Particularly relevant for mood disorders recent work has sho

Particularly relevant for mood disorders, recent work has shown that the neuromodulator, dopamine, decreases visually-evoked BOLD signals at the same time as increasing the local cerebral blood flow, suggesting a disproportionate increase in energy metabolism (Zaldivar et al., 2014). Thus, as neuromodulator concentrations change throughout the day and during different behaviours (or with drug treatment), the relationships between the BOLD signal, blood flow, energy metabolism and neuronal activity can be profoundly altered. If these relationships are not constant across all selective serotonin reuptake inhibitors regions and neuronal pathways, then the employment of a control task or stimulus will never be sufficient to exclude the possibility that neurovascular coupling differences are in fact responsible for BOLD signal differences seen with any other task, stimulus or brain region of interest. The most promising approach for separating neuronal activity contributions to the BOLD signal from neurovascular coupling or metabolic contributions, is to directly measure electrical activity, using techniques such as EEG (Vitali et al., 2015), and/or blood flow, using techniques such as arterial spin labelling (Wang et al., 2011). Performed alongside fMRI, data from these techniques can reveal whether the relationship between neuronal activity and blood flow remains constant across participant groups, brain regions and behavioural tasks. The addition of even one of these techniques can vastly improve the interpretation of fMRI data. For example, if an increased BOLD response is accompanied by an increased EEG response to the same task, this is a good indication that the BOLD signal increase reflects an increase in neuronal activity. If, on the other hand, data from the two techniques contradict each other, this would indicate that additional factors, such as neurovascular coupling differences, are contributing to the BOLD signal difference between groups. Several labs and clinicians are beginning to implement direct measures of neural activity or blood flow alongside fMRI, with promising results (e.g. Dinstein et al., 2012; and see Wang et al., 2011 and Vitali et al., 2015 for thorough reviews). For fMRI studies investigating the brain mechanisms of drug action in particular, it may be useful to employ some additional techniques to dissociate the pharmacological effects on neural activity and neurovascular coupling. Murphy and Mackay (2011) provide a thoughtful discussion of the promising directions and general caveats for fMRI analysis of drug action. One interesting approach is time-series analysis; examining BOLD signal changes during the precise time-scale over which the drug is known to act. If these changes correspond to the time-scale of behavioural changes, then it is likely that they reflect some changes in neuronal processing. This method involves acute administration of the drug, which has the advantage that a placebo control group can also be examined, without the ethical concerns of longer-term placebo treatment. A behavioural difference between the drug and placebo group (as has recently been observed for acute fluoxetine treatment; Capitão et al., 2015), would lend confidence to the idea that the drug causes neural activity differences, which could theoretically be detected by fMRI. Of course, whether the BOLD signal differences that are observed accurately represent these neural activity differences, still depends on whether the relationship between neural activity, blood flow and energy use is the same between drug and placebo groups. Along similar lines, if BOLD signal changes in response to longer-term drug treatment are predictive of clinical improvement (e.g. Cullen et al., 2016, discussed in Section 4), this suggests that the BOLD signal changes are likely reflecting neurobiological changes that contribute to the observed behavioural changes. Again, confidently ascribing these BOLD changes to neuronal activity changes still depends on excluding the possibility that the drug directly interferes with neurovascular coupling. Bourke and Wall (2015) provide a good review of strategies through which such confounding factors in the interpretation of pharmacological fMRI can be mitigated.