Arguments in favor of surgery at an earlier age (under 5years) are the following: 1) reduction in the number of visits to the doctor, which are required for the follow-up of a minor with perforated eardrum; 2) hypoacusis and privation of aquatic activities with affect on the quality of life; 3) higher incidence of severe secondary complications due to chronic otitis media in younger children; 4) better cochlear reserve at younger ages, with greater potential to restore and preserve hearing; 5) limitation of the damage that chronic infection can cause to other structures in the middle ear; and 6) auditory loss that alters the development and the quality of academic activities [11-16]. Despite this arguments, the recommendation to delay tympanoplasty, generally until six years of age, is widely accepted. In support of said recommendation are the following aspects: alterations in the healing process or re-perforation of the graft by repetitive infections of the superior respiratory tracts; unpredictable function of the Eustachian tubes; immunological immaturity; possibility of spontaneous resolution; difficulty of post-operative care; lack of confidence of the part of the parents in the procedure; and equivalence of the perforation to the function of the ventilation tube [2,13,14]. There is evidence that tympanoplasty is more successful in children over six years of age [1,2,4,6,10,12-14], although other studies have not demonstrated significant differences in this respect [1,3,17].

In addition to age, other factors have been suggested as interfering with the success of the surgery: the size and site of the perforation; cause of the perforation; active infection at the time of surgery; the state of the auditory ossicles and of the mucosa of the middle ear; state of the contralateral ear (as a measure of the function of the Eustachian tube), and the surgical technique utilized [17]. The presence of adenoid tissue, whether it be hypertrophic or chronically infected, has been proposed as a prognostic factor; however, no statistically significant association has been demonstrated, above all in short-term results [1,2,4,18]. A factor that could have interfered in the result was the definition of success that was used. Children with craniofacial syndromes, with or without cleft palate, present a different evolution; for these children, the results reported in the literature are controversial [19,20].

Only five patients (10.4%) presented craniofacial dysmorphias; three cases (6.2%) corresponded to sequelae of uni- or bilateral cleft lip or cleft palate, and two (4.1%) to submucosal cleft palate (data not shown). The median age of the group with successful outcome was 11years of age (range: 5–16years), whereas that of the group with a failed outcome was nine years of age (range: 6–15years). The median age of onset of symptoms was five years of age for both the successful- and failed- outcome groups (range: 1–15years and 1–7years, respectively).

The values of the β coefficients are shown in Table3. Because they did not represent differences, the following three variables were excluded from the partial model: age at time of surgery, surgical technique (onlay vs. underlay), and the presence of otorrhea at the time of surgery

Although all 12 patients included in the validation had a probability of success >0.425 (the value selected as being best for the cut-off level), two patients (16.7%) had a poor evolution (Table4). The values of the validation were obtained in the following manner (data from the first patient (Table4) is used in the example): the onset of symptoms was after six years of age, corresponding to a value of 1 for the variable x₁ (Table4); this value was multiplied by the value of the corresponding coefficient β for that variable (onset of symptoms) in Table3, i.e., β = −1.330. For the variable, contralateral ear, the patient’s ear was normal, giving a value of 1, which was then multiplied by the value of β for contralateral ear (0.777). For the variable, adenoidectomy, the multiplication product, 1 × -0.123, corresponded to the fact that the patient had undergone adenoidectomy prior to the tympanoplasty, with −0.123 being the coefficient for adenoidectomy in the model. Substituting the products into equation of probability in logistic regression, the resulting value was 0.78.

To our knowledge, this is the first study to validate an index that predicts the result of tympanoplasty in children. In general terms, the characteristics of the patients were similar to those treated in other hospitals. For the population of the United States, the most frequent causes of perforation have been, first, a result of the placement of ventilation tubes and, second, chronic perforation due to infection [4]. However, other reports have mentioned a high frequency of perforations with infectious etiology [16], the frequency being equal to that reported for patients in this series.

It is important to point out that, in the present work, age at the time of surgery was excluded because this variable did not show sufficient weight in the model to predict the success of surgery. A limitation of this study would be the considering that the youngest patient included was five years of age. Using logistical regression analysis, Sckolnick et al. [3] postulated that, as age increased, the odds ratio of success diminished, with the cut-off point placed at nine years of age; that is, the best prognosis was for 1–5year-old patients, declining 9% for each year of age thereafter, until nine years of age when the success rate increased to a value similar to that found for those older than 16years of age. For this reason, it is possible to identify a group of patients, 7–12years of age, with a lower rate of success; the majority of the population studied was, on average, 7.1years old [3]. The present study is one of the few that evaluate the results of tympanoplasty in very young children. A study by Black et al. in 1994 included patients as young as two years of age; no statistically significant differences were encountered in the results of the surgery for these children [11]. In the only meta-analysis that individually evaluated articles that were cited in the current study, the authors showed that, in the majority (25 of 30) of these reports, no relation was found between age and the success of surgery. This contrasts with the results of the same meta-analysis, in which a statistically significant relation was found between this factor and a successful result. The authors of this meta-analysis attributed these discrepancies to the following: the majority of the individual studies analyzed either did not include patients younger than seven years of age, used a different definition of success, or had methodological errors, such as the size of the sample or the type of study (retrospective) [1]. In a study published in 2010, which reported (retrospective) the results of 132 tympanoplasties performed on children, ranging from 6–15years and divided into two groups (<8 and >8years), once again no statistically significant differences between age and successful result were found [18]. It is of note that, in that study, the definition of success was the same as the one used in the present study; one limitation of this definition is that the frequency of success is lower than when just one factor is considered.

Another important point to consider is the age of onset of chronic otitis media. In the literature, this is used in the age of the perforation. For example, patients with retained ventilation tube, who underwent tympanoplasty during the removal of the tube and who did not suffer changes due to chronic inflammation are said to be at age “zero” of the perforation. This population differs from those whose perforation occurred either after a process of otitis, or secondary to the extrusion of a tube, but for whom a determined period of time passed before the tympanoplasty was performed [4]. In that same study, the authors found a statistically significant relation between the age at which the ear was perforated and the result of the surgery; nevertheless, upon excluding the patients at age zero of the perforation, this relation loses significance. In other studies, no relation was found between these factors [1,4,9]. The sample size of our study was small; however, the age of onset of chronic otitis media was associated with the success of tympanoplasty.

For patients with craniofacial dysmorphias (basically, the sequelae of cleft lip or cleft palate), the results of tympanoplasty are controversial. On one hand, Dornhoffer et al., in a series of 20 patients, with a total of 26 otological surgeries, concluded that tympanoplasty is a reasonable treatment for patients with sequelae of cleft lip or cleft palate. In that work and in the current study, the same three criteria of success were used and similar results were obtained [25]. In a retrospective study of 26 patients who had sequelae to cleft lip or cleft palate and who underwent tympanoplasty, no statistically significant difference was found in the results for integration of the graft, hearing, and the necessary equalizing of pressure in the Eustachian tube, as compared with the results for patients of the same age without cleft palate [21]. Despite the small sample size in our study, this variable was related with the prognoses of the patients, one reason in this study would be that the increased technical difficulty encountered in performing tympanoplasty on patients with craniofacial dysmorphia tends to negatively affect the outcome of the surgery.

One limitation of this study was that, with the seven variables included in the predictive model, it was possible to predict only 81% of the successful outcomes. In the validation study, all patients with a successful evolution could be predicted. Nevertheless, the specificity was only 67%; therefore, it was expected that this model could not predict 33% of the failures. In the present study, 17% of the patients who had been predicted to have successful outcomes had, in fact, unfavorable post-surgical evolutions. However, the predictive model produced better results than did analyzing each variable separately. The model that included the three outcome criteria was the most equilibrated of these models was the one that incorporated all three parameters of success as the outcome criterion, because it predicted a greatest number of failures without affecting the percentage of predicted successes.