Are the conclusions supported by the evidence?

Many people might be confused about the term ‘placebo’ that is used in Baxendale et al's study.¹ The paper clearly refers to the low-intensity-light arm as receiving placebo treatment, and the clinical trial registration (http://clinicaltrials.gov/show/NCT01028456) also indicates that the low-intensity group is receiving a placebo. However, this has some implications for the interpretation of the results.

If the low-intensity arm is indeed a placebo, the active treatment group did not differentiate from placebo and this is, therefore, a negative study. If, however, the low-intensity arm is receiving an active treatment then there is no placebo group and we cannot determine whether any changes in symptoms were due to the treatment or would have occurred by chance.

The conclusions that light therapy may ‘be an effective treatment for symptoms of low mood in epilepsy at lower intensities than those typically used to treat seasonal affective disorder’ cannot be supported by the findings of this study, since there was not an adequate control group. Further, the authors acknowledge that a number of non-specific factors may account for any improvements in depression and anxiety and all participants received relaxation. I strongly suspect that the fact that the participants had their eyes open during relaxation does not negate the effects on anxiety that relaxation training might have. In addition, most of the improvement in both groups (particularly on the depression subscale) had occurred before they were exposed to the intervention, i.e. at T ₂.

The clinical trial registration indicates that the control arm should have been receiving 100 lux for 30 min a day and the active arm 10 000 lux for 30 min a day. The study suggests that both arms received 20 min of light per day, with the control arm receiving an intensity of 2000 lux. It is not clear why the intensity was increased.

The attrition rate was high in both groups: 18/45 (40%) in the control arm and 15/46 (32.6%) in the active arm. Five patients in the active arm had an increase in seizures or required their medication to be increased (compared with two patients in the control arm). In the other paper emerging from this study,² the authors caution about using bright light in this population because ‘it may result in an increase in seizures for some’. None of this caution is evident in the paper published in the British Journal of Psychiatry. Indeed, there is not a single mention of adverse effects, despite them being reported elsewhere.

The analysis does not appear to have been intention-to-treat, and the results are only reported for those patients that completed the trial. This is a significant weakness when the authors have reported the possibility of adverse effects in other journals and when the attrition rates are relatively high. It is not clear why this intervention in an epilepsy population is treated with some reservations, yet it is reported much more favourably when there are some improvements in a secondary outcome measure reflecting some aspects of mental health (anxiety and depressive symptoms) which occurred before the intervention.