Data and subject heterogeneity and data sharing: keys to translational success in spinal cord injury research?

Spinal cord injury (SCI) is a highly devastating and complex injury with many secondary consequences. Finding a treatment for SCI has been a rollercoaster ride through exciting peaks and sobering valleys. As a matter of fact, there are still no robust and reliable clinical treatments to minimize or repair spinal cord damage. The reasons are manifold and in this opinion piece, we will argue that subject heterogeneity and a lack of transparency in reporting findings are potential contributors to the challenge of finding and translating treatments from the bench to the bedside. Let’s, for example, look at age at the time of SCI as a variable since individuals can suffer an SCI throughout their entire life span. The average age at the time of SCI increased from around 30 years in the 1970s to the median age of 50.5 years in the USA (Jain et al., 2015). This is significant as age is a key determinant of functional recovery following most injuries including SCI. To increase the translational relevance of experimental rodent models, researchers are increasingly examining age as a biological variable in SCI pathophysiology, neuroplasticity, and recovery. For example, our most recent publications report on experiments utilizing both 4- and 14-month-old mice to better model the SCI demographic. One of our main findings is that despite a well-established and observed age-associated decrease in antioxidant defense, older animals responded worse, relative to young animals, to antioxidant treatment after SCI (Stewart et al., 2022b). A second, albeit unrelated and almost unpublished observation, is that immunoglobulin G (IgG) levels within the spinal cord increase in an age- and sex-dependent manner (Stewart et al., 2022a). As we describe below, the latter was originally slated to be buried in the often overlooked and inaccessible confines of supplemental data. Collectively, these findings illustrate how the introduction of demographic heterogeneity (age) complicates but also sophisticates SCI research and highlights the value of open, transparent reporting of unrelated (i.e., incidental) data. A few years ago, Dr. A. Stewart, a postdoctor in the Gensel laboratory set out to test a supposedly straightforward hypothesis. It is well known that endogenous antioxidants and free-radicle scavengers, such as glutathione (GSH), decrease with age. SCI also reduces GSH. Augmenting GSH levels through therapeutic treatment with the GSH precursor, N-acetylcysteine-amide (NACA), improves outcomes in young rodent models of SCI (Patel et al., 2014). Therefore, he hypothesized that the age-dependent decrease in antioxidant capacity in older animals would exacerbate reactive oxygen species damage and impair functional recovery after SCI. Consistent with this hypothesis we observed significant decreases in GSH coincident with increased oxidative damage in aged (14-month-old) relative to young (4-month-old) mice after SCI (Stewart et al., 2022b). The administration of NACA (for 3 days after SCI) was sufficient to restore GSH levels and reduce oxidative damage in both 4- and 14-month-old mice (Stewart et al., 2022b). Despite the therapeutic reversal of age-dependent oxidative damage, NACA treatment did not improve functional or anatomical outcomes in 14-month-old SCI mice. In fact, we observed an unpredictable trend toward toxicity with NACA treatment in older SCI mice with no indication of toxicity in young SCI mice (Stewart et al., 2022b). Our observation with NACA treatment highlights the challenge and importance of modeling clinical heterogeneity in basic science. Despite a well-reasoned hypothesis and biological evidence of therapeutic efficacy (reduced oxidative damage), age confounded functional treatment efficacy in an unpredictable way. Our observation is not the result of anecdotal effects specific to NACA. Previously we observed unforeseen age-divergent responses to treatment where the same dose and treatment (in this case with the mitochondrial-based treatment, 2,4-dinitrophenol) resulted in significantly worse outcomes in 4-month-old SCI mice and significantly improved outcomes in 14-month-old SCI mice (Stewart et al., 2021). The lesson from these published findings is that limiting studies to a single age or sex (sex discussed previously in Stewart et al. (2020)) may mask unforeseen effects across different demographic populations. As the age at which individuals suffer an SCI changes, it is important to consider the translational relevance of the preclinical models utilized. In the process of obtaining data for our NACA manuscript, we observed an interesting, incidental finding outside the scope of our original hypothesis. While performing western blot on spinal cord samples, we observed that IgG (examined as a control for the specificity of other primary antibodies) was elevated in an age-, sex-, and injury-dependent manner. Specifically, IgG was significantly increased in older female and male mice after SCI but not in younger mice, with older female mice having significantly more IgG than any other group (Stewart et al., 2022a). Our original plan was to report these observations as supplemental data within the NACA manuscript. However, through discussions with an editor of the journal, we agreed that given the relatively sparse data available with aged animals, these observations warranted an independent report; one that was subsequently accepted and published (Stewart et al., 2022a). The implications of this “incidental” finding are numerous. Increased IgG may be indicative of an increased propensity for autoimmunity with age after SCI, be associated with increased myelin and cellular debris clearance, drive age-dependent inflammatory responses, and be associated with other secondary injury and repair processes as we discussed (Stewart et al., 2022a). Only time will tell whether the “incidental” findings of age-associated IgG levels or the hypothesis-driven conclusion of age-associated impairments with NACA treatment is of higher impact and significance for the field. Potentially of greater implications, is the question of how many incidental findings go undiscovered – lost in the black holes of supplemental figures, or, even more commonly, remain hidden in notebooks and data sheets within a laboratory. A similar fate is common to so-called “negative” data. This includes findings where the experimental groups do not differ from the controls, or in the case of truly negative treatment approaches (e.g., that produce side effects). Thus, these data remain unpublished as they do not fit within the hypothesis testing design’s straightforward, predicted, and homogenous outcomes. The lack of reporting all data and the limited transparency in the current publication model contribute to the common challenges of data bias and dark data which accumulates billions of dollars in investment and is a massive waste of research efforts (Callahan et al., 2017). Although this appears like an easily addressed problem, one must recognize the increased burden and challenge of publishing incidental and negative data. For example, it is very common that repetitions and more control experiments are requested by journal editors and reviewers. The result is that a research group moves on to greater and new adventures rather than investing in an unpredictable future or an incidental finding that is potentially beyond their own expertise. One can only imagine what information could be extracted using advanced bioinformatics if all the dark data from such findings were available. It becomes quickly obvious that knowledge translation can be improved dramatically. One way to view incidental findings or “negative” results is to consider these data as heterogeneous pieces of information. Incidental findings, then, are a form of data heterogeneity. Our recent publications highlight the impact of introducing just the smallest amount of subject heterogeneity (two age groups instead of one) into a research study. In our opinion, the subject heterogeneity of two age groups, or both sexes, provides an incremental step toward successful translation; surely the inclusion of two age groups better models the subject heterogeneity observed clinically? Can the inclusion and reporting of data heterogeneity, especially of unrepresented subjects such as varied ages and both sexes, also provide a step toward successful translation? If we consider the limited data heterogeneity in the publications of basic science research, in contrast to the clinical reality where, differences in age, sex, genetic background, and medical history are the norm, a major challenge in translation becomes obvious. But how can we increase data heterogeneity without changing the infrastructure and limiting productivity and career advancements? The current reward (and promotion) system is biased toward hypothesis-driven, homogeneous data reporting and publications. This needs to evolve by rewarding the publication of incidental or negative data as well as the introduction of heterogeneity into the laboratory. The first steps to achieve this would be designated funding and a process to release all research findings without the requirement for further experiments or judgment of the outcome. One approach to increase transparency in research is to release individual subject data in open-data repositories. This approach would reduce the current publication bias and reduce the waste of research efforts and funds. An example data repository is the Open Data Commons for SCI (ODC-SCI.org, Callahan et al., 2017), which is based on a community effort and regularly updated based on user feedback. In the ODC-SCI, data can be uploaded and then published (with a digital object identifier, (DOI)) in a Findability, Accessibility, Interoperability, and Reusable (FAIR)-share manner (Wilkinson et al., 2016) and protected by a CC-BY license (Fouad et al., 2020). This provides the community with curated and publicly available data, including a data dictionary and metadata information, enabling the community to interpret and reuse the data. Considering the pending data management and publication mandate from National Institutes of Health (NOT-OD-21-013) and other funding bodies, sharing data on a suitable portal will become even more important in the future, helping to make research more transparent, streamlined, and effective. Collectively, the inclusion of subject heterogeneity and the reporting of data heterogeneity in basic science research may be important steps in the successful translation of research efforts into real improvements in human health (Figure 1).Figure 1: The inclusion and reporting of subject and data heterogeneity in basic science research increases the translational relevance of preclinical rodent studies.ODC-SCI: Open Data Commons for SCI (ODC-SCI.org); SCI: spinal cord injury. Created with BioRender.com.Both Dr. Fouad and Dr. Gensel serve as part of the leadership team for the ODC-SCI. This work was supported by NIH, Nos. NIH R01 NS116068 (to JCG) and NIH T32 NS077889 (to OHW). C-Editors: Zhao M, Liu WJ, Yu J; T-Editor: Jia Y

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.