Masculinizing e ects on otoacoustic emissions and auditory evoked potentials in women using oral contraceptives

Transcription

1 Hearing Research 142 (2000) 23^33 Masculinizing e ects on otoacoustic emissions and auditory evoked potentials in women using oral contraceptives Dennis McFadden * Department of Psychology and Institute for Neuroscience, Mezes Hall 330, University of Texas, Austin, TX 78712, USA Received 21 April 1999; received in revised form 10 December 1999; accepted 11 December 1999 Abstract The otoacoustic emissions (OAEs) and auditory evoked potentials (AEPs) measured in two separate large scale studies were examined retrospectively for potential differences between those women using, and those not using, oral contraception (OC). Fourteen dependent variables were examined, all of which exhibited substantial sex differences. For 13 of those 14 dependent variables, the means for the users of OC were shifted away from the means of the non-users in the direction of the males. Specifically, for four different measures of OAE strength, for seven of eight measures of AEP latency or amplitude, and for two cognitive tests (mental rotation and water level), the means for the users of OC were located intermediate to those of the non-users of OC and the males. Few of these differences between users and non-users of OC achieved statistical significance, but the near universality of the direction of the difference suggests that oral contraceptives do produce a weak masculinizing effect on some auditory structures. These weak masculinizing effects appear to run contrary to the facts that the levels of both free testosterone and estradiol are lower in women using OC than in normal-cycling women. Past findings on auditory sex differences may have underestimated those sex differences. ß 2000 Elsevier Science B.V. All rights reserved. Key words: Otoacoustic emission; Auditory evoked potential; Oral contraceptive; Masculinization; Estrogen 1. Introduction Otoacoustic emissions (OAEs) are sounds produced by the cochlea that propagate back into the external ear canal where they can be recorded using a small, sensitive microphone system (see Probst et al., 1991 for a review). OAEs are presumed to be a harmless by-product of the system of cochlear ampli ers (Davis, 1983) that line the basilar membrane and contribute to the remarkable sensitivity of hearing. Considerable circumstantial evidence exists to suggest that prenatal exposure to high levels of androgens diminishes the strength of the cochlear ampli ers, and thereby the strength of the OAEs a person exhibits. Among the evidence are these facts (for citations and the details of the argument, see McFadden and Pasanen, 1998,1999; McFadden, 1998): females have more and stronger OAEs than * Tel.: +1 (512) ; Fax: +1 (512) ; mcfadden@psy.utexas.edu males, and this is true for infants and children as well as for adults; females have better hearing sensitivity than males; females with male co-twins have weaker OAEs than females with female co-twins; and homosexual and bisexual females have weaker OAEs than heterosexual females. Less information exists about the possible auditory e ects of high androgen levels postnatally. One report (McFadden et al., 1998) suggests that decreasing androgen levels in an adult male may lead to more spontaneous OAEs (SOAEs) (that is, a more female-like cochlea) but other interpretations of this single case study are possible. Several investigators (Vermeulen and Thiery, 1982; Gaspard et al., 1983; Bancroft et al., 1991) have reported that the level of plasma free testosterone is higher in normal-cycling women than in women using oral contraception (OC). If high androgen levels do have a suppressing e ect on the cochlear ampli ers of adults, then one would expect weaker OAEs (and reduced hearing sensitivity) in normal-cycling women than in women using OC / 00 / $ ^ see front matter ß 2000 Elsevier Science B.V. All rights reserved. PII: S (00)

2 24 D. McFadden / Hearing Research 142 (2000) 23^33 Two of our recent experiments (both designed for a di erent purpose) involved large numbers of young women, making comparisons possible between those subjects using OC and those not. In one experiment (McFadden and Pasanen, 1998,1999), the measures were SOAEs and click-evoked OAEs (CEOAEs), both of which are presumed to originate from structures in the cochlea associated with the cochlear ampli er. In the other experiment (McFadden and Champlin, 2000), the measures were auditory evoked potentials (AEPs) of various time delays, commonly characterized as the auditory brainstem response (ABR), auditory middle-latency response (MLR) and auditory long-latency response (LLR). These three AEPs are generally presumed to originate from successively later stages along the chain of auditory centers connecting the cochlea to the cortex, although the sequence of peaks does not map directly onto a hierarchy of structures. Some AEPs are known to show sex di erences (Hall, 1992). To the extent that these sex di erences also are based, at least in part, on di erences in androgen level, AEPs might also exhibit di erences in users and non-users of OC. Accordingly, a retrospective examination of both the OAE and AEP experiments was undertaken. 2. General method Information about use of OC was extracted from a questionnaire completed by all subjects in both experiments, and, in the AEP experiment, also by questions asked at the time of scheduling the test session. In the OAE experiment, all female subjects were tested without regard to their menstrual cycles. In the AEP experiment, subjects using OC were also tested without regard to their cycles, but procedures were implemented to test the normal-cycling subjects during the midluteal phase of their cycles, when the levels of both estrogen and progesterone are high (and according to Elkind-Hirsch et al. (1994) when the ABR response is most di erent from that of males). Speci cally, the normal-cycling subjects were scheduled only for days 16^26 following the onset of their last menstrual period. For this report, only the data from heterosexual females and heterosexual males are included from the two experiments. (Using the data from all female subjects would have added eight homosexual and bisexual women to the group using OC from the OAE experiment and 10 homosexual and bisexual women from the AEP experiment). The four heterosexual women taking OC injections in the AEP experiment also were eliminated from this study. So many di erent brands and types of OC were being taken by the users of OC that it was impossible to analyze the data separately by brand or type of preparation. In the OAE experiment, about 44% of the subjects were using a combination product in which the doses of estradiol and progestin are constant across the cycle (rather than phasic products in which the dose of one or the other varies across the cycle), in the AEP experiment, about 57% were using a combination product, and across both experiments, about 53% were using a combination product. No subject was using a progestin-only product. The average age of the users and non-users of OC was essentially identical. In the OAE experiment, the users of OC were about 0.1 year younger than the non-users (21.6 vs. 21.7), and in the AEP experiment, the users were about 0.1 year older (21.3 vs. 21.2). 3. OAEs 3.1. Method In the OAE experiment, measurements were obtained from both ears for both SOAEs and CEOAEs. This yielded a measure of SOAE number, two measures of SOAE strength and a measure of CEOAE strength for each ear. Procedures for extracting these values can be found in the original papers (McFadden and Pasanen, 1998,1999). Brie y, overall SOAE power is the sum of the individual powers of all SOAEs in an ear and is expressed in db sound pressure level (db SPL re 20 WPa); power per SOAE is the sum of the individual powers of all SOAEs in an ear divided by the number of SOAEs contributing, expressed in db SPL; CEOAE amplitude is the rms amplitude of a 20-ms segment of the CEOAE waveform beginning 6 ms after the presentation of the click, also expressed in db SPL 1. Only the CEOAE data taken with the highest click level were used here. For purposes of comparing subjects using and not using OC, a two-ear average was calculated for each subject for each OAE measure. All subjects had scores in both ears for SOAE number and CEOAE strength, but because there is no meaning to SOAE strength for any ear having no SOAEs, subjects with no SOAEs in one or both ears could not be used for the statistical tests on the measures of SOAE strength. Accordingly, the number of subjects contributing to the statistical tests reported below on SOAE strength was smaller than for the other comparisons. For statistical analysis, the measure SOAE number was transformed 1 Technically, level is the correct term for describing values expressed in db SPL. The term power is used here intentionaly as a reminder that the SOAE detection algorithm operated on the power spectrum, and the calculations of SOAE strength were done entirely in power units, with only the nal results being converted into db.

3 D. McFadden / Hearing Research 142 (2000) 23^33 25 Table 1 Means (in bold), S.E.M. (in italics) and statistical analyses of overall di erences between non-users and users of oral contraceptives for various measures of OAEs Measure N Non-users Users Di erence t value P value Males E ect sizes Non-user/user Non-user/males SOAE number 36/21/ Overall SOAE power 29/17/ Power/SOAE 29/17/ CEOAE amplitude 36/21/ Two-ear averages used throughout. N indicates the number of non-users/users/males having usable data for both ears for the measure under consideration. For SOAE number, data were log-transformed prior to statistical analysis. For CEOAE amplitude, only the data for the highest click level were used. The P values shown are for two-tailed tests. as ln (X+1), where X is the mean number of SOAEs in the two ears of a subject Results In Table 1 are shown the mean values for the various OAE measures for the subjects using and not using OC. Also shown are the values of the test statistic, and the associated probabilities, for unpaired t tests of the various comparisons between the groups using and not using OC. As can be seen, no comparison even approached statistical signi cance, but every di erence was in the direction of weaker values for the subjects using OC. That is, the subjects using OC were invariably shifted away from the normal-cycling subjects in the male direction. For comparison, the male means are shown to the right in this table 2. E ect sizes are shown at the far right of Table 1 for the di erence between the users and non-users of OC and for the di erence between the non-users of OC and the males. These e ect sizes were calculated by dividing the di erence between the means of the two relevant groups by the square root of the weighted mean of the two variances for those groups. For comparisons 2 The term masculinization is used here to denote these shifts in the male direction seen for the users of OC. Although this term is descriptively accurate (and provides economy of communication), it also carries the implication that the mechanisms underlying these shifts are well-known and understood, and that is decidedly not the case. Although considerable circumstantial evidence suggests that androgenic mechanisms are involved in the e ects observed (at least for OAEs; see McFadden and Pasanen, 1999 for the logic of the argument), the details are not yet known, and it is possible that anti-estrogenic elements may be involved as well. The use of the term masculinization here is not meant to rule out such multiple contributions to the shifts observed. The term defeminization carries approximately equal descriptive power perhaps, but that term is typically reserved for discussions of complex, multiply determined behaviors, not for simple physiological measures varying along a single dimension, as is true for the measures discussed here. of the present sort, Cohen (1992) suggested that e ect sizes of 0.2, 0.5 and 0.8 could be regarded as small, medium and large, respectively. (Table IV of McFadden and Pasanen (1999) contains the corresponding e ect sizes for these sex di erences when the users and nonusers of OC are pooled.) Not shown in Table 1, but worthy of mention in passing, is the fact that, for the users of OC, the two measures of SOAE strength were both slightly greater in the left ear than the right ear, unlike the nding here for the non-users, or for the users and non-users combined (McFadden and Pasanen, 1999). There was no corresponding reversal of ears in the measures SOAE number and CEOAE strength (or in any of the measures for the males). This pattern of results could exist if the contraceptive drugs acted to reduce the strength of strong SOAEs (which are typically in the right ear) more than the strength of weak SOAEs. An informal visual examination of the data suggested that this may be the correct explanation. 4. AEPs 4.1. Method In the AEP experiment, an averaged waveform was obtained using both the ipsilateral and contralateral pair of electrodes and with the insert earphone placed in both the left and right ears, for a total of four averaged waveforms per subject per stimulus level (70 and 35 db normal hearing level or nhl) per type of AEP (ABR, MLR and LLR). For each type of AEP, estimates of latency were obtained from each of those four averaged waveforms for each of several peaks in the waveform, and for some waveform peaks, estimates of amplitude were also obtained. The peaks were identi ed by judges, each of whom had at least 2 years experience

4 26 D. McFadden / Hearing Research 142 (2000) 23^33 Table 2 Means (in bold), S.E.M. (in italics) and statistical analyses of overall di erences between non-users and users of oral contraceptives for various waves of the ABR, MLR and LLR, and for two cognitive tasks Users Condition N Nonusers Di erence t value P value Males E ect sizes Non-users/ users Non-users/ males ABR wave-i amplitude, 70-dB clicks 44/32/ ABR wave-v latency, 70-dB clicks 44/34/ ABR wave-v amplitude, 70-dB clicks 44/34/ ABR wave-v latency, 35-dB clicks 44/33/ ABR wave-v amplitude, 35-dB clicks 44/33/ MLR wave-n a amplitude, 35-dB clicks 22/22/ LLR Wave-N 1 amplitude, 70-dB clicks 47/35/ LLR wave-n 2 latency, 70-dB clicks 46/35/ Mental-rotation test 86/57/ Water-level test 86/57/ For the entries under non-users and users, each subject contributed a four-way mean of her left and right brain values (see text) except for ABR wave I, where entries are means of ipsilateral left and ipsilateral right only (no contralateral data included). Latencies are in ms and amplitudes are in nv. N indicates the number of non-users/users/males having usable data for both the left and right brain for the measure under consideration. For the mental-rotation and water-level tests, the entries are in units of average number correct and average error in degrees, respectively, and the subjects are from the AEP and OAE experiments combined. The P values shown are for two-tailed tests. E ect size is the absolute value of the di erence in the means of the two groups divided by the square root of the weighted means of the variances of the two groups. For every comparison except LLR wave N 2, the users of oral contraceptives were shifted in the male direction. in judging AEP waveforms. The judges were blind to whether the subjects were users or non-users of OC (or male). For each measure of latency and amplitude, an overall mean was calculated over the four averaged waveforms 3. Across the three types of AEP and the two click levels, there were a total of 19 such fourway averages of latency and amplitude possible for each subject. For each of those 19 AEP measures, the basic sex di erence was examined by comparing the four-way mean for the heterosexual women not using OC with the corresponding mean for the heterosexual males. A two-tailed t test and an e ect size were calculated for each variable. When the K level was corrected for the large number of comparisons being made, few of these 19 comparisons achieved statistical signi cance, but for eight of the 19 comparisons, the e ect size for the sex di erence was 0.5 or greater (characterized as a medium-sized e ect by Cohen, 1992). Those eight AEP 3 Wave I of the ABR was an exception. For it, only the two ipsilateral averaged waveforms were used to calculate overall means, on the grounds that there is no meaning to a contralateral response at the level of the eighth nerve. measures were the ones chosen for examination for differences between the women using OC and the normalcycling women. The only subjects included in the analysis conducted on a particular variable were those for whom estimates could be obtained from all four averaged waveforms for that variable. In general, the AEP measures showing substantial sex di erences were associated with early evoked responses Results The same basic pattern of results was obtained for the AEP measures as for the OAE measures. For all but one of the eight comparisons examined, the mean latency or amplitude for the women using OC was shifted away from that of the normal-cycling women in the direction of the males. In Table 2 are shown the fourway means for the users and non-users of OC for each of those eight AEP measures that exhibited a substantial sex di erence. Also shown are the values for the unpaired t tests, their associated probabilities and the e ect sizes for each of the comparisons between users and non-users of OC. In addition, e ect sizes are shown

5 D. McFadden / Hearing Research 142 (2000) 23^33 27 for the basic sex di erence for each of the measures. As was true for the OAE data in Table 1, few of the comparisons between users and non-users of OC approached statistical signi cance, although some of the e ect sizes were moderately large. However, for every measure except latency of wave N 2 of the LLR (the longest latency wave measured), the mean values for the users of OC were shifted toward those of the males (whose four-way means are shown to the right in Table 2). Thus, masculinization e ects of the sort seen in the OAE data for users of OC are also present in the AEP data. There is a suggestion in the AEP data that the masculinization e ects may be strongest for the earlier (lower level) potentials and for the weaker click stimulus. The reason for such trends is unclear, but it may be that the cochlear ampli ers contribute proportionally more to those AEPs than to the later AEPs or to the AEPs generated by stronger clicks 4. The pattern of e ects seen in Table 2 for the ABR waves (shorter latency and greater amplitude for females than males) is in accord with the directions of the sex di erences routinely observed for these measures (Hall, 1992). This di erence in ABR latency persists even when the sex di erence in head size is taken into account (Trune et al., 1988; Don et al., 1993), suggesting that the latency di erence is attributable to neural di erences in auditory structures at the level of the midbrain and below. For LLR wave N 1, the amplitude was greater in the females than the males, just as for wave V of the ABR. For MLR wave N a and LLR wave N 2, however, the sex di erences were opposite to those seen in ABR, with the latency being longer and the amplitude being smaller in the females than in the males. The issue of sex di erences has not been extensively studied for the later AEPs (Hall, 1992), and neither con rmations nor contradictions of these reversals of the direction of the sex di erence were found in the literature. Elkind-Hirsch et al. (1992a) monitored changes in the ABR through the course of the menstrual cycle, both in nine users of OC and in nine normal-cycling women, but no statistical comparisons between the two groups were reported. Examination of the data in Table 2 of Elkind-Hirsch et al. (1992a) suggests the existence of a small masculinizing e ect of the sort reported here, but 4 When a faster rate of click presentation was used to measure the ABR (Fast ABR), the mean latency and amplitude of wave V for the users of OC were also shifted toward the males, but those measures were judged to be too similar to regular ABR to constitute additional evidence, and consequently were not included in Table 2. With Fast ABR, the e ect sizes for the sex di erence were 0.68 and 0.86 for wave-v latency and amplitude, respectively, and the e ect sizes for the di erence between users and non-users of OC were 0.19 and 0.08 for wave-v latency and amplitude, respectively. it is unlikely that those di erences in latency would have been signi cant given the small numbers of subjects involved. (A comment in Elkind-Hirsch et al. (1994) appears to con rm the absence of statistical signi cance.) The di erence between users and non-users in the Elkind-Hirsch et al. data was about 1/3 of the di erence reported in Table 2 here for comparable stimulus conditions. Elkind-Hirsch et al. (1992a) also asserted that there were no signi cant di erences in wave-v amplitude between conditions or subjects. Samani et al. (1987) did compare the ABRs of users and non-users of OC. Two rates of click presentation were used and waves I^VII were examined. For most of the comparisons, the 20 users of OC had shorter latencies than the 10 non-users (the opposite direction of e ect from that reported here), and, like here, most of those di erences were not statistically signi cant. The reason for this discrepancy with the present data is not clear, but it does appear as if Samani et al. did not test their 10 normal-cycling women during the midluteal phase, as was done here. 5. Cognitive tests 5.1. Method Included as part of the questionnaire administered in both the OAE and AEP experiments were two cognitive tests that routinely show large sex di erences. One taps the subject's ability to mentally rotate shapes in threedimensional space (Vandenberg and Kuse, 1978). The subject's task is to determine for each test gure which two of a set of four comparison gures portray the same object from di erent perspectives, and which two portray di erent objects. Males typically perform better on this task than females. Here, the subject was given 3 min to respond to 10 such multiple-choice items. Scoring gave two points when both answers were correct, no points if one answer was correct and one incorrect, and one point if only one answer was given and it was correct. A subject's overall score was simply the sum of these points across the 10 items. The second test required that the subject draw a line across a gure of a water glass to indicate the surface of the uid inside the glass. The glasses were tilted at 20³, 40³ and 60³ to the left and to the right of vertical. One each of these six stimuli was provided. Scoring involved calculating the (absolute) magnitude of error from horizontal in degrees and averaging those values across the six items. Males are typically more accurate at this task than females, and this di erence persists even when the subjects are experienced with vessels containing uids (Thomas et al., 1973). Because these two cognitive tests were identical in the

6 28 D. McFadden / Hearing Research 142 (2000) 23^33 two experiments, the data were combined across experiments Results The average values for the two cognitive tests are shown at the bottom of Table 2, along with the various summary statistics and the two e ect sizes. Of most importance here is the fact that for both tests, the users of OC had scores intermediate to those of the normalcycling women and the males (whose means are shown at the far right in Table 2). The same pattern of results was obtained for the water-level test when the proportion of subjects having an average score greater than 3³ was used as the dependent variable. These cognitive tests reveal that the masculinization of the users of OC apparently was not limited to their auditory systems, although, again, the shifts were small and the individual comparisons did not achieve statistical signi cance. For both cognitive tests, the sex di erence and the di erence between users and non-users of OC were greater in the AEP experiment than in the OAE experiment. This may be related to the fact that, in the OAE experiment, all subjects were tested without regard to their menstrual cycles, but in the AEP experiment, the normal-cycling women were tested during the midluteal phase of their cycle when women are maximally di erent from men according to the data of Hampson (1990a,b). 6. Lack of independence When interpreting these results, it is necessary to take into consideration that the various measures were not independent. As reported in McFadden and Pasanen (1999), the correlations between the various OAE measures ranged from 0.73 (for CEOAE amplitude with power per SOAE) to 0.97 (for overall SOAE power with power per SOAE). For the various AEPs considered here, the range was 0.04 (for ABR wave-v latency at 35 db with MLR wave-n a amplitude at 35 db) to 0.66 (for ABR wave-v latency at 70 db with ABR wave-v latency at 35 db). Accordingly, the fact that 13 of the 14 dependent variables considered here showed the same direction of e ect is less compelling than would be true had the various measures been independent. In an attempt to resolve the uncertainty about the meaning of these patterns of results, simulations were performed 5. The purpose of the simulations 5 Thanks go to Professor J.C. Loehlin for suggesting a simulation of this sort. was to use the data available to estimate the likelihood that the pattern of results obtained in the two experiments occurred by chance. The strategy was to sort the subjects randomly into `users' and `non-users' and compare the patterns of results obtained with the patterns of results actually present in the data. For these simulations, the data for the users and non-users of OC were pooled (within an experiment), and random samples of the size of the users group (21 in the OAE experiment and 36 in the AEP experiment) were drawn from that pool. For each dependent variable listed in Tables 1 and 2, an e ect size was calculated between the group sampled and the remainder of the female subjects in that pool (thereby simulating users and non-users of OC). The absolute values of those individual e ect sizes were averaged across all the auditory and cognitive measures (six measures in the OAE experiment and 10 in the AEP experiment), and that mean of absolute values was compared to the mean of the absolute values of the e ect sizes actually obtained between users and non-users of OC in that experiment (which was in the OAE experiment and in the AEP experiment). This procedure was implemented times and a tally obtained of the number of samples for which the mean e ect size for the random partitioning of subjects was larger than the mean e ect size actually observed in the data. The actual mean e ect size was equaled or exceeded 6046 times for the OAE experiment, but only 192 times for the AEP experiment. These correspond to implied probability values of and 0.019, respectively. A similar result was obtained when the simulations used only the auditory and not the cognitive measures (four measures in the OAE experiment and eight in the AEP experiment). The mean of the absolute values of e ect size actually obtained was exceeded in the random partitionings 4023 times in the OAE experiment but only 189 times in the AEP experiment, corresponding to implied probability values of and 0.019, respectively. That is, for the AEP experiment, the actual partitioning of subjects into users and non-users of OC did yield rather a rare pattern of results, but for the OAE experiment, it did not. Because these simulations took into account whatever dependences existed among the various measures in each experiment, the simulations stand as compelling evidence that the obtained pattern of results in the AEP experiment was rather unlikely to have occurred by chance, but the obtained pattern of results in the OAE experiment was not particularly unlikely. Again, it should be recalled that the group di erences observed in the OAE experiment may be underestimates because those measurements were obtained without regard to the menstrual cycle.

7 D. McFadden / Hearing Research 142 (2000) 23^ Discussion In summary, a retrospective examination of data collected in two large scale experiments on the auditory system has revealed that four OAE measures, seven (of eight) AEP measures and two cognitive measures all exhibited the same trend. The means for the users of OC were shifted away from those of the normalcycling women and in the direction of the males. Although all of the shifts were small, the consistency in the direction of the shifts strongly suggests the existence of a weak masculinizing e ect of OC. One implication of the present ndings is that past attempts to assess the magnitudes of the sex di erences in these various measures very likely underestimated their true magnitudes, by an amount related to the number of users of OC included in each particular experiment. Another implication of these ndings is that, in the future, experimenters desiring to study the neural mechanisms underlying these various auditory measures apparently will need to be attentive to the variable of OC use. Generally, the AEP responses that showed substantial sex di erences here (and accordingly the ones examined for OC di erences in Table 2) were ones associated with early AEPs, and thus, lower structures in the auditory system. It is important to appreciate that the absence of sex di erences and of di erences between users and non-users of OC in the AEPs of longer latency is not conclusive evidence that di erences of some sort do not exist in higher auditory centers. The apparent absence of e ect could simply be attributable to the well-known fact that evoked potentials, especially the later potentials, do not represent responses from single, successive, individual locations in the brain (e.g. Hall, 1992). Rather, each peak in the AEP is a sum in time and space across synchronous, or nearly synchronous, rings originating from multiple neural sites. Accordingly, sex or group di erences existing at one site might be smeared or obliterated by rings at another, collateral site that does not exhibit the same sex or group di erences, or exhibits them in the opposite direction. There are a number of possible ways to interpret the present data. Because this was only a retrospective, correlational study, technically all that legitimately can be concluded is that there exists some weak correlation between the use of OC and the presence of a slightly masculinized auditory system. However, simple intuition suggests that such a rigidly technical interpretation would be too conservative in this case, and that there is good reason to suspect the existence of a causal relationship between use of OC and the masculinization e ects reported here. Consider rst the likelihood that some of the primary factors contributing to whether or not a college-age woman is using OC are surely highly situational, and subject to change on relatively short notice. For example, the beginning of a romantic involvement could lead a non-user of OC to become a user within a matter of days, or the converse might occur. And because a airs of the heart are so unpredictable, it is fully believable that chance played a major role in determining which speci c women in the present study were users of OC and which not at the time of testing. In addition, the research literature suggests that 30^60% of young, new users of OC discontinue use within the rst year because of side-e ects of various sorts (e.g. Trussell and Kost, 1987; Kusseling et al., 1995; Rosenberg et al., 1995). Taken together, these factors suggest that were all the same subjects tested again at some point in the near future, the breakdown of who was and who was not using OC would not be the same. Assuming that the measures would still show the same direction of e ect as observed here, it appears reasonable to presume that the present ndings represent more than a simple correlation and, thus, reasonable to expect that the weak masculinization e ects observed will ultimately prove to be causally related to the use of OC. Certainly the existence of a causal relationship deserves to be among the rst issues explored in any true experiments done in the future on this general topic. Much of the following discussion about possible explanations for the di erences between users and nonusers of OC is based on the assumption that the oral contraceptives caused the di erences between the two groups Hormone levels As mentioned in Section 1, for some years now, this laboratory has been interested in some apparent e ects of androgens on the auditory system, particularly on OAEs. We have suggested elsewhere that di erential exposure to androgens prenatally may explain such facts as the normal sex di erences in OAEs and hearing sensitivity, and the weaker OAEs in females having male co-twins than in females having female co-twins (McFadden, 1993; McFadden et al., 1996). The idea is that the cochlear ampli ers (Davis, 1983) are permanently weakened by prenatal exposure to high levels of androgens, and the sex di erences in OAEs and hearing sensitivity are both consequences of that weakening. The e ects have been presumed to be permanent because the sex di erences in OAEs exist in infants and children (reviewed in McFadden, 1998) even though there is no known sex di erence in childhood androgen levels after about 2 months of age (Smail et al., 1981). The possible existence of additional, postnatal androgen e ects on the cochlear ampli ers of adults is implied by the fact that the SOAEs of a single adult male engaged in estrogen therapy appeared to shift in the

8 30 D. McFadden / Hearing Research 142 (2000) 23^33 female direction (more SOAEs) as androgen levels fell (McFadden et al., 1998). We have emphasized the role of androgens, and minimized the contributions of estrogens in our theorizing, because both the menstrual cycle and pregnancy seem to have so little e ect on OAEs (Bell, 1992; Haggerty et al., 1993; Burns et al., 1993). From this line of argument, one might be tempted to speculate that di erences in androgen levels were responsible for the OAE and AEP di erences reported here between the users and non-users of OC; speci cally, that the shift of these measures in the male direction in the users of OC is attributable to their having higher levels of androgens than the normal-cycling subjects. Contradicting that interpretation of the present results is the nding that the levels of plasma free testosterone are lower in users of OC than in normal-cycling women (Vermeulen and Thiery, 1982; Gaspard et al., 1983; Bancroft et al., 1991). Accordingly, the users of OC should have exhibited `hyper-feminine' not masculinized responses. Apparently our simple androgen story cannot easily be extended to include the present results. An interesting logical possibility is that the nding of lower levels of plasma free testosterone in users of OC is not a consequence of users producing less testosterone than non-users, but rather of users binding more of the testosterone they do produce or aromatizing more of it to estradiol. Intracellular estradiol is known to be a powerful masculinizing agent. Under this logical possibility, perhaps some auditory neurons in the central nervous system are being masculinized more than usual, which is in accord with our previous explanation of our various OAE results. There is some evidence in accord with this suggestion in the ABR literature. The work of Elkind-Hirsch et al. (1992a,b, 1994) implicated estrogen as the relevant hormone controlling ABR latency. For example, wave-v latency was the most male-like when estrogen levels were high, both in normal-cycling women (Elkind-Hirsch et al., 1992a) and in women receiving hormone replacement therapy after premature ovarian failure (Elkind-Hirsch et al., 1992b). The simultaneous presence of progesterone seems to o set this e ect of estrogen in both types of subjects. Wave-V latency for the normal-cycling women was most di erent from that of men during the midluteal phase of the menstrual cycle. The explanation offered for these e ects by Elkind-Hirsch et al. was that high levels of estradiol decrease the speed of neural conduction (perhaps through increased synthesis of GABA), and thereby increase the latency of wave V, but that the presence of progesterone interrupts these e ects of estradiol. The marked sex di erence in wave- V latency was attributed to males aromatizing larger quantities of testosterone into estradiol (Elkind-Hirsch et al., 1994), which is known to be a potent masculinizing agent in many neurons in the central nervous system. Under this explanation, then, masculinization of AEPs is not due to testosterone level per se, but to one of its intracellular derivatives. If users of OC aromatized more testosterone into estradiol in certain auditory neurons than non-users of OC, the masculinization effects reported here might be the result. A nding by Coleman et al. (1994) appeared to contradict this Elkind-Hirsch et al. explanation. These experimenters ovariectomized female rats and then administered estrogen replacement of varying magnitude. The typical result was that the latencies to the various waves of the ABR were longer in the ovariectomized animals than in the controls (that is, ovariectomy led to masculinization of the ABR), but the administration of estrogen reversed this e ect. In fact, the groups of animals given estrogen following ovariectomy commonly had latencies that were even shorter (`more feminine') than those of the controls. The implication is that high levels of estrogen increase the speed of neural conduction, just the opposite of the Elkind-Hirsch et al. suggestion. In speaking to the apparent discrepancy, Coleman et al. (1994) suggested that there may be a nonmonotonic e ect of estrogen level on wave-v latency, and Elkind-Hirsch et al. were working with larger relative doses than they were. It is tempting to accept this explanation because the Elkind-Hirsch et al. account has considerable appeal, especially given the present outcomes. Whatever the ultimate resolution of this apparent contradiction, it is clear that any attempt to explain the e ects of the menstrual cycle (or the use of OC) on the ABR will have to be consistent with the Elkind-Hirsch et al. (1992a) nding that the level of estradiol was low and relatively constant throughout the month in users of OC, as compared to the considerable monthly variation in estradiol shown by the normal-cycling women Body temperature Body temperature is known to a ect AEPs (Hall, 1992). Speci cally, the lower the body temperature, the longer the latency and smaller the amplitude, at least for ABR waves I and V. Although body temperature was not monitored in the two experiments on which the present study was based, other research suggests that this factor is not likely to be the explanation for the di erences reported here. In normal-cycling women, the core body temperature is about 0.2^0.6³C higher in the luteal phase (when progesterone and estrogen levels are both high) than in the follicular phase (reviewed by Rogers and Baker, 1997; also see Kattapong et al., 1995). Similarly, in users of OC, body temperature is 0.2^0.3³C higher dur-

9 D. McFadden / Hearing Research 142 (2000) 23^33 31 ing the latter part of the 21 days of drug use (the pseudoluteal phase) than during the following week when no OC is taken and menstruation occurs (however, this temperature di erential in users of OC apparently is greatest during sleep and diminishes substantially after awakening (Kattapong et al., 1995)). Of most importance here is the fact that users of OC appear to have day-time core body temperatures that are about 0.1³C higher than those of normal-cycling women during corresponding phases of the menstrual cycle (extracted from Figs. 1 and 2 of Kattapong et al., 1995). In our AEP experiment, the normal-cycling women were tested during the midluteal phase of their cycles, but the users of OC were tested at random relative to their cycles. Taken together, these facts suggest that at the time of testing, the body temperature for the normal-cycling women probably averaged about 0.05^0.10³C lower than that for the users of OC. Accordingly, one would expect the normal-cycling women to have slightly longer ABR latencies and slightly smaller ABR amplitudes than the users of OC, but the exact opposite was observed (see Table 2). Thus, it appears that a di erence in core body temperature is not the basis for the di erences in AEPs seen for users and non-users of OC reported here. Note that even if the direction of e ect was not wrong, the magnitude of the latency di erence between users and non-users of OC was too large to be explained by the presumed di erences in body temperature in the two groups. The standard estimate is that ABR latency will shift by 150^200 Ws for a temperature change of 1³C (Don et al., 1993; Hall, 1992). If the temperature di erence between users and non-users of OC was about 0.05^0.10³C, as estimated above, that would produce a latency di erence of 15^20 Ws at most (assuming a linear function relating latency to temperature), and Table 2 reveals that the latency difference for ABR wave V between users and non-users of OC was about 70^210 Ws across the two click levels. Out of concern about the possibility that a di erent average body size might produce a di erent average body temperature in the two groups of women (e.g. Adam, 1989), the (self-reported) heights and weights were compared for the users and non-users of OC in the AEP experiment. The normal-cycling women were about 0.23 cm shorter and about 1.8 kg heavier than the users of OC, and neither di erence approached statistical signi cance. According to the data of Adam (1989), the 1.8-kg weight di erence would have amounted to a lower body temperature in the normalcycling women of about 0.012³C which is in the wrong direction to account for the latency di erences observed, and even had the direction been correct, that temperature di erence would have produced an insigni cant di erence in latency Secondary factors The speci c question about body temperature raises a general point. One class of possible explanations for the weak masculinization e ects reported here is that they stem from what might be thought of as secondary factors related to use of OC. That is, it is believable that OC produces some of its e ects on the body and brain through more or less direct actions of the hormones included in the OC tablet, whereas other e ects are produced indirectly by factors that are themselves the result of hormonal actions. Core body temperature is presumably an example of the latter. Because neural conduction time is temperature-dependent, a change in core temperature could lead to changes in latency in some of the waves of the AEP without there being a direct, local action of the hormones themselves on the auditory neurons involved in producing those AEPs. Although the speci c possibility of body temperature now seems unlikely, logically, a similar argument could be made for any number of secondary factors, such as blood pressure (Heintz et al., 1996), cerebrospinal uid pressure, metabolism, availability of neurotransmitters, changes in e erent pathways in the auditory central nervous system, etc. (see Reinberg et al., 1996). Any comprehensive search for possible explanations of the present results should include such secondary factors Hearing sensitivity Hearing sensitivity is known to vary through the menstrual cycle of normal-cycling women, with sensitivity being about 5 db worse during menses than during the midluteal phase (e.g., Swanson and Dengerink, 1988). Presumably the locus of this e ect is beyond the cochlear ampli ers because SOAEs change only minimally through the cycle (Bell, 1992; Haggerty et al., 1993). These monthly uctuations in hearing sensitivity are much reduced for users of OC (Swanson and Dengerink, 1988). In the AEP experiment, where the normal-cycling women were tested during the midluteal (most sensitive) phase, the average hearing sensitivity of the normal-cycling women was probably slightly better than that of the users of OC. The relationship between this fact and the results observed is not clear, however, because Jerger and Johnson (1988) reported that ABR latency was essentially unchanged by even relatively large di erences in hearing sensitivity (actual hearing losses) or in the strength of the click stimulus used. This also suggests that any incipient, undetected hear-

10 32 D. McFadden / Hearing Research 142 (2000) 23^33 ing losses existing within groups should not have contributed substantially to the results observed. from L. Strother, C. Scott, A. Harkrider, M. Dean, J. Marler, K. Hill, M. Dittrich and D. Simmons. 8. Summary Fourteen dependent variables collected across two experiments were examined for evidence of di erences between users and non-users of OC. All of these measures exhibited moderate to large sex di erences. For 13 of these measures, the means for the users of OC lay intermediate to the means for the normal-cycling women and the males. One parsimonious explanation is that OC use produces a masculinization of certain auditory structures. The mechanism(s) producing this masculinization are unclear, but possibly relevant is the fact that the levels of plasma free testosterone are di erent in users of OC than in normal-cycling women (Vermeulen and Thiery, 1982; Gaspard et al., 1983; Bancroft et al., 1991). Although the direction of this particular di erence appears to contradict the present ndings, it may be that an androgen derivative such as estradiol is higher in some cell populations in the users of OC than in the normal-cycling women. Of special interest to the writer is the fact that use of OC cannot account for the stronger OAEs seen in heterosexual women than in homosexual and bisexual women (McFadden and Pasanen, 1998,1999). Logically, if plasma free testosterone did have the e ect of suppressing the cochlear ampli ers, and thereby weakening OAEs, then the lower testosterone levels in users of OC (Vermeulen and Thiery, 1982; Gaspard et al., 1983; Bancroft et al., 1991), and the greater proportion of OC users among the heterosexuals, might have been the basis for the stronger OAEs measured in the group of heterosexual women. However, for OAEs, the di erences between users and non-users of OC were all in the wrong direction and were too small to account for the di erences between heterosexual and non-heterosexual women. Acknowledgements This work was supported by research Grant DC from the National Institute on Deafness and other Communication Disorders. All experimental procedures were approved in advance by the UT Institutional Review Board. C.A. Champlin and E.G. Pasanen provided crucial technical assistance and comments on a preliminary draft. J.C. Loehlin and C. Meston also commented on the draft version. For the OAE experiment, assistance was received from S. Oropeza, L. Ramirez, S. Bratcher, C. Furche, A. Phan and N. Callaway. For the AEP experiment, assistance was received References Adam, K., Human body temperature is inversely correlated with body mass. Eur. J. Appl. Physiol. 58, 471^475. Bancroft, J., Sherwin, B.B., Alexander, G.M., Davidson, D.W., Walker, A., Oral contraceptives, androgens, and the sexuality of young women: II. The role of androgens. Arch. Sex. Behav. 20, 121^135. Bell, A., Circadian and menstrual rhythms in frequency variations of spontaneous otoacoustic emissions from human ears. Hear. Res. 58, 91^100. Burns, E.M., Campbell, S.L., Arehart, K.H., Keefe, D.H., Long-term stability of spontaneous otoacoustic emissions. Abst. Assoc. Res. Otolaryngol. 16, 98A. Cohen, J., A power primer. Psych. Bull. 112, 155^159. Coleman, J.R., Campbell, D., Cooper, W.A., Welsh, M.G., Moyer, J., Auditory brainstem responses after ovariectomy and estrogen replacement in rat. Hear. Res. 80, 209^215. Davis, H., An active process in cochlear mechanics. Hear. Res. 9, 79^90. Don, M., Ponton, C.W., Eggermont, J.J., Masuda, A., Gender di erences in cochlear response time: An explanation for gender amplitude di erences in the unmasked auditory brain-stem response. J. Acoust. Soc. Am. 94, 2135^2148. Elkind-Hirsch, K.E., Stoner, W.R., Stach, B.A., Jerger, J.F., 1992a. Estrogen in uences auditory brainstem responses during the normal menstrual cycle. Hear. Res. 60, 143^148. Elkind-Hirsch, K.E., Wallace, E., Stach, B.A., Jerger, J.F., 1992b. Cyclic steroid replacement alters auditory brainstem responses in young women with premature ovarian failure. Hear. Res. 64, 93^ 98. Elkind-Hirsch, K.E., Wallace, E., Malinak, L.R., Jerger, J.F., Sex hormones regulate ABR latency. Otolaryngol. Head Neck Surg. 110, 46^52. Gaspard, U.J., Romus, M.A., Gillain, D., Duvivier, J., Demey-Ponsart, E., Franchimont, P., Plasma hormone levels in women receiving new oral contraceptives containing ethinyl estradiol plus levonorgestrel or desogestrel. Contraception 27, 577^590. Haggerty, H.S., Lusted, H.S., Morton, S.C., Statistical quanti- cation of 24-hour and monthly variabilities of spontaneous otoacoustic emission frequency in humans. Hear. Res. 70, 31^49. Hall, J.W., III, Handbook of Auditory Evoked Responses. Allyn and Bacon, Boston, MA. Hampson, E., 1990a. Estrogen-related variations in human spatial and articulatory-motor skills. Psychoneuroendocrinology 15, 97^ 111. Hampson, E., 1990b. Variations in sex-related cognitive abilities across the menstrual cycle. Brain Cogn. 14, 26^43. Heintz, B., Schmauder, C., Witte, K., Breuer, I., Baltzer, K., Sieberth, H.-G., Lemmer, B., Blood pressure rhythm and endocrine functions in normotensive women on oral contraceptives. J. Hypertens. 14, 333^339. Jerger, J., Johnson, K., Interactions of age, gender, and sensorineural hearing loss on ABR latency. Ear Hear. 9, 168^176. Kattapong, K.R., Fogg, R., Eastman, L.F., E ect of sex, menstrual cycle phase, and oral contraceptive use on circadian temperature rhythms. Chronobiol. Int. 12, 257^266. Kusseling, F.S., Wenger, N.S., Shapiro, M.F., Inconsistent contraceptive use among female college students: Implications for intervention. J. Am. Coll. Health 43, 191^195.