Editorial Spotlight: Sandra Carvalho

Sandra Carvalho is an Assistant Professor with habilitation in Psychology and Neuroscience in the Department of Basic Psychology at the School of Psychology, University of Minho, Portugal. A clinical psychologist by training, she completed her PhD in Clinical Psychology at the University of Minho and subsequently undertook postdoctoral training in neurorehabilitation at Harvard Medical School and Spaulding Rehabilitation Hospital in Boston.

Over the past 20 years, her research has examined the relationship between brain function, cognition, emotion, and human behavior. Combining behavioral and cognitive assessment with EEG, event-related potentials, and non-invasive brain stimulation—including TMS, tDCS, and tACS—she investigates the mechanisms underlying neuroplasticity and clinical response.

Her work aims to translate experimental neuroscience into personalized interventions by identifying markers that can guide stimulation targets, parameters, timing, and integration with cognitive and behavioral therapies. She has applied this approach to obsessive-compulsive disorder, Tourette syndrome, chronic pain, major depression, attention-deficit/hyperactivity disorder, and stroke.

---

As a researcher, you have extensive experience with neuromodulation interventions to improve mental health and cognitive function, which is such an exciting and rapidly evolving field. What do you see as the biggest promise for this therapeutic approach, and what are the biggest challenges?

Over the past 19 years, I have worked with neuromodulation techniques in both research and clinical contexts. Throughout this journey, I have become increasingly convinced of two things: neuromodulation can produce meaningful effects, but the response is also profoundly individual. Even among people with the same diagnosis, we observe considerable variability in the magnitude, timing, and duration of the response. This is understandable because no two brains—and no two personal or clinical histories—are exactly alike. Factors such as baseline brain activity, previous experiences, disease trajectory, concurrent treatments, and the individual’s cognitive and emotional state may all influence how the brain responds to stimulation.

State dependency is particularly important. The effect of stimulation depends not only on where and how we stimulate, but also on the functional state of the brain at that particular moment. This means that we cannot continue to assume that the same dose and stimulation parameters will be appropriate for every patient. We need to understand the appropriate intensity, duration, frequency, and repetition of the dose; how many sessions are required; whether and when maintenance or booster sessions are needed; and, crucially, when stimulation should be stopped.

This final question is especially important because more stimulation is not necessarily better. Plasticity is not always adaptive, and stimulation may interact with homeostatic and metaplastic mechanisms in complex ways. One of the field’s central challenges will therefore be to identify the optimal therapeutic window: the point at which stimulation promotes learning and functional reorganization without becoming unnecessary, ineffective, or potentially counterproductive. We need to determine not only the optimal duration of each stimulation session, but also the ideal interval for rest and consolidation between sessions, the number of treatment cycles required, and the timing of any subsequent reinforcement. The brain has a natural tendency to return to previously established patterns of functioning, so inducing an immediate change is not sufficient. We must learn how to consolidate that change and support long-term learning.

Another phenomenon that has particularly captured my attention is the delayed response to treatment. In my work with people with major depression and chronic pain, I have observed that improvement does not always occur during the treatment period itself. With tDCS, for example, some patients may show a clearer response only a few days or after the stimulation sessions have ended. This raises important scientific and clinical questions. If outcomes are assessed too early, we may incorrectly classify someone as a non-responder. We therefore need to understand the trajectories of response more fully and identify which patients are likely to respond immediately, gradually, or only after a delay.

One of neuromodulation’s greatest strengths is that it is not only a therapeutic intervention, but also a powerful tool for studying the brain. By influencing the activity of specific regions and distributed networks, techniques such as transcranial magnetic stimulation and transcranial electrical stimulation allow us to investigate causal relationships between brain activity, cognition, emotion, and behavior. Repeated stimulation may also support functional—and potentially structural—plasticity, making it especially promising for neurorehabilitation. The objective is not simply to modulate a circuit or reduce a symptom, but to help the brain reorganize and strengthen adaptive patterns that support meaningful and lasting functional recovery.

I therefore believe that the future lies in personalized, state-dependent, and adaptive neuromodulation. Open-loop systems deliver stimulation according to parameters established in advance, without continuously considering the brain’s ongoing state. Closed-loop systems can instead use real-time information from EEG, event-related potentials, fMRI, or other physiological and behavioral measures to determine when and how stimulation should be delivered. These systems could eventually help us deliver the right intervention, at the right dose, to the right neural target, and at the right moment for each person.

This possibility has been a longstanding focus of our research. In 2017, we reported a proof-of-concept surface EEG–tDCS closed-loop system in humans, in which an algorithm detected predefined changes in EEG activity and successfully triggered stimulation. More recently, we investigated individually tailored transcranial alternating current stimulation using each participant’s endogenous event-related P3 response. The stimulation frequency and task timing were adjusted to individual electrophysiological characteristics, illustrating how neural markers might be used to align stimulation with the cognitive process we aim to modulate.

The remaining challenges are substantial. We need reliable biomarkers, robust real-time signal processing, reproducible algorithms, and larger, adequately powered clinical trials. We also need longer follow-up periods that can capture immediate, gradual, and delayed responses. Most importantly, we must establish whether neurophysiological changes translate into durable improvements in everyday functioning and quality of life. For me, the central promise of neuromodulation is therefore not simply our increasing ability to stimulate the brain. It is our growing capacity to listen to the brain, understand its individual dynamics, and adapt our interventions accordingly to promote effective and lasting rehabilitation.

What aspects of Open Science do you feel are most important in your field, and where do you see the next steps for openness in this research area?

For me, the most important principle of Open Science is transparency across the entire research cycle. This is particularly critical in neuromodulation because results can be influenced by numerous methodological decisions: the stimulation target, intensity, duration, frequency, electrode or coil placement, number and spacing of sessions, the brain state during stimulation, concurrent interventions, participant characteristics, outcome selection, and analytical strategy. If these decisions are not described fully, it becomes difficult to interpret a study, reproduce its methods, or understand why apparently similar protocols produce different findings.

We are currently witnessing an unprecedented acceleration in scientific publication. Rapid dissemination can be extremely valuable, particularly when it facilitates timely access to new knowledge and encourages collaboration. However, speed and volume do not necessarily equate to scientific progress. The pressure to publish quickly can contribute to fragmented studies, small samples, insufficient methodological detail, selective outcome reporting, and conclusions that extend beyond the evidence. Findings may circulate widely before they have been adequately scrutinized or independently replicated. Open access to an article is important, but making a publication freely available does not, by itself, make the underlying research transparent, reproducible, or methodologically robust.

Open Science should help address these problems. Prospective trial registration, preregistration of hypotheses and primary outcomes, publication of detailed protocols and statistical analysis plans, and the use of Registered Reports can make the distinction between confirmatory and exploratory analyses clearer. They can also reduce undisclosed methodological flexibility and selective reporting. In clinical research, changes from a registered protocol are sometimes necessary, but they should be explained transparently. Equally important is the publication of negative, null, and unexpected findings. If we publish only positive results, we create a distorted evidence base, overestimate treatment effects, and risk unnecessarily exposing additional participants to interventions or research questions that have already been investigated.

Transparency is also essential when reporting the intervention itself. Neuromodulation studies should provide sufficient information about stimulation parameters, equipment, electrode or coil positioning, dose, treatment schedule, concurrent tasks, adverse events, and participant characteristics to allow other groups to understand and reproduce the protocol. Access to de-identified data, analysis code, computational models, and methodological materials can further support verification, secondary analysis, and collaboration. The FAIR principles—making research outputs findable, accessible, interoperable, and reusable—provide an important framework for this work.

Nevertheless, openness also requires careful judgment. Clinical, neuroimaging, and electrophysiological datasets may contain highly sensitive or potentially identifiable information. Responsible data sharing cannot mean placing all data online without adequate safeguards. It must be supported by appropriate informed consent, ethical oversight, careful de-identification, secure repositories, controlled-access procedures when necessary, and recognition of the communities and researchers who generated the data. We must also avoid creating a model of Open Science that benefits primarily well-resourced institutions while transferring the work and risks of data production to less-resourced researchers or clinical populations. Appropriate attribution, reciprocity, and equitable opportunities for collaboration are fundamental.

The next step is therefore to embed openness into the design of research rather than treating it as an administrative task at the point of publication. This means planning for transparency, reproducibility, and responsible data sharing before recruitment begins. It also means harmonizing methods and outcome measures across laboratories, supporting multicenter studies and independent replication, involving patients in defining meaningful outcomes, and rewarding researchers for sharing protocols, datasets, code, and rigorous negative findings—not only for producing novel positive results.

I am strongly supportive of Open Science, but I believe we must adopt a substantive rather than superficial definition of openness. Open Science should not simply help us publish more, or publish faster. It should allow us to understand precisely how knowledge was produced, assess its limitations, test whether findings are reproducible, and ensure that publicly available evidence is sufficiently reliable to inform future research and, ultimately, clinical care. Its real value lies in making science not only more accessible, but also more rigorous, accountable, inclusive, and trustworthy.

As an Academic Editor for PLOS ONE, you have facilitated fair and thorough peer review processes and thoughtfully handled research reporting the results of clinical trials, which provide crucial clinically relevant information that can impact health care decisions. How do you approach evaluating these manuscripts and overseeing peer review?

 Throughout my career, I have served extensively as both an editor and a reviewer, handling or reviewing several hundred manuscripts. This experience has reinforced my awareness of the responsibility involved: editorial decisions must be independent, transparent, consistent, and based on the quality of the research rather than on the authors’ institution, seniority, country, or the direction of the results.

When evaluating a clinical-trial manuscript, I begin with the integrity of the study. I examine whether the research question is clearly defined and clinically relevant, whether the design and sample size are appropriate, and whether ethical approval, informed consent, and prospective trial registration are documented. I compare the reported primary and secondary outcomes with the trial registration and protocol and assess the statistical methods, participant flow, missing data, protocol deviations, adverse events, and potential sources of bias. Reporting guidelines such as CONSORT are extremely valuable, but completing a checklist cannot replace careful scientific and ethical judgment.

I also consider whether the interpretation is proportionate to the evidence. Statistical significance is not the same as clinical relevance, and a non-significant result does not necessarily mean that a study is uninformative. Effect sizes, uncertainty, potential harms, study limitations, and generalizability must all be considered. This is especially important in clinical research because overstated conclusions can influence subsequent studies, clinical expectations, and potentially health care decisions.

For peer review, I try to select reviewers with complementary expertise—for example, in the clinical condition, the intervention, trial methodology, or statistical analysis—while carefully considering independence and potential conflicts of interest. I look for reviews that are rigorous, constructive, and focused on matters that genuinely affect the validity, transparency, or interpretation of the study.

My role as an editor is not simply to count positive and negative recommendations. Reviewers may reasonably emphasize different aspects of a manuscript or offer conflicting advice. I evaluate each report critically, distinguish essential revisions from optional suggestions, reconcile contradictory recommendations, and communicate a clear and fair path forward to the authors. Even when a manuscript cannot be accepted, I believe the decision should be respectful, transparent, and clearly supported by scientific reasoning.

Ultimately, I aim to ensure that well-conducted studies are evaluated fairly regardless of whether their findings are positive, negative, or unexpected. For me, the central questions are whether the research was conducted ethically, whether the methods and reporting are sufficiently rigorous and transparent, and whether the conclusions are genuinely supported by the evidence.

---

Disclaimer: Views expressed by contributors are solely those of individual contributors, and not necessarily those of PLOS.