypoxia in surgically treated bladder cancer (BC) is not well described. We studied the role of hypoxia in two independent series of urothelial bladder cancers treated with radical cystectomy.

METHODS

279 patients from the University Hospital Network (UHN), Toronto, Canada, and Turku University, Finland were studied. Hypoxia biomarkers (HIF1-α, CAIX, GLUT-1) and proliferation marker Ki-67 were analyzed with immunohistochemistry using defined tissue microarrays. Kaplan-Meier methods and Cox proportional hazards regression models were used to investigate prognostic role of the factors.

RESULTS

In univariate analyses, strong GLUT-1 positivity and a high Ki-67 index were associated with poor survival. In multivariate model containing clinical prognostic variables, GLUT-1 was an independent prognostic factor associated with worse disease-specific survival (HR 2.9, 95% CI 0.7-12.6, Wald p = 0.15 in the Toronto cohort and HR 3.2, 95% CI 1.3-7.5, Wald p = 0.0085 in the Turku cohort).

CONCLUSION

GLUT-1 is frequently upregulated and is an independent prognostic factor in surgically treated bladder cancer. Further studies are needed to evaluate the potential role of hypoxia-based and targeted therapies in hypoxic bladder tumours.

Analysis of the intra- and intertumoral heterogeneity of hypoxia in pancreatic cancer patients receiving the nitroimidazole tracer pimonidazole

Background:

Hypoxia is thought to be an adverse feature of pancreatic cancer, but direct measurement in patients is technically challenging. To address this, we characterised the intra/interpatient heterogeneity of hypoxia in surgical specimens from patients who received the 2-nitroimidazole tracer pimonidazole pre-operatively.

Methods:

Pimondazole was given intravenously 16–20 h before pancreatectomy, and the extent and intratumoral heterogeneity of hypoxia determined by image analysis applied to multiple tissue blocks stained by immunohistochemistry. Intra/interpatient heterogeneity was estimated by variance component analysis.

Results:

Pimonidazole staining was analysed in 10 tumours. The extent of labelling varied amongst patients (0–26%), with a broader range of hypoxia in the epithelial (1–39%) compared with the stromal (1–13%) compartments. Variance component analysis demonstrated greater inter- than intrapatient variability of hypoxia, and that multiple (4–5) tumour sections are required to provide a consistent evaluation of its extent in individual tumours.

Conclusions:

There is significant intra- and intertumoral heterogeneity of hypoxia in pancreatic cancers, and these do not appear to be generally more hypoxic than other cancer types. This study establishes the feasibility to assess hypoxia in pancreatic cancer patients using pimonidazole, but questions the reliability of measurements made using a single tissue section.

The role of elective para-aortic lymph node irradiation in patients with locally advanced cervical cancer

AIMS

Pelvic lymph node positivity in cervical cancer is known to be an adverse prognostic factor and is associated with an elevated risk of clinically occult para-aortic lymph node metastases. The purpose of this study was to examine the benefit of elective para-aortic lymph node radiotherapy (PART) in patients with no clinical or radiographic evidence of para-aortic lymph node metastases receiving concurrent cisplatin chemotherapy.

MATERIALS AND METHODS

Patients treated with radiotherapy and concurrent cisplatin for cervical cancer from 1999 to 2009 were identified in two prospective databases. All patients received external beam pelvic radiotherapy (PRT) to a median dose of 50 Gy concurrently with weekly cisplatin 40 mg/m(2). This was followed by pulse dose rate intracavitary brachytherapy to a median dose of 40 Gy. Patients at high risk of occult para-aortic metastases also received PART to a median dose of 40 Gy.

RESULTS

There were 228 patients suitable for analysis; the median follow-up was 4.6 years. The addition of PART to PRT was not associated with a significant difference in disease-free survival (hazard ratio 1.1, confidence interval 0.7-1.8, P = 0.75) or overall survival (hazard ratio 1.6, confidence interval 0.9-2.7, P = 0.11) on multivariate analysis. There was no significant difference in the rate of para-aortic relapse with PART versus PRT (hazard ratio 2.01, confidence interval 0.79-5.12, P = 0.14). The 3 year grade 3-4 late toxicities were 11% for the PART group versus 8% for PRT (hazard ratio 1.39, confidence interval 0.58-3.37, P = 0.47).

CONCLUSIONS

These results suggest that cervical cancer patients treated with radiotherapy and concurrent cisplatin do not benefit from elective PART.

Abstract P2-10-33: Mitotic Component of Grade Can Distinguish Breast Cancer Patients at Greatest Risk of Local Relapse

Background: With the recent recognition of many different molecular subtypes of breast cancer a desire to more specifically categorize tumors to allow tailoring of treatment to individual patients has developed. This has largely involved the development of molecular tests rather than the re-examination of current pathologic criteria. We wanted to evaluate standard pathologic features to determine their ability to predict for local recurrence.

Materials and Methods: Slides were retrieved for review from 280 of 769 women who had participated in a trial of tamoxifen with or without breast irradiation between December 1992 and June 2000 and for whom outcome data up to 18 years was available. All women were 50 years of age or older at the time of enrollment and had T1 or T2 node negative breast cancer. The cases for which slides were obtained were representative of the whole group. The slides were reviewed by two breast pathologists (SJD and NAM). Several features were evaluated; modified Nottingham histologic grade and its components- degree of tubule formation, nuclear pleomorphism and mitotic count. Mitotic component of grade was calibrated to the microscopic field size used. The presence of lymphatic/vascular space invasion was also scored. A statistical analysis was performed to relate these pathologic features to local recurrence at up to 18 years.

Results: The strongest predictor of local recurrence was the mitotic component of the Nottingham histologic grade with 5.7% for mitotic score 1/3 (n = 200), 19.6% for mitotic score 2/3 (n = 37) and 19.8% for mitotic score 3/3 (n = 43)(Gray's p-value = 0.0021). Overall grade was also able to predict for local recurrence with 2.6% for Grade 1 (n = 49), 10.6% for Grade 2 (n = 162) and 17.9% for Grade 3 (n = 71)(Gray's p-value=0.026). However, neither architecture (0% vs. 9.5% vs. 9.8%, Gray's p-value=0.74) nor degree of nuclear pleomorphism (0% vs. 7.9% vs. 11.5%, Gray's p-value=0.37), the other components of histologic grade, showed a statistically significant difference for recurrence. The presence or absence of endothelial lined space invasion was also found to be not statistically different (9.3% vs. 13%, Gray's p-value=0.55).

Conclusion: Within this cohort of tamoxifen treated T1 and T2 breast cancer patients 50 years of age or older, mitotic index could stratify women into groups with high and low risk of recurrence. If validated this may be a useful way of allocating patients to different treatment groups. Additional validation studies are planned on similar groups of patients.

Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr P2-10-33.

Plasma osteopontin as a biomarker of prostate cancer aggression: relationship to risk category and treatment response

BACKGROUND

High plasma osteopontin (OPN) has been linked to tumour hypoxia, metastasis, and poor prognosis. This study aims to assess whether plasma osteopontin was a biomarker of increasing progression within prostate cancer (PCa) prognostic groups and whether it reflected treatment response to local and systemic therapies.

METHODS

Baseline OPN was determined in men with localised (n=199), locally recurrent (n=9) and castrate-resistant, metastatic PCa (CRPC-MET; n=37). Receiver-operating curves (ROC) were generated to describe the accuracy of OPN for distinguishing between localised risk groups or localised vs metastatic disease. We also measured OPN pre- and posttreatment, following radical prostatectomy, external beam radiotherapy (EBRT), androgen deprivation (AD) or taxane-based chemotherapy.

RESULTS

The CRPC-MET patients had increased baseline values (mean 219; 56-513 ng ml(-1); P<0.0001) compared with the localised, non-metastatic group (mean 72; 12-438 ng ml(-1)). The area under the ROC to differentiate localised vs metastatic disease was improved when OPN was added to prostate-specific antigen (PSA) (0.943-0.969). Osteopontin neither distinguished high-risk PCa from other localised PCa nor correlated with serum PSA at baseline. Osteopontin levels reduced in low-risk patients after radical prostatectomy (P=0.005) and in CRPC-MET patients after chemotherapy (P=0.027), but not after EBRT or AD.

CONCLUSION

Plasma OPN is as good as PSA at predicting treatment response in CRPC-MET patients after chemotherapy. Our data do not support the use of plasma OPN as a biomarker of increasing tumour burden within localised PCa.

Use of genetic instability to predict biochemical recurrence in intermediate-risk prostate cancer

42

Background: Biomarkers of local and systemic recurrence are needed to individualize patient risk categories and better define treatment. We hypothesized that genomic instability, as measured by percent genome alteration (PGA), can predict for biochemical failure in intermediate- risk prostate cancer.

Methods: High-resolution array comparative genomic hybridization (arrayCGH) was used to identify PGA in frozen biopsies from 120 intermediate-risk prostate cancer patients. Our cohort included 39 T1c tumors, 78 T2 tumors and 2 T3 tumors. The Gleason score was 6 in 32 tumors, 7 in 82 tumors and 8–9 in 6 tumors. PSA ranged from 2.1–33 (median 8.0). Patients were treated with intensity-modulated radiotherapy (IMRT) with doses of 75.6–79.8 Gy in 1.8–2Gy fractions, or 60–66 Gy in 3 Gy fractions.. Twenty-five percent of patients also received neoadjuvant-concurrent bicalutamide (150mg po od). Biochemical failure, defined by Phoenix criteria or the initiation of salvage therapy, was observed in 35 patients after median follow-up of 5.4 years (range 0.9–8.8).

Results: Array CGH showed variable PGA ranging from <1% to 35% (median 6.7%). PSA and the use of hormonal therapy independently influenced biochemical relapse, and formed a baseline clinical model to which PGA was added. PGA was found to be a strong predictor of biochemical relapse (p<0.0001) independent of the clinical prognostic factors (pre-treatment PSA, Gleason score and T-category). PGA was also found to be associated with unique tumour suppressor and oncogene gene loci clusters involved in genetic stability (e.g. loss of PTEN, p53, RB, NKX3.1, ATM, PARP-1 and gain of c-MYC; validated by in situ FISH).

Conclusions: This is the first report to show genetic instability can independently predict for biochemical recurrence in intermediate-risk prostate cancer. Current studies are associating specific gene loci regions with clinical outcome. Our results could provide a way forward for individualized medicine for non-indolent prostate cancer based on initial daignostic biopsy material. Supported by Prostate Cancer Canada, The Terry Fox Foundation and the Canadian Cancer Society.

No significant financial relationships to disclose.

High-dose chemotherapy and auto-SCT in elderly patients with Hodgkin's lymphoma

Our purpose was to assess efficacy and toxicity of high-dose chemotherapy (HDCT) and ASCT in patients with relapsed and refractory Hodgkin's lymphoma (HL) aged 60 years and older and compare the results with a group of younger HL patients treated in a similar manner. We identified 15 consecutive patients, with HL aged 60 years and older who underwent HDCT (etoposide 60 mg/kg+ melphalan 160 mg/m(2)) and ASCT at our institution from May 2001 to March 2008. The results were compared with a cohort of 157 younger HL patients treated in a similar manner from January 1999 to December 2006. After a median follow-up of 2.5 years, PFS at 3 years after ASCT was 73% (95% confidence interval (CI) 37-90) for the older group and 56% (95% CI 46-64) for the younger group (P=0.45); OS after ASCT was 88% (95% CI 39-98) for the older group and 84% (95% CI 75-90) for the younger group (P=0.80). No transplant-related deaths were seen. Our study suggests that ASCT is feasible for selected elderly patients with HL, giving similar results to younger patients in terms of survival and toxicity.

Amplification of the Prolactin Receptor Gene in Mammary Lobular Neoplasia

Background: Lobular carcinoma in situ (LCIS) has long been considered a marker of increased risk of cancer in both breasts. However, recent studies have shown that it can behave as a non-obligate precursor lesion as well. The biology and natural history of LCIS still remains ill defined in part because it is a challenging lesion to study as it usually does not have an identifiable gross appearance and is only recognized in fixed tissue specimens.Material and Methods: Using array comparative genomic hybridization (aCGH) we have analyzed regions of amplification found in LCIS and adjacent invasive lobular carcinoma (ILC) in a series of thirteen cases of archival patient samples from our institution. Degenerate oligonucleotide primed (DOP) PCR was performed for whole genome amplification of the extracted DNA from microdissected tissue samples prior to microarray analysis. Analysis of microarray data was performed using Significance Analysis of Microarrays (SAM). Of the ten most amplified genes in LCIS (highest SAM scores), one was selected for quantitative real time PCR (Q-PCR) validation due to the limited amount of material available from these cases. Q-PCR validation was performed on samples from 8 cases of LCIS and invasive lobular carcinoma and 12 archival cases of ductal carcinoma in situ (DCIS) and invasive ductal carcinoma (IDC) for comparison.Results: Amongst the 10 genes with the highest SAM scores, the prolactin receptor gene (PRLR) was selected for further Q-PCR validation in our limited samples. Amplification of PRLR was confirmed by Q-PCR in 4/8 (50%) of cases of LCIS and 4/8 (50%) cases of ILC, compared to 0/12 (0%) cases of DCIS and 3/12 (25%) cases of IDC. When LCIS and ILC were combined into one group, there was more amplification of the prolactin receptor gene when compared to the DCIS and IDC group (p= 0.01). The level of amplification between the two groups also differed in the range of copy number values, which was lower in the ductal group (0.78-1.58, n=24) compared to the lobular group (0.92-3.68, n=16) (p&lt;0.05).Conclusion: We have identified the prolactin receptor as a potential molecular target in lobular neoplasia using array comparative genomic hybridization. In contrast, we have shown that the prolactin receptor may not be as important for the progression of ductal lesions. These results support the view that lobular and ductal carcinoma evolve along separate pathways. Validation of the expression of PRLR in a larger number of LCIS cases is warranted.

Citation Information: Cancer Res 2009;69(24 Suppl):Abstract nr 4151.

Dynamics of micronuclei in rat skin fibroblasts after X irradiation

In a previous study, we demonstrated DNA damage, expressed as micronuclei, in binucleate dermal fibroblasts obtained from human skin 2-9 weeks after fractionated radiotherapy. Here we assessed micronuclei in X-irradiated skin fibroblasts from 9-14-week-old female Lewis rats as a function of time after a single dose of radiation to determine the lifetime of such damage in the skin. After irradiation with 5, 10, 15 and 18 Gy, formation of micronuclei at 1 day or 2 months postirradiation increased up to about 10 Gy, with evidence for a plateau at higher doses. The time course of micronuclei present in the skin fibroblasts demonstrated a plateau region (approximately 20 days after 18 Gy and about 2 months after 10 Gy) before the number of micronuclei started to decline. Residual micronuclei were observed for more than 1 year after irradiation. Monomicronucleated cells predominated in fibroblasts from nonirradiated skin, whereas in fibroblasts from irradiated skin, multimicronucleated cells predominated and persisted (together with monomicronucleated cells) in the residual levels of damage at late times. The results suggest that DNA damage in dermal fibroblasts can be assayed by the micronucleus assay in samples from irradiated skin up to 1 month after irradiation for doses up to at least 10 Gy. Further studies are needed to define the dose-response relationship in detail.

Is there a relationship between repopulation and hypoxia/reoxygenation? Results from human carcinoma of the cervix

Long overall treatment times are detrimental for cure by radiotherapy and it has been argued that this may be due to repopulation occurring during the course of treatment. However, attempts to predict treatment outcome in relation to tumour proliferation, using pretreatment measurements of kinetic parameters such as Tpot or labelling index (LI) have not met with great success. One possible reason is that hypoxia/reoxygenation is linked to the growth of the tumour and its ability to repopulate. Data from studies in animal models have provided support for this possibility. We made measurement of tumour hypoxia, reoxygenation during treatment and pretreatment measurements of both Tpot and LI in groups of patients with cervix carcinoma undergoing radical radiation treatment. The data show a relationship between pretreatment pO2 measurements and treatment outcome, but reoxygenation did not show any association with treatment outcome. There was no significant association between pretreatment kinetic parameters and treatment outcome, nor was there any evidence of a relationship between pretreatment kinetic parameters and pO2. In the small group of 28 patients whose tumours underwent measurements of both pretreatment kinetic parameters (Tpot, LI) and reoxygenation, there was no relationship between these two sets of measurements. There was also no evidence that a combination of kinetic and reoxygenation measurements could be predictive of treatment outcome.

High interstitial fluid pressure (IFP) and hypoxia as biomarkers of cisplatin chemoradiation response in advanced cervix cancer

5584

Background: Chemoradiotherapy (CRT) has been shown to improve survival compared to RT alone for locally advanced cervix cancer. The tumour microenvironment in cervix cancer is also known to influence disease progression and response to treatment. In this prospective study, pre-treatment tumour hypoxia and interstitial fluid pressure (IFP) were examined as potential biomarkers of improved treatment effectiveness in a cohort of patients treated with definitive radiation alone, or with the addition of concurrent cisplatin CRT. Methods: Between April 1994 and January 2006, 309 eligible patients with cervix cancer were entered into a prospective study of hypoxia and IFP prior to definitive treatment. Patients accrued between 1994 and 1999 (n = 115) were treated with RT alone, and those accrued between 2000 and 2006 (n = 194) received RT and concurrent weekly cisplatin CRT. Clinical characteristics were similar between the two cohorts except pelvic lymph node metastases were more frequent in the CRT cohort (42% vs. 19%, p = 0.0005), likely due to changes in definition of positive nodes on CT or MR imaging. The median follow-up was 3.6 years (2.9 years for CRT and 7.8 years for RT). Results: The use of CRT improved outcome in hypoxic tumors compared to RT (57% 3-yr DFS vs. 42%, p = 0.045) with a trend to improved DFS in patients with high IFP tumors (57% 3-yr DFS vs. 44% for RT alone, p = 0.056). A strong interaction was identified between IFP and treatment (CRT vs. RT, p = 0.007). After correcting for the effects of clinical prognostic factors, cisplatin significantly improved DFS in the high IFP group (p = 0.02), and showed a trend towards improved DFS in those with hypoxia (p = 0.1). In patients with both high IFP and hypoxia a similar benefit of cisplatin was seen. Conclusions: Patients with high IFP and hypoxic tumours may selectively benefit from the addition of concurrent cisplatin CRT to their treatment regimen. This observation is contrary to an effect of high IFP on impaired drug delivery, but may reflect reduced DNA repair under hypoxic conditions, which could facilitate cell killing with both radiation and cisplatin chemotherapy.

No significant financial relationships to disclose.

Osteosarcoma in adolescents and adults: survival analysis with and without lung metastases

BACKGROUND

Survival data are still limited and controversial about adult patients with osteosarcoma as older group of patients has mostly been excluded from the historical trials.

PATIENTS AND METHODS

Patients with osteosarcoma, from 1986 to 2003, in a single center, were reviewed. Survival according to a cutoff age of 40 was studied. Patients with lung metastases were identified. Variables at first lung involvement including time to first lung metastases, multiplicity and size of the metastatic lesions and use of chemotherapy were all analyzed.

RESULTS

A total of 247 patients, with age range of 14-77 years, were reviewed. Five-year survival is 66% with no difference between patients <40 or >or=40 years. Eighty-five patients, with either synchronous or metachronous lung involvement, have 3-year postlung metastases survival (PLMS) of 30%. Forty-seven patients (55.3%) underwent lung resection with 3-year PLMS of 38% compared with 16% for nonoperated patients (P = 0.00023). Patients who developed lung metastases within a year and have fewer than four lung lesions have better PLMS (P < 0.0001 for both).

CONCLUSIONS

Older patients have identical survival to pediatric population and should have a similar management approach. Complete metastectomy is the key issue for prolonged survival. Time to lung metastases and number of lung lesions are the most important prognostic factors.

Local control with conventional and hypofractionated adjuvant radiotherapy for ductal carcinoma in-situ: is there a difference in outcome?

Abstract #1162

Purpose&#x2028; Trials of adjuvant whole breast radiotherapy (WBRT) following breast-conserving surgery for invasive breast cancer indicate equivalent efficacy and morbidity for conventional and hypofractionated treatment. WBRT for ductal carcinoma in-situ (DCIS) improves local control, however an optimal dose fractionation remains undefined. Our group policy allowed for the use of both schedules, therefore we compared local control in women with DCIS following breast-conserving surgery.&#x2028; Methods&#x2028; Two hundred and sixty six patients treated between January 1999 and December 2004 with conventional (50 Gy/25 fractions) or hypofractionated (42.4 Gy/16 daily fractions or 40 Gy/16 fractions + 12.5 Gy/5 fraction boost) radiotherapy schedules to the whole breast after breast-conserving surgery for DCIS were identified and retrospectively reviewed. Treatment schedules were evaluated for outcomes in terms of ipsilateral breast recurrence.&#x2028; Results&#x2028; The median age at diagnosis 56.7 years (range 32.2-83.8 years), and 82% (218 patients) presented with mammographic abnormality. Median tumour size was 1.5 cm (range 0.4-10 cm), grade 3 tumours were present in 105 women (39%), and comedonecrosis in 145 (55%). Fifty-one patients (19%) had multifocal disease. Median follow-up was 45.1 months (range 0.7-106.8 months). Prognostic features were well matched in both groups.&#x2028; One-hundred and four patients (39%) were treated with conventional and 162 (61%) with hypofractionated WBRT. There were a total of 14 confirmed local recurrences in the previously radiated breast. Six recurrences occurred in the 104 (5.8%) patients treated with a conventional schedule and 8 of 162 (4.9%) patients treated with a hypofractionated schedule. Actuarial risk of recurrence at 4 years was 7% with hypofractionated WBRT and 5% with a standard schedule (p=0.85).&#x2028; &#x2028; Eight patients recurred with DCIS and 6 with invasive carcinomas, all 14 were treated surgically with adjuvant treatment as indicated clinically, 1 patient with clinical recurrent disease refused biopsy and further conventional treatment. Univariate analysis showed a trend to increased risk of recurrence with grade 3 tumours (10% at 4 years vs. 3% for grade1/2, p=0.053) and presence of comedonecrosis (8% vs. 2%, p=0.096), but not multi-focal disease (7% vs. 5%, p=0.78).&#x2028; Conclusions&#x2028; Hypofractionated adjuvant WBRT schedules following breast-conserving surgery for DCIS have comparable local control rates to a conventional radiation schedule. Hypofractionated WBRT is more convenient for patients, has equivalent morbidity and should be considered in this patient group.

Citation Information: Cancer Res 2009;69(2 Suppl):Abstract nr 1162.

Utility of subsequent conventional dose chemotherapy in relapsed/refractory transplant-eligible patients with diffuse large B-cell lymphoma failing platinum-based salvage chemotherapy

Up to 60% of patients with relapsed/refractory diffuse large B-cell lymphoma (DLBCL) do not respond to second-line (salvage) chemotherapy and hence are not offered autologous hematopoietic cell transplantation (AHCT). The utility of further salvage chemotherapy in an attempt to proceed with AHCT remains undefined. The authors reviewed 201 patients with DLBCL relapsed/refractory to anthracycline-based chemotherapy who received first-line salvage chemotherapy containing cis-platinum. Of the 120 non-responders to first-line platinum-based salvage chemotherapy, 73 received second-line salvage chemotherapy. The response rate to second-line salvage chemotherapy was 14%. Factors predicting lack of response were progression on primary therapy (p = 0.007), abnormal lactate dehydrogenase findings (p = 0.0027) and tumor bulk (p = 0.013) at second progression. Eight patients who responded received AHCT and appeared to have comparable survival to those transplanted after one salvage regimen. The authors conclude that the utility of second-line salvage chemotherapy is low, and that it is best reserved for patients demonstrating initial anthracycline sensitivity and low tumor burden.