Authors' reply

We agree that phosphene thresholds are typically defined at the 50% response rate level, although it should be recognised that the setting of a threshold is an arbitrary process. A number of our participants had thresholds near and approaching the maximum stimulator output and use of a lower level of threshold acceptance allowed for a more precise estimation of their visual cortical excitability. Importantly, given the comparability between stimulus response plots of controls and patients with dementia with Lewy bodies (Fig. 2), ¹ it is unlikely that use of a 25% cut-off for threshold adversely affected our findings.

Dr Brigo highlights the issue of non-response to the stimulation and that this may be as a result of causes other than insufficient stimulation strength. Indeed, phosphene perception, or lack of, may not necessarily originate in the visual cortex but may depend on higher visual areas or indeed non-visual areas as well as recurrent processing. ^2,3 However, the reasons that Dr Brigo presents to explain the non-response - including imprecision in finding the optimal position for stimulation delivery over the occiput, greater depth of the primary visual cortex leading to reduced magnetic field strength at the level of the cortex, and use of the figure-of-eight coil - are actually arguments supporting the assumption that failure to respond in some individuals is due to insufficient current stimulation to the neural locus responsible for phosphene elicitation.

In our study we sampled nine equally spaced scalp sites, giving good symmetrical cover of the occiput; this was a compromise between precision and limiting the experiment duration in a vulnerable patient group. The figure-of-eight coil has been frequently used in phosphene research (e.g. Kammer et al ⁴ ) and was chosen because of its spatial accuracy; larger, diffuse-field coils could theoretically activate areas external to the visual areas of interest or indeed induce retinal phosphenes. In addition, we would contend that the transcranial magnetic stimulation (TMS) methodologies we employed meant that we had comparable and, in some cases, better rates of phosphene response compared with other studies in young healthy individuals.

Dr Brigo indicates potential differences in the lower and upper visual cortical activation with TMS and certainly our data of greater phosphene elicitation in the lower visual fields supports this. Our use of the adjusted phosphene threshold ratio to control for group differences in atrophy also accounted for skull thickness, although whether the positions we chose for these measurements directly related to the precise locus of stimulation on the visual cortex, we agree, is a methodological limitation. The use of magnetic resonance-guided stereotactic coil placement, for example, would help with this issue and allow for more precise threshold determination.

As suggested by Dr Brigo we performed an analysis only on those participants who responded to TMS (controls, n = 17; patients, n = 17) and the findings were in line with our main analyses: there were no significant differences between the controls and patients for phosphene threshold (controls: median 64.0% (IQR = 32.5%); patients: median 67.0% (IQR = 20.0%); U = 139.5, P = 0.87) and phosphene response rate (controls: median 6.0 (IQR = 7.0); patients: median 8 (IQR = 5); U = 112.5, P = 0.27). Correlations between the Neuropsychiatric Inventory hallucinations subscale score in patient responders and the phosphene excitability measures (phosphene threshold, Kendall's t = —0.28, P = 0.15; phosphene response rate, t = 0.46, P = 0.02) were in the same direction as the main analysis, although less significant owing to the smaller sample and the fact that the four patients who did not respond to TMS at the maximum stimulator output had significantly less severe and frequent visual hallucinations compared with patient responders (Mann-Whitney U-test 16.5, P<0.001). Clearly, the lack of phosphene response (regardless of cause) is associated with fewer visual hallucinations and thus we would argue that inclusion of non-responders in our analyses is essential in providing a more holistic understanding of the underlying aetiology of this symptom in dementia with Lewy bodies.