Latent subgroup analysis of a randomized clinical trial through a semiparametric accelerated failure time mixture model

Change plane model averaging for subgroup identification

Central to personalized medicine and tailored therapies is discovering the subpopulations that account for treatment effect heterogeneity and are likely to benefit more from given interventions. In this article, we introduce a change plane model averaging method to identify subgroups characterized by linear combinations of predictive variables and multiple cut-offs. We first fit a sequence of statistical models, each incorporating the thresholding effect of one particular covariate. The estimation of submodels is accomplished through an iterative integration of a change point detection method and numerical optimization algorithms. A frequentist model averaging approach is then employed to linearly combine the submodels with optimal weights. Our approach can deal with high-dimensional settings involving enormous potential grouping variables by adopting the sparsity-inducing penalties. Simulation studies are conducted to investigate the prediction and subgrouping performance of the proposed method, with a comparison to various competing subgroup detection methods. Our method is applied to a dataset from a warfarin pharmacogenetics study, producing some new findings.

Subgroup Identification and Regression Analysis of Clustered and Heterogeneous Interval-Censored Data

Clustered and heterogeneous interval-censored data occur in many fields such as medical studies. For example, in a migraine study with the Netherlands Twin Registry, the information including time to diagnosis of migraine and gender was collected for 3975 monozygotic and dizygotic twins. Since each study subject is observed only at discrete and periodic follow-up time points, the failure times of interest (i.e., the time when the individual first had a migraine) are known only to belong to certain intervals and hence are interval-censored. Furthermore, these twins come from different genetic backgrounds and may be associated with differential risks for developing migraines. For simultaneous subgroup identification and regression analysis of such data, we propose a latent Cox model where the number of subgroups is not assumed a priori but rather data-driven estimated. The nonparametric maximum likelihood method and an EM algorithm with monotone ascent property are also developed for estimating the model parameters. Simulation studies are conducted to assess the finite sample performance of the proposed estimation procedure. We further illustrate the proposed methodologies by an empirical analysis of migraine data.

Set-regression with applications to subgroup analysis

Regression is a commonly used statistical model. It is the conditional mean of the response given covariates μ ( x ) = E ( Y | X = x ) . However, in some practical problems, the interest is the conditional mean of the response given the covariates belonging to some set A. Notably, in precision medicine and subgroup analysis in clinical trials, the aim is to identify subjects who benefit the most from the treatment, or identify an optimal set in the covariate space which manifests treatment favoritism if a subject's covariates fall in this set and the subject is classified to the favorable treatment subgroup. Existing methods for subgroup analysis achieve this indirectly by using classical regression. This motivates us to develop a new type of regression: set-regression, defined as μ ( A ) = E ( Y | X ∈ A ) which directly addresses the subgroup analysis problem. This extends not only the classical regression model but also improves recursive partitioning and support vector machine approaches, and is particularly suitable for objectives involving optimization of the regression over sets, such as subgroup analysis. We show that the new versatile set-regression identifies the subgroup with increased accuracy. It is easy to use. Simulation studies also show superior performance of the proposed method in finite samples.

Instrumental variable estimation of early treatment effect in randomized screening trials

The primary analysis of randomized screening trials for cancer typically adheres to the intention-to-screen principle, measuring cancer-specific mortality reductions between screening and control arms. These mortality reductions result from a combination of the screening regimen, screening technology and the effect of the early, screening-induced, treatment. This motivates addressing these different aspects separately. Here we are interested in the causal effect of early versus delayed treatments on cancer mortality among the screening-detectable subgroup, which under certain assumptions is estimable from conventional randomized screening trial using instrumental variable type methods. To define the causal effect of interest, we formulate a simplified structural multi-state model for screening trials, based on a hypothetical intervention trial where screening detected individuals would be randomized into early versus delayed treatments. The cancer-specific mortality reductions after screening detection are quantified by a cause-specific hazard ratio. For this, we propose two estimators, based on an estimating equation and a likelihood expression. The methods extend existing instrumental variable methods for time-to-event and competing risks outcomes to time-dependent intermediate variables. Using the multi-state model as the basis of a data generating mechanism, we investigate the performance of the new estimators through simulation studies. In addition, we illustrate the proposed method in the context of CT screening for lung cancer using the US National Lung Screening Trial data.

Sentinel Lymph Node Biopsy: Indications and Technique

Sentinel lymph node biopsy is a key tool in the care of many patients with melanoma. The indications for the procedure have gradually become clearer over the 3 decades since the technique was developed. For appropriately selected patients, it carries enormous significance. Although it is a minimally invasive procedure, it does carry some risk. It is also a multidisciplinary procedure, requiring knowledge and experience from several specialties including nuclear medicine, surgery, and pathology.

A New Compromise Design Plan for Accelerated Failure Time Models with Temperature as an Acceleration Factor

An accelerated life test of a product or material consists of the observation of its failure time when it is subjected to conditions that stress the usual ones. The purpose is to obtain the parameters of the distribution of the time-to-failure for usual conditions through the observed failure times. A widely used method to provoke an early failure in a mechanism is to modify the temperature at which it is used. In this paper, the statistically optimal plan for Accelerated Failure Time (AFT) models, when the accelerated failure process is described making use of Arrhenius or Eyring equations, was calculated. The result was a design that had only two stress levels, as is common in other AFT models and that is not always practical. A new compromise plan was presented as an alternative to the widely used “4:2:1 plan”. The three-point mixture design proposed specified a support point in the interval that was optimal for the estimation of the parameters in AFT models, rather than simply the middle point. It was studied in comparison to different commonly used designs, and it proved to have a higher D-efficiency than the others.

Inferring latent heterogeneity using many feature variables supervised by survival outcome

In cancer studies, it is important to understand disease heterogeneity among patients so that precision medicine can particularly target high-risk patients at the right time. Many feature variables such as demographic variables and biomarkers, combined with a patient's survival outcome, can be used to infer such latent heterogeneity. In this work, we propose a mixture model to model each patient's latent survival pattern, where the mixing probabilities for latent groups are modeled through a multinomial distribution. The Bayesian information criterion is used for selecting the number of latent groups. Furthermore, we incorporate variable selection with the adaptive lasso into inference so that only a few feature variables will be selected to characterize the latent heterogeneity. We show that our adaptive lasso estimator has oracle properties when the number of parameters diverges with the sample size. The finite sample performance is evaluated by the simulation study, and the proposed method is illustrated by two datasets.

Practice-Changing Developments in Stage III Melanoma: Surgery, Adjuvant Targeted Therapy, and Immunotherapy

In this article, we will focus on the practice-changing developments for stage III melanoma, from the use of the sentinel node (SN) biopsy to complete lymph node dissection (CLND) and upcoming adjuvant therapies. MSLT-1 (Multicenter Selective Lymphadenectomy Trial-1) was the first and only prospective randomized controlled trial to examine whether the SN biopsy has any notable melanoma-specific survival benefit (primary endpoint). MSLT-1 randomly assigned 2,001 patients to undergo either wide local excision (WLE) and an SN biopsy or WLE and nodal observation. Two prospective randomized controlled trials have examined the potential benefit for immediate CLND versus delayed CLND after sequential observation. Both the DECOG-SLT and MSLT-2 trials failed to demonstrate a notable benefit for immediate CLND; therefore, sequential follow-up with ultrasonography and a delayed CLND in the case of relapse should be considered the new standard of care. The CheckMate 238 study demonstrated a notable benefit for adjuvant nivolumab in terms of 18-month relapse-free survival (RFS) rates compared with high-dose adjuvant ipilimumab. Single-agent adjuvant BRAF inhibition has been examined and failed to improve RFS. However, the COMBI-AD study did demonstrate a substantial benefit for combination BRAF and MEK inhibition for patients with BRAF-mutated resected stage IIIA to IIIC melanoma.

Estimating heterogeneous treatment effects for latent subgroups in observational studies

Individuals may vary in their responses to treatment, and identification of subgroups differentially affected by a treatment is an important issue in medical research. The risk of misleading subgroup analyses has become well known, and some exploratory analyses can be helpful in clarifying how covariates potentially interact with the treatment. Motivated by a real data study of pediatric kidney transplant, we consider a semiparametric Bayesian latent model and examine its utility for an exploratory subgroup effect analysis using secondary data. The proposed method is concerned with a clinical setting where the number of subgroups is much smaller than that of potential predictors and subgroups are only latently associated with observed covariates. The semiparametric model is flexible in capturing the latent structure driven by data rather than dictated by parametric modeling assumptions. Since it is difficult to correctly specify the conditional relationship between the response and a large number of confounders in modeling, we use propensity score matching to improve the model robustness by balancing the covariates distribution. Simulation studies show that the proposed analysis can find the latent subgrouping structure and, with propensity score matching adjustment, yield robust estimates even when the outcome model is misspecified. In the real data analysis, the proposed analysis reports significant subgroup effects on steroid avoidance in kidney transplant patients, whereas standard proportional hazards regression analysis does not.

Subgroup analysis with semiparametric models toward precision medicine

In analyzing clinical trials, one important objective is to classify the patients into treatment-favorable and nonfavorable subgroups. Existing parametric methods are not robust, and the commonly used classification rules ignore the fact that the implications of treatment-favorable and nonfavorable subgroups can be different. To address these issues, we propose a semiparametric model, incorporating both our knowledge and uncertainty about the true model. The Wald statistics is used to test the existence of subgroups, while the Neyman-Pearson rule to classify each subject. Asymptotic properties are derived, simulation studies are conducted to evaluate the performance of the method, and then method is used to analyze a real-world trial data.

Predicting disease Risk by Transformation Models in the Presence of Unspecified Subgroup Membership

Some biomedical studies lead to mixture data. When a discrete covariate defining subgroup membership is missing for some of the subjects in a study, the distribution of the outcome follows a mixture distribution of the subgroup-specific distributions. Taking into account the uncertain distribution of the group membership and the covariates, we model the relation between the disease onset time and the covariates through transformation models in each sub-population, and develop a nonparametric maximum likelihood based estimation implemented through EM algorithm along with its inference procedure. We further propose methods to identify the covariates that have different effects or common effects in distinct populations, which enables parsimonious modeling and better understanding of the difference across populations. The methods are illustrated through extensive simulation studies and a real data example.

Role of sentinel lymph node biopsy as a staging procedure in patients with melanoma: A critical appraisal

Worldwide, sentinel node (SN) biopsy for accurate staging is now part of the standard work-up of patients with melanomas ≥1.0 mm Breslow thickness, as it is for staging patients with breast cancer. Nuclear medicine imaging and surgical techniques have evolved to such a degree that a SN can be identified and removed in virtually every patient. Nevertheless, some opposition to a routine SN biopsy remains, perhaps due to a failure to appreciate the serious implications of incomplete or inaccurate staging. Guided by a critical appraisal of the available evidence, this review elucidates the definition of an SN, discusses the sensitivity and specificity of the information it provides, emphasises that it is a minor staging procedure that can lead to improved survival when followed by appropriate therapy, and explains the necessarily unconventional and complex design of the only randomised trial that addresses this subject. It also describes other benefits and risks of an SN biopsy and outlines its role in current melanoma management.

Sentinel lymph node biopsy followed by lymph node dissection for localised primary cutaneous melanoma

BACKGROUND

Melanoma is the leading cause of skin cancer-associated mortality. The vast majority of newly diagnosed melanomas are confined to the primary cutaneous site. Surgery represents the mainstay of melanoma treatment. Treatment strategies include wide excision of the primary tumour and sentinel lymph node biopsy (SLNB) to assess the status of the regional nodal basin(s). SLNB has become an important component of initial melanoma management providing accurate disease staging.

OBJECTIVES

To assess the effects and safety of SLNB followed by completion lymph node dissection (CLND) for the treatment of localised primary cutaneous melanoma.

SEARCH METHODS

We searched the following databases up to February 2015: the Cochrane Skin Group Specialised Register, the Cochrane Central Register of Controlled Trials (CENTRAL) in The Cochrane Library (2015, Issue 1), MEDLINE (from 1946), EMBASE (from 1974), and LILACS ((Latin American and Caribbean Health Science Information database, from 1982). We also searched the following from inception: African Index Medicus, IndMED of India, Index Medicus for the South-East Asia Region, and six trials registers. We checked the reference lists of included and excluded studies for further references to relevant randomised controlled trials (RCTs). We searched ISI Web of Science Conference Proceedings from inception to February 2015, and we scanned the abstracts of major dermatology and oncology conference proceedings up to 2015.

SELECTION CRITERIA

Two review authors independently assessed all RCTs comparing SLNB followed by CLND for the treatment of primary localised cutaneous melanoma for inclusion. Primary outcome measures were overall survival and rate of treatment complications and side effects.

DATA COLLECTION AND ANALYSIS

Two review authors independently extracted and analysed data on survival and recurrence, assessed risk of bias, and collected adverse effect information from included trials.

MAIN RESULTS

We identified and included a single eligible trial comparing SLNB with observation and published in eight different reports (from 2005 to 2014) with 2001 participants. This did not report on our first primary outcome of overall survival. The study did report on the rate of treatment complications. Our secondary outcomes of disease-specific and disease-free survival, local recurrence and distant metastases were reported. There were 1347 participants in the intermediate-thickness melanoma group and 314 in the thick melanoma group.With regard to treatment complications, short-term surgical morbidity (30 days) in 1735 participants showed no difference between SLNB and observation (risk ratio [RR] 1.11; 95% confidence interval [CI] 0.9 to 1.37) for wide excision of the tumour site but favoured observation for complications related to the regional nodal basin (RR 14.36; 95% CI 6.74 to 30.59).The study did not report the actual 10-year melanoma-specific survival rate for all included participants. Instead, melanoma-specific survival rates for each group of participants: intermediate-thickness melanoma (defined as 1.2 to 3.5 mm) and thick melanomas (defined as 3.50 mm or more) was reported.In the intermediate-thickness melanoma group there was no statistically significant difference in disease-specific survival between study groups at 10 years (81.4 ± 1.5% versus 78.3 ± 2.0%, hazard ratio [HR] 0.84; 95% CI 0.65 to 1.09). In the thick melanoma group, again there was no statistically significant difference in disease-specific survival between study groups at 10 years (58.9.3 ± 4.1% versus 64.4 ± 4.6%, HR 1.12; 95% CI 0.77 to 1.64). Combining these groups there was some heterogeneity (I² = 34%) but the total HR was not statistically significant (HR 0.92; 95% CI 0.74 to 1.14). This study failed to show any difference for its stated primary outcome.The summary estimate for disease-free survival at 10 years favoured SLNB over observation in participants with intermediate-thickness and thick melanomas (HR 0.75; 95% CI 0.63 to 0.89).With regard to the rate of local and regional recurrence as the site of first recurrence, a benefit of SLNB uniformly existed in both groups of participants with intermediate-thickness and thick melanomas (RR 0.56; 95% CI 0.45 to 0.69). This is in contrast with a uniformly unfavourable effect of SLNB with regard to the rate of distant metastases as site of first recurrence, in both groups of participants with intermediate-thickness and thick melanomas (HR 1.33; 95% CI 1.03 to 1.72).

AUTHORS' CONCLUSIONS

We contacted the trial authors querying the lack of data on overall survival which was the primary outcome of their important study. They stated "there are numerous additional analyses that have yet to be reported for the trial". We expect that overall survival data will be available in a future update of this review.Disease-free survival and rate of local and regional recurrence favoured SLNB in both groups of participants with intermediate-thickness and thick melanomas but short-term surgical morbidity was higher in the SLNB group, especially with regard to complications in the nodal basin.The evidence for the outcomes of interest in this review is of low quality due to the risk of bias and imprecision of the estimated effects. Further research may have an important impact on our estimate of the effectiveness of SLNB in managing primary localised cutaneous melanoma. Currently this evidence is not sufficient to document a benefit of SLNB when compared to observation in individuals with primary localised cutaneous melanoma.

No survival benefit for patients with melanoma undergoing sentinel lymph node biopsy: critical appraisal of the Multicenter Selective Lymphadenectomy Trial-I final report

BACKGROUND

Sentinel lymph node biopsy (SLNB) was developed in the hope that it would improve outcomes for patients with melanoma. SLNB is an area of discussion and controversy in melanoma medicine. The final trial results of the Multicenter Selective Lymphadenectomy Trial (MSLT-I) have now been published and the authors suggest their long-term results 'clearly validate the use of sentinel-node biopsy in patients with intermediate-thickness or thick primary melanomas'. An accompanying editorial states that MSLT-I is a practice-changing trial.

CONCLUSIONS

However, critical appraisal of MSLT-I data does not support the claims of the final report. On the contrary, MSLT-I failed to demonstrate that there is a significant treatment-related difference in the 10-year melanoma-specific survival rate in the overall study population. Furthermore, there was no improvement in overall or melanoma-specific survival of the intermediate-thickness group (1·2-3·5 mm). Completion lymphadenectomy can result in complications in about a third of patients, with a rate of clinically significant lymphoedema following axillary or groin dissection of 5-10%. Unnecessary lymphadenectomy can therefore have a major effect on patient quality of life. The evidence provided by Morton et al. does not support the claim that sentinel lymph node biopsy followed by lymphadenectomy in patients with positive sentinel nodes should be the standard of care in patients with melanoma. Readers are encouraged to check with registration sites to make sure declared primary outcomes are fairly reported. Post-hoc analyses are at best exploratory and cannot be used to form the principal conclusions of a trial.

Multicenter Selective Lymphadenectomy Trial-I confirms the central role of sentinel node biopsy in contemporary melanoma management: response to 'No survival benefit for patients with melanoma undergoing sentinel lymph node biopsy: critical appraisal of the Multicenter Selective Lymphadenectomy Trial-I final report'

Sentinel lymph node (SLN) biopsy has become a standard procedure for many patients with melanoma and is recommended in numerous national and professional melanoma guidelines. The Multicenter Selective Lymphadenectomy Trial (MSLT-1) confirms earlier large database studies and prospective clinical trials in demonstrating the independent and unequalled prognostic value of the SLN. It also demonstrates the ability of biopsy-directed management to provide effective regional disease control with the least possible morbidity. These benefits are not in question and provide ample justification for the procedure, even without evidence of a survival benefit. However, MSLT-1 also provides strong evidence of a substantial reduction in the risk of melanoma death for patients with intermediate thickness melanomas who harbour occult nodal metastases at the time of presentation. Denying appropriately selected patients with melanoma the opportunity to undergo SLN biopsy is no longer reasonable or acceptable.

Challenges and solutions to pre- and post-randomization subgroup analyses

Subgroup analyses are commonly performed in the clinical trial setting with the purpose of illustrating that the treatment effect was consistent across different patient characteristics or identifying characteristics that should be targeted for treatment. There are statistical issues involved in performing subgroup analyses, however. These have been given considerable attention in the literature for analyses where subgroups are defined by a pre-randomization feature. Although subgroup analyses are often performed with subgroups defined by a post-randomization feature—including analyses that estimate the treatment effect among compliers—discussion of these analyses has been neglected in the clinical literature. Such analyses pose a high risk of presenting biased descriptions of treatment effects. We summarize the challenges of doing all types of subgroup analyses described in the literature. In particular, we emphasize issues with post-randomization subgroup analyses. Finally, we provide guidelines on how to proceed across the spectrum of subgroup analyses.

A semiparametric linear transformation model to estimate causal effects for survival data

Semiparametric linear transformation models serve as useful alternatives to the Cox proportional hazard model. In this study, we use the semiparametric linear transformation model to analyze survival data with selective compliance. We estimate regression parameters and the transformation function based on pseudo-likelihood and a series of estimating equations. We show that the estimators for the regression parameters and transformation function are consistent and asymptotically normal, and both converge to their true values at the rate of n ^-1/2, the convergence rate expected for a parametric model. The practical utility of the methods is confirmed via simulations as well as an application of a clinical trial to evaluate the effectiveness of sentinel node biopsy in guiding the treatment of invasive melanoma.

Final trial report of sentinel-node biopsy versus nodal observation in melanoma

BACKGROUND

Sentinel-node biopsy, a minimally invasive procedure for regional melanoma staging, was evaluated in a phase 3 trial.

METHODS

We evaluated outcomes in 2001 patients with primary cutaneous melanomas randomly assigned to undergo wide excision and nodal observation, with lymphadenectomy for nodal relapse (observation group), or wide excision and sentinel-node biopsy, with immediate lymphadenectomy for nodal metastases detected on biopsy (biopsy group). Results No significant treatment-related difference in the 10-year melanoma-specific survival rate was seen in the overall study population (20.8% with and 79.2% without nodal metastases). Mean (± SE) 10-year disease-free survival rates were significantly improved in the biopsy group, as compared with the observation group, among patients with intermediate-thickness melanomas, defined as 1.20 to 3.50 mm (71.3 ± 1.8% vs. 64.7 ± 2.3%; hazard ratio for recurrence or metastasis, 0.76; P=0.01), and those with thick melanomas, defined as >3.50 mm (50.7 ± 4.0% vs. 40.5 ± 4.7%; hazard ratio, 0.70; P=0.03). Among patients with intermediate-thickness melanomas, the 10-year melanoma-specific survival rate was 62.1 ± 4.8% among those with metastasis versus 85.1 ± 1.5% for those without metastasis (hazard ratio for death from melanoma, 3.09; P<0.001); among patients with thick melanomas, the respective rates were 48.0 ± 7.0% and 64.6 ± 4.9% (hazard ratio, 1.75; P=0.03). Biopsy-based management improved the 10-year rate of distant disease-free survival (hazard ratio for distant metastasis, 0.62; P=0.02) and the 10-year rate of melanoma-specific survival (hazard ratio for death from melanoma, 0.56; P=0.006) for patients with intermediate-thickness melanomas and nodal metastases. Accelerated-failure-time latent-subgroup analysis was performed to account for the fact that nodal status was initially known only in the biopsy group, and a significant treatment benefit persisted.

CONCLUSIONS

Biopsy-based staging of intermediate-thickness or thick primary melanomas provides important prognostic information and identifies patients with nodal metastases who may benefit from immediate complete lymphadenectomy. Biopsy-based management prolongs disease-free survival for all patients and prolongs distant disease-free survival and melanoma-specific survival for patients with nodal metastases from intermediate-thickness melanomas. (Funded by the National Cancer Institute, National Institutes of Health, and the Australia and New Zealand Melanoma Trials Group; ClinicalTrials.gov number, NCT00275496.).