Translational Research in Lung Transplantation: How Do We Get From Mouse to Human?
Article first published online: 7 DEC 2011
© 2011 American Society of Transplantation and the American Society of Transplant Surgeons
American Journal of Transplantation
Volume 12, Issue 2, pages 281–282, February 2012
How to Cite
Lederer, D. J. (2012), Translational Research in Lung Transplantation: How Do We Get From Mouse to Human?. American Journal of Transplantation, 12: 281–282. doi: 10.1111/j.1600-6143.2011.03858.x
- Issue published online: 27 JAN 2012
- Article first published online: 7 DEC 2011
- Received 30 August 2011, revised 30 September 2011 and accepted for publication 12 October 2011
In this issue of the American Journal of Transplantation, a report by Neujahr et al. examines whether seven putative protein biomarkers in bronchoalveolar lavage (BAL) fluid obtained during the first year after lung transplantation were associated with clinically relevant outcomes in a cohort study of 40 transplant recipients at a single US center (1). Two of the biomarkers, CXCL9 and CXCL10, were associated with increased rates of graft loss and bronchiolitis obliterans syndrome (BOS). These chemokines are strong chemoattractants for activated T and NK cells (2) and have previously been shown to play important roles in acute and chronic lung allograft rejection in cross-sectional studies of both animals and humans (3,4). Together with these previous studies, the authors’ work suggests that CXCL9 and CXCL10 might be biomarkers of allograft injury that potentially could predict the future risk of BOS.
Yet, the true implications of the authors’ work are difficult to determine. These chemokines are universally expressed in tissue injury–repair and their presence in the lung may, therefore, be a nonspecific signal of airway or lung injury. In addition, lung allograft rejection can occur through CXCR3-independent pathways (5), suggesting that these chemokines show us only one part of a bigger picture.
The main problem, however, stems from the authors’ approach to studying humans. Biologists have historically chosen to study genetically homogeneous cells, tissues and organisms within a tightly regulated environment, thereby decreasing variability and measurement error. Simple hypothesis tests performed upon data acquired from a small number of repeated experiments in such a setting are arguably sufficient to permit cautious causal inference. Because humans cannot be “controlled” in the usual sense, simple hypothesis tests are uninformative in observational studies of humans and a different approach to handling variability and error is required.
Humans are a heterogeneous species, differing from one another by age, gender, genetic composition, comorbidities, medication and illicit substance use, social and economic influences and countless other characteristics. When one or more of these factors (e.g. immunosuppressive drug regimen) might be associated with an exposure of interest (BAL chemokine concentrations) and be an alternative cause of the outcome of interest (allograft rejection), then this factor (now a potential confounder) must be carefully measured and accounted for. In this study, the authors considered potential confounders individually, a widely used but unconvincing method (6), rather than employing a counterfactual (causal) model building strategy (7), which might have provided evidence against confounding. Thus, although there may have been demonstration of “statistical significance,” a p value of 0.05 is not a litmus test for the truth and should not distract us from more elusive threats to validity.
The impact of bias must also be considered in observational studies. The authors’ primary measure of exposure was the area under the curve of chemokine concentrations over a years’ time, an unproven measure that may have been influenced by the number of bronchoscopies performed and the indication for each bronchoscopy. This approach gives equal weight to early and late measurements and similarly ignores differences between brief high concentrations and longer-lived low concentrations. Most importantly, early decrements in lung function might have prompted additional bronchoscopies that misclassified exposure differentially by outcome. Such a situation could easily have led to the detection of an association between a BAL biomarker and BOS when none truly exists.
Studies that ignore or inadequately handle confounding and bias are the norm in the field of lung transplantation, but this need not be the case. We must move our field beyond single-center studies that treat humans like lab mice and instead embrace a large-scale “systems” approach that relies heavily on established epidemiologic principles (8). This is no small feat. It requires a multicenter approach to adequately control for confounding and to minimize selection bias. We must standardize methods (including quality control) across centers for spirometry, BAL and transbronchial lung biopsy and we are in dire need of a reliable BOS classification scheme that has construct validity for biological measures and predictive validity for important outcomes. Once established, the next step is to garner support from funding agencies to develop networks similar to the Immune Tolerance Network and the Clinical Trials in Organ Transplantation projects, which have largely (but not completely) ignored lung transplantation. The onus is on us to prove that we are ready for just such an investment by showing that our methods are sound. Think about what Neujahr could have done if the considerable infrastructure required to successfully perform translational research in lung transplant recipients were already in place. Let's not let him down.
The author of this manuscript has no conflicts of interest to disclose as described by the American Journal of Transplantation.