The past and future of ‘reported outcomes’ in studies on chemotherapy neuropathy

The manuscript by Park et al. in this issue of Annals of Oncology [1] reports on potential clinical and genetic prognostic factors of chemotherapy-induced peripheral neuropathy (CIPN). Predicting CIPN in individual patients, before symptoms manifest, is a critical goal that has not been achieved in the field. Even at the cohort level (a lesser challenge) available CIPN studies tend to disagree, regarding which predictors (genetic and otherwise) are significant. Therefore, additional clinical data are critically needed. We congratulate Park et al. on achieving this step: to put ICON7, a large, high-quality clinical trial, onto the ‘map’ of CIPN research.

Park et al. also report genetic analyses. Of the single nucleotide polymorphisms (SNPs) tested, none was significantly associated with CIPN when tested alone. A combined end point that the authors call ‘additive polymorphisms’ found that SNPs in the genes MAPT and GSK3B were associated with clinician- and patient-reported CIPN, respectively. (Of note, Park et al. appear to use the term ‘additive’ here not to refer to the use of a standard additive model combining the genotype information of two alleles but instead to a sum score of alleles across multiple loci per gene.) It remains somewhat unclear to the reader, whether this analysis was developed post hoc, how many score-versus-phenotype comparisons were allowed, and how the multiple testing problem was approached. The potential of ICON7 for a future comprehensive genetic analysis is strong. While the present study reports a carefully planned first step in regard to genetic testing that was supported by a biological rationale (for the selection of the small number of SNPs tested), the authors (or their collaborators) may be working already on more comprehensive analyses, such as genome-wide association study (GWAS) arrays and/or exome- or genome sequencing, which will allow re-testing of all genetic variants claimed previously (by other studies) to be associated with CIPN and some responsible unbiased discovery. Results from such work would be welcomed.

In the present manuscript, Park et al. emphasize a comparison of clinician reported outcomes (CROs) and patient reported outcomes (PROs), a distinction that takes a generation of researchers back in time. When one of the authors of this editorial started his career about three decades ago, clinical trials were designed to have CROs as the primary end point. Over time, however, it became apparent that physicians tend to underreport symptoms experienced by patients [2]. Trials, published many years to decades ago, showed that PROs identified a higher incidence and severity of treatment-related toxicities, than did CROs, for example, for symptoms such as oral mucositis [3] and chemotherapy-induced nausea [4]. The clinical use of PROs have recently been linked to improved quality of life measures [5] and even overall survival improvements [6]. The superiority of PROs have been extended to the field of CIPN [7, 8], where, for instance, a recent genetics study (using massively parallel ‘nextgen’ whole exome sequencing) selected patients for analysis solely on the basis of PROs [9]. While we might argue that the absence of a diagnostic gold standard for CIPN has kept the field from settling the PRO versus CRO issue in a statistically compelling way, a shortcoming that Park et al. did not remedy either, PROs have won most arguments and are now supported by majority consensus, while CROs have, largely, been relegated to the past.

What, however, is the future of ‘reported outcomes’ in CIPN and, for clinical oncology symptom control research, more broadly?

New ideas tend to emerge in very dynamic subcultures of society before their adoption into the field of medicine. Considering this premise, we could, for instance, understand the dismissal of CROs, in favor of PROs, as part of the cultural shift from medical paternalism to the patient-as-a-partner philosophy that changed medicine in the wake of the 1970s liberalization of Western cultures: CROs represented paternalism, while PROs better represented a new generation of physicians who strived for more patient autonomy.

What is the social trend of our time, the 2010s, that will (and should) affect ‘reported outcomes’ in CIPN research? In short, it is the widespread availability of data in society and our ability (or that of our now-trainees) to access and analyze it with ever better sophistication. The future generation of researchers will be well versed in ‘big data’ analysis and personally empowered by fluency in writing program code as the lingua franca of the exploring human mind.

Which ‘Reported Outcome’ will match the expectations of this coming generation? Let’s just call them ‘hashtag RO,’ or #RO. #ROs are just that: all the bits of outcome data in a machine-readable format with annotations that are friendly to human researchers, so that they can decipher them easily. #RO will replace the currently all-too-common practice of using ‘Hidden (or hideous) Reported Outcomes’ (HROs). HROs are data that are incompletely documented and hide behind a curtain of interpretation (often following the narrative bias of a particular field). The genomic data sharing policy instituted by NIH shows how some parts of the scientific enterprise already lead the way in #RO. The symptom control field is poised to join #RO.

Let us consider the enclosed study by Park et al. as an example of how #ROs might be helpful. Some readers might be puzzled for instance, whether the comparison between CROs and PROs as presented by the authors was affected by a systematic bias, because CROs were reportedly used in ICON7 to alter therapy decisions, while PROs were not: patients with high CROs (but normal PROs) had their chemotherapy dose reduced, while patients with high PROs (but normal CROs) continued with the same chemotherapy timing and dose. If this occurred only in few patients (which the reader cannot know), the impact is probably small, but if it occurred frequently, it might explain the relative lack of high CRO scores compared with the relative abundance of high PRO scores, which is a key argument by Park et al. #ROs, if available, would provide a straightforward answer to any reader interested in clarifying this issue.

Today, most studies in the oncology symptom control field, even those that are most expertly planned and executed, represent incremental progress. Great value may result from data sharing with a broader group of colleagues in the field. While a single study may, on its own, neither be paradigm shifting nor a final answer to a long-standing challenge, the potential for insight may exponentially increase when a growing number of studies can be interpreted together, which can be achieved through #ROs. The current crop of CIPN studies may serve as an example because (nearly) all of them collect similar clinical and PRO data and many have (or will), in addition, carry out well-established genotyping methods such as GWAS or exome or genome sequencing, but most still lack #RO. Let’s not wait until a future generation explains why HROs were obsolete in the year 2017. Rather, lets support authors by defining criteria for publication that allow investigators to succeed in their careers by helping the entire field to prosper.

Funding

National Institute of Nursing Research (NINR) of the National Institutes of Health (NIH) (R01NR15259).

Disclosure

The authors have declared no conflicts of interest.