Abstract
Dear Sir In a recent issue of Cephalalgia, Bohotin and colleagues (1) present further data in support of their position that the visual cortex is hypoexcitable in patients with migraine. They make a number of observations about data analysis in earlier studies, including our own (2). We are concerned about some of these observations, and we here offer some suggestions for future work in this area.
Single-pulse transcranial magnetic stimulation over the occiput can result in the perception of phosphenes, which are probably due to the direct activation of neurones in the primary visual cortex (3), or possibly underlying white matter (4). A common strategy in neurophysiological research into migraine has been to establish the strength of the applied magnetic field required for phosphene perception. This is termed the phosphene threshold, and is usually expressed as a percentage of maximal output of the stimulating device used.
Interpretation of phosphene threshold data is complicated by several factors, as we have previously observed (5). One of the most important is variability in the anatomy of the occipital region. In standard transcranial magnetic stimulation (TMS) protocols, coil position is referenced to the inion, with the assumption that the calcarine fissure lies in the same mid-sagittal plane. In fact, the two are frequently displaced because of variation in the anatomy of the occipital bone (6, 7). Furthermore, the inion itself is difficult to define in up to 30% of patients (7). A certain number of participants in most published studies of phosphene thresholds in migraine do not perceive phosphenes even when stimulated at the maximum output of the stimulator. One compelling reason for this being the case is that the correct point for stimulation may not have been found in these participants. It is manifestly not possible, within the ethical and practical constraints of clinical research, to map phosphene thresholds over every square millimetre of the occipital skull. In our work (2), we report both the percentages of patients and controls perceiving phosphenes, and a statistical analysis of phosphene threshold that includes only those participants in whom phosphenes were actually induced. We are confident that this is an appropriate way of communicating our findings.
Bohotin and colleagues (1) appear to take an opposing view. They are of the opinion that participants in whom phosphenes cannot be induced should be assigned a phosphene threshold of 100%. This seems equivalent to an argument that insufficient magnetic field strength is available from the stimulator to induce phosphenes in those participants. As far as we are aware, there is no published evidence that bears directly on this question. Oddly, in their paper, the range of phosphene thresholds for healthy volunteers is given as 18–87%, which cannot be correct if they have replaced missing values with 100% throughout their data. This renders interpretation of their analysis rather difficult. This issue notwithstanding, their approach brings some statistical risks. First, the resultant dataset is unlikely to be normally distributed, as it will contain a large percentage of 100% values. The assumptions of standard parametric tests, such as the analysis of variance that they used, may therefore not be met. Second, it may act to obscure important differences in phosphene threshold in the subset of participants who do perceive phosphenes.
Rather than dwell on statistical uncertainties, it seems important to examine how TMS methodology may be improved. It is now possible to position a TMS coil stereotactically with reference to a structural magnetic resonance image of the individual participant's brain (8). This is clearly important in that it removes any uncertainty about where the primary visual cortex may lie relative to skull landmarks. It has the additional benefit that the distance between the face of the coil and the surface of the cortex may be calculated. Skull thickness is another anatomical reason for variability in phosphene thresholds, and previous studies have simply not been able to take it into account. Using stereotactic coil positioning would permit the use of skull thickness as a covariate in statistical analyses of phosphene threshold, so that, in effect, the confound introduced by irrelevant individual anatomical variation could be filtered out.
There are also other TMS-based methods available that circumvent the need for the reporting of phosphenes. We have shown that the suppression of visually presented letter stimuli by a TMS stimulus to the primary visual cortex is reduced in migraine with aura (9). This finding has been replicated by Aurora and her colleagues (personal communication) and is consistent with a functional hyperexcitability of the visual cortex. Brighina and colleagues (10) have used a very ingenious repetitive TMS protocol to examine the perception of illusory contours, also finding evidence for cortical hyperexcitability in migraine. Both methods overcome the subjectivity inherent in phosphene-based studies.
We are strongly of the opinion that TMS has an important role to play in future studies of cortical function in migraine. It is clear, however, that both methodological rigour and appropriate statistical sophistication will be required.