Competing Risk Research Articles

546 - Cytomegalovirus (CMV) Infection and Disease Incidence and Risk Factors Across Diverse Hematopoietic Cell Transplantation (HCT) Platforms Using a Standardized Monitoring and Treatment Approach: A Comprehensive Evaluation From a Single Institution

CMV remains a major complication of allogeneic HCT, with over 40% of at risk recipients (donor (D) and/or recipient (R) seropositive) experiencing infection in the first 100 days post-HCT. At the NIH, all HCT recipients are monitored weekly from day 0-100 with blood CMV quantitative PCR (qPCR) with a uniform approach to preemptive therapy across HCT protocols. This offers a unique opportunity to comprehensively evaluate the cumulative incidence (CI) of and risk factors for post-HCT CMV infection and disease across different T cell manipulation approaches (posttransplantation cyclophosphamide (PTCy), proximal serotherapy, ex vivo T cell depletion (TCD), and calcineurin/mTOR inhibitor (CNI/mTORi) based) and different graft sources (marrow (BM), peripheral blood (PBSC), and cord blood (CB)). This is a retrospective study of 364 consecutive patients transplanted between Aug 10, 2011 and Feb 10, 2017. CMV infection was defined as 2 qPCRs between 3.08-4.11 log10 IU/mL in 1 week, 1 qPCR value >4.11 log10 IU/mL, or sufficient clinical suspicion for CMV disease to prompt therapy. Competing risks (CR) included death, relapse, and graft failure, with analyses including and not including systemic steroids for GVHD or engraftment syndrome as an additional CR. Recipient/donor/HCT characteristics are shown in Figure 1. Among at risk recipients, the 100 day CI of CMV infection was 46% (34% when systemic steroid use was included as a CR). The CI of CMV infection varied mainly by recipient CMV serostatus (Figure 2, P < .0001), with 100 day estimates of 2% (D/R −/−), 6% (+/−), 38% (−/+), and 40% (+/+). With regard to graft T cell manipulation, the 100 day CI of CMV infection, with (Figure 3, P = .029) and without steroids (p = NS) as a CR, respectively, was 25% and 42% (CNI/mTORi), 37% and 42% (serotherapy), 40% and 54% (PTCy), and 50% and 60% (TCD). CB was associated with 100 day CI of CMV infection of 64%, higher than BM at 39% and PBSCs at 30%, P = .002. There was a tendency toward lower 100 day CI of CMV infection for recipients of higher T cell doses when systemic steroid use was included as a CR, P = .07. Among CMV seropositive recipients, CMV therapy duration was longer with CMV seronegative donors (median 28 days) compared to CMV seropositive (median 20 days), P = .03. The median duration of CMV therapy was longer at 36 days for CB HCT, compared to PBSC (21 days) and BM (29 days), P = .03. CMV disease developed in 12 patients in the 100 days post HCT at a median of 27 days: 75% in the setting of GVHD, 83% in seropositive recipients, 75% involving gut, and occurring in .5% of PTCy, .4% of CNI/mTORi, .3% of serotherapy, and 0% of CB and TCD HCTs. In conclusion, CMV infection incidence varies based on recipient CMV serostatus, T cell manipulation strategy, and graft source, with longer treatment duration with seronegative donors and/or CB recipients. CMV disease rates before day 100 are similar across platforms and occur predominantly in the setting of GVHD.Figure 2Cumulative incidence of CMV infection in the first 100 days after HCT, by donor (D) and recipient (R) CMV serostatus.View Large Image Figure ViewerDownload Hi-res image Download (PPT)Figure 3Cumulative incidence of CMV infection in the first 100 days after HCT, by T cell manipulation strategy, with comparison using Gray's method of post-transplantation cyclophosphamide (PTCy), serotherapy (proximal ATG or alemtuzumab), ex vivo T cell depletion (TCD), and calcineurin inhibitor/mTOR inhibitor (CNI/mTOR)-based approaches.View Large Image Figure ViewerDownload Hi-res image Download (PPT)

Outcomes for Women with Minimal-Volume Ductal Carcinoma In Situ Completely Excised at Core Biopsy.

Overdiagnosis and overtreatment of ductal carcinoma in situ (DCIS) are concerns, especially for women with low-volume, screen-detected DCIS. This study aimed to evaluate the outcomes for such patients. Women who had minimal-volume DCIS (mDCIS, defined as DCIS diagnosed by core biopsy but with no residual disease on the surgical excision) treated with breast-conserving surgery from 1990 to 2011 were identified. Ipsilateral and contralateral breast events (IBE and CBE) were compared by competing-risk (CR) analysis. Kaplan-Meier (KM) estimates and log-rank tests were used to evaluate covariates. The study identified 290 cases of mDCIS. The median age of the patients was 53years. Radiation therapy (RT) was performed for 27.6% and endocrine therapy for 16.2% of the patients. The median follow-up period was 6.8years. Overall, the IBE rates were 4.3% at 5years and 12.3% at 10years. Among the women not receiving RT, the 5- and 10-year IBE rates (5.4 and 14.5%) were higher than the CBE rates (1.8 and 2.7%). Among those receiving RT, the IBE rates (1.5 and 6.0%) were lower than the CBE rates (4.1 and 15.6%). The women receiving RT trended toward significantly lower IBE rates (p = 0.07). Age, grade, and endocrine therapy were not significantly associated with IBE risk. Among the patients with mDCIS who did not receive RT, the IBE risk was substantially higher than the CBE risk, demonstrating that even DCIS of very low volume is associated with clinically relevant disease. The finding that the IBE risk was greater than the CBE risk supports current strategies that treat DCIS as a precursor rather than a risk marker. Women with mDCIS are not at negligible risk for IBE in the absence of adjuvant therapy.

Oncological and surgical outcome after treatment of pelvic sarcomas.

Background and objectivesTreatment of pelvic tumors remains challenging due to complex anatomy, poor oncological outcome and high complication rates. We sought to investigate the long-term oncological and surgical outcome of these patients.MethodsBetween 1980 and 2012, 147 patients underwent surgical treatment for pelvic sarcoma. Histological diagnosis was Chondrosarcoma in 54, Ewing’s Sarcoma/PNET in 37, Osterosarcoma in 32 and others in 24 patients. Statistical analysis for the evaluation of oncological and surgical outcome was performed by applying Cox proportional hazards regression and Fine-Gray regression models for competing risk (CR) endpoints.ResultsThe estimated overall survival (OS) to death was 80%, 45% and 37% at 1, 5 and 10 years, respectively. Univariate analyses revealed a statistically significant unadjusted influence of age age (p = 0.038; HR = 1.01), margin (p = 0.043; HR = 0.51) and grade (p = 0.001; HR = 2.27) on OS. Considering the multivariable model, grade (p = 0.005; HR = 3.04) and tumor volume (p = 0.014; HR = 1.18) presented themselves as independent prognostic factors on OS. CR analysis showed a cumulative incidence for major complication of 31% at 5 years. Endoprosthetic reconstruction had a higher risk for experiencing a major complication (p<0.0001) and infection (p = 0.001).ConclusionsPelvic resections are still associated with a high incidence of complications. Patients with pelvic reconstruction and high volume tumors are especially at risk. Consequently, a cautious decision-making process is necessary when indicating pelvic reconstruction, although a restrictive approach to pelvic reconstruction is not necessarily reasonable when the other option is major amputation.

Estimating cumulative live-birth rates after IVF treatment with Kaplan–Meier and competing risk methods

Objective(s)To explore the use of competing risk (CR) as compared to the commonly used Kaplan–Meier (KM) methodology in estimating the cumulative live-birth rate (CLBR) after IVF Treatment in a context of high dropout rates and informative censoring. Study designWe compare the KM and CR methodologies for estimating 2-year CLBR in a retrospective cohort of 2779 patients undergoing 5002 embryo transfers over a period of 9 years, from 2000 to 2008, at KKIVF Centre. ResultsWe observed a total of 1105 LB (39.8%), and a dropout rate of 44.2% (1228 patients). The overall CLBR is lower with CR compared with KM method (39% vs 52%) after up to nine embryo-transfer cycles over a period of two years. The highest CLBR was achieved for ovulation disorders (57% vs 49%, KM vs CR) followed by male factors (54% vs 43%, KM vs CR), with poorer outcomes from patients with decreased ovarian reserve (37% vs 16%, KM vs CR) and endometriosis (36% vs 25%, KM vs CR). As dropouts in our cohort are generally older and more likely to have poorer ovarian reserves, the CR method, which accounted for these dropouts, is likely to give more meaningful estimation of IVF success rates. Conclusion(s)The CR method should be considered as a useful alternative in deriving CLBR for IVF treatment where dropout rates are high and when informative censoring is involved.

Cemented total hip arthroplasty revisions in patients of eighty years and older.

It is often a difficult decision whether it is safe to perform revision hip surgery in a patient of 80 years and older. Therefore we evaluated the results of cemented revisions in these elderly patients. Clinical data, radiographs and complications of 49 consecutive cup and/or stem revisions in 48 patients were prospectively collected. The average age of the patients at surgery was 84 years (range, 80-92). We performed Kaplan-Meier (KM) analysis and also a competing risk (CR) analysis because in this series the presence of a competing event (i.e. death) prevents the occurrence of endpoint rerevision. Twenty-nine patients (30 hips) died without rerevision during follow-up and their data was included. The average follow-up of the 16 surviving patients was eight years (range, six to 13). Six re-operations were performed, of which three were re-revisions. Eight-year survivorship was 91.6% (95% confidence interval (CI) 76-97%) for endpoint re-revision for any reason. With the CR analysis we calculated that due to the increasing number of competing events, the KM analysis overestimates the failure rate with 32% for this endpoint. The average Harris hip score improved from 49 to 74. Mortality within three months after surgery was 6%. One postoperative fracture occurred and six hips dislocated. Cemented revisions can provide satisfying results in patient of 80 years and older with acceptable survivorship and complication rates.

Estimating risk of venous thromboembolism in patients with cancer in the presence of competing mortality

In studies on cancer-associated venous thromboembolism (VTE), patients not only are at risk for VTE but also may die from their underlying malignancy. In this competing-risk (CR) scenario, we systematically compared the performance of standard (Kaplan-Meier estimator [1-KM]), log-rank test, and Cox model) and specific CR methods for time-to-VTE analysis. Cancer patients (1542) were prospectively followed for a median of 24 months. VTE occurred in 112 (7.3%) patients, and 572 (37.1%) patients died. In comparison with the CR method, 1-KM slightly overestimated the cumulative incidence of VTE (cumulative VTE incidence at 12 and 24 months [1-KM vs. CR]: 7.22% vs. 6.74%, and 8.40% vs. 7.54%, respectively). Greater bias was revealed in tumor entities with high early mortality (e.g., pancreatic cancer, n = 99, 24-month cumulative VTE incidence: 28.37% vs. 19.30%). Comparing the (subdistribution) hazard of VTE between patients with low and high baseline D-dimer, the Cox model yielded a higher estimate than the corresponding CR model (hazard vs. subdistribution hazard ratio [95% CI] 2.85 [1.92-4.21] vs. 2.47 [1.67-3.65]). For this comparison, the log-rank test yielded a higher test statistic and smaller P-value than Gray's test (χ(2) on 1 degree of freedom: 29.88 vs. 21.34). In patients with cancer who are at risk for VTE and death, standard and CR methods for time-to-VTE analysis can generate differing results. For 1-KM, the magnitude of bias is a direct function of competing mortality. Consequently, bias tends to be negligible in cancer patient populations with low mortality but can be considerable in populations at high risk of death.

Flexible parametric estimation technique for a competing risks model with unobserved heterogeneity: a Monte Carlo study

We analyse a flexible parametric estimation technique for a competing risks (CR) model with unobserved heterogeneity, by extending a local mixed proportional hazard single risk model for continuous duration time to a local mixture CR (LMCR) model for discrete duration time. The state-specific local hazard function for the LMCR model is per definition a valid density function if we have either one or two destination states. We conduct Monte Carlo experiments to compare the estimated parameters of the LMCR model, and to compare the estimated parameters of a CR model based on a Heckman–Singer-type (HS-type) technique, with the data-generating process parameters. The Monte Carlo results show that the LMCR model performs better or at least as good as the HS-type model with respect to the estimated structure parameters in most of the cases, but relatively poorer with respect to the estimated duration-dependence parameters.

Identification of liver transplant candidates with hepatocellular carcinoma and a very low dropout risk: Implications for the current organ allocation policy

It has been shown that patients with hepatocellular carcinoma (HCC) meeting the United Network for Organ Sharing T2 (Milan) criteria have an advantage in comparison with patients without HCC under the current organ allocation system for liver transplantation (LT). We hypothesized that within the T2 HCC group, there is a subgroup with a low risk of wait-list dropout that should not receive the same listing priority. This study evaluated 398 consecutive patients with T2 HCC listed for LT with a Model for End-Stage Liver Disease exception from March 2005 to January 2011 at our center. Competing risk (CR) regression was used to determine predictors of dropout. The probabilities of dropout due to tumor progression or death without LT according to the CR analysis were 9.4% at 6 months and 19.6% at 12 months. The median time from listing to LT was 8.8 months, and the median time from listing to dropout or death without LT was 7.2 months. Significant predictors of dropout or death without LT according to a multivariate CR regression included 1 tumor of 3.1 to 5 cm (versus 1 tumor of 3 cm or less), 2 or 3 tumors, a lack of a complete response to the first locoregional therapy (LRT), and a high alpha-fetoprotein (AFP) level after the first LRT. A subgroup (19.9%) that met certain criteria (1 tumor of 2 to 3 cm, a complete response after the first LRT, and an AFP level ≤ 20 ng/mL after the first LRT) had 1- and 2-year probabilities of dropout of 1.3% and 1.6%, respectively, whereas the probabilities were 21.6% and 26.5% for all other patients (P = 0.004). In conclusion, a combination of tumor characteristics and a complete response to the first LRT define a subgroup of patients with a very low risk of wait-list dropout who do not require the same listing priority. Our results may have important implications for the organ allocation policy for HCC.

Modeling low-dose mortality and disease incubation period of inhalational anthrax in the rabbit

There is a need to advance our ability to conduct credible human risk assessments for inhalational anthrax associated with exposure to a low number of bacteria. Combining animal data with computational models of disease will be central in the low-dose and cross-species extrapolations required in achieving this goal. The objective of the current work was to apply and advance the competing risks (CR) computational model of inhalational anthrax where data was collected from NZW rabbits exposed to aerosols of Ames strain Bacillus anthracis.An initial aim was to parameterize the CR model using high-dose rabbit data and then conduct a low-dose extrapolation. The CR low-dose attack rate was then compared against known low-dose rabbit data as well as the low-dose curve obtained when the entire rabbit dose–response data set was fitted to an exponential dose–response (EDR) model. The CR model predictions demonstrated excellent agreement with actual low-dose rabbit data. We next used a modified CR model (MCR) to examine disease incubation period (the time to reach a fever >40°C). The MCR model predicted a germination period of 14.5h following exposure to a low spore dose, which was confirmed by monitoring spore germination in the rabbit lung using PCR, and predicted a low-dose disease incubation period in the rabbit between 14.7 and 16.8 days.Overall, the CR and MCR model appeared to describe rabbit inhalational anthrax well. These results are discussed in the context of conducting laboratory studies in other relevant animal models, combining the CR/MCR model with other computation models of inhalational anthrax, and using the resulting information towards extrapolating a low-dose response prediction for man.

Empirical comparison of methods for analyzing multiple time-to-event outcomes in a non-inferiority trial: a breast cancer study

BackgroundSubjects with breast cancer enrolled in trials may experience multiple events such as local recurrence, distant recurrence or death. These events are not independent; the occurrence of one may increase the risk of another, or prevent another from occurring. The most commonly used Cox proportional hazards (Cox-PH) model ignores the relationships between events, resulting in a potential impact on the treatment effect and conclusions. The use of statistical methods to analyze multiple time-to-event events has mainly been focused on superiority trials. However, their application to non-inferiority trials is limited. We evaluate four statistical methods for multiple time-to-event endpoints in the context of a non-inferiority trial.MethodsThree methods for analyzing multiple events data, namely, i) the competing risks (CR) model, ii) the marginal model, and iii) the frailty model were compared with the Cox-PH model using data from a previously-reported non-inferiority trial comparing hypofractionated radiotherapy with conventional radiotherapy for the prevention of local recurrence in patients with early stage breast cancer who had undergone breast conserving surgery. These methods were also compared using two simulated examples, scenario A where the hazards for distant recurrence and death were higher in the control group, and scenario B. where the hazards of distant recurrence and death were higher in the experimental group. Both scenarios were designed to have a non-inferiority margin of 1.50.ResultsIn the breast cancer trial, the methods produced primary outcome results similar to those using the Cox-PH model: namely, a local recurrence hazard ratio (HR) of 0.95 and a 95% confidence interval (CI) of 0.62 to 1.46. In Scenario A, non-inferiority was observed with the Cox-PH model (HR = 1.04; CI of 0.80 to 1.35), but not with the CR model (HR = 1.37; CI of 1.06 to 1.79), and the average marginal and frailty model showed a positive effect of the experimental treatment. The results in Scenario A contrasted with Scenario B with non-inferiority being observed with the CR model (HR = 1.10; CI of 0.87 to 1.39), but not with the Cox-PH model (HR = 1.46; CI of 1.15 to 1.85), and the marginal and frailty model showed a negative effect of the experimental treatment.ConclusionWhen subjects are at risk for multiple events in non-inferiority trials, researchers need to consider using the CR, marginal and frailty models in addition to the Cox-PH model in order to provide additional information in describing the disease process and to assess the robustness of the results. In the presence of competing risks, the Cox-PH model is appropriate for investigating the biologic effect of treatment, whereas the CR models yields the actual effect of treatment in the study.

Model for end-stage liver disease-sodium and survival benefit in liver transplantation

There are currently no studies calculating the survival benefit of liver transplantation (LT) according to model for end-stage liver disease-sodium (MELD-Na) and based on the competing risk (CR) method. We enrolled consecutive adult patients with chronic end-stage liver disease entering the waiting list (WL) for primary LT (WL group=337) and undergoing LT (LT group=220) in the period 2006-2009. Two independent multivariable regressions (WL and LT models) were created to measure the prognostic power of MELD-Na with respect to MELD. For the WL model, both Cox and CR multivariable analyses were performed. Estimates were finally included in a Markov model to calculate 3-year survival benefit. WL Cox model: MELD-Na (P<0.0001) and MELD (P<0.0001) significantly predicted survival. WL CR model: MELD-Na (P=0.0045) and MELD (P=0.0109) significantly predicted survival. LT Cox model: MELD-Na (P=0.7608) and MELD score (P=0.9413) had not correlation with survival. Benefit model: MELD and MELD-Na had an overlapping significant impact on 3-year survival benefit; CR method determined a significant decrease in 3-year life expectancy (LE) estimations. MELD-Na and MELD scores similarly predicted 3-year LT survival benefit, but the gain in LE is significantly lower when a CR method is adopted.

Estimating Loss to Follow-Up in HIV-Infected Patients on Antiretroviral Therapy: The Effect of the Competing Risk of Death in Zambia and Switzerland

BackgroundLoss to follow-up (LTFU) is common in antiretroviral therapy (ART) programmes. Mortality is a competing risk (CR) for LTFU; however, it is often overlooked in cohort analyses. We examined how the CR of death affected LTFU estimates in Zambia and Switzerland.Methods and FindingsHIV-infected patients aged ≥18 years who started ART 2004–2008 in observational cohorts in Zambia and Switzerland were included. We compared standard Kaplan-Meier curves with CR cumulative incidence. We calculated hazard ratios for LTFU across CD4 cell count strata using cause-specific Cox models, or Fine and Gray subdistribution models, adjusting for age, gender, body mass index and clinical stage. 89,339 patients from Zambia and 1,860 patients from Switzerland were included. 12,237 patients (13.7%) in Zambia and 129 patients (6.9%) in Switzerland were LTFU and 8,498 (9.5%) and 29 patients (1.6%), respectively, died. In Zambia, the probability of LTFU was overestimated in Kaplan-Meier curves: estimates at 3.5 years were 29.3% for patients starting ART with CD4 cells <100 cells/µl and 15.4% among patients starting with ≥350 cells/µL. The estimates from CR cumulative incidence were 22.9% and 13.6%, respectively. Little difference was found between naïve and CR analyses in Switzerland since only few patients died. The results from Cox and Fine and Gray models were similar: in Zambia the risk of loss to follow-up and death increased with decreasing CD4 counts at the start of ART, whereas in Switzerland there was a trend in the opposite direction, with patients with higher CD4 cell counts more likely to be lost to follow-up.ConclusionsIn ART programmes in low-income settings the competing risk of death can substantially bias standard analyses of LTFU. The CD4 cell count and other prognostic factors may be differentially associated with LTFU in low-income and high-income settings.

Hepatocellular Carcinoma Patients Are Advantaged in the Current Liver Transplant Allocation System

Patients with hepatocellular carcinoma (HCC) within Milan criteria receive priority on the liver transplant waiting list (WL) and compete with non-HCC patients. Dropout from the WL is an indirect measure of transplant access. Competing risks (CR) evaluation of dropout for HCC and non-HCC patients has not previously been reported. Patients listed between 16 March 2005 and 30 June 2008 were included. Probability of dropout was estimated using a CR technique as well as a Cox model for time to dropout. Overall, non-HCC patients had a higher dropout rate from the WL than HCC patients (p < 0.0001). This was reproducible throughout all regions. In Cox regression, tumor size, model for end-stage liver disease (MELD) score and alpha fetoprotein (AFP) were associated with increased dropout risk. Multivariable analysis with CR showed that MELD score and AFP, were most influential in predicting dropout for HCC patients. The index of concordance for predicting dropout with the CR was 0.70. HCC patients appear to be advantaged in the current allocation scheme based on lower dropout rates without regard to geography. A continuous score incorporating MELD, AFP and tumor size may help to prioritize HCC patients to better equate dropout rates with non-HCC patients and equalize access.

Competing risk analysis of events 10 years after revascularization

Objectives. To evaluate the influence of competing risk (CR) non-cardiac death during long-term follow-up of revascularized patients on the interpretation of the cardiac outcomes. Methods. Retrospectively, we compared outcomes estimated with the Kaplan-Meier and the cumulative incidence function (CIF) methods after a median 10.8 years follow-up in 1 234 consecutive patients (594 CABG, 640 PCI) undergoing first time non-emergent revascularization in a community cohort. Results. Overall 301 (24.4%) patients died (27.3% in the CABG vs. 21.7% in the PCI group, p=0.02). The causes of death were cardiac (10.3%) and non-cardiac (14.1%). CR analysis showed a similar probability of cardiac death (CIF 0.10 (95% CI 0.092, 0.18) vs. 0.093 (0.07, 0.12)) in the CABG and PCI treated patients, respectively. The probability for acute myocardial infarction (CIF 0.12 vs. 0.16 p<0.001), congestive heart failure (CIF 0.15 vs. 0.09 p=0.007) in the CABG and PCI group respectively, differed. The differences were also statistically significant after multivariate adjustment for the competing risks of death. For all outcomes the Kaplan-Meier method overestimated risk estimates. Conclusions. The competing risk adjusted probability for cardiac death, but not other cardiac endpoints are comparable in patients treated with either CABG or PCI after very long-term follow-up. The risk for all-cause death was mainly predicted by the occurrence of non-cardiac diseases.

A competing risks approach for time estimation of household WEEE disposal

The recent growth in the number of electrical and electronic devices is viewed as one the priority waste streams in European Union waste management policy. This paper presents the findings of a survey to study domestic habits with respect to Waste Electrical and Electronic Equipment (WEEE) in Spain. A specific problem when performing this estimation arises from the fact that consumers quite often store old appliances at home when they are no longer used. Focusing on four different types of appliance, survival analysis (SA) is used to study both the usage span and the reasons for no longer using each device. The time that the discarded products were kept at home before being disposed of was studied using competing risks (CR) analysis. The results of the analysis provide information on the distribution of the studied variables for the different outcomes as well as the influence exerted by the socio-demographic variables considered. Relations between these characteristics and the storage time of the appliances before disposal emerge based on survey data. For instance, the CR model finds that the storage time of the some appliances (i.e. refrigerator) is related to these social-demographics factors. However, other appliances (i.e. microwave oven) are less influenced by these factors. The attitude and motivation of the respondents to the survey as regards the End-of-Life of appliances were also analysed. A majority of respondents do not store discarded appliances at home. The first reason for storing appliances at home is the possibility of it being useful in the future and the second that the respondents did not know what to do with them.

PMD15 An Unbiased Method of Pharmaceutical Cost Analysis

Pharmacoeconomic evaluation can assess costs by using controlled trials. Statistical analysis of costs follows a normal (N), lognormal (LN), or survival (S) model. Each model assumes different statistical properties that may introduce bias. If one pharmaceutical is less expensive and more effective than another, then no incremental cost-effectiveness ratio is calculated. Therefore, economically attractive pharmaceuticals may not be identified if a biased method of cost analysis is used. OBJECTIVE: The purpose of this study was to (1) evaluate potential bias in methods of cost analysis and (2) demonstrate an unbiased method. METHODS: Cost models were considered unbiased if they accounted for (1) independent variables, (2) disease-specific death, (3) unrelated death (4) dropouts, and (5) informative censoring. Potential differences in statistical significance were demonstrated by simulating 12-month cost data from the GUSTO study that compared use of streptokinase and tissue plasminogen activator (TPA). We simulated n = 10,000 from a gamma distribution for each therapy, and 5% incidences of death, unrelated death, or dropouts. RESULTS: N and LN only account for independent variables. Kaplan-Meier (KM) does not account for independent variables, but competing risk (CR) accounts for all factors. For streptokinase and TPA, mean actual costs were $24,575 and $24,990; mean simulated costs were $24,575 and $24,990. Costs of streptokinase versus TPA were significantly different under N (mean difference [95% confidence interval] =$415 [277–554], p > 0.001), and under LN ($431 [289–547], p > 0.001) For S models, costs were not significantly different under KM (hazard ratio for costs [95% confidence interval] = 0.94 [0.89–1.01], p = 0.08), or CR (0.96 [0.91–1.01], p = 0.09). The robustness of results and goodness of fit depend upon the actual distribution of costs. CONCLUSION: A competing risk model is an unbiased and comprehensive method of cost analysis. Further research is required to assess the clinical importance of alternate methods of cost analysis.