Duration and recurrence of otitis media with effusion in children from birth to 3 years: prospective study using monthly otoscopy and tympanometryBMJ 1997; 314 doi: https://doi.org/10.1136/bmj.314.7077.350 (Published 01 February 1997) Cite this as: BMJ 1997;314:350
- Sarah C Hogan, audiological scientista,
- Kenneth J Stratford, senior research scientista,
- David R Moore, reader in physiologya
- Correspondence to: Ms Hogan email@example.com
- Accepted 9 September 1996
Objective: To monitor the natural course of otitis media with effusion.
Design: Prospective, longitudinal assessment of the state of the middle ear by otoscopy and tympanometry at monthly intervals from birth to 3 years.
Setting: Domiciliary visits to family homes.
Subjects: 95 full term infants born between August 1991 and November 1993.
Main outcome measures: Observed and simulated data (Monte Carlo) for the duration of single episodes of otitis media with effusion.
Results: 17 of the children had unilateral or bilateral otitis media with effusion for more than half of their first three years of life. Thirty three of the 95 children had tympanograms suggestive of otitis media with effusion at more than a third of observations; the remaining 62 had such tympanograms at less than a third of observations. The data of each group were described by a first order Markov model, yielding a mean duration of unilateral effusion episodes of 5-6 weeks in both groups; the mean duration of bilateral effusion was 6 and 10 weeks in the low and high incidence groups, respectively. However, the main difference between the groups was the time spent between episodes of effusion: effusion free periods were, on average, three times longer in the children who experienced less otitis media with effusion.
Conclusion: Children who are susceptible to otitis media with effusion tend to have more separate episodes of effusion rather than an increased overall duration of episodes. Such children are primarily distinguished by the likelihood with which they acquire the disease than by their ability to recover from it.
Otitis media with effusion, a broadly defined condition that includes secretory, non-suppurative, and serous otitis, as well as acute otitis media,1 is highly prevalent among young children. Effusions produce a conductive hearing loss that averages 20 dB but can be as great as 50 dB.2 A major problem in determining the sequelae of otitis media with effusion is the assumptions made about the occurrence and duration of otitis media with effusion. The fluctuating nature of otitis media with effusion, its occurrence in either or both ears with variable time courses, and the medical management of identified cases all make the analysis of data documenting the natural course of otitis media with effusion difficult. In general, previous studies have examined the natural course in one of three ways: by basing the study on high levels of medical contact,3 by monitoring effusion after acute otitis media,4 5 or by regular prospective cohort sampling.678
The studies that have investigated the duration of otitis media with effusion after an acute infection defined onset as the day of diagnosis or treatment.4 5 The state of the middle ear was monitored either until the effusion had resolved or to the point of surgical intervention or other arbitrary cut off point. These studies have therefore been unable to specify events, beyond the numbers of effusions in a selected population, that persisted during predetermined intervals.
Otitis media with effusion is often asymptomatic and, thus, only prospective studies have the potential to monitor accurately “silent” effusions. Several Scandinavian studies of the natural course of otitis media with effusion have sampled paediatric populations at intervals of three months or more and have commented on the dynamic nature of the condition.678 The costs of more frequent monitoring have limited the number of studies with a temporal resolution of less than three months. However, to gain an understanding of the natural course of otitis media with effusion that is both precise and unbiased more regular sampling is desirable. We investigated the time course of otitis media with effusion by examining its duration, laterality, and recurrence during the first three years of life by frequent monitoring and mathematical modelling.
Subjects and methods
Between 1991 and 1993, 112 full term (38-42 weeks' gestation) healthy infants in the Oxford area who had uneventful birth histories were recruited soon after birth, without regard to sex or socioeconomic or ethnic group. The health visitors of each of five general practices gave parents an introductory letter, to which they replied if they were interested in participating. The study had ethics committee approval, and parental consent was obtained at the outset. Infants were examined about every month for the first three years of life. Each ear was examined otoscopically to check that there was no obstruction of the canal. A Lucas Grason Stadler GSI 33 middle ear analyser was used to obtain tympanograms at 226 Hz from both ears when infants were not distressed or unhappy.9 Tests in which a tympanogram was unobtainable because of a lack of cooperation or the accumulation of wax were disregarded. The tympanograms were categorised as type A (compliance ≥0.2 ml, pressure ≥-150 daPa (decaPascals)), type C (compliance ≥0.2 ml, -400 daPa< pressure ≤-151daPa), or type B (compliance<0.2 ml, pressure ≤-400 daPa). If type B was recorded at either or both ears for three successive months the parents were asked to consult their general practitioner. Particular attention was paid to instances of otitis media with effusion after an acute infection as the administration of antibiotics or antimicrobial drugs may have influenced the duration of the effusion.
For modelling purposes types A and C represented an effusion free state and type B an instance of otitis media with effusion. At each observation an infant was, therefore, in one of four states: binaural otitis media with effusion (B type tympanograms in both ears), effusion in left ear, effusion in right ear, or no otitis media with effusion in either ear. Given the assumption that the probability of an infant being in any state depended only on the state at the previous observation, a stochastic model (first order, four state Markov process) of the time course of otitis media with effusion was developed. A state transition matrix (T), which defined the probability of all possible transitions between each state, was computed from the observed data.10 This matrix therefore provided the information needed to calculate the duration of episodes of monaural and binaural otitis media with effusion on the basis of likelihood (see fig 1). The mean duration for each state (d) was computed from the diagonals of T, using the relation d=1/(1-pnn), where pnn was the probability of a particular state remaining unchanged from one observation to the next.
Two model forms were investigated. The first assumed that the time course of otitis media with effusion was similar in all children and that the variations in the observed data were simply the result of sampling. The second assumed several different time courses, based on the overall prevalence of otitis media with effusion (see Results). Monte Carlo techniques were used to assess whether the models were consistent with the observed data. Simulation programmes used a random number generator optimised for long sequences of uncorrelated, pseudorandom numbers.11
Of the 112 infants who were enrolled, 95 had (at 31 January 1995) participated for more than a year, 67 for more than two years, and 31 for more than three years. Only 17 infants (15%) left the study during the three years. Three infants had bilateral grommets (ventilation tubes) at 12-23 months. Their ears were categorised as B when tympanometry indicated that the grommet was not patent and there was a concurrent, flat tympanogram. For infants followed up to 1, 2, and 3 years of age the modal value for the number of home visits was 11 (range 3-13 visits), 21 (14-24), and 32 (26-35) respectively. The mean interval between visits was 33.6 (SD 10.4) days with a modal value of 28 days.
Preliminary modelling was used to compare the duration of all effusions in children who had a history of acute otitis media associated with effusion (clinically diagnosed and treated with antimicrobial agents) with all effusions of children without such a history. Mean durations of otitis media with effusion (bilateral or unilateral) did not differ between the two groups (5 weeks for unilateral effusion, 9 weeks for bilateral effusion in those with acute otitis media associated with effusion, 5 weeks for unilateral effusion, 8 weeks for bilateral effusion in those without such disease). No distinction was made between children according to whether they had acute otitis media in the remainder of the analysis.
The state transition matrix for the entire observed dataset (95 children) is presented schematically in figure 1. These data show that (a) there was a low probability of changing between unilateral states (from the left to the right ear being affected or the other way round), (b) there was a comparatively high probability of remaining in a bilateral state (either free of effusion or with bilateral effusion), and (c) about one half of the unilateral states reverted to normal (effusion free) after just one observation.
Analysis of this T matrix gave the following mean durations for each state: 22 weeks for no effusion, 5 weeks for effusion in the right or left ear, and 8 weeks for bilateral effusion (fig 1).
The observations were modelled by age (in years) for the first three years. The mean duration of unilateral (6 weeks) and bilateral (9 weeks) effusions did not change with age. Strikingly, however, the mean effusion free period in each of the first and second years (23 weeks) was less than half of that in the third year (51 weeks).
The data were further analysed on a child by child basis, and the amount of time that each child was affected by otitis media with effusion was computed. This was simply the number of positive observations of otitis media with effusion (unilateral or bilateral) divided by the total number of observations for that child (fig 2). There was a broad spread of observed prevalence, with an appreciable minority of children (17) having the disease more than half of the time. Monte Carlo simulations were performed to assess whether this variation would be expected from a first order Markov model. The observations had a broader spread than the simulated data (fig 2), although the means of the two sets were the same. This suggested that the distribution of observed episodes of otitis media with effusion across individual children was not random.
The data were divided, initially, into two groups on the basis of whether children had otitis media with effusion for more (n=33; “high”) or less (n=62; “low”) than the mean affected time (34%). Each group was analysed using the same Markov modelling techniques. The state durations and probabilities for the two groups (fig 3) differed from those for the whole sample (fig 1). The high incidence group had a greater chance of remaining with bilateral effusion than did the low incidence group, showing a 1.7-fold increase in the duration of bilateral effusion. Interestingly, there was no difference between the groups in the duration of unilateral otitis media with effusion, suggesting that unilateral otitis media with effusion may differ qualitatively from the bilateral form. However, the main difference between the two groups was the mean time between episodes of otitis media with effusion (10 weeks compared with 31 weeks). Thus, the high incidence group were primarily distinguished by the likelihood with which they acquired the disease rather than by their ability to recover from it.
To examine further the hypothesis that children who are susceptible to otitis media with effusion tend to have more frequent episodes of the disease, the data were divided into four groups (of about equal size) by the proportion of time spent with otitis media with effusion (fig 4). The interval between episodes decreased noticeably as the proportion of time with otitis media with effusion increased. By contrast, there was a small associated increase in the mean duration of each bilateral episode. Slightly more than one quarter of the children formed the group with the lowest prevalence of effusion. For these children the time free of effusion lasted, on average, more than 60 weeks and was then interrupted by either a unilateral or bilateral effusion of comparatively short duration (about 4 weeks) before they returned to the effusion free state. At the other end of the range a group of children who had otitis media in more than 45% of tests had an average of 9 weeks free of effusion, followed by either unilateral (averaging 6 weeks) or bilateral effusion (averaging 10 weeks), and then by another comparatively short time free of effusion (9 weeks). When the data are considered in this way the cyclical and dynamic course of the effusion process can be fully appreciated
We found that the mean durations for episodes of unilateral and bilateral otitis media with effusion were 5 and 8 weeks, respectively. Nearly one fifth of children had either unilateral or bilateral otitis media with effusion for more than half of their first three years of life. An increased susceptibility to the disease was more likely to be associated with more frequent episodes than with an overall increased duration of individual episodes.
Despite the prevalence of otitis media with effusion, there have been few attempts to analyse the duration of episodes or the interval between episodes. This information may be important for understanding the consequences of the disease and for defining treatment procedures. For example, otitis media with effusion in infancy has been associated with a reduced level of binaural unmasking in later childhood,12 13 presumably leading to a reduced ability to detect target sounds in noisy environments. It would seem desirable to know what aspect of otitis media with effusion (type, laterality, duration, associated hearing impairment, etc) leads to this problem, partly to target treatment to the appropriate children and partly to help in understanding how early hearing loss may affect neural development in general.14
Another treatment issue is raised by the specific findings that the average length of otitis media with effusion episodes is 1-2 months and that children who are more susceptible tend to have more frequent rather than longer episodes. Clinical practice in the United Kingdom is that children detected with otitis media with effusion should undergo a period of “watchful waiting” before surgical intervention. If this practice is maintained our results suggest that examination about every six months should detect cases of recurrent otitis media in children susceptible to effusion. Children who are less susceptible would not normally be expected to have a recurrence during this time. Our analysis suggests that children under 2 years of age are more susceptible to recurrent effusion than children in their third year. This result differs from that of an American study, in which children under 2 years old were estimated to have a greater (fourfold) persistence of effusion after acute otitis media than older children.4 However, the clinical implication from both studies is that children under 2 years of age are likely to have more otitis media with effusion than older children and that particular attention should be given to infants.
The finding that unilateral otitis media with effusion did not differ in duration between the high and low groups suggests a possible qualitative difference between unilateral and bilateral effusions. It may be, for example, that the site or mode of acquisition of the disease vary, with bilateral effusions deriving from systemic factors and unilateral effusions deriving from more peripheral and local disease. This idea is highly speculative, but it may be of interest to examine the causes of unilateral and bilateral otitis media with effusion separately to see whether any of the hypothesised risk factors—for example, breast feeding, birth weight—are more commonly associated with either laterality.
This prospective, longitudinal study analysed the duration and recurrence of otitis media with effusion by monthly monitoring with otoscopy and tympanometry
A first order Markov model described the observed data adequately
The model predicted that episodes of unilateral and bilateral otitis media with effusion have a mean duration of 5 and 6-10 weeks, respectively
An increased susceptibility to otitis media with effusion was primarily associated with more frequent episodes
We thank Drs Martin Nowak and Lindi Wahl for their help and advice on the modelling aspects of this study. We also thank Mr Martin Burton and Dr Keith Dorrington for their helpful criticism of the manuscript. The study would not have been possible without the generous, voluntary cooperation of the families who participated. We would like to express our greatest thanks to them.
Funding: Medical Research Council, Hearing Research Trust, and the Wellcome Trust.
Conflict of interest: None.