Evidence Transcranial Direct Current Stimulation Can Improve Saccadic Eye Movement Control in Older Adults
Round 1
Reviewer 1 Report
Some general comments:
How long the effects last? Is there any long-lasting benefits?
The fact that in the follow-up study did not meet the sample size..does this affect the results?
Why the sample size was based on the study on the young adults?
Why use 10-20 EEG system over MRI to determine the positioning of anodal electrode? Wouldn’t increasing the electrode size to a (5 x 7 cm) reduce precision [ see line 72 & line 93]
Since all the participants are male, maybe do a study with females to see if same effect is observed.
Since the participants completed five blocks of eye movements in a sequential order-even though trials within were randomized- could the median reaction time still be affected by practice effects?
Since this is a repeated measures study, even though a 7 day period separating active and offline experiments, could there still be lasting effects that carried over to the 2nd experiment ?
Is the research design appropriate?
I'm not familiar with Antoniades et al. (2013) on which the experimental design was based.
But lines 171-173 indicate that while 120 trails were repeated for both pro and antisaccades 3 antisaccade blocks were repeated consecutively. By contrast, prosacade blocks did no occur consecutively. Ideally, the presentation of pro and ant saccades should also follow a crossover design.
A minor point is that statistical analysis might have been simplified by use of a multilevel linear model for repeated-measures, which is not affected by sphericity.
Are the methods adequately described?
Yes.
Are the results clearly presented?
Table 1 Error rates are clear, but the overall layout of the table is confusing. Otherwise the results are well documented. Please clear up.
Are the conclusions supported by the results?
In part, yes. The results suggest "online" or engagement of saccadic eye movements during tDCS may induce improved saccadic performance. However, that 3 of 16 study recruits could not perform the required saccadic exercise suggests the therapeutic application may be restricted.
More importantly, the low final sample size of just 10 certainly places a restriction on the implied potential therapeutic relevance.
Potential oculomotor therapeutic benefit to older adults is significant.
While study participant size was low, the crossover design lends weight to the results.
However, the overall design (based on Antoniades et al. (2013)) may bias outcome as noted, and that 3 of 16 (~19%) of recruits could not perform the task is anther potential limitation.
On line 324 they state they should - or someone should - replicate on a larger population pool.
If the Antoniades et al. (2013) adopted is regarded as sound, then the paper has good merit. However, while the latter author justifies the protocol sequencing (https://www.sciencedirect.com/science/article/pii/S0042698913000357), the current study would be elevated by also assessing results of a reversed vision of the Antoniades et al. (2013) approach.
Style comments.
-Fig 3 - remove horizontal lines - clutters graph
Author Response
Point 1: How long the effects last? Is there any long-lasting benefits?
Response 1: Our study was not designed to assess long-term effects, but we agree that this would be important to assess through future research. We analysed saccade block and did not find any evidence of tDCS-related performance changes across the timeframe that we assessed. We now specify this at the start of the Results and also mention in the Discussion (third paragraph) the need for future research to assess the duration of the effects.
Point 2: The fact that in the follow-up study did not meet the sample size..does this affect the results?
Response 2: We have made it clear in the Discussion that replication in a larger sample will be needed. As pointed out in the Discussion, the effect size was fairly large and Bayes factor indicates that the data are over four times more likely under H1 than under H0. Nonetheless, we ensured that interpretation was done in the context of the small sample size.
Point 3: Why the sample size was based on the study on the young adults?
Response 3: This is the only study to date reporting saccadic eye movement control benefits from tDCS. We have clarified this in the revised Methods (at the end of the Statistical Analyses section).
Point 4: Why use 10-20 EEG system over MRI to determine the positioning of anodal electrode? Wouldn’t increasing the electrode size to a (5 x 7 cm) reduce precision [ see line 72 & line 93]
Response 4: EEG, instead of MRI, makes the tDCS protocol more clinically feasible (by using a quicker and less expensive method of determining electrode positioning). The larger electrode does indeed mean a less focal (more widespread) electric field, which we think suits both the use of EEG instead of MRI and the widespread (diffuse) nature of brain aging. These points are described in the last paragraph of the Introduction, and are reinforced at the start of the Discussion. We edited these points for clarity.
Point 5: Since all the participants are male, maybe do a study with females to see if same effect is observed.
Response 5: We have added a comment about this in the Discussion (3rd paragraph).
Point 6: Since the participants completed five blocks of eye movements in a sequential order-even though trials within were randomized- could the median reaction time still be affected by practice effects?
Response 6: There was no evidence of practice effects across the saccade blocks, and saccade block did not interact with the stimulation condition in either experiment. We now state this at the start of the results section. Thank you for the suggestion.
Point 7: Since this is a repeated measures study, even though a 7 day period separating active and offline experiments, could there still be lasting effects that carried over to the 2nd experiment ?
Response 7: Experiment was a between-subjects factor. tDCS condition (sham vs. active) was a within-subjects factor with a 7 day washout. We agree that carryover effects are possible. To address this, we counterbalanced the order of the tDCS condition (sham vs. active) across participants within each experiment. In the revised manuscript we ensured that this is clearly stated within the Method section, and we also commented on the possibility of carryover effects watering down differences between the active versus sham conditions (see Discussion, third paragraph).
Point 8: Is the research design appropriate? I'm not familiar with Antoniades et al. (2013) on which the experimental design was based. But lines 171-173 indicate that while 120 trails were repeated for both pro and antisaccades 3 antisaccade blocks were repeated consecutively. By contrast, prosacade blocks did no occur consecutively. Ideally, the presentation of pro and ant saccades should also follow a crossover design. If the Antoniades et al. (2013) adopted is regarded as sound, then the paper has good merit. However, while the latter author justifies the protocol sequencing (https://www.sciencedirect.com/science/article/pii/S0042698913000357), the current study would be elevated by also assessing results of a reversed vision of the Antoniades et al. (2013) approach.
Response 8: This is an interesting suggestion. We were not involved in designing the internationally standardised antisaccade protocol (Antoniades et al., 2013), which was designed by highly knowledgeable experts. In addition, the eye movement protocol used was identical to Chen and Machado (2017), which allows comparison across the studies. We can consider adopting a crossover design for saccade type in future studies, although we do think that standardization of protocols is important. We have added this point to the Discussion (third paragraph).
Point 9: A minor point is that statistical analysis might have been simplified by use of a multilevel linear model for repeated-measures, which is not affected by sphericity.
Response 9: We adopted the same statistical method as used in Chen and Machado (2017), which allows comparison across the studies. Good idea to consider using a multilevel linear model for repeated-measures in the future, thanks.
Point 10: Are the results clearly presented? Table 1 Error rates are clear, but the overall layout of the table is confusing. Otherwise the results are well documented. Please clear up.
Response 10: Thank you for pointing this out. We have fixed the formatting errors.
Point 11: Are the conclusions supported by the results? In part, yes. The results suggest "online" or engagement of saccadic eye movements during tDCS may induce improved saccadic performance. However, that 3 of 16 study recruits could not perform the required saccadic exercise suggests the therapeutic application may be restricted. More importantly, the low final sample size of just 10 certainly places a restriction on the implied potential therapeutic relevance. Potential oculomotor therapeutic benefit to older adults is significant. While study participant size was low, the crossover design lends weight to the results. However, the overall design (based on Antoniades et al. (2013)) may bias outcome as noted, and that 3 of 16 (~19%) of recruits could not perform the task is anther potential limitation. On line 324 they state they should - or someone should - replicate on a larger population pool.
Response 11: Yes, some of the participants (3 out of the full sample of 26) were excluded due to an inability to perform the antisaccade task. This equates to 11.5%. In the Discussion we point out the importance of replication in a larger cohort and discuss interpretation of the results in the context of the small sample size. In response to your points, we have added a limitation about >10% not being able to perform the antisaccade task (see Discussion, third paragraph).
Point 12: Fig 3 - remove horizontal lines - clutters graph
Response 12: Done.
Thank you so much for all of your comments and suggestions. Very helpful!
Reviewer 2 Report
The authors of this paper, using anodal stimulation of the DLPC in controls aged 65-71 years , showed that an ongoing stimulation decreased anti-saccade latencies better than testing after a Stimulation of 10 min. No effects were found in the error rate during anti-saccades and in the prosaccade parameters.
The paper is written fluently and the methods including statistics are sound. Literature is cited accurately. The number of controls is not very high and different groups are used in the off and on stimulation paradigm. The stimulation side was randomized and not compared within one subject.
Points of concern:
Due to the low number of participants some questions could not be answered, e.g. was the stimulation side relevant with respect to the right-or left handednesss.
Define the anti-saccade errors: normally there are type I and II errors. What were the results?
What is given in the bracket in table 1, the standard deviation? Please indicate.
Figure 3 should be much more detailed comparing data for pro- and anti-saccades as well as for the direction.
Was there an effect over time or in between the blocks in the off- and on-paradigm.
In the discussion it remains unclear what really causes the decrease in latency without changing the error rate. Please be more specific and present the data in a more general context,e.g. in comparison to the transmagentical stimulation data and lesion studies.
Author Response
Point 1: Due to the low number of participants some questions could not be answered, e.g., was the stimulation side relevant with respect to the right- or left handedness?
Response 1: According to the Measurement of Handedness (Chapman & Chapman, 1987), all participants in both experiments were right-handed except one (ambidextrous) in the offline experiment (this is reported in the Participants section). Thus, we cannot analyse stimulation side with respect to handedness. We have added in the Discussion (third paragraph) a comment about additional research being needed to determine whether the findings generalize to left handers.
Point 2: Define the anti-saccade errors: normally there are type I and type II errors. What were the results?
Response 2: Consistent with Kanai et al. (2012) and Chen and Machado (2017), the antisaccade errors analysed are reflexive errors (i.e., erroneous stimulus-directed saccades, which to our knowledge are not normally broken down into type I and type II). We have clarified this is the revised manuscript.
Point 3: What is given in the bracket in Table 1, the standard deviation? Please indicate.
Response 3: Yes, standard deviation. Thank you for pointing out this omission, which we have now added to the Table 1 Note.
Point 4: Figure 3 should be much more detailed comparing data for pro- and anti-saccades as well as for the direction.
Response 4: Thank you for this suggestion. We have added to Figure 3 the other two performance metrics (pro-saccade RTs and reflexive error rates during the anti-saccade task).
Point 5: Was there an effect over time or in between the blocks in the off- and on- paradigm?
Response 5: No, saccade block did not interact with the stimulation condition (p > .2 in all cases). We have now added a statement about this in the first paragraph of the results section.
Point 6: In the discussion it remains unclear what really causes the decrease in latency without changing the error rate. Please be more specific and present the data in a more general context, e.g., in comparison to the TMS data and lesion studies.
Response 6: The trend for improved reflexive error rates during the antisaccade task in the online experiment approached significance (p = .051) and the effect was of a medium size (dz = -0.578). In the revised manuscript, this is more readily apparent thanks to your suggestion to add reflexive error rates to Figure 3 (thanks for that!). In addition, we have revised our wording in the Discussion to ensure that it is clear that we are not claiming improvements in latency but not error rate; on the contrary, both latency and error rate tended to improve in the online experiment. Thank you for helping us to clarify this. In addition, we added comparisons with TMS and lesion studies in the Discussion.
Reviewer 3 Report
The manuscript describes two experiments designed to test the idea that anodal tDCS over frontal regions can improve oculomotor function in elderly people. Experiment 1 uses an offline stimulation protocol and report no effect of tDCS on prosaccades or antisaccades. Experiment 2 uses an online protocol and reports a statistically significant reduction in antisaccade latency from the tDCS compared to sham conditions which had a medium - large effect size. There was also a trend towards reduced error rate in the antisaccade task which had a medium effect size. There were no significant effects in the prosaccade task.
The rationale for the study is clearly explained and the methods and analyses are appropriate. I felt the reporting of effect sizes and Bayes factors greatly helped me understand the effects being reported and overall the results seem clear-cut and convincing. Overall I think this is an interesting study and see no major issues. Clearly it would be better if the online experiment was adequately powered, so it is unfortunate that it was not possible to match the original sample size. Given the small sample I think it might be helpful to report the results of a post-hoc power calculation. By my estimate the online study had a power of .74, which isn’t that far off the standard power of .8 used for apriori calculations. I think this might reassure the reader that the effect reported in the online condition is real.
The introduction claims that saccade control is reduced with age and implies that this has important clinical consequences. However, I think this line of argument would be stronger if the authors could cite some evidence that small but statistically significant changes in saccade latencies are meaningful from a clinical perspective. For example, is there any evidence that seniors have problems scanning busy intersections and an increased risk of a car accident while driving?
I would also like to see a bit more discussion about the mechanism by which tDCS selectively facilitated antisaccade production. As far as I can see, there is no explicit discussion of the cognitive or neural mechanism that might account for this facilitation. There is some generic discussion of non-oculomotor functions, but this is all rather vague and a bit unsatisfying. One possibility is that tDCS disrupts the computation of the salience of the onset associated with making an antisaccade, so the competition between the saccade vector triggered by the onset of the antisaccade and the desired saccade vector is resolved more quickly. This explanation would be consistent with TMS studies showing that stimulating over FEF seems to disrupt salience representations (e.g. Zenon, Filial, Duhamel & Oliver, 2010, Lane et al., 2010) and makes it easier to disengage attention from invalid peripheral cues (Smith Jackson & Rorden 2005).
Related to the previous point, there should be some discussion of how the current work relates to that of Kanai et al., 2012. They examined a similar question in young adults and found a rather different pattern of effects. Notably, anodal stimulation facilitated contralateral prosaccade latency and had no effect on antisaccades. The authors should offer a more thorough explanation of why the results are so different in the two studies
A final minor point. The authors describe their work as ‘seminal’. I find this a little bit presumptuous, especially given the caveats they note in the discussion. The work is interesting and potentially important, but I feel it is premature to describe it as seminal. I suggest revising the title to remove the words “Seminal Evidence”
Author Response
Point 1: Given the small sample I think it might be helpful to report the results of a post-hoc power calculation. By my estimate the online study had a power of .74, which isn’t that far off the standard power of .8 used for apriori calculations. I think this might reassure the reader that the effect reported in the online condition is real.
Response 1: Yes, the online study had a power of .74. As suggested, we have added this post hoc power calculation result to the Discussion (third paragraph). Thank you for this suggestion.
Point 2: The introduction claims that saccade control is reduced with age and implies that this has important clinical consequences. However, I think this line of argument would be stronger if the authors could cite some evidence that small but statistically significant changes in saccade latencies are meaningful from a clinical perspective. For example, is there any evidence that seniors have problems scanning busy intersections and an increased risk of a car accident while driving?
Response 2: Good idea! We have added relevant references to the Introduction (first paragraph).
Point 3: I would also like to see a bit more discussion about the mechanism by which tDCS selectively facilitated antisaccade production. As far as I can see, there is no explicit discussion of the cognitive or neural mechanism that might account for this facilitation. There is some generic discussion of non-oculomotor functions, but this is all rather vague and a bit unsatisfying. One possibility is that tDCS disrupts the computation of the salience of the onset associated with making an antisaccade, so the competition between the saccade vector triggered by the onset of the antisaccade and the desired saccade vector is resolved more quickly. This explanation would be consistent with TMS studies showing that stimulating over FEF seems to disrupt salience representations (e.g. Zenon, Filial, Duhamel & Oliver, 2010, Lane et al., 2010) and makes it easier to disengage attention from invalid peripheral cues (Smith Jackson & Rorden 2005).
Response 3: Thank you for these suggestions. We have edited the second paragraph of the Discussion to incorporate these ideas and add relevant references. We took care not to be too specific about the relevant brain regions given the diffuse nature of the tDCS.
Point 4: Related to the previous point, there should be some discussion of how the current work relates to that of Kanai et al., 2012. They examined a similar question in young adults and found a rather different pattern of effects. Notably, anodal stimulation facilitated contralateral prosaccade latency and had no effect on antisaccades. The authors should offer a more thorough explanation of why the results are so different in the two studies
Response 4: Yes, it is true that Kanai et al. (2012) reported prosaccade benefits (associated with an online prosaccade protocol). We have now referred to this in the Discussion; thank you for suggesting this improvement. Kanai et al. (2012) did find improvements in antisaccade performance (reduced reflexive errors; see Figure 4 panels A and C), which was an important motivation for the current study. In addition, there was a numerical trend for improved antisaccade latencies (see Figure 3 panels A and C). In the revised manuscript, we have made this clearer. Differences in the results could potentially relate to differences in the population under study (young versus older adults) or the size of the electrode (small versus large). Given the large electrodes used in the current study, relative to Kanai, and the likely electric field distribution, it seems unsurprising to us that effects were not contralaterized (with respect to the anodal electrode) in the current study, which we now specify. Thank you for assisting us to improve the clarity of our manuscript.
Point 5: The authors describe their work as ‘seminal’. I find this a little bit presumptuous, especially given the caveats they note in the discussion. The work is interesting and potentially important, but I feel it is premature to describe it as seminal. I suggest revising the title to remove the words “Seminal Evidence”
Response 5: We have removed “Seminal” from the title. We retained the word “Evidence” as we feel that the title is more accurate with this word. We also edited out “seminal” elsewhere in the manuscript.
