Introduction

Transcranial direct current stimulation (tDCS), a noninvasive neuromodulation technique, has putative potential to treat obesity as even single sessions have been found to suppress food craving and eating ([1, 2] and [3] for review of studies). The typical target for stimulation is the dorsolateral prefrontal cortex (DLPFC), a region involved in the inhibitory cognitive control of eating and possible suppression of food reward [3]. Scientific validation of tDCS to treat obesity is hampered, however, by the ability of participants to distinguish real from control conditions [2, 4,5,6,7,8]. This problem may also underlie inconsistent and null results in craving and eating studies using tDCS [3, 9, 10]. The traditional control condition involves administering current to the same area of the head as the real condition, but only very briefly at the beginning and end of the session. The real condition involves administering current throughout the typically 20-min session [1, 4]. Consequently, there are more physical sensations (e.g., itching, tingling, burning), and changes such as increased fatigue and decreased attention, during real vs. control tDCS [10,11,12]. These experiences have led participants and patients to correctly guess real from control conditions at a rate well above chance when administered 2 mA current, [4, 5], the intensity typically administered in food craving and eating studies, and even when administered 1.5–1 mA current [2, 6, 7].

Believing that one has received real vs. control tDCS is a problem for validation of the technique because positive outcomes could be the result of expectation or placebo effects, not stimulation-induced neural changes [13]. In a previous study in our lab, participants with overweight and obesity craved and ate significantly less food when told they were going to receive real tDCS, even when they actually received the control condition [14]. In sum, the traditional control has made it challenging to determine if reports of reduced food craving and eating are due to neuromodulation or expectation effects. Therefore, the goal of this study was to more rigorously evaluate the effect of tDCS on food craving and eating by using a control method that would be difficult to distinguish from the real condition. This was achieved by administering current for the entirety of a session, just as in the real condition, but over a cortical location that was not expected to affect appetite.

Methods

Participants

Twenty-eight participants enrolled in “Introduction” to Psychology classes at the University of Alabama at Birmingham (UAB) participated for research credit (19F/9 M; 19 Black, 8 White, 1 Asian). The mean age was 21, 18–41, and mean body mass index (BMI) was 34.0, SD = 7.05, 26–50. Exclusion criteria were implanted biomedical devices, history of brain trauma or brain surgery, schizophrenia, bipolar disorder, eating disorder, suicidal ideation, in a weight-loss program, pregnancy, breastfeeding, illicit drug use, intent to stop or start a prescription drug known to influence appetite, and allergy to ingredients in the test foods. Inclusion criteria were BMI ≥ 25 and age 18–55.

Transcranial direct current stimulation (tDCS)

Participants received real and control tDCS in counterbalanced order with at least 48 h between sessions. Both real and control conditions delivered 2 mA current for 20 min. For the real condition, electrodes were affixed over the dorsolateral prefrontal cortex (DLPFC), the most common target in appetite studies, using an anode-right/cathode-left montage (F4/F3 of EEG 10–20 system). For the control condition, electrodes were placed over the sensorimotor cortex (SMC) with an anode-left/cathode-right montage (C3/C4). The SMC was chosen over other piloted sites because (a) stimulation produced the most similar physical sensations to DLPFC stimulation, (b) it was reasonably posterior to the DLPFC without producing phosphenes, (c) it did not require more head straps than DLPFC stimulation, and most importantly, (d) the SMC was not known to affect appetite. A 1-channel stimulator (TCT Research, Hong Kong) with 40 × 60 mm electrodes was used.

Measures

Body mass index (BMI), hunger, and demographics

Height and weight were measured and BMI was calculated as kg/m2. A hunger scale assessed hunger and fullness using a visual analog scale from 1 to 10. Anyone reporting levels outside 4–7 (neutral) was rescheduled. Sex, age, and race were collected via electronic survey.

Psychological questionnaires

The Barratt Impulsiveness Scale-11 assessed degree of attention, motor, and planning impulsiveness. The Palatable Eating Motives Scale assessed frequency of eating tasty foods for motives unrelated to hunger. The Dutch Eating Behavior Questionnaire-Restraint assessed dieting intent and behavior. The Binge-Eating Scale assessed a range of binge-eating severity to control for possible subthreshold binge-eating disorder. The Short Suggestibility Scale assessed propensity to believe messages and to be persuaded by others. Scores on most of these measures were previously found to affect tDCS outcomes [1].

Physical Sensations Scale (PSS) ratings

A paper/pen PSS was designed and used to assess presence and intensity of various physical sensations, general discomfort, attentiveness, and fatigue from a scale of 0–10 in four evenly spaced time intervals during each tDCS session.

Food craving task

Participants rated 24 images on a computer for general liking and for wanting the food “right now if it was available” on a 1–5 scale. Wanting scores for any food with a liking score of one or two were removed from analyses to avoid floor effects due to not liking the food in the first place. A mean score was derived across items representing each of the four food types: sweets, fatty proteins, carbohydrates, and mixed macronutrients (e.g. pizza, nachos). Difference scores (pre- minus post-tDCS ratings) were used in analyses.

Eating test

Participants were left in private room for 12 min with a generous amount of pre-measured Double-Stuf Oreo® cookies, Skittles® candies, and Lay’s Classic® potato chips, and a small bottle of water. They were instructed to at least taste each food in order to complete a palatability rating sheet. This was a ruse for the actual test which was to measure amount consumed. They were instructed to eat as much food as they wanted because all remaining food would be discarded for sanitary reasons. Remaining food was weighed and converted to kilocalories.

Real vs Control Interview (RCI)

An in-person RCI was designed and conducted after the last tDCS session. The RCI asked participants to guess whether they had been given 2 “real”, 2 “fake”, or 1 “real” and 1 “fake” tDCS session and to provide a confidence rating for their guess on a Likert-like scale. Participants then gave reasons for their guess in an open-ended fashion. Participants who guessed that both sessions were real were asked to guess again after considering that one was actually real and the other fake.

Procedure

During consenting, participants were read aloud the section stating, “tDCS has been shown to decrease, increase, or have no effect on food craving and eating”. Prior to each visit they were asked to come into the lab not too hungry or full. This was checked with the hunger scale. On the first visit they were measured for a BMI and completed the demographics and baseline questionnaires electronically. Then (and on both visits), they completed the pre-stimulation food craving task and were administered real or control tDCS, completed the PSS, and then the post-stimulation craving task. On the second visit, participants underwent the same procedures but this time they received the alternate tDCS condition. They were then interviewed with the RCI, and debriefed.

Statistical analysis

Outcome measures were inspected for normality; there were no outliers. Repeated Measures ANOVAs determined differences between pre- and post-stimulation food craving and possible tDCS condition interactions. RMANOVA also assessed between-condition differences in kcals eaten, and in PSS ratings. For all RMANOVAs, age, BMI, and baseline trait scores were entered separately as covariates, and sex and ethnicity as between-subject factors. Chi-square (X2) determined if number of participants guessing conditions as real or fake differed from chance. Alpha was set at 0.05 for significance and partial eta squared (Ƞp2) determined effect sizes with 0.01, 0.06, and 0.14 denoting cutoffs for small, medium, and large effects, respectively.

Results

Effect of the tDCS conditions on food craving and eating

Food craving ratings did not differ for any of the food types following tDCS. There was no tDCS condition interaction. An exception is that when BMI was controlled, craving for sweets was suppressed by the control condition (SMC-targeted stimulation) but not the real condition (DLPFC-targeted stimulation); p = 0.02, Ƞp2 = 0.19. Controlling for demographics or psychological traits did not change these results. As for amount consumed, participants ate a total of 378.2 ± 48.5 kcals after DLPFC stimulation, vs. 410.6 ± 40.3 kcals after SMC stimulation, but this difference was not significant (p = 0.31, ns). Controlling for demographics, BMI, or psychological trait scores revealed no significant tDCS condition effects on eating.

Effect of tDCS conditions on PSS ratings

There were no differences in mean discomfort (2.04 vs. 2.01), attention (6.02 vs. 5.74), or fatigue (2.97 vs. 2.6) ratings between real and control conditions, respectively, all ns. In addition, as shown in Fig. 1, there were no tDCS condition differences at any of the four time points during the sessions for general discomfort (a), fatigue (b), or attention (c). Ratings at each time point within conditions for these experiences also did not differ. Finally, as shown in Fig. 2, there was no difference between tDCS conditions on the frequency of individual kinds of physical sensations reported.

Fig. 1
figure 1

Mean Physical Sensations Scale (PSS) ratings of discomfort (a), fatigue (b), and attention (c) during transcranial direct current stimulation (tDCS) targeting the sensorimotor cortex (SMC) vs. dorsolateral prefrontal cortex (DLPFC) as the control and real condition, respectively

Full size image
Fig. 2
figure 2

Mean number of times that participants reported sensations on the Physical Sensations Scale (PSS) during transcranial direct current stimulation (tDCS) targeting the sensorimotor cortex (SMC) vs. dorsolateral prefrontal cortex (DLPFC) as the control and real condition, respectively

Full size image

Guessing tDCS conditions as real or control

Fifty percent of the participants guessed that both conditions were real (R/R), and 50% guessed that one was real and the other was the control (R/C). No participant guessed that both conditions were controls. The R/R and R/C guessers did not differ in degree of confidence behind their guess (ratings 7.21 ± 0.50 vs. 6.54 ± 0.50, ns). When the R/R guessers were asked to consider that only one of the conditions was real and one was control, 8 of 14 guessed DLPFC stimulation was real, and 6 of 14 guessed SMC was real, a non-significant difference (X2 = 0.79, ns). Ninety-six percent of the participants identified physical sensations during stimulation as the primary reason for their guess. Importantly, when PSS ratings for R/R and R/C guessers were compared, no differences were observed.

Discussion

This study assessed the ability of tDCS targeting the DLPFC to suppress food craving and eating when treatment expectations were minimized by reducing the discernibility of real and control conditions. Under these conditions, DLPFC stimulation failed to suppress these variables. This is in contrast to previous reports of DLPFC tDCS reducing craving and eating, including reports from our lab [1, 2] and others′ (see [3]) when using the traditional-control. However, the findings were consistent with reported null findings [1, 3, 9, 10]. Since 71% of participants were unable to guess real from control conditions, confidence is increased that the control method minimized the confounding effects of expectation. This is a significant improvement over rates in most studies that have published this statistic [2, 4,5,6,7,8]. Further, those who guessed receiving one real/one control session did not rate physical sensations during stimulation —or other experiences— differently between the two conditions.

A limitation of the study is that participants did not also receive a traditional sham condition to directly evaluate the advantage of the constant-current alternate-location control. However, we are confident that DLPFC tDCS failed to reduced cravings because baselines ratings did not differ from post-stimulation ratings. The control condition, SMC stimulation, also did not reduce craving, except for sweets when covarying for BMI. That SMC, not DLPFC stimulation, decreased this craving strengthens confidence in the null effects of DLPFC stimulation on craving. However, for food consumption, we had no baseline measure (though most studies use amount eaten after the traditional-control as the value for comparing to real tDCS [1,2,3]). Hence, it is possible that a true suppression on consumption by DLPFC stimulation was masked by a similar decrease in consumption by SMC stimulation, relative to the amount participants might have eaten following a traditional-control session. However, if participants with higher BMIs are able to discern real from traditional-control tDCS, expectation effects would incline them to eat less after real vs. control tDCS. In this study, expectation effects were diminished. While this strengthens confidence in the null effect of DLPFC tDCS on eating, future investigations that include a traditional-control condition are needed to conclude the superiority of SMC stimulation as a control for obesity tDCS studies.

Another study limitation is that we cannot know if the ambiguous information regarding the effects of tDCS on appetite (read to the participants from the consent form) had any additive effect on the predominant inability to guess real from control conditions. However, this procedure is also a strength of the study in that it minimized the possibility that preconceived expectations of tDCS could mask any significant outcomes. Since a high BMI was a criterion, the participants naturally may have suspected that tDCS was meant to decrease, not increase, their craving and eating. If they came into the study with this expectation, however, it did not appear to influence the results. This might not have been the case if the traditional control had been used, as it would have been easier to discern real from control conditions. Another strength and unique feature from past tDCS studies on appetite is that our sample was predominantly Black and included mostly BMIs in the obesity and severe obesity (class II and III) categories. This made our sample very relevant to populations most in need of obesity treatments [15].

Conclusion

Overall, this study did not find within-participant differences between the effects of tDCS aimed at the DLPFC and at the SMC on craving or amount of food eaten. The lone exception was a significant suppression of craving for sweets by SMC stimulation when controlling for BMI. Additionally, the results support the use of active alternate-site stimulation as a control method for tDCS research. The method’s primary advantage over the traditional sham control is demonstrated by our finding that participants were largely unable to identify it as a control condition, thereby allowing expectation effects to be better controlled. Together, these findings question if the DLPFC is the optimal target for suppressing craving and eating. They also might explain disparate and null findings reported by others targeting this brain region [3, 9, 10]. Whether the SMC is a better tDCS target to alter appetite remains to be determined through additional investigations.