Among the ten participants, four were low-dose pill users, and nine were university students at the beginning of the study. The mean age was 22.1 years, and the average length of the menstrual cycle during the experiment was 30.86 days. Participants’ mean sleep duration and sleep midpoint, and mean PSQI global score (PSQIG) and SJL, obtained from the questionnaire and MCTQ administered at the beginning of the experiment, respectively, are presented in Table 1. The mean QP value for a year, calculated using a chi-square periodogram, was 280,476.06.

When the correlations between the sleep duration, sleep midpoint, length of menstrual cycle, PSQIG, SJL, and QP values were evaluated (Table 2), no correlations were found between most variables. However, a strong negative correlation (r =  − 0.867, P = 1.15 × 10−3, Pearson’s correlation coefficient) was observed between the sleep midpoint and QP value and a positive correlation was observed between the menstrual cycle days and SJL (r = 0.638, P = 4.71 × 10−2, Pearson’s correlation coefficient). Furthermore, a positive correlation was found between the sleep midpoint and the PSQIG (r = 0.619), although the difference was not statistically significant.

Figure 2a shows representative graphs depicting BBT fluctuations over a year. Large individual differences in BBT were found; however, a biphasic pattern of low- and high-temperature phases was observed in all participants. Classifying the BBT into five menstrual stages (Fig. 1), and comparing the deviation from the average BBT of each participant yielded the following observations: − 0.09 ± 0.19 °C in the menstrual phase, − 0.09 ± 0.22 °C in follicular phase I, − 0.08 ± 0.20 °C in follicular phase II, + 0.05 ± 0.21 °C in luteal phase I, and + 0.14 ± 0.21 °C in luteal phase II, with significant stage-specific variation (P < 0.001, one-way ANOVA, F (4,3405) = 175.943). In a stage-by-stage comparison, the BBT in luteal phases I and II were significantly higher than those in the menstrual phase and follicular phases I and II (P < 0.001, Tukey’s test). BBT in luteal phase II was significantly higher than those in luteal phase I (P < 0.001, Tukey’s test; Fig. 2b). Additionally, the maximum and minimum values were extracted from the average BBT at each menstrual stage to ascertain BBT amplitude during each individual's menstrual cycle (Table 1). The BBT amplitude positively correlated with the length of the menstrual cycle and SJL (r = 0.730, P < 0.05 for BBT amplitude and length of menstrual cycle; r = 0.713, P < 0.05 for BBT amplitude and SJL; Pearson’s correlation coefficient, Table 2).

Menstrual variations in basal body temperature. a Representative graphs showing basal body temperature fluctuations over a year. The purple lines represent the basal body temperature, and the yellow lines represent the day of ovulation. b Violin plots showing the deviation of the average basal body temperature of each participant at each phase of the menstrual cycle. The white circles represent the mean values. M menstrual phase, F1 follicular phase I, F2 follicular phase II, L1 luteal phase I, L2 luteal phase II. Differing letters (a, b, and c) between groups indicate significant differences (P < 0.001, Tukey’s test)

Figure 3a depicts actograms representing the sleep–wake rhythm over a year. A large overall variation in the waking and sleeping times of the participants was observed. For sleep duration and midpoint, the deviation from the mean was determined for each participant. Sleep duration, sleep midpoint, and QP values were classified into five menstrual stages and compared.

Menstrual variations in sleep–wake rhythm. a Representative graphs showing basal body temperature fluctuations over a year. The left panel represents an example of high sleep–wake rhythm variability and low QP value, and the right panel represents an example of low sleep–wake rhythm variability and high QP value. The vertical axis shows the date, and the horizontal axis shows the time. The black bars represent the sleep state, and the white bars represent the waking state. The density of the black bars indicates the depth of sleep. The yellow line represents the day of ovulation. b, c Violin plots showing the deviation of the average (b) sleep duration and (c) sleep midpoint of each participant at each phase of the menstrual cycle. d Violin plots showing QP values at each phase of the menstrual cycle. The white circles represent the mean values. M menstrual phase, F1 follicular phase I, F2 follicular phase II, L1 luteal phase I, L2 luteal phase II, QP quasi-peak. Differing letters (a and b) between groups indicate significant differences (P < 0.05, Tukey’s test)

Sleep duration was + 4 ± 116 min in the menstrual phase, + 4 ± 117 min in follicular phase I, − 8 ± 103 min in follicular phase II, + 4 ± 117 min in luteal phase I, and − 4 ± 104 min in luteal phase II, with no significant differences due to variations in the menstrual phase (P > 0.05, one-way ANOVA, F (4,3509) = 1.965; Fig. 3b).

The sleep midpoint was + 0.03 ± 1.24 h in the menstrual phase, + 0.12 ± 1.34 h in follicular phase I, − 0.06 + 1.28 h in follicular phase II, + 0.06 ± 1.35 h in luteal phase I, and − 0.11 ± 1.53 h in luteal phase II. Thus, the sleep midpoint showed significant variation by stage (P < 0.001, one-way ANOVA, F (4,3509) = 3.409). Based on the menstrual phase, the sleep midpoint was advanced during follicular phase I, delayed during follicular phase II, advanced during luteal phase I, and delayed during luteal phase II. Statistically, the sleep midpoint in follicular phase I was significantly earlier (P < 0.01, Tukey’s test) than that in luteal phase II (Fig. 3c).

The QP values were 914.43 ± 144.00 in the menstrual phase, 975.51 ± 173.34 in follicular phase I, 976.04 ± 181.04 in follicular phase II, 915.93 ± 141.40 in luteal phase I, and 916.96 ± 163.55 in luteal phase II, indicating significant stage-specific variation (P < 0.001, one-way ANOVA, F (4,590) = 4.939). The QP values in follicular phases I and II were significantly higher than those in the menstrual phase and luteal phases I and II (P < 0.05, Tukey’s test; Fig. 3d).