We would like to thank Tsikas and colleagues for their comments on our article Dietary Weight Loss, Exercise, and Oxidative Stress in Postmenopausal Women: A Randomized Controlled Trial (1).
In their letter, they state, in our view, erroneously, that the utility of 8-isoprostaglandin F2α, malondialdehyde (MDA) and 4-hydroxy-2-nonenal (HNE) as markers of free radical–induced oxidative stress is limited because they are also produced enzymatically from arachidonic acid by COX. This statement is questionable as Morrow and colleagues have demonstrated that COX either does not produce 8-isoprostaglandin F2α or produces such low quantities that they do not impact 8-isoprostaglandin F2α levels (2). The other markers MDA and HNE require further study before making any definitive statement on this topic.
Tsikas and colleagues also comment on the role of artifactual formation in stored biospecimens and describe lower analyte concentration in plasma collected at the end of two of their clinical studies, compared with levels collected at the beginning of the study, suggesting that decay is occurring during specimen collection or storage. However, we would expect levels of 8-isoprostaglandin F2α levels to increase with artifactual oxidation. The trial groups decreased instead, which might suggest other analytic issues.
We agree that plasma samples must be collected very carefully. We have collected blood directly into EDTA tubes on ice water, followed by processing within 90 minutes and then storage at −80°C. However, our stability studies have shown no changes in 8-isoprostaglandin F2α concentrations during the collection period. Importantly, we have stored plasma at −80°C, and in repeated assays of pooled plasma controls over a 7-year period, demonstrated plasma F2-isoprostane stability over time (3, 4). However, we recognize that serum requires further study for a complete characterization of its stability during collection and storage.
Tsikas et al. suggest that the decrease in plasma F2-isoprostanes might be due to the difference in the samples' age at the time of analysis. However, the blocked randomization design, a method used to ensure a balance in sample size across groups over time (5), of the parent study means that participants were recruited at the same rate into each arm of the study (6). Thus, on average, the samples in each arm were stored for the same length of time. Therefore, given the use of the generalized estimating equations modification of linear regression to account for intraindividual correlation over time in the parent study, sample storage time would not meaningfully change the results, as it would be proportional across each arm.
Finally, Tsikas and colleagues also suggest that the lower decrease in the control group may be due to differences in nutrition, making them unsuitable for a placebo group. We are unclear about this underlying point. In the parent study, intervention groups randomized to the diet intervention were advised to alter their diet, unlike the control group. A previous report based on this study found that, unlike participants randomized to the intervention arms, no statistically significant changes in fat or other nutrient intake occurred in participants randomized to the control arm, comparing baseline with 12-month intake (6). As the study was designed to investigate the role of dietary weight loss and exercise on markers of oxidative stress, the use of this control group is appropriate.
We appreciate the opportunity to discuss these important issues in more depth.
See the original Letter to the Editor, p. 485
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.