We appreciate the opportunity to respond to a recent editorial in the Journal by TearLab’s Chief Medical Officer, Chief Scientific Officer, and others who have financial involvement with the company criticizing our peer-reviewed manuscript in the same issue. In that article, we demonstrated that there was no correlation (among 186 patients) between longitudinal change in osmolarity (as measured by the TearLab osmometer) and changes in either corneal fluorescein staining or in patient symptoms, 2 widely accepted outcome measures in dry eye disease. As ocular surface disease investigators, we certainly understand the difficulties in assessing patient responses in dry eye, including in clinical trials. Although we were early adopters of the TearLab osmometer to aid in the diagnosis of some cases of mild dry eye, our recent data suggest that this technology may have limited value in monitoring patient response to therapy. Our findings are in accord with other recent studies suggesting a lack of correspondence between tear osmolarity and other clinically relevant measures of dry eye disease, as well as a very recent study that found an inverse correlation between osmolarity and symptoms in Sjögren syndrome patients.
We conducted our study because the usefulness of tear osmolarity in managing dry eye disease has not thus far been demonstrated in any large study. We agree that dry eye parameters can fluctuate over time, sometimes independently of one another, and that this can cause challenges, including discrepancy in findings among studies and inconsistent results in randomized trials. However, these shortcomings seemingly can also apply to tear osmolarity measurements. A key and perhaps primary goal of dry eye therapy is relief of patient symptoms; another is improvement in ocular surface health, for example, as measured by surface staining. Yet our data demonstrate that nearly half the patients who had improved symptoms showed increases in tear osmolarity and the other half showed decreases in tear osmolarity. The same 50–50 split was seen when evaluating patients with improvements in surface staining over time.
We find the methodological critiques levied against our peer-reviewed article in the aforementioned editorial misleading for several reasons. First, the editorial authors make numerous references to our study (clearly described in our article as a retrospective, observational cohort study) as a “trial” and, after erroneously doing so, impugn the study for a lack of “controls,” setting up a straw-man argument. Second, it is both clinically and statistically valid to examine a cohort of dry eye patients and assess correlations among changes in clinically relevant signs and symptoms. Indeed, the editorialists reiterate findings from a small TearLab-funded study they conducted in which they also examined a cohort of dry eye patients with no control group. Third, the editorialists actually refer to the TearLab-funded study as “the unbiased data,” directly implying ours must be biased. Fourth, the editorial authors state that whereas their data are somehow immune to regression to the mean, our findings are fully attributable to that phenomenon. In fact, there are several reasons why that is highly unlikely, among them: (1) regression to the mean is primarily of concern when extreme cut-points are chosen, which was not the case in our study, which included a typical clinic population ranging from mild to severe disease; (2) we presented data on subgroups based on osmolarity cut-points (308 mOsm/L and 314 mOsm/L) and still observed no correlation between changes in osmolarity and either symptoms or staining; and (3) we also presented data on a number of disparate subgroups based on various cut-points and reported consistent results: no correlation between change in osmolarity with change in either corneal staining or symptoms.
Finally, the editorialists’ comparison of tear osmolarity with glycated hemoglobin (HbA1c) in monitoring of diabetes is inaccurate. The primary reason glycated hemoglobin is used for monitoring chronic hyperglycemia and disease control is because of its strong proven associations with microvascular and macrovascular complications of diabetes. However, in contrast to the editorialists’ contention, there are limited and contradictory data on the usefulness of osmolarity measurements for assessing disease progression or improvement in any long-term or large study. In fact, the authors and their company have emphasized the substantial fluctuations of osmolarity measurement in dry eye patients —entirely the opposite of glycated hemoglobin measures that tend to be quite stable, and yet are responsive over time to therapy. Indeed, tear osmolarity, as measured by the TearLab device, can show significant variance even in the same eye at the same visit.
The example of diabetes actually could be an apt one to emphasize the apparent limitations of tear osmolarity in disease monitoring in dry eye. For example, there is little doubt that diabetes is mostly the result of insulin deficiency (type I) or insulin resistance (type II). Similarly, there is agreement that hyperosmolarity plays an important role in the development of dry eye disease and surface epitheliopathy. However, insulin levels are not a useful index for monitoring diabetic disease severity, demonstrating that the identification and ability to measure a factor, even one that plays a central role in disease pathogenesis, does not necessarily translate into that factor being a reliable biomarker for disease progression.
We believe that a fact-based and balanced editorial can serve a positive function in the healthy exchange of ideas, including on provocative topics. Although we recognize that strong convictions, amplified by commercial ties, can raise passions, we believe that the tenor of the referenced editorial was unnecessarily noncollegial and inflammatory. Several of the same authors recently made a similar indictment of other investigators’ independent work that had also been mildly critical of the TearLab osmolarity technology. We believe that the Journal should have managed these transgressions more effectively.
All would agree that our literature should provide a venue for a healthy, reasoned, and dispassionate discourse on controversial topics. In that context, and because we believe we are all working to improve the health of our patients, we hope that we can continue our debates on the merits of various technologies in a collegial and evidence-based manner.