|Home | About | Journals | Submit | Contact Us | Français|
This study is designed to measure the degree to which spiral ganglion cell (SGC) survival in the left and right ears is similar in profoundly hearing-impaired human patients with symmetric (right/left) etiology and sensitivity. This is of interest because a small difference between ears would imply that one ear could be used as a control ear in temporal bone studies evaluating the impact on SGC survival of a medical intervention in the other ear.
Forty-two temporal bones from 21 individuals with bilaterally symmetric profound hearing impairment were studied. Both ears in each individual were impaired by the same etiology. Rosenthal’s canal was reconstructed in two dimensions and segmental and total SGCs were counted. Correlation analysis and t-tests were used to compare segmental and total counts of left and right ears. Statistical power calculations illustrate how the results can be used to estimate the effect size (right/left difference in SGC count) that can be reliably identified as a function of sample size.
Left counts (segmental and total) were significantly correlated with those in the right ears (p<0.01) and the coefficients of determination for segments 1 to 4 and total count were respectively 0.64, 0.91, 0.93, 0.91 and 0.98. The hypothesis that mean segmental and total counts of right and left are the same could not be rejected by paired t-test.
The variance in the between-ear difference across the temporal bones studied indicates that useful effect sizes can be reliably identified using subject numbers that are practical for temporal bone studies. For instance, there is 95% likelihood that an interaural difference in SGC count of approximately 1000 cells associated with a treatment/manipulation of one ear will be reliably detected in a bilaterally-symmetric profound hearing loss population of temporal bones from approximately 10 subjects.
Spiral ganglion cells (SGCs), the first-order neurons of the auditory system, have been the subject of many studies. For example, the number of these cells in normal-hearing ears (Guild, 1932) and in hearing-impaired ears with different etiologies (Nadol et al., 1989) has been reported. One question that has not been directly studied is whether the difference in the number of SGCs between ears of an individual is small when the hearing sensitivity is similar. This is of interest because a small difference would imply that one ear could be used as a control ear in temporal bone studies evaluating the impact on SGC survival of a treatment, insult or some other factor experienced only by the other ear.
When matched for hearing sensitivity, the relatively large, across-subject variation in the SGCs counted leads one to question whether a similar variation holds across ears in the same subject. For temporal bones from subjects with normal-hearing sensitivity while living, Guild (1932) documented a range of 23,193 to 39,114 total SGCs (N=10) and Otte et al. (1978) reported approximately 20,000 to 37,000 SGCs (estimated from their Figure 3; N=8). The range can be even larger in the case of subjects who suffered significant hearing impairment. In 16 temporal bones from subjects with documented pure tone threshold average greater than 70 dB HL during life, Hinojosa et al. (1983) reported counts of total SGCs ranging from 0 to 25,873 SGCs (N=16). If the relatively large across-subject variability in SGC counts seen in both normal and hearing-impaired ears is also exhibited between ears in the same subject, the use of one ear in each subject as a control for the other would provide relatively weak statistical power.
In spite of this potential problem, it is not unusual for investigators to use one ear as a control for the other in temporal bone studies. For instance, Khan et al. (2005) studied the effect of cochlear implantation on SGC survival by assuming both ears of an implantee had the same number of SGCs before implantation and compared the SGC count of the implanted ear with that of the contralateral ear. Fayad et al. (2006) made the same assumption and evaluated the effect of multi-channel cochlear implants on different elements of temporal bone histopathology (including SGC count) by comparing the implanted with the non-implanted ear. Except for Nadol et al. (1989) who reported a correlation of 0.87 between the SGC counts of the right and left ears of 27 subjects (not selected for symmetric hearing sensitivity), there is little basis for assuming the SGC count for one ear can serve as an effective control for the opposite ear.
The present study evaluates the degree to which the SGC counts in the left and right temporal bones of 21 subjects who were profoundly hearing impaired during life are similar. We also demonstrate how the results can be used in statistical power calculations to help guide the design of experiments using one ear of each temporal bone pair as a control for the other.
The 42 temporal bones from 21 subjects (Table 1) selected for study included a subset of bones from the 93 temporal bones of 66 profoundly hearing impaired individuals studied by Nadol et al (1989) meeting the following criteria: (1) the right and left temporal bones of each subject are available, (2) same etiology for hearing loss in both ears in each subject and (3) bilaterally symmetric profound hearing impairment documented by a statement in the medical history or by audiometric test results (bilateral pure tone average ≥ 90 dB with maximum 10 dB difference between ears at each test frequency). Five subjects were born with congenital hearing impairment and twelve cases had sudden or rapid progressive sensorineural hearing loss so that in 17 cases the period between hearing loss (pure tone average >25 dB) and profound hearing loss (pure tone average ≥ 90 dB) was not greater than one year. Although the etiology of hearing impairment varied across subjects, within each individual the etiology was the same for both ears. While pure tone audiometry was not available for 11 of the 21 cases, the medical history associated with each of these cases described the patient’s hearing loss as profound and symmetric. For 10 cases, audiometric results were available (see Table 2) and document a pure tone average hearing level of 90 dB or greater that was bilaterally symmetric (maximum threshold difference between ears of 10 dB at each frequency eliciting a response).
The temporal bones were removed after death, fixed in Heidenhain Susa solution or 10% buffered formalin and decalcified in ethylene diamine tetra acetic acid (EDTA), and embedded in celloidin (Schuknecht, 1968). The temporal bones were sectioned at a thickness of 20 µm in the horizontal plane and every tenth section was stained with hematoxylin and eosin and mounted on a glass slide. Rosenthal’s canal was reconstructed two-dimensionally (Figure 1) by a method described by Schuknecht (1953) and Otte et al. (1978). All spiral ganglion cells with visible nuclei were counted on every tenth section. SGC counts for the four segments identified in Figure 1 were computed by adding all counts across slides from the same segment. Each segmental count was multiplied by ten to account for unmounted sections, and again multiplied by a correction factor to account for doubly counted spiral ganglion cells (Konigsmark, 1970; Nadol, 1988). The total spiral ganglion cell count was computed as the sum of the four segmental counts (Nadol, 1988; Nadol et al., 1989).
Correlation analysis was used to assess the strength of association between the SGC counts of the right and left ears and the t-test was used to determine whether the mean segmental and total SGC counts of the right and left ears differed significantly. Power calculations (based on the statistics associated with the counts made for each of the 21 subjects) are presented to illustrate how the results can be used to estimate the effect size (difference in SGC count between test and control ears) that can be reliably identified as a function of sample size (number of subjects).
Table 1 shows that the 21 cases were impaired by a variety of etiologies including congenital malformations such as Mondini and CHARGE associated cochlear malformation, and acquired diseases such as viral labyrinthitis. The subjects included 11 males and 10 females ranging in age from 0.6 to 92 years at the time of death. There was a wide range of the duration of deafness that varied between a few months in cases 13 and 3 to more than 80 years in cases 7 and 10.
While the medical history of each patient described a symmetric and profound hearing loss, audiometric test results were available in only 10 cases and are presented in Table 2. One of the case selection criteria applied to those subjects with acoustic test results was a symmetric hearing loss: a maximum 10 dB difference in hearing threshold between ears at each frequency eliciting a response. This criterion is easily applied to cases 11 and 16 where response thresholds were measured at each frequency tested. In the other eight cases with audiometric results, the subjects did not respond at the maximum acoustic intensity (indentified by the subscript to the no-response [NR] entries) at some frequencies. In these cases both the response threshold and no-response entries were symmetric across ears and the hearing loss was considered symmetric. For example, the thresholds measured for 250 and 500 Hz in case 8 were within 10 dB across ears and no response was elicited in either ear for frequencies 1,000 to 8,000 Hz for acoustic intensities up to 100 dB (NR100).
Table 3 gives the segmental and total SGC counts for each subject’s right and left ears. The range of SGC counts across subjects for both segmental and total counts is large. In Case 16, not a single SGC was identified in any section of either ear, but 18,108 SGCs were counted in the left ear of case 6. Even in cases impaired by the same etiology, the number of SGCs may differ significantly across subjects. Case 2 and 21 both suffered an impairment of recent origin from neomycin ototoxicity, but the across-subject difference in the total SGC count was more than 5000 for both the left and right ears. It should be mentioned that the counts represent the number of SGCs at the time of death which was sometimes long after the last audiological measurements. For example this period for case 11 was 25 years but in case 16, in which there were no SGCs at the time of the patient’s death, the hearing threshold measured 65 dB at 500 Hz one and half years before death. Interestingly, in case 13, the number of SGCs was fewer on the side in which the cochlea had 1.5 turns where 20 SGCs were counted whereas the other side with 1 turn had 2190 SGCs.
Mean right and left SGC counts (with standard deviations) are plotted by segment in Figure 2. The means for the right and left ear in each segment and for the total were very similar. In general, the segment II counts tend to be higher than the other segment counts.
Table 4 lists the differences between the left and right SGC counts (segmental and total) for each subject. The differences within a column tend to be evenly distributed between negative (right count > left count) and positive (left count > right count), and the column means are relatively small suggesting that right-ear counts are not consistently larger or smaller than left-ear counts. The results in Table 5 show that the mean difference in SGC count associated with each column of Table 4 is not significantly different than zero (all p-values>0.05). This is consistent with the left-ear and right-ear counts being similar within a subject as can be seen in the scatter plots with regression lines plotted in Figure 3.
Analyzing the results by whether or not the subject’s hearing status was documented audiologically did not reveal significant differences between the two subpopulations for total SGC count or for the counts made for segments II–IV. For example, the mean magnitude of the total left-right count differences for the subjects with audiological data (555 SGCs) was not significantly different from the mean for subjects without audiological measures (692 SGCs; t=−0.57, df=17, p=0.57). In the case of segment I (cochlear base), however, the mean magnitude of the left-right count difference was significantly greater (t=2.89; df=14; p=−.01) for subjects without audiological documentation (481 SGCs) than for subjects with documentation (199 SGCs).
The correlations between left-ear and right-ear SGC counts are high (R2>0.89) and very significant (all p-values < 0.0001) for the total count and all segmental counts except for segment I. The coefficient of determination (R2 =0.645) for the segment I counts is lower but significant (p<0.0001) and the range of counts across subjects is considerably smaller than the other segments (see Figure 3). When the results for subjects with and without audiological documentation were analyzed separately, the high (R2>0.89) and significant (p<0.0001) correlations found in the combined-subjects analyses were maintained for each subgroup for total count and segmental counts II–IV. In the case of segment I, the correlation between the left-ear and right-ear counts was substantially weaker for the subgroup without audiological documentation (R2=0.39; p=0.038) than the subgroup with documentation (R2 =0.90; p<0.0001). When the segmental counts are combined to produce a total count and the subject subgroups combined, the correlation between the right and left ears is especially strong with the variance in one ear accounting for 98% of the count variance found in the opposite ear.
The relatively small mean differences and high correlation between the left and right SGC counts are both consistent with a research strategy using one ear as a control for the other. However, the relatively large standard deviations listed in Table 5 for the differences and the degree to which some individual points diverge from the diagonal lines in Figure 3 emphasize the desirability of using statistical power analysis to guide the selection of sample sizes. Figure 4 illustrates how to evaluate the statistical power (sensitivity) implied by the results using plots of minimum effect size (minimum mean difference between the SGC counts of the control and treatment ears required to reach significance) as a function of sample size (number of subjects for whom right and left temporal bones are available) for a set of standard parameters.
Consider, for example, the bottom panel of Figure 4 that represents computations based on our data for total SGC counts. In this panel: the error variance was estimated by the standard deviation of 845 listed in Table 5; α=0.05 (a 5% chance that differences greater than the minimum effect size are due to chance); and power=0.95 (95% likelihood a real difference will be recognized as significant). Thus if we have a sample of 10 temporal bone pairs, moving vertically from the x-axis at the 10-subjects position to the filled circle and then horizontally, we intersect the y-axis just above 1000 SGCs. This means that a treatment effect (difference between the SGC counts for the control and treatment ears) greater than 1000 is very likely to be detected and found to be significant in a sample of 10 subjects drawn from a population like the one represented by this study.
The results presented for the 21 subjects with bilaterally symmetric profound hearing loss indicate that the within-subject between-ear differences in SGC counts were relatively small. An example set of power analysis calculations lead to the conclusion that realistic sample sizes (e.g., 10 to 20) are sufficient to reliably detect treatment effects on the order of 1,000 SGCs in temporal bone populations similar the one studied here.
One question that arises when these results are to be applied to a different set of subjects satisfying the same selection criteria used in this study is whether our population is representative of the general population of bilaterally symmetric profoundly impaired individuals. If it is, then using the statistics associated with this study’s population will likely be appropriate when predicting the power of a different set of subjects selected using the same criteria.
Figure 5 plots the distribution of total SGC counts made in this study together with counts made in other laboratories for normal and profoundly hearing-impaired subjects. It is not surprising that the SGC counts reported by Guild (1932) and Hinojosa and Marion (1983) for normal-hearing subjects are typically larger than those for the hearing-impaired subjects. The distribution of counts reported by Hinojosa and Marion (1983) and by Fayad and Linthicum (2006) for subjects with profound hearing loss are very similar and do not differ from one another statistically (t=1.02, df=27, p=0.32) even though the Fayad and Linthicum (2006) subjects underwent cochlear implantation. The maximum count for the left and right ears of the current study are consistently lower than the Hinojosa and Marion (1983) and Fayad and Linthicum (2006) populations and result in a significant difference between our population and the Hinojosa and Marion (1983) population (e.g., t=−3.37, df=34, p=0.002).
Our working hypothesis is that the difference between our population and the populations of Fayad and Linthicum (2006) and Hinojosa and Marion (1983) are real and not an artifact of different methods. Each of the three studies used the same standard techniques of temporal bone processing developed by Schuknecht (Schuknecht, 1968; Schuknecht, 1953). One difference between the current study and the studies of Fayad and Linthicum (2006) and Hinojosa and Marion (1983) is that we counted cells with a visible nucleus and they counted cells with a visible nucleolus. Because of relative large size of the nucleus as compared to the nucleolus, we used a correction factor of 0.68 compared to the 0.90 correction factor for the nucleolus. When one considers the rationale and theory associated with the two correction factors and the evidence for their validity (Konigsmark, 1970; Nadol, 1988), we see no reason to suspect that the difference in the distribution of total SGC counts between the studies is due to methodological differences. Thus, it is possible that our population of bilaterally symmetric profoundly impaired subjects under-represent subjects with larger numbers of SGCs. This would be especially troubling if the within-subject across-ear count difference tended to increase with increasing number of SGCs. This is not the case in our data set. In the case of total SGC count, for instance, the linear regression slope for between-ear difference versus right-ear SGC count is −0.0033 and not significantly different from zero (p=0.90), Thus, it is unlikely that the difference between our results and results from a population more like Hinojosa and Marion (1983) would alter the standard deviations associated with the counts shown in Table 5 and, therefore, the power calculations illustrated in Figure 4.
As shown in Table 4, the greatest difference in total count between ears was 2170 in case 13 (CHARGE syndrome) with a Mondini dysplasia in both ears. Although CHARGE syndrome (Coloboma of the eye, Heart defect, Atresia of the choanae, Retardation of growth or development, Genital hypoplasia, Ear malformation) is primarily associated with cochlear hypoplasia (Arndt et al., 2010), other concomitant asymmetric malformations such as hypoplasia and aplasia of the VIII th nerve, dysplasia of the internal auditory canal and oval window atresia have been described in these patients (Amiel et al., 2001; Arndt et al., 2010). This relatively large difference between right and left ears possibly could be explained by asymmetric hypoplasia of auditory nerve.
The coefficient of determination for segment I counts (0.64) was substantially lower than the other segments (all greater than 0.90). This lower correlation in segment I may be explained by the SGCs of segment I being more susceptible to bilaterally nonsymmetric insults like acoustic or other trauma in addition to the major etiology listed in Table 1.
One finding illustrates a challenge for ear selection for cochlear implantation. In case 13 with Mondini dysplasia associated with CHARGE syndrome, the left cochlea is 1.5 turns and the right less than 1 turn. It is interesting that the side with the shorter cochlea has 2130 more SGCs than the side with the longer cochlea. If this patient had been considered for cochlear implantation, typically, the longer cochlea (by CT scan) would be selected; not the ear with the larger SGC count.
Histological examination of 21 pairs of temporal bones from bilateral-symmetric profoundly-impaired ears found relatively small within-subject between-ear differences in counts of spiral ganglion cells. Results also suggest that realistic sample sizes (e.g., 10 to 20 subjects) are sufficient to reliably detect unilateral treatment effects on the order of 1,000 SGCs in temporal bone populations similar to the one studied when one ear is used as a matched control for the treated ear.
This work was supported by grant R01-DC00152 from the National Institute of Deafness and Other Communication Disorders.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.