Thirty-eight athletes with ACL injuries participated in the study. All study participants completed questionnaires defining their ACL injury, last menstrual period, prior knee injury, school, and type of birth control used (if any). Each athlete provided a 30-mL saliva sample within 72 hours of injury. Saliva samples were stored in a sealed container and frozen at −20°C. The samples were then shipped on dry ice to the Oregon Regional Primate Research Center (Beaverton, OR) for analysis. Physical examination, magnetic resonance imaging, or surgery confirmed the injury in all subjects.

One of the 38 athletes reported having a hysterectomy; her data were excluded due to inadequate information about hormone-replacement therapy. Consequently, study analyses included data from 37 patients: 21 provided both saliva samples and menstrual histories, 10 provided only saliva samples, and 6 provided only menstrual histories.

Because the previous literature indicated mixed results regarding the correlation between saliva and serum sex-hormone levels,23–28 we first performed an analysis to establish whether such a correlation existed in our sample. Thirteen control samples from uninjured females were obtained to test the correlation between saliva and serum sex-hormone levels. Salivary progesterone was assayed by routine radioimmunoassay, and E₂ was assayed with a modification of the 3rd Generation Double Antibody Estradiol assay (Diagnostic System Laboratories, Webster, TX) at the Oregon Regional Primate Research Center.

To test our null hypothesis that the ACL injuries were not correlated with the athletes' menstrual cycles, we performed a Monte Carlo simulation to generate multiple pseudocontrol groups. If the ACL injuries were not correlated with the menstrual cycle, the injuries would occur randomly throughout the cycle. That is, the probability of injury would follow a uniform distribution over the menstrual cycle. Using Stata computer software (version 6, Stata Corp, College Station, TX), we simulated 50 hypothetical control groups, each with 100 subjects. The timing of an ACL injury assigned to each hypothetical subject was determined by a uniform distribution to ensure that the probability of injury in each 2-day interval was the same for all intervals. We then compared the injured group with each computer-generated group. We tested whether the probability of injury was the same for each 2-day interval between the injured athletes' group and a group of 100 hypothetical subjects.

The correlations between saliva and serum estrogen and progesterone were 0.73 (α = .01) and 0.72 (α = .01), respectively. The correlation between the self-reported last menstrual period at the time of injury and the actual salivary and progesterone levels was 95% for the 21 athletes who provided this information (Table). Because salivary sex-hormone levels confirmed self-reported menstrual history in all but one case, we were able to use the measured sex-hormone levels to place athletes who did not provide menstrual histories in the appropriate phase of the menstrual cycle at the time of injury. Among all 37 athletes for whom data were analyzed, 25 injured their ACLs during the follicular phase, 1 during the ovulatory phase, and 11 during the luteal phase of the menstrual cycle. Six athletes injured their ACLs while on oral contraceptives, of whom 5 sustained their injury during the follicular phase and 1 during the luteal phase. Only 1 of the athletes on oral contraceptives reported her cycle day at the time of injury (day 2 of menses). Interestingly, the E₂ and P concentrations were very low in this athlete (E₂ = 0.73 pg/mL, P = 58 pg/mL) and 3 others on oral contraceptives (E₂ range, 0.16–0.20 pg/mL; P range, 15–29 pg/mL), suggesting that they were all injured around the time of menses.

Among the 27 athletes who self-reported their menstrual histories, 10 sustained injuries during the few days before and the first 2 days after the onset of menses (Figure 1). The frequency of observed injury for days 1–2 of the menstrual cycle was significant at α = .05 for all 50 comparisons between the injured athletes and our computer-simulated subject groups. This implies that the high frequency of the ACL injury observed in this interval was not due to random chance. For days 25–26, 27–28, and 7–8, the athletes' incidence of ACL injury was lower than the probability determined by a uniform distribution at α = .10 in 33, 34, and 36 comparisons, respectively, out of 50 total simulations. For the other intervals, fewer than 10 comparisons out of 50 were significant at α = .10, indicating that during these intervals, ACL injuries were likely to be uncorrelated with the menstrual phases.

Therefore, a significantly greater number of ACL injuries occurred on days 1 and 2 of the menstrual cycle. Salivary sex-hormone levels correlated with self-reported cycle day.

This is the first study to confirm self-reported menstrual histories with salivary sex-hormone profiles at the time of ACL injury. We found that 26 of 37 athletes tore their ACLs during the follicular phase of the menstrual cycle. Among athletes who self-reported their menstrual histories, 10 of these 27 injuries occurred during the few days before and the 2 days after the onset of menses (Figure 1). The levels of E₂ and P are both low at this time (Figure 2). This hormonal condition contrasts with the follicular phase during which P is low and E₂ peaks sharply before ovulation, and it follows the midluteal phase during which E₂ and P are both elevated for several days. These results are consistent with a report of no correlation between ACL injury and the general category of “luteal” phase,3 as well as 2 other reports indicating that injury is more likely during the late luteal and early follicular phases of the cycle.21,22

We did not obtain information on the typical lengths of the athletes' menstrual cycles. Although the “normal” menstrual cycle lasts 28 days, 3 of our 37 athletes (8% of our sample) sustained injuries after day 28. The probability that ACL injuries occurred during the prolonged menstrual interval (>28 days) was no different from the probability determined by a uniform distribution. That is, we could not reject the null hypothesis that the injuries occurring during this time were due to random chance.

We chose to use computer-simulated subjects because there was no well-defined control group. To qualify to be in a legitimate experimental control group, subjects would have needed the same menstrual cycles as the injured females but different distributions of injury among the cycles' phases. Therefore, instead of performing an experimental-control group comparison, we tested a simpler null hypothesis that the ACL injuries occurred randomly in each day of the menstrual cycle. Each computer-generated subject had an equal chance (0.0357) of injury in each day of her menstrual cycle. That is, these subjects had the same menstrual cycles as the injured subjects, but the probability of injury was different. We rejected our null hypothesis and found that ACL injuries occurred most frequently during the early menstrual cycle.

Past studies of the correlation between salivary and serum sex-hormone measurement have yielded conflicting results.24–28 However, more recent supersensitive, double-antibody techniques for measuring sex hormones in saliva have shown good correlation between saliva and serum levels.23 One group has even recommended using saliva to obtain hormone profiles in patients with difficult venous access.24 Using saliva to obtain sex-hormone profiles of athletes fits well in the athletic arena because little planning or equipment is needed to obtain and store the saliva. A simple ziplock-type bag works well to hold the saliva and can be placed immediately on ice and transferred to a freezer soon thereafter or upon return from a road trip.

Our findings depend heavily on the accuracy with which we determined the day of the menstrual cycle at the time of injury. Such determinations are complicated by problems in obtaining blood from injured athletes in an athletic setting and the fact that a single measurement of one hormone cannot unequivocally define the day of the menstrual cycle. However, we successfully overcame these problems by measuring both E₂ and P in saliva. This approach is effective because elevated (at or near the typical highest concentration) E₂ is characteristic of the follicular phase (with an E₂ spike occurring at ovulation), elevated P is characteristic of the luteal phase, and low E₂ and P are characteristic of menses. Menstrual phases defined in this way correlated well (>95%) with the self-reported menstrual histories (see Table), thereby confirming the accuracy of our athletes' recollections of their menstrual cycles. We, therefore, have confidence that all but 1 or 2 of our athletes accurately recalled the dates of their last menstrual periods before ACL injury.

Although we do not know why ACL injuries occur around the time of menses, our current research is focused on characterizing sex differences in ACL tissue remodeling. Cyclic changes in E₂ and P may alter expression of genes encoding tissue-remodeling enzymes and proteins, which, in turn, could favor either net tissue degradation or repair at specific times during the menstrual cycle. If a molecular basis for sex differences in ACL injury is found, treatments may be instituted to decrease the injury rates in females.

In conclusion, ACL injuries occurred most frequently on days 1 and 2 of menses, suggesting that ACL injury is not random but occurs more often around the time of menses, when circulating sex-hormone levels are low and after a time when both E₂ and P were elevated. Additionally, salivary sex-hormone profiling correlates well with serum profiling, and in this athletic population, adequately identified menstrual cycle phase at the time of injury. This is, therefore, an effective technique to identify correlations between injury and hormone patterns in athletes.