Better Hearing Australia Conference
Adelaide 7-11 August 1994
A model forecasting the prevalence in hearing loss in the Australian population over the next 20 years based on trends in decline in otoacoustic emission strength.
Eric L. LePage
Hearing Loss Prevention Research Unit
National Acoustic Laboratories,
126 Greville Street, Chatswood, NSW 2069.
While pure tone audiometric measures provide the first concrete signs of ear disability, the accumulation of damage can be tracked for years prior to this critical level of damage using otoacoustic emissions (Murray & LePage, 1993; LePage and Murray, 1993). We have translated the concept of the rise in cochlear damage to that of rates of ageing of the ear. In these terms presbycusis, or the normal ageing effect may be redefined as a minimal rate of ageing of the ear, or a decline in emission strength of 2 to 3 dB per decade. The Australian population over 40 years of age seems to be declining at about 3 dB per decade, however, many individuals are being tracked at declining at between 20 to 30 dB per decade. Prominent amongst this group are young people and musicians who regularly engage in heavy music or leisure noise exposure and who are already reporting symptoms of hearing loss by age 20. Using ABS figures for the projections for growth of Australian population in future, the rates of rise of numbers of individuals reaching critical values of ear damage are modelled, subject to various assumptions. A key output of the model is a curve described as a rate-of-ageing curve, plotted versus age which shows that the rates of damage accumulation is greatest in the first 20 years of life. This is in sharp contrast to the common experience of the presentation of symptoms for most people being later in life. This feature is difficult to explain unless people in their first 20 years, on average, are not currently being subjected to much higher rates of cochlear damage accumulation than during the corresponding period of their parents’ generation. If the model is realistic in its explanation of the current situation, the continuity which is inherent in the model suggests that an inexorable trend exists toward premature hearing loss in Australian young adults within ten years, as the very long term effects of accelerated ear damage eventually become manifest and are confirmed with pure tone audiometry.
Otoacoustic coherent emission strength (CES) is high (15-30dB) in young, healthy ears, but is low in ears damaged by noise or drugs which may not yet be accompanied by hearing loss (LePage and Murray, 1993). In the same study it was found that pure tone thresholds are typically within normal limits while CES values remain above a critical value of -2.5 dB SPL. Murray and LePage (1993) have shown that there is an overall decline in CES value versus age for Australian males and females.
The rates of decline towards the critical value may be used to estimate the number of Australians likely to suffer a hearing loss. By knowing the mean and standard deviation of each age range from our population estimates for CES it is possible to determine the proportion of persons in each age range who have emission strengths below the critical value. This fraction is used to scale the documented Australian population statistics for each age range to provide an estimate of the numbers of persons likely to be suffering a hearing loss.
We have developed the concept (Murray et al., 1994, this volume) that presbycusis (Macrae, 1984) and noise-induced hearing loss (NIHL) are both seen as late manifestations of an ageing process, the extent of which we characterise by the Coherent Emission Strength (CES dB SPL). “Young ears” have a high CES value while “old ears” have a low CES value. As with most other systems in the human body, ageing of the ears is not fixed nor strictly tied to a person’s age, but may occur at rates which are faster or slower. We assess the rate of ageing in terms of the rate of decline of CES. “Pure presbycusis” is here defined as the ageing process having a minimal rate of loss of outer hair cells (i.e. equated to a linear rate of decline of CES of -3 dB/decade) and which currently produces symptoms from middle-age onwards. NIHL is defined as the clinical end stage reached as a result of an accelerated rate of ageing of on average greater than 3dB/decade (i.e. <-3 dB/decade). Regular noise exposure can increase the instantaneous rate of decline tenfold or more, intermittently or for longer periods. Hence “excessive noise” can be equated to a sporadic or prolonged noise exposure resulting in an accelerated rate of CES decline exceeding 3 dB/decade.
Noise is not the only influence producing outer hair cell damage. For comparison, some antibiotics such as the aminoglycosides (e.g. Streptomycin, Gentamycin) and cytotoxic drugs (e.g. Cisplatinum) may produce much higher rates of destruction of cochlear outer hair cells amounting to one-third the length of the cochlea (e.g. 3000-4000 OHCs) in the ten day period describing the typical course (e.g. LePage, 1981). In terms of ageing this constitutes an accelerated rate exceeding 5000 dB/decade. When amino-glycosides are administered in conjunction with certain loop diuretics which facilitate entry into the cochlea, permanent outer hair cell damage may occur over a few hours, or about ten times faster again. One study on the effect of cisplatinum on brainstem responses and otoacoustic emissions was recently carried out in guinea pigs (McAlpine and Johnstone, 1990). The behavioural studies on ototoxic drugs have used hearing thresholds as the primary criteria for damage, but it is now seen that these measures would not have adequately gauged the outer hair cell damage.
The model focusses upon the apparent ‘dip’ in the age characteristic in the second decade which suggests an explanation in terms of an accelerated rate of decline in emission strength with age in the Australian population. This paper is about exploring this potential explanation and what its consequences are, if any, on the future incidence of hearing loss in Australia.
The model is implemented in two stages. The first stage is to derive a curve which describes how fast Australian ears have been ageing in the recent past. This curve is plotted against age for the population. In the second stage this curve is applied to the current age picture for the Australian population to project what it likely to happen in the future.
Assumptions of the model
- Presbycusis is not so much an end stage of hearing loss in the elderly, but a process of decline in net outer hair cell performance which can be modelled as a steady or linear decline of CES throughout life.
- The CES values determined within each age range are assumed to be normally distributed.
- The age characteristic obtained in the study from people all over Australia is assumed to represent that of the whole population.
- The number in each age range for our sample is assumed to be proportionately distributed throughout the country.
- The starting year for deriving of the rate-of-ageing characteristic is 1980.
- The age dependence of the rate of ageing characteristic is fixed over the term of the model (20 years), representing the net effect of all factors contributing to outer hair cell loss.
The first assumption is necessary for redefining the concept of ageing of hearing in terms of objective measurement of net OHC performance. The second appears to be justified for most age ranges on the basis of the distribution seen in the 6000 (approx.) records in our database as seen in Figure 1. The third assumption is necessary in the absence of other relevant information. The fourth assumption, that the age distribution is not significantly different in any one part of the country from any other part stems from the figures supplied by Australian Bureau Statistics (1994). The fifth concerns the year taken as time zero for the model. It is a tradeoff between going back as far in time as possible to try to make the assumption of “pure presbycusis” as realistic as possible (before any accelerated ear damage occurred), and keeping it as recent as possible to capture the most recent trend in case these should be subject to variation across generations. The last assumption makes specific allowance for any maturation process which e.g. leads to young people to tolerate very loud sounds less well as they progress through their twenties.
Figure 1. Frequency histograms showing age dependence of otoacoustic emission strengths (CES dB SPL) for the various age ranges (left margin) for a subset of the data from which all cases of identified ear pathology have been removed. The right margin shows numbers, means and standard deviations in each range. Note that the modal values for the 13-18 year group are as low as those for th 43-48 year group and that the distributions are to a first approximation normal.How does the rate-of-ageing characteristic describe risk for hearing loss? Individual susceptibility to hearing loss is here related to rates of decline of coherent emission strength (CES), and is found to vary widely between individuals and also to vary with time (Murray et al, 1994; this volume). Assuming hypothetically that all persons would display the same emission strength, then the rate-of-ageing characteristic would describe the average risk for hearing loss at each age because faster rates would reach the critical value after a short time interval. Therefore, if any individual ear’s emission strength is low to begin with, then the risk of susceptibility is still higher; i.e. hearing loss more imminent.
Figure 2. The output of model Stage 1 showing the rate-of-ageing characteristic which is derived from the NAL database to 1994 and which is used in Stage 2 (projection). This curve suggests that while presbycusis is a condition for the elderly, the damage accumulation is highest in the first two decades and tapers off during the third.
The first objective of the model is to determine the shape and size of the rate-of- ageing characteristic by a process of iteration so that the model mimics (fits) the shape of the age characteristic in our current database (labelled 1994). The age characteristic and the dip is modelled by initially assuming that at some time in the past, the CES vs age characteristic for the population was declining linearly at -3 dB/decade, i.e. “pure presbycusis”. For convenience the starting year for the onset of accelerated damage is taken as 1980. The result of this first stage of the model is a description of how the rate-of-ageing curve varies with age.
The age characteristic is plotted in two-year steps. Also every two years, 1980, 1982, 1984, and so on, the decline in the emission strength (CES value) is calculated, based on an initially arbitrary rate of ageing vs age curve. This process is repeated until 1994, when the resulting age characteristic is compared with the curve derived from our population measurements. The rate-of-ageing characteristic is then reshaped so as to minimise the difference between the projected and real age curve for 1994. The resulting curve is then taken as fixed for projections from 1994 on.
In the model, the age characteristic has been plotted it in two year steps. Then, in every two years, from 1980 onwards, 1980, 1982, etc., the decline in emission strength is calculated based on an assumed rate of ageing versus age curve. This process is repeated until 1994, when the age characteristic is compared with the curve obtained from our measurements.
The rate of ageing characteristic is then reshaped and the process repeated to minimise the difference between the projected age curve and the one that we get from our measurement. The result of the first stage is shown in Figure 2. This curve shows that individuals over 40 years old on average age at less than 4 dB/decade, but that individuals less than 20 years old age more than 6.5 dB/decade.
The fixed rate curve is then used to project future curves for mean CES vs age for the next twenty years 1994, 1996, …. 2004, …. 2014. The standard deviations vs age are assumed to be fixed to today’s values. The population in each 2 year bin is assumed to be normally distributed. ABS projections for Australia’s population are used in calculations of numbers of Australians affected over the next 20 years. The number of males and females whose CES values cross the critical value (associated with the onset of mild hearing loss) is plotted.
Figure 3 shows the Australian Bureau of Statistics population projections for the next 20 years. People in the post war “baby boomer” range are going to age more than the previous generation and this feature is the primary reason for the increase the population. The curves that the average age of the population is going to increase for males and females, but that females live longer than males.
Figure 3. Australian Bureau of Statistics projections for the numbers of Australians in each 2 year bin for the next 20 years. Most of the population increase will be in the range 45 to 65 years.RESULTS
To illustrate the model, the figures for Australian males are plotted against age. The starting output of the model ageing curve is linear, labelled ‘1980 “Pure Presbycusis”‘ (Figure 4A) and is plotted alongside the empirical curve for 1994 (Murray et al, 1994; this volume) which has the dip in it. The dashed line indicates the critical value for emission strength measured by the left ordinate. The lowest curve referred to the right ordinate indicates the number of males in the population (millions) in every two year bin, so this shows, as one expects, more older people to have declined to the critical value of emission strength than younger people.
Next, all the individuals in the population are “aged” by two years by recalculating the projected emission strengths for each two year bin based upon the rate of decline appropriate for each age bin from Fig. 3. Figure 4B shows, the result for 1982. Since the fastest rate of ageing is for the lowest age ranges most change is seen the projected CES curve, which is now no longer linear but is kinked. Note that there is no discernable change to the distribution of males affected with hearing loss at any age.
Figure 4, Panels A,B,C and D(see text).
This ageing process is applied a further six times and the result is shown in Figure 4C. Note that there is substantial agreement between the model curve and the empirical curve; agreement which was achieved only after iteratively arriving at the curve shown in Figure 3. The model underestimates the level of decline in the first ten years, because the decline in this age range is highly constrained by the fact that all newborns (on average) begin life with the same emission strength. Secondly, while the model could fit the empirical curve in the first few years, it cannot then cope with fitting the rest of the curve. The choice was made instead to limit the average rate of ageing in this age range to make its performance overall much more realistic. The output of the model, the lowest curve representing numbers affected now shows a discernable increase in the numbers of males affected, particularly at the lower age ranges, but the age range predominantly affected is still over 55 years of age.
Proceeding to the second stage, that of projection, the ageing process is continued in two year steps, operating on this curve for 1994. In the year 2000, for example, we see that the mean emission strength for the population lies substantially below the 1994 curve, so that many more individuals have reached the critical value. The lower thin line (representing the mean minus one standard deviation) has already crossed the critical value for ages above 20, indicating that more than 16 percent of all Australian males between ages 20 and 50 will have hearing problem by the year 2000. i.e. the number of affected males is growing relative to the number of older people and this is confirmed by the lowest curve. This means that across Australia between 100 and 200 thousand males in each two year age range would be affected.
Figure 5 (lead-up to 1994 values) and Figure 6 (future values 1994 to 2004) show the combined results of the model. In each are four panels, A, B, C and D, plotted against age. The top two panels show the mean value of emission strengths for the population and the lower two panels show the numbers of people affected with at least a mild hearing loss.
Figure 5. Top panels show the emission strength declines; bottom panels show adjusted prevalence to match 1994 figures, for males (left panels) and females (right panels)Where the mean emission strength values are already lower than the critical value this means more than 50 percent of the population with a mild hearing loss, and by implication many of these will have moderate and severe losses instead. By comparing Figures 5 and 6 (panels A) it is clearly seen that for any year, the emission strengths for males on average lie about 4 dB below emission strengths for females, a result which derives from our raw otoacoustic emission data. Australian females on average have less accumulated ear damage than Australian males, a result which fits very well with audiometric statistics accumulated for many years. Accordingly this means (panels C) that more males have hearing problems than females, many more in fact. The same rate of ageing curve in Figure 4 is used for both males and females and therefore the mean values of males being lower than for females implies that males reach critical values earlier than females (LePage and Murray, unpublished results).
Figure 6. Top panels show the emission strength declines; bottom panels show predicted prevalence (to 2014), for males (left panels) and females (right panels)The lower panels show the feature of the model which commands attention. It is that while the numbers of individuals affected with a hearing loss are relatively small up to 1994, the model predicts that there will be a steep rise in the numbers affected beyond the present. This follows from the fact that while only the tails of the distribution have crossed critical values to date, the downward progress of mean emission strengths towards the critical value means that there is expected to be a relatively rapidly rising function beyond 1994. In simple terms this means that the small increase in hearing problems in the young observable now it represents a “tip of the iceberg” effect and as it becomes manifest over the next few years, many more males will be affected than females.
The next significant feature of the model is that the rise in numbers will be occurring in the younger age group and that the target age group will, in 20 years time, be today’s teenagers. Taking a specific example from Fig. 6. lower panel, if we choose the age range for the model between 24 and 26, something like 200000 males, who will have at least a mild hearing loss by 2014 and something like 55000 females.
The model shows that males, on average, are going to reach the critical value before females. From 1980 to 1984 the model says that there has been a significant increase in the number of younger people reaching the initial symptoms of hearing loss. Table 1 shows the total number according to the model. There is about 3 million Australians total – just over 2 million males and about 0.7 million females – at least at the point of onset of symptoms of hearing loss and/or tinnitus, which again is fairly consistent with the current audiometric picture.
Table 1. Projections of total numbers affected (millions)
at the level of at least a mild hearing loss at each of the dates listed.
It was found in the process of developing the model that the shape of the curve in Figure 3 is very tightly constrained in order to achieve this level of agreement by the two curves in Figure 4C over most of the age range.
The implication of the model is that as time progresses, although there will be a gradual decline in emission strengths there will nevertheless be a steep rise in the numbers affected. It has been shown that it is possible to estimate the numbers of people having hearing problems. According to the model, the population currently affected with at least a mild hearing loss is about 2.5 million males and 0.7 million females, which is about 30 per cent of the population and here about 9 per cent of the population. The figure of 3.2 million is slightly larger than known Australian Statistics (about 2.5 million). However, our estimate actually may be more indicative of the number of Australians in the borderline condition emission strengths are critical but for which symptoms have not yet been recognised, perhaps because the better ear allows the individual to cope satisfactorily.
By the year 2000 the population will have gone up slightly, but the percentage of the population affected will be significantly increased by about 1.5 times and it will be increased by about the same ratio as females. The percentage affected will go up according to the model. After 10 years from now we are suggesting that 51 per cent of the male population will be affected and 15 per cent of the females. In 20 years time it could be as much as 78 per cent of the male population have at least a mild hearing loss and 25 per cent of the female population.
So, in summary, what the model has done is to take the current population and treat the apparent dip in the age curve quite seriously. As shown in our first paper (Murray et al, 1994, this volume), we have numerous examples of emission spectra which appear to represent very damaged ears in young people. In fact, they often look as damaged as we see in people who are known to have industrial hearing loss. We have looked at the coal industry (LePage et al., 1993), we have looked at the music industry, we have looked at quite a few industrial situations, and these people have emission spectra which have these characteristic patterns of damage. We find these characteristic damage patterns repeated in many, many young people that we look at.
So, we think that the dip is real and, if that is the case, then it suggests that we have a latent damage picture which has not begun to show up significantly in any previous audiometric measures (Carter et al, 1982, 1984) but which, nevertheless, is going to percolate through and start making its presence felt within 10 years. So, in essence, the prediction is that the ratio of young people to older people reporting a hearing loss will go up and perhaps go up quite substantially together with the multiple costs associated with treating the individuals affected.
A suggestion arising from this work is that the ear appears to have evolved to provide a long lifetime with very little hearing problems if we look after our hearing. The problem is that in Western society we may have developed a mode of living which basically uses up most of our lifetime allocation of outer hair cells in the first thirty to forty years.
Hearing loss prevention is a good example of being more valuable than the “cure” because there is currently no significant treatment for two of the major symptoms of hearing loss, viz. loss of selection and tinnitus. A key ingredient of any future prevention program must be the introduction of otoacousic emission screening to provide early warning of accelerated damage (LePage et al., 1993)..
Carter, N. L., Waugh, R.L., Keen, K., Murray, N.M. and Bulteau, V.G., (1982). Amplified music and young people’s hearing. Med. J. Aust. Aug.7, 1982, 125-128.
Carter, N., Murray, N, Khan, A. and Waugh, R., (1984). A longitudinal study of recreational noise and young people’s hearing. Aust. J. Audiol., 6, 45-53.
LePage, E.L. (1981). The role of nonlinear mechanical processes in mammalian hearing. Ph.D. Thesis, The University of Western Australia.
LePage, E.L. and Murray, N.M. (1993). Click-evoked otoacoustic emissions: comparing emission strengths with pure tone audiometric thresholds. Aust. J. Audiol., 15, 9-22.
LePage, E.L., Murray, N.M. and Macrae, J.H. (1993a). Otoacoustic emission assessment of ear damage in coal mine workers: Pilot study May – October 1992. National Acoustic Laboratories Commissioned Report No. 75.
LePage, E.L., Murray, N.M., Tran, K., and Harrap, M.J. (1993b). The ear as an acoustical generator: otoacoustic emissions and their diagnostic potential. Acoustics Australia. Vol. 21/3, 86-90.
Macrae, J. H. (1990). Noise-induced permanent threshold shift and presbyacusis. Aust. J. Audiol., 13, 23–29.
McAlpine, D. and Johnstone, B.M., (1990). The ototoxic mechanism of cisplatin. Hear. Res. 47, 191-204.
Murray, N.M, LePage, E.L. and Tran, K., (1994). Ageing Characteristics of the Australian population in terms of otoacoustic emission strengths: global and individual picture. (This volume).
Murray, N.M., and LePage, E.L. (1993). Age dependence of otoacoustic emissions and apparent rates of ageing of the inner in an Australian population. Aust. J. Audiol. 15, 59-70.
Last Modified: Tuesday, 30 November 2004