Abstract
Precision medicine trials and targeted therapies have shifted to the forefront of oncology. Although targeted therapies have shown initial promise, implementation across the broad landscape of oncology has many challenges. These limitations include an incomplete understanding of the functional significance of variant alleles as well as the need for clinical research and practice models that are more patient-centered and account for the complexity of individual patient tumors. Furthermore, successful implementation of targeted therapies will also be predicated on efforts to standardize the framework for patient management support. Here, we review current implementations of targeted therapies in precision oncology and discuss how “actionability” is defined for molecular targets in cancer therapeutics. We also comment on the growing need for bioinformatics tools and data platforms to complement advances in precision oncology. Finally, we discuss current frameworks for integrating precision oncology into patient management and propose an integrated model that combines features of molecular tumor boards and decision support systems. Mol Cancer Ther; 16(12); 2645–55. ©2017 AACR.
See related article by Pilié et al., p. 2641
Introduction
With the evolving landscape of medical oncology, focus has shifted away from nonspecific cytotoxic treatment strategies toward therapeutic paradigms more characteristic of precision medicine, whereby therapy is delivered to patients on the basis of unique patient clinical and molecular features. When applying precision medicine, the goal is to tailor diagnosis and treatment to each patient's individual biologic profile, while minimizing exposure to unnecessary or ineffective therapies. Technological advances in accessibility of patient and tumor genomics have improved understanding of tumor biology and led to enhancements in the ability to identify and target major molecular drivers of cancer. These developments have shifted precision oncology to the forefront of cancer treatment strategies.
Despite early successes of targeted cancer therapies, complexities in therapeutic development and application have been revealed, in many cases due to considerable genomic heterogeneity among tumor histology subtypes. Some of these complexities will need to be addressed by customized combination therapy or by moving targeted therapeutics from end-stage disease to earlier stages, when the disease has evolved to a lesser extent (1). Although significant advances in targeted therapeutics have been observed in some areas in medical oncology—notably in late-stage melanoma (2) and non–small cell lung cancer (3)—there are many areas that have not experienced similar progress, at least in part due to the paucity of biomarker-driven trials (4–6).
Limitations also exist with respect to clinical trial design, availability of biomarker data, and challenges in understanding how to use existing, yet un-annotated data in clinical practice. These obstacles have limited the application of targeted therapies and highlight the need for validated bioinformatics tools and data platforms that can help guide clinicians in patient management.
Here, we review the implementation of precision oncology trials to study targeted therapies. We also discuss considerations in describing the “actionability” of a molecular target, with respect to the biomarker–response association. Lastly, we comment on important topics relevant to the development of a standardized framework for integrating precision oncology trials and targeted therapies into patient management.
Clinical development pathway of targeted therapies
Targets of precision therapeutics may include aberrant products of altered genes, cell surface molecules differentially expressed in cancer, and molecules that regulate immune cell activity. Support for targeted therapies in oncology has been driven by several factors. First, the advent of next-generation sequencing (NGS) and the emergence of increasingly cost- and time-efficient genomic profiling methods have advanced our capability to develop novel clinical trial designs (7–9), and current turnaround times of three to four weeks (half of which is generally attributed to specimen collection) have made tumor molecular profiling clinically feasible in patient management (10).
Second, improved understanding of the molecular pathology of disease has aided in the capacity to develop therapeutics targeting oncogenic drivers. Initial efforts using matched therapeutic agents were driven by the perspective that molecular alterations were specific to tumor location and histopathology, and early clinical development of targeted therapies often followed that of standard-of-care cytotoxic chemotherapies. However, lack of evidence supporting this approach led to trial designs that match specific molecular alterations to therapeutic agents, independent of tumor cell origin (11–14). These considerations contributed to the birth of precision oncology trials. Nonetheless, adaptations have been necessary to address limitations of classical trial designs. These challenges include technical and practical concerns hindering patient selection and biomarker discovery (11).
Precision oncology trials
Several factors provided the impetus for precision oncology trials. First, while traditional cytotoxic chemotherapy targets common, generic disease mechanisms shared among tumors, tumor heterogeneity beyond random variation has been repeatedly demonstrated (11). Second, the use of a biomarker approach to tailor treatments to subsets of patients with the same tumor type became more common, particularly in programs that aimed to show a significant therapeutic effect in unselected patients where the prevalence of the biomarker was low (11). This created challenges in advancing a drug's development, especially if therapeutic effects were detectable in only a small patient population. As a result, early phase trials now feature selected patient populations to minimize the inclusion of patients unlikely to respond to novel treatments.
The evolution of the clinical drug development pathway has resulted in flexible trial designs, including “umbrella” and “basket” trials (5, 11, 15, 16). “Umbrella” trials, such as BATTLE-2 (NCT01248247), Lung-MAP (NCT02154490), and I-SPY (NCT01042379), assign patients with particular tumor histologies to treatment regimens specifically developed to target the tumor's oncogenic molecular pathway. The Leukemia & Lymphoma Society's (LLS) Beat AML Core Study is particularly unique as it will be the first-ever precision oncology trial for a blood cancer (NCT02927106). Newly diagnosed acute myeloid leukemia (AML) patients will be assigned to receive investigational therapies targeting particular hematologic malignancies based on genomic profiling.
Alternative approaches to “umbrella” trials are “basket” and “hybrid” trials. “Basket” trials, such as CUSTOM (NCT01306045) and SHIVA (NCT01771458) treat patients with specific agents targeting aberrant molecular pathways, independent of tumor origin. In this way, therapeutic agents may be considered effective across tumor types. Lastly, “hybrid” trials incorporate aspects of both “umbrella” and “basket” trials into one protocol and feature subtrials that target different molecular drivers within the same histology or the same molecular driver across different histologies (16). Several of these trials have been published, and trials such as UCSD PREDICT and the MD Anderson Phase I initiative demonstrate that approximately 25% to 30% of patients can be matched to therapy when larger NGS panels are used, as well as improved outcomes in matched versus unmatched patients, albeit in a nonrandomized setting (13, 14, 17).
Nonetheless, limitations do exist with flexible trial designs. The majority of trials feature monotherapies, yet treatment strategies incorporating drug combinations have provided benefit to patients with multiple genomic aberrations or advanced cancers (10). Approval of targeted drug combinations remains a challenge because of the potential for overlapping drug toxicities. Furthermore, the SHIVA trial notably concluded that off-label use of molecularly guided targeted therapies did not show improvement in progression-free survival over conventional therapies (18). However, features of the trial, including choice of inappropriately matched therapies, use of monotherapies ineffective in advanced cancers, and assignment of targeted treatments not based on molecular profiling, have been criticized (19, 20). Lessons learned from the trial have been important considerations in the design of later trials, and consequently, support for precision therapy remains encouraging.
Here, we briefly discuss representative “hybrid” and “basket” trials led by different government initiatives, pharmaceutical companies, and scientific organizations, including the National Cancer Institute's Molecular Analysis for Therapy Choice (NCI-MATCH) trial (21), the Signature Program (ref. 22; Novartis), the MyPathway trial (ref. 23; Genentech), and the American Society of Clinical Oncology (ASCO)-led Targeted Agent and Profiling Utilization Registry (TAPUR) study (ref. 24; Table 1).
Representative biomarker-based precision oncology “basket” and “hybrid” trials
NCI-MATCH.
NCI-MATCH is an ongoing trial that analyzes patient tumors for actionable mutations for which targeted treatments are available (NCT02465060; ref. 25). NCI-MATCH enrolls patients with advanced solid tumors and lymphomas who are no longer responding or never responded to standard treatment and evaluates whether treatment based on the molecular profile of disease, independent of tumor origin, provides clinical benefit (26). Prior to enrollment, NCI-MATCH provides a pretreatment biopsy to screen for eligibility in one of the treatment arms. Furthermore, if a patient does not match with the initial biopsy and a potential treatment match later becomes available, the patient may require rebiopsy to evaluate eligibility. Although this provides a great opportunity for biomarker exploration, it is limiting for patients with tumors not amenable to rebiopsy. Since opening with 10 treatment arms, NCI-MATCH has now expanded to 30 treatment arms.
Signature program.
The Signature Program includes exploratory signal-finding studies, with primary endpoints assessing the clinical benefit rate of molecularly guided targeted therapies (27). A key feature of the Signature Program is expedited protocol delivery, whereby physicians may request to receive a protocol if a patient is determined to have an actionable mutation in any tissue type.
To date, five Signature trials are active (NCT02160041, NCT02186821, NCT02187783, NCT01885195, and NCT01833169), with the remaining either completed or closed due to low enrollment. Preliminary data for 106 patients with CDK4/6 pathway aberrations and treated with ribociclib showed clinical benefit (complete/partial response or stable disease) in 19 patients at week 16, with preliminary antitumor activity additionally observed in 4 patients (28, 29). Further assessment of the correlation between clinical benefit and genomic mutation profiling is ongoing.
MyPathway.
MyPathway evaluates treatment regimens for advanced cancers in patients with specific molecular alterations related to HER2-overexpression, -amplification, or -activating mutations, epidermal growth factor receptor (EGFR)-activating mutation, BRAF-activating mutations, and potentially actionable Hedgehog pathway mutations (NCT02091141; ref. 23).
Preliminary findings observed objective response in 22 out of 118 patients across 9 tumor types, with observed clinical activity in patients with HER2-amplified bladder, biliary, and colorectal cancer, as well as patients with BRAF-mutated lung cancer (30, 31). Interim data have demonstrated activity in expanded cohorts of HER2-amplified or mutated patients with colorectal, urothelial, and biliary cancers, particularly in colorectal cancer, where an overall response rate and clinical benefit rate of 37.5% and 46.9% was observed, respectively (32–34).
TAPUR.
TAPUR supports targeted therapies by providing flexibility in tumor and blood sample selection for genomic profiling, as well as less-restrictive eligibility criteria for study enrollment (NCT02693535; refs. 24, 35). This design allows for enrollment from academic and private practice settings, thereby expanding the potential application of study findings to aid in providing insights in real clinical practice. The Syapse Precision Medicine Platform helps guide and automate the study operations workflow, integrate patient data and clinical outcomes, and support review of cases during molecular tumor boards.
Several pharmaceutical companies have agreed to participate and provide study drugs without any patient cost. Currently, these include 17 drugs targeting 15 targeted therapy pathways in advanced solid tumors and hematological malignancies, including multiple myeloma and B-cell non-Hodgkin lymphoma. TAPUR also plans to lower the enrollment age from 18 to 12 years in order to allow participation to adolescent patients for study drugs in which there is a defined adolescent dose.
Discussion
Limitations of targeted therapies and precision oncology trials
Although personalized cancer therapies are evolving into achievable standards in medical oncology, significant challenges exist. These include insufficient evidence to guide practical biomarker implementation, limited understanding of the functional effects of variant alleles, and a lack of consensus regarding the level of evidence required to select patients for personalized therapies.
First, treatment strategies for which reliable predictive biomarkers exist are limited, with evidence of foreseeable clinical benefit found for only select genomic alterations in a limited number of cancer types. Consequently, there is often little support for routine clinical implementation of biomarker-targeted therapies. Furthermore, trial-based studies on rare biomarkers may have high screening failure rates during patient selection.
Second, the development of targeted therapies requires strong evidence that genetic alterations within a particular disease have an impact on tumor behavior (36, 37). There are often limited data assessing the functional significance of variant alleles and lack of consensus as to what constitutes the minimum level of evidence necessary to define a variant allele as a potential biomarker in clinical practice (37). Improving our understanding of tumor biology and establishing these standards are crucial to continued progress in precision oncology.
Level of evidence required to define “Actionability”
The application of NGS has significantly improved the prospect of utilizing genetic screening as a practical method to identify genetic variants that may be associated with an increased risk of developing cancer. The benefit of these advances has been well documented in patients with BRCA1/2 mutations, and certain genetic variants have measurable effects on protein function that significantly increase the risk of developing breast cancer (38–40). The clinical applicability of the BRCA1/2 mutations has been difficult to replicate with other genetic variants that may represent potential molecular targets.
Furthermore, different criteria for treatment selection exist when considering patients with variants of unknown significance (VUS). As such, the risk of excluding patients who may still benefit from treatment exists. This variability in defining selection criteria can be seen across “umbrella” and “basket” trials. For example, the NCI-MATCH protocol establishes rules of evidence for treatment selection, while independent models for classifying the “actionability” of biomarkers have also been proposed (Table 2). NCI-MATCH's criteria, as well as the schemes proposed by Van Allen and colleagues (41), Vidwans and colleagues (42), Meric-Bernstam and colleagues (36), Sukhai and colleagues (43), and Carr and colleagues (44), provide individual levels of classification that describe the strength of evidence associating particular biomarkers with disease response in given tumor types. Overall, similarities exist between these classification schemes, and NCI-MATCH's criteria have demonstrated practicality and applicability in practice.
Comparison of representative classification schemes assessing the level of evidence for biomarker–response association
Despite the difficulty in designing trials around patients with VUS, there is utility in including these patients in trials. For example, while the TAPUR protocol does not explicitly exclude patients with VUS, these cases are encouraged to be sent to TAPUR's molecular tumor board to discuss whether rationale exists that the patient may derive clinical benefit from targeted therapies. Additional focus will be necessary, as the predictive value of genetic variants on associated disease risk, such as the example of BRCA1/2 variants and breast cancer risk, may be initially unknown without further investigation. Lastly, it is important to note the distinction between germline mutations, such as BRCA1/2, and somatic mutations, such as those in EGFR. Although the availability of data differs across the literature for germline and somatic mutations, ongoing efforts exist to bridge this gap and expand this area.
Framework for Precision Oncology
While defining the “actionability” of a molecular target is subjective, a clearer consensus is necessary for building a framework for implementing targeted therapies into clinical practice. This framework includes components such as bioinformatics tools, data repositories, and platforms for patient management support.
Bioinformatics tools
Increased understanding of the mutational landscape of cancer has been accompanied by difficulties in determining the functional relationships between molecular alterations and tumorigenesis. Although independent knowledge of the oncogenic potential of rare mutations exists for some genetic alterations, there is often insufficient evidence to support therapeutic approaches that target these alterations (45). As such, there is a growing need for bioinformatics tools that can shed light on unknown mechanisms of disease associated with specific genetic variants. Numerous resources have previously been reviewed, and representative tools are included in Table 3.
Representative prediction tools for classifying the functional significance of variant alleles
These tools include independent methods and combined methods that integrate the prediction scores of other tools. The methods differ in their prediction algorithms for annotating and scoring variants, with variance in parameters including sequence conservation and homology as well as protein structure and function. Several tools have been previously reviewed for performance (46, 47), with respect to relative and predictive utility. While individual tools outperformed others in different factors—all-around performance, prediction rate, sensitivity and specificity, etc.—a combined approach using multiple prediction tools provided benefits over individual methods, including a more evenly balanced sensitivity and specificity (46).
The availability of bioinformatics tools has greatly improved our understanding of the genomic landscape of cancer. Nonetheless, there is variability in how tools generate prediction models. For example, BRAF V600E has been demonstrated as a valuable target for BRAF and MEK tyrosine kinase inhibitors in anticancer treatments (48–50), yet it can be considered VUS by prediction tools under varying parameters. To address this, initiatives to create repositories and databases for variant-level data based on these bioinformatics tools are prevalent, including those by TCGA (NCI and NHGRI), cBioPortal for Cancer Genomics (Memorial Sloan-Kettering Cancer Center), My Cancer Genome (Vanderbilt-Ingram Cancer Center), COSMIC (Wellcome Trust Sanger Institute), and ClinVar (NCBI). However, many databases are limited by the lack of defined data on the functional significance of variants and a general inability to sufficiently mine the scientific literature for relevant information. Nonetheless, efforts have been promising. For example, My Cancer Genome's DIRECT database has catalogued 188 primary EGFR mutations in lung cancer and their associated drug responses (51). Although the initial objective was to catalogue relevant somatic mutations in lung cancer, the goal is to expand DIRECT to include all known mutations of clinical significance across various cancers. Given that VUS are evolving targets, efforts such as these will ultimately help clinicians better predict the functional significance of these variants.
Data sharing
Complementing the advancements in precision oncology is the need to develop platforms for data sharing. Although analytic technologies have experienced success in different industries, implementation in a clinical setting requires further adaptation. The American Association for Cancer Research's (AACR) Project Genomics, Evidence, Neoplasia, Information, Exchange (GENIE; ref. 52) and the Oncology Precision Network (OPeN; ref. 53)—formed by Intermountain Healthcare, Stanford Cancer Institute, Providence St. Joseph Health, and Syapse—are promising efforts that aim to aggregate and link cancer genomics data and clinical outcomes.
The gaps in applying mutational and mechanistic discoveries to derive clinically meaningful value in patient care underscore the importance of complementing the implementation of targeted therapies with efficient data compilation and sharing to help clinicians make informed decisions. Despite the remaining chasm, the efforts by GENIE and OPeN are promising steps forward and serve as valuable models in clinical practice.
Patient management support
Molecular tumor boards.
Precision oncology practice models centered around molecular tumor boards (MTB) have been valuable implementations in guiding personalized patient management. Recent experiences with MTBs have provided insight on their effectiveness in practice. Several experiences are discussed here, including those with the Northwestern OncoSET (Sequence, Evaluate, Treat) program and the University of California San Diego (UCSD) Moores Cancer Center.
The Northwestern OncoSET program—through collaboration between the Robert H. Lurie Comprehensive Cancer Center and Northwestern Medicine—launched in 2015 as a research initiative designed to provide patients with personalized cancer treatment options through a combined oncology and genomics approach (54). To support the patient management decision process, OncoSET has established the Lurie Cancer Center's MTB which, similar to existing tumor boards, brings together a multidisciplinary team that meets weekly to discuss cases and decide upon recommendations for patient treatment plans, which may involve treatment in a clinical trial or alternative therapies.
Two experiences with MTBs at the Moores Cancer Center have been reviewed—one presenting various cancer cases (55) and the second presenting breast cancer cases (56). Discussion of presented findings culminated in a consensus on a recommended course of treatment tailored for each patient, with the ultimate treatment decision left to the discretion of the treating physicians. Published findings from these MTB experiences have been positive. In the first study, among the 34 patients whose cases were presented, 3 of 11 evaluable patients achieved partial responses (55). In the second study, 17 of 43 patients were treated in congruence with the recommendations of the MTB, and 7 of the 17 achieved stable disease for at least 6 months or partial remission (56).
Three models developed by oncology pharmacy practitioners—at the University of Wisconsin, Indiana University, and Moffit Cancer Center—have also been described (57). Although the models vary, they share commonalities in benefitting from a multidisciplinary approach to precision oncology. Results from the MTB within Indiana University's Health Precision Genomics Program observed a significant difference in number of patients reaching favorable progression-free survival between those treated with genomically guided therapy (43.2%) and those treated with non–genomically guided therapy (5.3%; ref. 58). Overall, these experiences underscore the importance of expanding opportunities to participate in MTBs in clinical practice.
Decision support system.
An alternative platform for bridging the gap between identification of molecular targets and implementation of personalized cancer therapeutics has been developed within the Institute for Personalized Cancer Therapy (IPCT) at MD Anderson Cancer Center (36, 59). To aid in clinical decision-making, the precision oncology decision support (PODS) team provides clinicians with relevant information for assessing the “actionability” of variant alleles and identifying available matched therapies and clinical trials. Although public databases provide variant-level information about published drug associations, information about the functionality of molecular alterations and relevant clinical trials is often not readily available, leaving clinicians to filter through scientific literature and public databases. Consequently, the PODS platform seeks to streamline the treatment decision process by communicating curated information to clinicians.
Recent experiences with MTBs and decision support systems have demonstrated the effectiveness of these platforms as frameworks for supporting personalized patient management. Nonetheless, opportunities for adaptation exist. Figure 1 compares the workflow for these approaches. Although MTBs foster knowledge sharing, the responsibility still falls on clinicians to present cases and participate in reaching recommended treatment decisions. Decision support systems provide curated information to clinicians which, while alleviating the burden on clinicians, offers less opportunities for multidisciplinary review and discussion. For both approaches, there are associated time and resource commitments, which limit feasibility. One approach to address this has been the development of virtual decision support tools that provide clinicians with the functionality to input patient and disease features in a portal and view recommended course of treatments. While this provides flexibility, the limiting factor would be the availability of pre-prepared decision support.
Patient management workflow strategies. Patient management workflow strategies with an (A) independent review model, (B) molecular tumor board model, (C) decision support system model, and (D) integrated model. In A, clinicians filter through scientific literature and public databases to reach treatment decisions for patients. B and C provide clinicians with support for making informed treatment decisions. In B, clinicians present cases to molecular tumor boards and a recommended course of action is provided. In C, decision support systems provide “curated” information to help clinicians make treatment decisions. In D, extracted and curated data are reviewed by clinicians and, if necessary, with the support of molecular tumor boards.
Nonetheless, data-driven endeavors continue to drive advances in precision oncology. For example, Watson Genomics from Quest Diagnostics (60) is a significant pursuit that aims to help clinicians identify tailored treatments for their patients. The University of North Carolina's Lineberger Comprehensive Cancer Center has experimented with Watson in MTBs (61). Out of 1,000 patients, Watson independently identified the same treatments recommended by the MTB in 99% of cases. Furthermore, Watson identified new treatment strategies not found by clinicians in 30% of cases and has also been taught to read radiologic scans and molecular diagnostics to identify abnormalities and potentially actionable genetic mutations, highlighting a promising opportunity for implementation as part of standard of care.
Future of precision oncology: Integrated patient management model
Limitations exist with using either MTBs or decision support systems independently. For example, artificial-intelligence platforms can only yield value from published data. In these cases, clinical experience provides crucial insight. To address these concerns, we propose an integrated model that combines the strengths of both approaches, allowing for extracted and curated data to be reviewed by clinicians and, if necessary, with the support of MTBs (Fig. 1). This model provides clinicians with the advantage of utilizing bioinformatics support and benefitting from knowledge sharing among peers, so as to ensure that the information used for informed treatment decisions is as error-proof as possible.
Conclusion
Targeted therapies are widely regarded as the framework for future treatment paradigms in the precision oncology era. With improved understanding of the molecular pathways that drive tumor biology and the increasing availability of cost- and time-efficient NGS technologies, the reality of such a paradigm shift comes into sharper focus. Nonetheless, although success has been demonstrated in some areas within oncology, the overall treatment paradigm remains in its infancy, with many challenges yet to be addressed before its potential can be fully realized across the broad landscape of oncology.
A multitude of considerations has limited the study of biomarker data and the clinical application of potential molecular targets (42, 62). Although high-throughput technologies have allowed sequencing data to be generated quickly and conveniently, the challenge of determining the functional significance of variant alleles remains. Because of this, “actionability” of molecular targets is often defined differently. Consequently, there is a lack of consensus regarding the level of evidence required to select molecular targets for clinical application. Addressing these considerations is important, as implementation of targeted therapies necessitates that the right “actionability” is applied to the right patients.
Furthermore, molecular alterations often do not segregate by tumor origin. Patients with metastatic disease usually have multiple genomic alterations, making their tumors both unique and complex. These considerations suggest the need for customized combinations of therapy, which is a patient-centered paradigm that departs significantly from current drug-centered models. Finally, genomic testing and matching has been applied almost exclusively in patients with metastatic disease which has been, more often than not, heavily pretreated. Hence, many precision medicine studies have high attrition rates due to the patient condition deteriorating quickly, and, unsurprisingly, resistance rates are high in these patients with complicated tumors who are generally treated with matched monotherapy. Moving to earlier stages of the disease, a strategy that was dramatically successful in chronic myelogenous leukemia, warrants investigation in solid tumors as well.
Nonetheless, there is hope that ongoing and future precision medicine trials will continue to provide insight regarding the value of potential biomarkers and inform the development of targeted therapies. Additionally, efforts to standardize a framework for data sharing are being explored, while support from data-driven and artificial-intelligence platforms is expected to shape the future of patient management. Overall, continued progress in the study and development of targeted therapies and immunotherapies to treat cancer is anticipated. However, that success will depend on collaborative efforts to optimize the implementation of precision oncology, a willingness to change models of clinical research to fit the new realities unveiled by genomics, and establishment of a harmonized, innovative approach to patient care.
Disclosure of Potential Conflicts of Interest
Y.K. Chae is a consultant/advisory board member for Foundation Medicine, Guardant Health, Biodesix, and Counsyl. R. Kurzrock has ownership interest in CureMatch, Inc., reports receiving a commercial research grant from Genentech, Merck Serono, Pfizer, Sequenom, Foundation Medicine, and Guardant, has ownership interest (including patents) in Curematch, Inc., and is a consultant/advisory board member for Actuate Therapeutics, Xbiotech, and Roche. No potential conflicts of interest were disclosed.
- Received July 3, 2017.
- Revision received August 4, 2017.
- Accepted August 16, 2017.
- ©2017 American Association for Cancer Research.
References
Volume 16, Issue 12
