Noise and Hearing
The Relative Contributions of Occupational Noise and Aging in Individual Cases of Hearing Loss Robert A. Dobie, MD Department of Otolaryngology-Head and Neck Surgery, The University of Texas Health Science Center At San Antonio, San Antonio, Texas
It is the contention of this article that a noise-exposed individual’s total hearing loss should almost always be treated as the sum of at least two components: noiseinduced permanent threshold shift (NIPTS) and agerelated permanent threshold shift (ARPTS). If HTL refers to the hearing threshold level for a given frequency or pure-tone average: HTL = NIPTS
ABSTRACT A method is proposed for allocation of hearing handicap between noise and aging in individual cases when noise exposure level and duration are known or can be estimated. A recently published international standard (ISO1999,1990) provides statistical models for hearing threshold changes associated with aging and noise exposure. When an individual’s hearing threshold level exceeds the sum of the median levels expected given that individual’s age, gender, exposure level, and exposure duration, the appropriate allocation depends on the correlation between age-related and noise-induced changes. However, the differences in allocation between assumptions of perfect and absent correlation are small. Only very small errors result from calculating the allocation based on median expectations for noise and aging. In most cases, agerelated changes exceed noise-induced changes for the 0.5, 1, 2, and 3 kHz pure-tone average; for men age 65, this is true for all exposure levels below 100 dBA. (Ear Hear 13 1:19-27).
IN MEDICAL-LEGAL settings, otolaryngologists are often asked to determine, on a “more probable than not” basis, whether an individual has noise-induced hearing loss (NIHL). When the history clearly indicates potentially hazardous noise exposure, audiometric findings are “consistent with NIHL,” and the workup is otherwise negative, the diagnosis of NIHL is usually made with little if any consideration of whether the severity of the hearing loss is proportionate to the exposure level and duration. The contribution of aging to the total hearing loss is usually not considered unless attorneys or workers’ compensation boards specifically request it. Ear and Hearing, Vol. 13, No. 1, 1992
+ ARPTS
(1)
Several major demographic studies of NIHL (summarized by Johnson, 1978) have yielded estimates of NIPTS for different exposure levels and durations. Similarly, ARPTS has been estimated for both men and women as a function of age (for example, Robinson, 1988; Robinson & Sutton, 1979). For both NIPTS and ARPTS, both median and extreme (10th and 90th percentile) values are available. Thus, it is possible to estimate the total hearing loss to be expected for a population of workers of specified age and gender, exposed at specified levels and durations, by simply combining appropriate distributions for NIPTS and ARPTS (Johnson, 1988): HTLo.1 = NIPTSo., + ARPTSo,, (least susceptible lo%)* (2) HTLo.5 = NIPTSo.5+ ARPT& (median)
(3)
HTL0.9 = NIPTSo.9+ ARPTSo,9(most susceptible 10%) (4) When exposure level and duration are known, one can use these relationships to estimate the relative contributions of NIPTS and ARPTS for an individual. The definitions stated here are opposite to those used by Johnson (1988) and by the International Standards Organization (1990), both of whom define the 10th percentile as the group showing the greatest loss. I believe my usage corresponds better to the generally accepted understanding of percentile: ”a value on a scale of one hundred that indicates the percent of a distribution that is equal to or below it” (Webster’s New Collegiate Dictionary, 1973). The choice is arbitrary and does not affect the arguments in this article; if one substitutes 10th for 90th percentile, and vice versa, the article corresponds to ISO-1999 usage.
.
0196/0202/92/1301-0019$03.00/0 EARAND HEARING Copyright 0 1992 by Williams & Wilkins Printed in the U S A .
Noise and Aging
19
However, there are two reasons why we should first consider the case where exposure level is unknown: first, this is commonly the case; and second, this discussion will highlight the issues involved in allocation of hearing loss (HL) and hearing handicap (HH). METHOD
proportion of HL allocable to aging for that ear: HHA -- HHB(HLB- ARpTs) HLB +’ HHw(HLw
- ARPTS HLW
)
(8)
One set of commonly used formulae for ARPTS (Robinson & Sutton, 1979) is based on population studies from which
Exposure Levels Unknown A previous paper (Dobie, 1990) addressed the case where hearing loss had accrued in a particular interval, during which some potentially noxious event occurred (noise exposure, head injury, etc.). Even when the severity of loss attributable to the noxious event is unknown, it is fair to assume some ARPTS also occurred and was additive; the median (50th percentile) ARPTS is the most reasonable estimate in this circumstance. For example, if a noise-exposed worker (level unknown) has a 40 dB threshold, and would be expected based on age and gender to have a 20 dB threshold in the absence of noise exposure (50th percentile ARPTS), it is reasonable to conclude that half the total loss is due to occupational noise and half to aging. As Lebo and Reddell (1972) pointed out, allocation of HH is not so straightforward. Recall that the widely used AMA/ AAO-HNS (1979) HH method uses the pure-tone average of 0.5, 1, 2, and 3 kHz, with a “low fence” of 25 dB HL and a growth rate of 1.5% handicap per dB HL above 25 dB. If one simply “age-corrects” the audiogram, unfair conclusions result. A worker with a binaurally symmetrical 30 dB pure-tone average with an expectation of 15 dB from aging alone has a HH of 1.5 (30-25) = 7.5%. Age-correcting the audiogram yields a 15 dB pure-tone average with a HH of 0, and one would falsely conclude that the noise exposure was not responsible for any handicap. It would be equally illogical (and unfair) to conclude that because no handicap would have been present in the absence of the noise exposure, the noise exposure was responsible for all of the handicap. Lebo and Reddell’s method is reasonable: first allocate HL (in this case, equally between NIPTS and ARPTS), then apply the same proportions to HH (in this case, 3.75% NIPTS and 3.75% ARPTS = 7.5% total hearing handicap). Lebo and Reddell Qd not address the case of asymmetrical hearing loss. Because the AMA/AAO-HNS (1979) method gives the better ear 5 times the weight of the poorer ear, the contribution of each ear to the total HH must be considered. When hearing is symmetrical, each ear contributes equally. When one ear is worse than the other, its contribution (HHw) is larger than that of the better ear (HHB)by an amount equal to 1/6 the interaural difference (Dobie, 1990): HH
=
HHB + HHw
5MIB + MIw =
6
2
+
6
(7)
where MI = monaural impairment. The proportion of the overall HH allocable to aging, then, is obtained by multiplying each ear’s share of the HH by the 20
Dobie
Exposure Level Known The International Standards Organization (1990) has recently published a new standard, ISO- 1999, which provides formulae relating NIPTS to exposure level (85- 100 dBA), duration (0-40 yr), and percentile (5th to 95th). Furthermore, ISO-1999 gives two examples of ARPTS databases: their database A is identical to Robinson and Sutton (1979), r e p resenting a “highly screened” population, whereas their database B represents an unscreened population, some of whom probably had “unreported occupational or other noise exposure.” Finally, ISO- 1999 accepts an additive relationship between NIPTS and ARPTS, with a small correction, which becomes significant only when the simple sum (NIPTS + ARPTS) is greater than 40 dB: HTL
= NIPTS
ARPTS) + ARPTS - (NIPTS)( 120
(9)
Figures 1 and 2 show loth, 50th, and 90th percentile ARPTS and NIPTS, respectively, for the 0.5, 1, 2, and 3 kHz puretone average, according to ISO-1999, for specified values of gender, age, exposure level, and exposure duration. Corso (1980) suggested that, given median values for ARPTS and NIPTS, one could very simply allocate HTL for an ear, as follows (let the subscript i represent values for an individual patient, and the subscript 0.5 represents median population values):
(5)
MIe MIw - MIB HHw=
individuals with occupational and nonoccupational noise exposure and/or known otological disease have been excluded. yowever, they cannot be taken to reflect “pure” aging changes. Ward (1984, for example) suggests that ARPTS in modem society is itself really the sum of several components: (1) presbycusis (genetically determined), (2) sociocusis (wear and tear from “normal” sound levels), and (3) nosoacusis (effects of undiagnosed acquired diseases on the ear). It is important to recognize that Ward’s concept of nosoacusis encompasses whatever effects conditions such as diabetes, hypertension, hyperlipidemia, smoking, etc., may (or may not) have on hearing, because patients with these conditions were not excluded from the population surveys used to determine ARPTS.
If, at the median, ARPTS represented 25% of the total HTL, Corso would have allocated that proportion to aging for all individuals with the same age, gender, and exposure, regardless of severity of total loss (HTL). This approach makes intuitive sense at the median. Assume that a worker has an HTL of 40 dB and that median values for his age, exposure level, and exposure duration are 15 (NIPTS) and 25 dB (ARPTS). It seems appropriate to conclude that 25/40 = 62.5% of his HTL is allocable to ARPTS. Ear and Hearing, Vol. 13, No. 1, 1992
adding values for A and N which are equally distant from their means. Although equation 11 still holds, it is now the standard deviations, rather than the variances, which are additive: =
+
(13) In general, aH2will depend on the degree of correlation ( I ) between A and N: OH2
OH2
20
30
40
50
60
70
AGE (years)
eercenflle
20
30
40 50 60 AGE (years)
70
Figure 1. ARPTS as a function of age and gender, for the 0.5, 1, 2, and 3 kHz pure-tone average, for highly screened (a) and lessscreened (b) populations. In part b, the values stated in 180-1999 (+ and 0)have been fitted to quadratic functions. However, if another worker with identical age and exposure history has an HTL of 60 dB, it is not immediately clear that one should apply the same allocation (62.5%) to this loss, as Corso’s method suggests. On the contrary, the appropriate allocation must take into account not only the median values for NIPTS and ARPTS, but also their variances: if NIPTS, for example, had a very small variance compared to ARPTS, one might conclude that almost all the additional loss suffered by this second worker was due to ARPTS. At this point, it is necessary to consider the ways in which variances add, and the method which has been used to estimate the distributions of NIPTS. Consider two normally distributed variables A and N, with means pA and pN and variances uA2 and U N ~If. a third variable H is created by adding randomly selected values of A and N, the new variable’s distribution can be simply characterized in terms of the other two distributions (A, N, and H will be used interchangeably with ARPTS, NIPTS, and HTL to simplify mathematical expressions; ISO- 1999 actually models each of these variables as the sum of two half-normal distributions, but this does not affect the analysis that follows): pH (TH2
= PA =
(TA2
+ PN
(1 1)
+ ON2
(12) In the above example, A and N are uncorrelated: in creating H, a high value for A is as likely to be added to a low value for N as to a high value for N. Conversely, if A and N are perfectly correlated, a member of H is always created by Ear and Hearing, Vol. 13, No. 1,1992
=
(TA2
((TA
+
ON2
UN)’
+ 2r(TA(TN
(14)
Figure 3 shows the effect of varying correlation on hypothetical distributions of A, N, and H. These distributions are plotted on Gaussian coordinates, with percentile as the horizontal axis. The horizontal axis could equivalently be relabeled with evenly spaced z scores (number of standard deviations above or below the mean). On Gaussian coordinates, a normal distribution is a straight line whose slope is equal to the standard deviation. Note that in this example, in which A and N have equal slopes, the mean of H is unaffected by r, but the slopes vary from zero ( r = - 1) to twice the slope of A and N ( I = 1). Consider now the distribution of NIPTS. Both HTL (=NIPTS + ARPTS) and ARPTS have been measured, and their distributions can be estimated from these measurements. However, NIPTS can only be derived by subtraction of threshold shifts in non-noise-exposed (ARPTS) from noise-exposed (HTL) populations. As implied in equations 2 to 4 in the introduction, extreme values (10th and 90th percentiles) for NIPTS have been derived by subtraction of corresponding percentile values in the HTL and ARPTS distributions. If it is true that
then it must also be true that (16) However, this would require perfect correlation (r = 1) between NIPTS and ARPTS (see equation 13), a condition which seems quite unlikely. Indeed, it is appropriate to view the NIPTS data presented in ISO-1999 and elsewhere not as actual estimates of the distribution of NIPTS in individuals, but as distributions which when added to ARPTS distributions, yield good estimates of HTL distributions. Figure 4 presents this discussion graphically. Assume we measure distributions of H and A as shown, and must estimate the distribution of N. If r = 1, the mean and standard deviation of N can be simply obtained by subtraction. This is how NIPTS distributions have been obtained for ISO-1999 and other sources. However, if r = 0, the true standard deviation for N, based on equation 12, would be: ONIPTS
= UHTL -
UARF‘TS-
(TH = (17) Figure 4 shows the range of slopes possible for N for r between -1 and 1, in this example. What is the correlation between NIPTS and ARPTS? Although this is clearly unknown, it seems more likely to be positive than negative. If ARPTS includes “sociocusis,” due to sounds of everyday life, susceptibility to this component of ARPTS should be correlated with susceptibility to NIPTS. In addition, it seems likely that because at least the major site of action for both NIPTS and ARPTS is the cochlear hair cell,
Noise and Aging
21
b.
L
23
2.5
P 3 2 cn I-
2 1.5 N
9 1
0
0
10 20 30 DURATION (years)
0
40
10
20 30 DURATION (years)
40
d.
b
Percentlle
-
35
20
30
P
25
2 20
If
9
/j B
10 5
0
0
0
10
20 30 DURATION (ywrs)
40
10 20 30 DURATION (years)
0
40
Figure 2. NIPTS as a function of exposure duration for the 0.5, 1, 2, and 3 kHz pure-tone average. The parameter is exposure level (dBA). NIPTS values for 0 to 10 yr exposure have been estimated by a logarithmic function, according to SO-1999.
80 70
80 70
I
60
60
g
,
50
50
40
40
30 20 10
30
20
0
10
-1 0
-20
0
10
50
90
10
50
Figure 3. The effect of varying correlation between two random variables, A and N, when a new random variable H is formed by adding values from A and N. The values of A, N, and H are plotted as a function of percentile on Gaussian coordinates.
susceptibility may be shared as well. In the absence of any firm basis for estimating a degree of correlation, it seems most reasonable to assume a value between 0 and 1. As shown in the Appendix, the best estimates of the relative contributions of ARPTS and NIPTS are different, depending on the correlation between them. If r = 1, and H = A + N, 22
Dobie
90
Percenllle
Percenllle
Figure 4. The effect of varying correlation between A and N when the distributions of H and A are known, but the distribution of N must be derived (Gaussian coordinates).
where H is now a constant, A=pA+
(H - PA - P N ) ~ A UA
N=pN+
+ UN
(H - PA - P UA
+ ON
N ~ N
(18)
(19)
Ear and Hearing, Vol. 13, No. 1, 1992
Conversely, if r = 0,
(21) The population standard deviations can be obtained from the tables and formulae of ISO- 1999, which models NIPTS (and ARPTS) as two half-normal distributions, with slightly different variances for the upper and lower halves. For example, in the upper half of the ARPTS distribution, one can obtain uA from the 50th and 90th percentile values (recall that the z value for the 90th percentile is 1.282): UA
=
b.9
-b.5 -
1.282
AA
1.282
For r = 1, uN is similarly calculated:
For r = 0, UN is obtained by expanding equation 17: 1 1.282 (AA + AN)^
U N = - J
- (Ad2
(24)
The results of typical allocations using equations 18 to 24 are shown in Table 1. The examples include permutations of two exposure levels ( 100 and 90 dBA), two exposure durations (40 and 20 yr), and two ages (65 and 4 9 , using database A (male) values for ARPTS. Table 1 shows the percentage of HTL allocableto noise for three conditions for each exposureage combination: HTL = NIPTSo.5 + ARPTSo,s.When the individual’s HTL is equal to the median expected HTL, allocation is straightforwardand requires no assumption regarding correlation between NIPTS and ARPTS. HTL = NIPTSO.~ARPTSo.9, r = 1. When an individual’s HTL is above the median and we assume r = 1, allocation uses equations 18, 19, 22, and 23. HTL = NIF’TSo.9 = ARPT!h, r = 0. With HTL above the median and r = 0, we use equations 20,2 1,22, and 24... In every case, 90th percentile calculations assuming r = 0 result in a larger allocation for NIPTS than when r = 1 is assumed, although the differences in percent allocable to NIPTS are rather small (6-13%). In cases where HTL was closer to the 50th percentile, the difference would be even smaller approaching zero as HTL approaches the median. For four of the six examples, the 50th percentile allocation is between the two 90th percentile values. For the other two examples, the 50th percentile noise allocation is larger than either 90th percentile allocation. Because we don’t know r,
+
but expect it to be between 0 and 1, an attractive simplication suggests itself: use Corso’s method, allocating HTL according to the proportions of NIPTS and ARPTS present at the median, regardless of the actual HTL (as long as it is between the 5th and 95th percentiles, where ISO-1999 purports to be valid). To do so would be very unlikely to underestimate the true noise allocation; for two of the examples in Table 1 (those involving a 45-yr-old man), Corso’s method is likely to slightly overestimate the proportion of HTL due to NIPTS. These options will be illustrated in detail in the following case study. CASE STUDY
A 60-yr-old man retires after 30 yr in a workplace where his daily exposure was 90 dBA; use of hearing protection was negligible. He has had no significant nonoccupational noise exposure, and no otological cause can be found for his hearing loss other than noise and aging. His retirement audiogram (Fig. 5) shows a symmetrical sensorineural hearing loss consistent with this combination, with pure-tone average HTLs (0.5, 1, 2, and 3 kHz) of 30 dB in each ear. No previous audiograms are available. The loth, 50th, and 90th percentiles for NIPTS, given 30 yr at 90 dBA, are taken from Figure 2b: 2.9, 4.1, and 6.7 dB. The equivalent figures for ARPTS, from Figure la, are -0.8, 11.4, and 26.7 dB. The median expectation for HTL, given this combination of age, gender, exposure level, and exposure duration, would be 15.5 dB, of which 26% (4.1 t 15.5) would be allocable to NIPTS and 74% (1 1.4 + 15.5) to ARPTS. Corso’s method would apply the same percentage allocation to this case, although the actual HTL (30 dB) is nearly twice the median. If we assume r = 1, we proceed as follows, using the 50th and 90th percentile ISO- 1999 distribution data: bA
=
(JN =
26.7 - 11.4 = 11.93 1.282 6.7 - 4.1 = 2.03 1.282
~
A = 11.4 +
N = 4.1
Level Duration Age 50%ile 90%ile (r = 1) 90%ile (r = 0)
100 90 90
90
65 65 45 65 65 45
58% 52% 77% 24% 20% 43%
Ear and Hearing, Vol. 13, No. 1,1992
5170 46% 62% 19% 16% 27%
(27)
11.4 - 4.1)2.03 + (30 -11.93 + 2.03 = 6.21
(28)
142.4 (26.7 - 1 1.4 6.7 - 4. 1)2 - (26.7 - 1 1.4)2 = 52.5 bN2 = ( 1.282)2 =
(29)
+
Table 1. Percent of HTL allocable to noise.
40 20 20 40 20 20
+
Thus, for r = 1, we would allocate 21% of the HTL to NIPTS and 79% (23.79 f 30) to ARPTS. If r = 0, using the same ISO-1999 data, we calculate: (iA2
100 100
(30 - 11.4 - 4.1)11.93 = 23.79 11.93 2.03
62% 5w0 75% 25% 22% 36%
A=
(30)
(1 1.4)(52.5)+ (30 - 4. I)( 142.4) =22.0 142.4 52.5
(31)
4. I)( 142.4) + (30 - 11.4)(52.5) = 8.0 142.4 + 52.5
(32)
Noise and Aging
23
N=(
+
01
10 20
11
I
%-
....)
The Relative Contributions of Occupational Noise and Aging in Individual Cases of Hearing Loss Robert A. Dobie, MD Department of Otolaryngology-Head and Neck Surgery, The University of Texas Health Science Center At San Antonio, San Antonio, Texas
It is the contention of this article that a noise-exposed individual’s total hearing loss should almost always be treated as the sum of at least two components: noiseinduced permanent threshold shift (NIPTS) and agerelated permanent threshold shift (ARPTS). If HTL refers to the hearing threshold level for a given frequency or pure-tone average: HTL = NIPTS
ABSTRACT A method is proposed for allocation of hearing handicap between noise and aging in individual cases when noise exposure level and duration are known or can be estimated. A recently published international standard (ISO1999,1990) provides statistical models for hearing threshold changes associated with aging and noise exposure. When an individual’s hearing threshold level exceeds the sum of the median levels expected given that individual’s age, gender, exposure level, and exposure duration, the appropriate allocation depends on the correlation between age-related and noise-induced changes. However, the differences in allocation between assumptions of perfect and absent correlation are small. Only very small errors result from calculating the allocation based on median expectations for noise and aging. In most cases, agerelated changes exceed noise-induced changes for the 0.5, 1, 2, and 3 kHz pure-tone average; for men age 65, this is true for all exposure levels below 100 dBA. (Ear Hear 13 1:19-27).
IN MEDICAL-LEGAL settings, otolaryngologists are often asked to determine, on a “more probable than not” basis, whether an individual has noise-induced hearing loss (NIHL). When the history clearly indicates potentially hazardous noise exposure, audiometric findings are “consistent with NIHL,” and the workup is otherwise negative, the diagnosis of NIHL is usually made with little if any consideration of whether the severity of the hearing loss is proportionate to the exposure level and duration. The contribution of aging to the total hearing loss is usually not considered unless attorneys or workers’ compensation boards specifically request it. Ear and Hearing, Vol. 13, No. 1, 1992
+ ARPTS
(1)
Several major demographic studies of NIHL (summarized by Johnson, 1978) have yielded estimates of NIPTS for different exposure levels and durations. Similarly, ARPTS has been estimated for both men and women as a function of age (for example, Robinson, 1988; Robinson & Sutton, 1979). For both NIPTS and ARPTS, both median and extreme (10th and 90th percentile) values are available. Thus, it is possible to estimate the total hearing loss to be expected for a population of workers of specified age and gender, exposed at specified levels and durations, by simply combining appropriate distributions for NIPTS and ARPTS (Johnson, 1988): HTLo.1 = NIPTSo., + ARPTSo,, (least susceptible lo%)* (2) HTLo.5 = NIPTSo.5+ ARPT& (median)
(3)
HTL0.9 = NIPTSo.9+ ARPTSo,9(most susceptible 10%) (4) When exposure level and duration are known, one can use these relationships to estimate the relative contributions of NIPTS and ARPTS for an individual. The definitions stated here are opposite to those used by Johnson (1988) and by the International Standards Organization (1990), both of whom define the 10th percentile as the group showing the greatest loss. I believe my usage corresponds better to the generally accepted understanding of percentile: ”a value on a scale of one hundred that indicates the percent of a distribution that is equal to or below it” (Webster’s New Collegiate Dictionary, 1973). The choice is arbitrary and does not affect the arguments in this article; if one substitutes 10th for 90th percentile, and vice versa, the article corresponds to ISO-1999 usage.
.
0196/0202/92/1301-0019$03.00/0 EARAND HEARING Copyright 0 1992 by Williams & Wilkins Printed in the U S A .
Noise and Aging
19
However, there are two reasons why we should first consider the case where exposure level is unknown: first, this is commonly the case; and second, this discussion will highlight the issues involved in allocation of hearing loss (HL) and hearing handicap (HH). METHOD
proportion of HL allocable to aging for that ear: HHA -- HHB(HLB- ARpTs) HLB +’ HHw(HLw
- ARPTS HLW
)
(8)
One set of commonly used formulae for ARPTS (Robinson & Sutton, 1979) is based on population studies from which
Exposure Levels Unknown A previous paper (Dobie, 1990) addressed the case where hearing loss had accrued in a particular interval, during which some potentially noxious event occurred (noise exposure, head injury, etc.). Even when the severity of loss attributable to the noxious event is unknown, it is fair to assume some ARPTS also occurred and was additive; the median (50th percentile) ARPTS is the most reasonable estimate in this circumstance. For example, if a noise-exposed worker (level unknown) has a 40 dB threshold, and would be expected based on age and gender to have a 20 dB threshold in the absence of noise exposure (50th percentile ARPTS), it is reasonable to conclude that half the total loss is due to occupational noise and half to aging. As Lebo and Reddell (1972) pointed out, allocation of HH is not so straightforward. Recall that the widely used AMA/ AAO-HNS (1979) HH method uses the pure-tone average of 0.5, 1, 2, and 3 kHz, with a “low fence” of 25 dB HL and a growth rate of 1.5% handicap per dB HL above 25 dB. If one simply “age-corrects” the audiogram, unfair conclusions result. A worker with a binaurally symmetrical 30 dB pure-tone average with an expectation of 15 dB from aging alone has a HH of 1.5 (30-25) = 7.5%. Age-correcting the audiogram yields a 15 dB pure-tone average with a HH of 0, and one would falsely conclude that the noise exposure was not responsible for any handicap. It would be equally illogical (and unfair) to conclude that because no handicap would have been present in the absence of the noise exposure, the noise exposure was responsible for all of the handicap. Lebo and Reddell’s method is reasonable: first allocate HL (in this case, equally between NIPTS and ARPTS), then apply the same proportions to HH (in this case, 3.75% NIPTS and 3.75% ARPTS = 7.5% total hearing handicap). Lebo and Reddell Qd not address the case of asymmetrical hearing loss. Because the AMA/AAO-HNS (1979) method gives the better ear 5 times the weight of the poorer ear, the contribution of each ear to the total HH must be considered. When hearing is symmetrical, each ear contributes equally. When one ear is worse than the other, its contribution (HHw) is larger than that of the better ear (HHB)by an amount equal to 1/6 the interaural difference (Dobie, 1990): HH
=
HHB + HHw
5MIB + MIw =
6
2
+
6
(7)
where MI = monaural impairment. The proportion of the overall HH allocable to aging, then, is obtained by multiplying each ear’s share of the HH by the 20
Dobie
Exposure Level Known The International Standards Organization (1990) has recently published a new standard, ISO- 1999, which provides formulae relating NIPTS to exposure level (85- 100 dBA), duration (0-40 yr), and percentile (5th to 95th). Furthermore, ISO-1999 gives two examples of ARPTS databases: their database A is identical to Robinson and Sutton (1979), r e p resenting a “highly screened” population, whereas their database B represents an unscreened population, some of whom probably had “unreported occupational or other noise exposure.” Finally, ISO- 1999 accepts an additive relationship between NIPTS and ARPTS, with a small correction, which becomes significant only when the simple sum (NIPTS + ARPTS) is greater than 40 dB: HTL
= NIPTS
ARPTS) + ARPTS - (NIPTS)( 120
(9)
Figures 1 and 2 show loth, 50th, and 90th percentile ARPTS and NIPTS, respectively, for the 0.5, 1, 2, and 3 kHz puretone average, according to ISO-1999, for specified values of gender, age, exposure level, and exposure duration. Corso (1980) suggested that, given median values for ARPTS and NIPTS, one could very simply allocate HTL for an ear, as follows (let the subscript i represent values for an individual patient, and the subscript 0.5 represents median population values):
(5)
MIe MIw - MIB HHw=
individuals with occupational and nonoccupational noise exposure and/or known otological disease have been excluded. yowever, they cannot be taken to reflect “pure” aging changes. Ward (1984, for example) suggests that ARPTS in modem society is itself really the sum of several components: (1) presbycusis (genetically determined), (2) sociocusis (wear and tear from “normal” sound levels), and (3) nosoacusis (effects of undiagnosed acquired diseases on the ear). It is important to recognize that Ward’s concept of nosoacusis encompasses whatever effects conditions such as diabetes, hypertension, hyperlipidemia, smoking, etc., may (or may not) have on hearing, because patients with these conditions were not excluded from the population surveys used to determine ARPTS.
If, at the median, ARPTS represented 25% of the total HTL, Corso would have allocated that proportion to aging for all individuals with the same age, gender, and exposure, regardless of severity of total loss (HTL). This approach makes intuitive sense at the median. Assume that a worker has an HTL of 40 dB and that median values for his age, exposure level, and exposure duration are 15 (NIPTS) and 25 dB (ARPTS). It seems appropriate to conclude that 25/40 = 62.5% of his HTL is allocable to ARPTS. Ear and Hearing, Vol. 13, No. 1, 1992
adding values for A and N which are equally distant from their means. Although equation 11 still holds, it is now the standard deviations, rather than the variances, which are additive: =
+
(13) In general, aH2will depend on the degree of correlation ( I ) between A and N: OH2
OH2
20
30
40
50
60
70
AGE (years)
eercenflle
20
30
40 50 60 AGE (years)
70
Figure 1. ARPTS as a function of age and gender, for the 0.5, 1, 2, and 3 kHz pure-tone average, for highly screened (a) and lessscreened (b) populations. In part b, the values stated in 180-1999 (+ and 0)have been fitted to quadratic functions. However, if another worker with identical age and exposure history has an HTL of 60 dB, it is not immediately clear that one should apply the same allocation (62.5%) to this loss, as Corso’s method suggests. On the contrary, the appropriate allocation must take into account not only the median values for NIPTS and ARPTS, but also their variances: if NIPTS, for example, had a very small variance compared to ARPTS, one might conclude that almost all the additional loss suffered by this second worker was due to ARPTS. At this point, it is necessary to consider the ways in which variances add, and the method which has been used to estimate the distributions of NIPTS. Consider two normally distributed variables A and N, with means pA and pN and variances uA2 and U N ~If. a third variable H is created by adding randomly selected values of A and N, the new variable’s distribution can be simply characterized in terms of the other two distributions (A, N, and H will be used interchangeably with ARPTS, NIPTS, and HTL to simplify mathematical expressions; ISO- 1999 actually models each of these variables as the sum of two half-normal distributions, but this does not affect the analysis that follows): pH (TH2
= PA =
(TA2
+ PN
(1 1)
+ ON2
(12) In the above example, A and N are uncorrelated: in creating H, a high value for A is as likely to be added to a low value for N as to a high value for N. Conversely, if A and N are perfectly correlated, a member of H is always created by Ear and Hearing, Vol. 13, No. 1,1992
=
(TA2
((TA
+
ON2
UN)’
+ 2r(TA(TN
(14)
Figure 3 shows the effect of varying correlation on hypothetical distributions of A, N, and H. These distributions are plotted on Gaussian coordinates, with percentile as the horizontal axis. The horizontal axis could equivalently be relabeled with evenly spaced z scores (number of standard deviations above or below the mean). On Gaussian coordinates, a normal distribution is a straight line whose slope is equal to the standard deviation. Note that in this example, in which A and N have equal slopes, the mean of H is unaffected by r, but the slopes vary from zero ( r = - 1) to twice the slope of A and N ( I = 1). Consider now the distribution of NIPTS. Both HTL (=NIPTS + ARPTS) and ARPTS have been measured, and their distributions can be estimated from these measurements. However, NIPTS can only be derived by subtraction of threshold shifts in non-noise-exposed (ARPTS) from noise-exposed (HTL) populations. As implied in equations 2 to 4 in the introduction, extreme values (10th and 90th percentiles) for NIPTS have been derived by subtraction of corresponding percentile values in the HTL and ARPTS distributions. If it is true that
then it must also be true that (16) However, this would require perfect correlation (r = 1) between NIPTS and ARPTS (see equation 13), a condition which seems quite unlikely. Indeed, it is appropriate to view the NIPTS data presented in ISO-1999 and elsewhere not as actual estimates of the distribution of NIPTS in individuals, but as distributions which when added to ARPTS distributions, yield good estimates of HTL distributions. Figure 4 presents this discussion graphically. Assume we measure distributions of H and A as shown, and must estimate the distribution of N. If r = 1, the mean and standard deviation of N can be simply obtained by subtraction. This is how NIPTS distributions have been obtained for ISO-1999 and other sources. However, if r = 0, the true standard deviation for N, based on equation 12, would be: ONIPTS
= UHTL -
UARF‘TS-
(TH = (17) Figure 4 shows the range of slopes possible for N for r between -1 and 1, in this example. What is the correlation between NIPTS and ARPTS? Although this is clearly unknown, it seems more likely to be positive than negative. If ARPTS includes “sociocusis,” due to sounds of everyday life, susceptibility to this component of ARPTS should be correlated with susceptibility to NIPTS. In addition, it seems likely that because at least the major site of action for both NIPTS and ARPTS is the cochlear hair cell,
Noise and Aging
21
b.
L
23
2.5
P 3 2 cn I-
2 1.5 N
9 1
0
0
10 20 30 DURATION (years)
0
40
10
20 30 DURATION (years)
40
d.
b
Percentlle
-
35
20
30
P
25
2 20
If
9
/j B
10 5
0
0
0
10
20 30 DURATION (ywrs)
40
10 20 30 DURATION (years)
0
40
Figure 2. NIPTS as a function of exposure duration for the 0.5, 1, 2, and 3 kHz pure-tone average. The parameter is exposure level (dBA). NIPTS values for 0 to 10 yr exposure have been estimated by a logarithmic function, according to SO-1999.
80 70
80 70
I
60
60
g
,
50
50
40
40
30 20 10
30
20
0
10
-1 0
-20
0
10
50
90
10
50
Figure 3. The effect of varying correlation between two random variables, A and N, when a new random variable H is formed by adding values from A and N. The values of A, N, and H are plotted as a function of percentile on Gaussian coordinates.
susceptibility may be shared as well. In the absence of any firm basis for estimating a degree of correlation, it seems most reasonable to assume a value between 0 and 1. As shown in the Appendix, the best estimates of the relative contributions of ARPTS and NIPTS are different, depending on the correlation between them. If r = 1, and H = A + N, 22
Dobie
90
Percenllle
Percenllle
Figure 4. The effect of varying correlation between A and N when the distributions of H and A are known, but the distribution of N must be derived (Gaussian coordinates).
where H is now a constant, A=pA+
(H - PA - P N ) ~ A UA
N=pN+
+ UN
(H - PA - P UA
+ ON
N ~ N
(18)
(19)
Ear and Hearing, Vol. 13, No. 1, 1992
Conversely, if r = 0,
(21) The population standard deviations can be obtained from the tables and formulae of ISO- 1999, which models NIPTS (and ARPTS) as two half-normal distributions, with slightly different variances for the upper and lower halves. For example, in the upper half of the ARPTS distribution, one can obtain uA from the 50th and 90th percentile values (recall that the z value for the 90th percentile is 1.282): UA
=
b.9
-b.5 -
1.282
AA
1.282
For r = 1, uN is similarly calculated:
For r = 0, UN is obtained by expanding equation 17: 1 1.282 (AA + AN)^
U N = - J
- (Ad2
(24)
The results of typical allocations using equations 18 to 24 are shown in Table 1. The examples include permutations of two exposure levels ( 100 and 90 dBA), two exposure durations (40 and 20 yr), and two ages (65 and 4 9 , using database A (male) values for ARPTS. Table 1 shows the percentage of HTL allocableto noise for three conditions for each exposureage combination: HTL = NIPTSo.5 + ARPTSo,s.When the individual’s HTL is equal to the median expected HTL, allocation is straightforwardand requires no assumption regarding correlation between NIPTS and ARPTS. HTL = NIPTSO.~ARPTSo.9, r = 1. When an individual’s HTL is above the median and we assume r = 1, allocation uses equations 18, 19, 22, and 23. HTL = NIF’TSo.9 = ARPT!h, r = 0. With HTL above the median and r = 0, we use equations 20,2 1,22, and 24... In every case, 90th percentile calculations assuming r = 0 result in a larger allocation for NIPTS than when r = 1 is assumed, although the differences in percent allocable to NIPTS are rather small (6-13%). In cases where HTL was closer to the 50th percentile, the difference would be even smaller approaching zero as HTL approaches the median. For four of the six examples, the 50th percentile allocation is between the two 90th percentile values. For the other two examples, the 50th percentile noise allocation is larger than either 90th percentile allocation. Because we don’t know r,
+
but expect it to be between 0 and 1, an attractive simplication suggests itself: use Corso’s method, allocating HTL according to the proportions of NIPTS and ARPTS present at the median, regardless of the actual HTL (as long as it is between the 5th and 95th percentiles, where ISO-1999 purports to be valid). To do so would be very unlikely to underestimate the true noise allocation; for two of the examples in Table 1 (those involving a 45-yr-old man), Corso’s method is likely to slightly overestimate the proportion of HTL due to NIPTS. These options will be illustrated in detail in the following case study. CASE STUDY
A 60-yr-old man retires after 30 yr in a workplace where his daily exposure was 90 dBA; use of hearing protection was negligible. He has had no significant nonoccupational noise exposure, and no otological cause can be found for his hearing loss other than noise and aging. His retirement audiogram (Fig. 5) shows a symmetrical sensorineural hearing loss consistent with this combination, with pure-tone average HTLs (0.5, 1, 2, and 3 kHz) of 30 dB in each ear. No previous audiograms are available. The loth, 50th, and 90th percentiles for NIPTS, given 30 yr at 90 dBA, are taken from Figure 2b: 2.9, 4.1, and 6.7 dB. The equivalent figures for ARPTS, from Figure la, are -0.8, 11.4, and 26.7 dB. The median expectation for HTL, given this combination of age, gender, exposure level, and exposure duration, would be 15.5 dB, of which 26% (4.1 t 15.5) would be allocable to NIPTS and 74% (1 1.4 + 15.5) to ARPTS. Corso’s method would apply the same percentage allocation to this case, although the actual HTL (30 dB) is nearly twice the median. If we assume r = 1, we proceed as follows, using the 50th and 90th percentile ISO- 1999 distribution data: bA
=
(JN =
26.7 - 11.4 = 11.93 1.282 6.7 - 4.1 = 2.03 1.282
~
A = 11.4 +
N = 4.1
Level Duration Age 50%ile 90%ile (r = 1) 90%ile (r = 0)
100 90 90
90
65 65 45 65 65 45
58% 52% 77% 24% 20% 43%
Ear and Hearing, Vol. 13, No. 1,1992
5170 46% 62% 19% 16% 27%
(27)
11.4 - 4.1)2.03 + (30 -11.93 + 2.03 = 6.21
(28)
142.4 (26.7 - 1 1.4 6.7 - 4. 1)2 - (26.7 - 1 1.4)2 = 52.5 bN2 = ( 1.282)2 =
(29)
+
Table 1. Percent of HTL allocable to noise.
40 20 20 40 20 20
+
Thus, for r = 1, we would allocate 21% of the HTL to NIPTS and 79% (23.79 f 30) to ARPTS. If r = 0, using the same ISO-1999 data, we calculate: (iA2
100 100
(30 - 11.4 - 4.1)11.93 = 23.79 11.93 2.03
62% 5w0 75% 25% 22% 36%
A=
(30)
(1 1.4)(52.5)+ (30 - 4. I)( 142.4) =22.0 142.4 52.5
(31)
4. I)( 142.4) + (30 - 11.4)(52.5) = 8.0 142.4 + 52.5
(32)
Noise and Aging
23
N=(
+
01
10 20
11
I
%-
....)
