Does 400 mg of Caffeine Improve Vigilance Across the Menstrual Cycle? What a 12-Athlete Pilot Actually Shows

Short answer: In a small randomized, double-blind crossover experiment, twelve trained women completed placebo and 400 mg oral caffeine testing during three menstrual-cycle phases. Caffeine was associated with better performance on a one-minute visual cancellation task and shorter responses on a simple computerized reaction-time task. The overall caffeine effects were statistically significant: vigilance F(1,11) = 14.11, p = .003 and reaction time F(1,11) = 5.24, p = .04. The reported condition-by-phase interaction for reaction time was not significant (p = .39). Meanwhile, time to exhaustion did not significantly improve with caffeine (p = .52), and the overall caffeine effect on jump height did not meet the usual .05 significance level (p = .08). These are task-specific group findings, not an individual impairment test, and no participant’s blood caffeine concentration was quantified.

The evidence is Ben Hsen and colleagues’ 2026 primary experiment in Nutrients, Effects of Caffeine Supplementation on Cognitive Function and Neuromuscular Performance Across Menstrual Cycle Phases in Eumenorrheic Female Athletes: A Randomized, Double-Blind Placebo-Controlled Pilot Trial. Read the original empirical paper (DOI 10.3390/nu18101512). A central interpretive distinction is that an average benefit of caffeine across phases is not proof that caffeine works better in one menstrual phase than another. Nor does a laboratory task establish that a particular person was alert, safe to drive, or affected by a measured drug concentration.

What precise scientific question did the investigators ask?

The investigators asked whether a single dose of caffeine changed cognitive and neuromuscular performance in trained athletes, and whether any such changes depended on the early-follicular, late-follicular, or mid-luteal stage of the menstrual cycle. Those are related but different questions. A treatment can have a statistically supported overall effect even if its interaction with menstrual phase is not demonstrated. Conversely, a favorable comparison in one phase cannot automatically be treated as evidence that the phase modifies the treatment effect.

This distinction is particularly important in forensic toxicology, where a broad statement such as “caffeine improved attention” might be taken out of context. The experiment used specific performance tests under scheduled conditions in a narrowly selected athlete sample. It did not evaluate crash avoidance, complex driving, standardized impairment classifications, witness accuracy, or chemical caffeine concentrations in forensic blood specimens. Its results should be interpreted at the level actually measured.

How many independent people were studied?

Twelve women, not seventy-two independent participants. The paper reports that twenty-five female athletes initially volunteered and that twelve met the selection criteria for analysis. Their mean reported age was 24.4 years, with a standard deviation of 2.7 years. They trained roughly five times each week and competed in a weekly match. The selection excluded contraception, irregular cycles, illnesses or injuries that could interfere with testing, and disclosed exposures or conditions potentially obscuring caffeine effects.

Each of the twelve participants could contribute six experimental visits: caffeine and placebo at each of three menstrual-cycle phases. The six visits improve within-person comparison but do not create seventy-two biologically independent individuals. Likewise, ten computer reaction-time trials, repeated sprint bouts, and many digits on a vigilance sheet are measurements nested within a person. Inference about a larger population cannot count these repeats as if each came from a newly recruited athlete.

The investigators did not perform a formal prospective sample-size calculation. With twelve participants, a failure to identify an interaction cannot be interpreted as proof that the true phase-specific effects are identical. A small pilot may detect a relatively pronounced within-person result while having inadequate precision for subtle interactions, heterogeneous drug response, or uncommon adverse outcomes.

What drug dose was actually administered?

During each active session, participants swallowed a capsule containing 400 mg of caffeine, with a matching placebo capsule in the comparison condition. Testing began approximately one hour after ingestion. This was a fixed absolute dose rather than a tailored milligram-per-kilogram dose. Because body mass varied, the paper reported average phase-associated equivalents of about 6.9 to 7.1 mg/kg. The nominal administered mass is known from the protocol; an internal or blood caffeine concentration is not.

The study instructed athletes to avoid caffeine-containing products for 24 hours before testing, and compliance was confirmed verbally. That procedure standardized recent self-reported intake but was not chemical verification of caffeine abstinence. There was no serial caffeine assay, individual plasma concentration-time curve, calculated caffeine elimination half-life, or directly established point at which a participant’s blood contained a particular caffeine level.

The design was randomized and double-blinded with respect to caffeine versus placebo. The investigators describe multiple menstrual cycles and roughly seven-to-ten-day intervals between visits to accommodate separate phase-specific conditions and washout. These controls reduce some biases without making the tests equivalent to a naturalistic workplace, driving, or emergency-response study.

How were the menstrual phases established?

The authors did more than label phases by calendar date alone. Participants kept menstrual diaries, ovulation was assessed using urine test strips, and serum estrogen and progesterone were measured to corroborate menstrual-cycle staging. The early-follicular, late-follicular, and mid-luteal labels therefore describe reproductive-hormone contexts supported by actual measurements. They do not describe directly measured caffeine pharmacokinetics.

The serum-hormone values are important because they make the phase definitions more tangible. They must not be conflated with drug concentrations: estrogen expressed in pg/mL and progesterone in ng/mL are reproductive-hormone observations, not measured caffeine or caffeine-metabolite levels. The researchers did not establish a serum hormone-by-caffeine concentration-response relationship or demonstrate which hormone caused any particular behavioral result.

What were the reported phase markers and effective nominal doses?

Values reorganized from the original paper’s descriptive Table 1 and presented as reported mean ± SD. The same twelve participants contributed repeated phase assessments. This table does not report caffeine in any biological specimen.

What did the vigilance test actually measure?

The vigilance measure was a one-minute paper-and-pencil visual cancellation task, not a long-duration psychomotor vigilance battery. Athletes scanned a printed field of 600 digits organized across 36 lines and crossed out occurrences of two designated digits, 9 and 6, as quickly and accurately as possible. The score was the number of correctly identified target digits. A higher score can reflect more successful visual search and response execution under this particular short, timed protocol.

The main caffeine-versus-placebo condition effect was significant, F(1,11) = 14.11, p = .003, partial η² = .56. The menstrual-phase comparison was not significant (reported p = .26). Figure 2 depicts caffeine-condition scores exceeding corresponding placebo bars across the three phases. This is the strongest cognitive finding in the paper on its stated statistical test, but it is not a chemical test and not a measure of sustained vigilance for an entire shift or overnight drive.

A one-minute cancellation task can be sensitive to scanning speed, strategy, familiarity, handwriting or motor execution, as well as attention. The experiment does not isolate all those components. It also does not determine whether caffeine reduces extremely rare catastrophic misses in traffic or industrial monitoring. A group benefit on a short test does not establish that any identified person would recognize a sudden hazard in an actual event.

How much did caffeine change reaction time?

The computerized reaction-time task asked participants to press a keyboard key when a blue square appeared. The program averaged ten valid responses after ten familiarization trials. Responses below 150 milliseconds and above 800 milliseconds were excluded by protocol. This creates a defined simple visual response measure rather than a comprehensive judgment, divided-attention, or driving performance assessment.

The reported overall treatment comparison favored caffeine: F(1,11) = 5.24, p = .04, partial η² = .32. The condition-by-menstrual-phase interaction, however, was F(2,22) = 0.98, p = .39. Figure 3 shows shorter mean reaction times with caffeine in the plotted conditions, with source figure markers for some within-phase comparisons. The figure is not an independent license to declare a definitive hormonal modification of caffeine’s treatment effect.

The source graph depicts differences on a scale measured in milliseconds, but the primary narrative does not provide a complete participant-level distribution or an exact adjusted treatment effect with a confidence interval. Accordingly, it would be misleading to digitize a rough bar height and present it as a validated individual response prediction. Moreover, excluding very fast or very slow trials may narrow the observed tail of performance; a study evaluating real hazard response would need to investigate consequential lapses and decision errors separately.

Why is an overall caffeine effect different from a phase interaction?

An overall drug effect asks whether the average response differs between caffeine and placebo when results are considered over the tested phases. An interaction asks whether the magnitude or direction of that difference itself changes with menstrual phase. The first can be present while the second is not detectable. This study supplies that exact pattern for reaction time: statistically supported caffeine main effect but a nonsignificant drug-by-phase interaction.

It is not statistically valid to argue “the comparison was significant in phase A but not in phase B, therefore caffeine works differently in A and B” without directly testing the between-phase difference in the caffeine effect. Nor does p = .39 prove equivalence. Demonstrating practical equivalence would require prespecified equivalence margins and adequate precision; neither is established by a nonsignificant interaction in a small exploratory study.

The paper’s abstract describes the reaction-time response as lacking a significant condition-by-phase interaction, while some discussion and concluding language characterizes certain responses as phase-dependent. For forensic interpretation the model comparison controls: a phase-specific causal effect is not established by the reported reaction-time interaction.

What happened to jump performance?

The counter-movement jump assessment used an optical measurement system while athletes jumped from a standardized position with hands on hips. The authors reported a significant overall menstrual-phase factor, F(2,22) = 4.19, p = .03, with higher placebo performance during the late-follicular phase than during the other two phases in their reported follow-up comparisons.

By contrast, the overall caffeine-versus-placebo main comparison was F(1,11) = 3.85, p = .08, and the condition-by-phase interaction was F(2,22) = 2.22, p = .13. Both exceed a conventional .05 cutoff. The authors describe a favorable within-mid-luteal caffeine comparison in Figure 4, but that narrower observation should not be promoted into a confirmed general caffeine effect or a proven menstrual-phase interaction.

A study can contain an attractive graph, a promising subgroup pattern, and an overall treatment test that is not significant. Those statements are not necessarily contradictions if the comparisons answer different hypotheses, but they carry different evidentiary weight. Reporting only the visually favorable bar or the favorable within-phase marker would obscure the nonsignificant overall caffeine main effect and interaction.

Did caffeine extend cycling time to exhaustion?

No statistically significant benefit was detected. After a standardized warm-up, the time-to-exhaustion protocol required cycling at 90% of a previously determined individual peak-power target until cadence fell below a specified portion of the participant’s initial selected value. This task measures tolerance of an intense exercise condition, not an individual’s ability to drive, remain awake through a shift, or avoid a collision.

The original model reported no significant effect of caffeine (F(1,11) = 0.44, p = .52), no menstrual-phase effect (F(2,22) = 0.49, p = .62), and no condition-by-phase interaction (F(2,22) = 0.09, p = .91). Saying that the study “proved caffeine does nothing for endurance” would be too strong. It established only that the specified test in twelve participants did not detect those differences with the reported analyses.

The protocol was demanding, the sample was small, and individual response variation is incompletely described. These features affect statistical power and transferability. Importantly, neither a null endurance test nor a positive cancellation test is a suitable substitute for an actual toxicological result from a specimen collected in a case.

What did the repeated-sprint experiment add?

A separate cycle-ergometer protocol consisted of ten five-second maximal sprints, each separated by twenty-five seconds of passive rest. Mechanical peak power was estimated in watts by the ergometer system. The source reported a strong decline across the repeated sprint sequence: F(9,99) = 74.5, p < .01 for the sprint-time factor. It also reported a menstrual-phase factor, F(2,22) = 7.04, p < .01, and a nonsignificant reported interaction, F(2,22) = 0.02, p = .98.

The authors’ figure-based follow-up comparisons describe higher peak power during late-follicular caffeine trials in later sprints, and higher tenth-sprint power with caffeine across phases. These observations can coexist with a strong within-session fatigue pattern and a nonsignificant reported general interaction. They should not be turned into a single claim that menstrual phase determines the response to caffeine or that caffeine prevents fatigue under all conditions.

The study did not report direct assays of muscle caffeine, catecholamine release, cerebral caffeine, or blood caffeine around the sprints. Watt values are mechanical task outputs. A sprint power difference cannot identify the amount of caffeine circulating in an individual athlete, the dose absorbed, or the physiological mechanism responsible for the observed change.

Which effects were significant, nonsignificant, or left unmeasured?

This table reorganizes selected results reported in the original article. “Not significant” applies only where a statistical test was performed. “Not measured” is a separate scientific category and must not be rewritten as p > .05.

What was actually measured in blood—and what was not?

The investigators report serum estrogen and progesterone as evidence supporting phase classification. A serum reproductive-hormone measurement is not a quantitative toxicology measurement of caffeine. The protocol also used ovulation-test urine strips. Those tests concern endocrine status, not chemical identification of caffeine or its metabolites in urine.

There was no reported laboratory assay of parent caffeine or paraxanthine in forensic venous whole blood, separated plasma, serum, saliva, or urine. The 400 mg capsule is the administered mass, not the concentration that later circulated in any compartment. An individual’s absorbed dose, distribution, metabolic rate, and elimination kinetics are not reconstructed by dividing the capsule mass by body weight.

A legal venous whole-blood sample contains red cells and plasma together. Serum and plasma are separated fluid fractions; an oral-fluid specimen is yet another matrix. Even when the same parent drug is measured in several matrices, concentrations are not interchangeable without empirical validation. Here the situation is simpler: the study did not report quantitative caffeine concentrations in any of those drug-testing matrices. The observed task effect should remain identified as a task effect.

Could this trial establish menstrual-phase-specific impairment?

No. The participants were healthy trained athletes completing experimental performance tests, not drivers selected for suspected impairment, patients with a specific clinical syndrome, or persons involved in a known accident. There was no prespecified impairment classification, no forensic decision threshold, no adverse-event adjudication, and no identification of an individual whose legal capacity changed because of caffeine.

Menstrual phase was a within-person research factor. A statistically nonsignificant interaction does not authorize an assertion that phase is irrelevant to every individual, but it also does not support attributing a particular behavior to cycle phase. The study is too small and narrow to develop or validate an individual probability of impairment from hormones, caffeine dose, a cancellation score, or a reaction-time measurement.

In litigation, a reported group response would need to be evaluated against the actual person, specimen collection chronology, validated analytical procedures, independent behavioral evidence, alternative explanations, and the precise proposition being offered. This study is not a case-specific reconstruction model.

How should a forensic reviewer separate dose, concentration, and performance?

These questions occupy three different evidentiary layers. Dose is how much caffeine the protocol supplied. Concentration would require analysis of a defined biological specimen at a recorded time using a validated measurement method. Performance is what the person did during a specified test. This study measured the first and the third, but not a quantitative caffeine concentration in forensic blood or other drug matrices.

That distinction also protects against reverse inference. Better average speed on a simple task does not demonstrate that the caffeine dose was completely absorbed, that a particular pharmacokinetic curve applied, or that more complicated real-world responses would improve. Conversely, a null exercise response does not show that no caffeine was present or pharmacologically active. Different tasks engage different processes, and the study did not measure a common chemical intermediary linking them.

For background on why a result cannot migrate between biological compartments, consult the site’s published guide to biological matrices. For the distinction between observed physiological performance and analyte-specific identification, see its published confirmatory drug-testing guide. For the additional questions raised by administered dose, absorption and time, see the published pharmacokinetics guide.

What makes the trial informative despite its small size?

The within-person randomized, blinded comparison is more informative than simply comparing unrelated women who happen to be in different menstrual phases. Phase confirmation used several sources of information, including measurable reproductive hormones. The investigators employed defined cognitive and exercise protocols and reported both favorable and nonsignificant outcomes. Those features give the experiment scientific value as a controlled exploratory comparison.

The work is also informative because the positive and negative results diverge. One-minute visual scanning and simple reaction time favored caffeine, but time to exhaustion did not; the caffeine main effect on jump height did not meet p < .05; and the model did not demonstrate reaction-time phase modification. Thus, an apparently broad claim that caffeine improves “performance” fragments into several narrower propositions. Good forensic reasoning preserves that fragmentation instead of translating a subset of positive findings into a global behavioral conclusion.

What methodological limitations constrain the conclusions?

First, there were only twelve analyzed athletes after twenty-five initially volunteered. The pilot was not powered in advance for interactions or complex subgroup comparisons. Second, participants were trained young adults with regular cycles who met restrictive inclusion criteria. Results may not generalize to people with disordered sleep, chronic disease, irregular cycles, different habitual caffeine consumption, or different baseline fitness.

Third, one fixed 400 mg dose was used, resulting in different relative mg/kg exposures. The study did not sample caffeine concentrations or directly test whether hormonal changes altered caffeine metabolism. Fourth, each test captures a specific performance construct. A short cancellation exercise and ten-trial button-press task do not comprehensively index vigilance during night driving, divided attention, or emergency response. Fifth, multiple outcomes and multiple phase-specific plotted comparisons make it especially important to distinguish a primary model result from an isolated favorable follow-up contrast.

Sixth, the paper reports some phase-specific descriptions despite nonsignificant interaction tests, and results figures provide group means rather than full individual trajectories and pre-specified clinical margins. Finally, compliance with caffeine abstention was verified verbally; nutritional, sleep-related, and psychological influences could not be eliminated completely. Together, these issues call for replication rather than either dismissal of the study or overconfidence in its application.

What should an attorney, expert, or clinician avoid claiming?

A reviewer should not claim that menstruation or a particular phase determines whether caffeine produces a measurable cognitive benefit. The published reaction-time interaction did not substantiate that claim. Nor should the reported p = .08 jump comparison be labeled conventionally significant merely because the source called it borderline or because one subplot included a favorable symbol. Distinguishing the reported analyses is necessary to avoid selection of only positive results.

The trial also cannot establish a universal safe or unsafe caffeine cutoff for driving, the plasma or whole-blood caffeine level following 400 mg, an individual clearance rate, a diagnostic test of impairment, or a causal explanation for any incident. It says nothing directly about THC, amphetamine, ethanol, or another drug simply because some drug studies also use reaction-time outcomes. A defined source result is not a transferable pharmacological or legal threshold.

What is the most defensible bottom line?

In this twelve-athlete placebo-controlled crossover pilot, a single 400 mg oral caffeine capsule improved performance on a brief digit-cancellation vigilance test and a simple reaction-time task on the reported overall statistical comparisons. The trial did not identify a significant reaction-time caffeine-by-menstrual-phase interaction and did not detect a caffeine benefit in the time-to-exhaustion model. Jump and repeated-sprint findings require particularly careful separation of model tests and phase-specific plotted follow-ups.

The forensic lesson is not that caffeine “prevents impairment.” It is that controlled research may support a specific mean task difference while leaving individual drug concentrations, critical real-world performance, and causation unanswered. If the question is the behavior of a particular person at a particular time, the study supplies context—not the missing individual evidence.

Seven scientific questions about this caffeine experiment

1. Were twelve or seventy-two people studied?

Twelve independent athletes were included in the analysis. Six possible experimental sessions per person yield repeated within-person observations, not seventy-two independent human participants. The same caution applies to ten button presses and ten sprints per session: many measurements do not make a large biological sample.

2. Did caffeine improve sustained attention at every menstrual phase?

The one-minute cancellation task favored caffeine overall, F(1,11) = 14.11 and p = .003, and the source figure depicts an advantage at each phase. But a brief cancellation score is not a direct measure of eight-hour or all-night vigilance, and the phase interaction for every possible attentional measure was not independently validated.

3. Was reaction time demonstrably more responsive to caffeine during the late-follicular phase?

Not on the reported drug-by-phase interaction: F(2,22) = 0.98, p = .39. The overall caffeine main effect on reaction time was significant, but separate favorable figure markers do not establish a significant difference between the treatment effects at different phases.

4. Did 400 mg caffeine significantly improve countermovement jump height?

The reported caffeine main-effect test was p = .08, not significant by the standard .05 benchmark. Menstrual phase had an overall p = .03 association with jump results, and one within-phase caffeine comparison was described as favorable. Those findings should not be merged into a proven broad caffeine-by-phase interaction.

5. Was cycling endurance improved even though the sprint chart looked better?

The separate time-to-exhaustion test had a caffeine main effect p = .52 and interaction p = .91, neither statistically significant. The repeated-sprint protocol addressed short maximal bursts and fatigue over ten sprints. A finding in one exercise format cannot substitute for a null result in another.

6. Were caffeine concentrations or metabolites measured in venous blood or urine?

No quantitative caffeine or paraxanthine assay is reported. Researchers measured serum reproductive hormones and used urine ovulation tests to confirm menstrual phase. The administered 400 mg dose and mg/kg calculations are not forensic concentrations and do not establish a concentration-to-task relationship.

7. Can the findings determine whether one driver or witness was impaired?

No. There was no direct driving study, individual impairment classification, legal concentration threshold, or case-event investigation. The findings describe group averages in particular laboratory tasks under specified doses and endocrine conditions; they cannot retrospectively resolve an individual person’s attention or causal contribution to an event.

Original scientific source

Ben Hsen I, Hamdi S, Ceylan Hİ, Erriahi S, de Giorgio A, Dergaa I, Bragazzi NL, Bouzid MA. Effects of Caffeine Supplementation on Cognitive Function and Neuromuscular Performance Across Menstrual Cycle Phases in Eumenorrheic Female Athletes: A Randomized, Double-Blind Placebo-Controlled Pilot Trial. Nutrients. 2026;18(10):1512. Original primary article DOI: 10.3390/nu18101512.