Customize your JAMA Network experience by selecting one or more topics from the list below.
Paradise JL, Bluestone CD, Colborn DK, et al. Adenoidectomy and Adenotonsillectomy for Recurrent Acute Otitis Media: Parallel Randomized Clinical Trials in Children Not Previously Treated With Tympanostomy Tubes. JAMA. 1999;282(10):945–953. doi:10.1001/jama.282.10.945
Author Affiliations: Departments of Pediatrics (Dr Paradise and Mss Colborn and Bernard), Otolaryngology (Dr Bluestone), and Audiology (Mr Smith), Children's Hospital of Pittsburgh, Pittsburgh, Pa; Departments of Pediatrics (Dr Paradise and Ms Bernard) and Otolaryngology (Drs Paradise and Bluestone and Mr Smith), School of Medicine, and Department of Biostatistics, Graduate School of Public Health (Dr Rockette and Ms Kurs-Lasky), University of Pittsburgh.
Context Adenoidectomy and adenotonsillectomy are commonly performed in US children
to reduce the occurrence of persistent or recurrent otitis media, but evidence
supporting the efficacy of the operations is limited.
Objectives To test the efficacy of adenoidectomy and adenotonsillectomy in children
with persistent or recurrent otitis media who had not previously undergone
tube placement and to compare the relative efficacy of adenoidectomy alone
vs adenotonsillectomy in such children.
Design Two parallel randomized clinical trials.
Setting and Participants A total of 461 children aged 3 to 15 years were enrolled at Children's
Hospital of Pittsburgh, Pa, between April 1980 and April 1994. Four hundred
ten children were observed for up to 3 years.
Interventions Children without recurrent throat infection or tonsillar hypertrophy
(304 enrolled; 266 followed up) were randomized to either an adenoidectomy,
adenotonsillectomy, or control group; children who had such conditions (157
enrolled; 144 followed up) were randomized to an adenotonsillectomy or control
Main Outcome Measures Occurrence rate of episodes of acute otitis media by treatment group
and estimated proportion of time with otitis media.
Results In both trials, most subjects were eligible because of recurrent acute
otitis media, with or without persistent otitis media with effusion. A total
of 47 children assigned to surgical treatment groups had no surgery. The efficacy
of surgery in both trials was modest and limited mainly to the first follow-up
year. The largest differences in that year were found in the 3-way trial between
the adenotonsillectomy group and the control group: mean annual rate of episodes
of acute otitis media, 1.4 vs 2.1 (P<.001); and
mean estimated percentage of time with otitis media, 18.6% vs 29.9% (difference,
11.3%; 95% confidence interval, 4.4%-18.2%; P=.002).
Perioperative and postoperative complications or other adverse events occurred
not infrequently, especially among subjects undergoing adenotonsillectomy
Conclusions Our study showed limited and short-term efficacy of both adenoidectomy
and adenotonsillectomy; given the risks, morbidity, and costs of these procedures,
these data suggest that neither operation should ordinarily be considered
as a first surgical intervention in children whose only indication is recurrent
acute otitis media.
Adenoidectomy and adenotonsillectomy (T&A) are the most commonly
performed major surgical operations among US children; in 1994 an estimated
140,000 US children younger than 15 years underwent adenoidectomy, and an
estimated 286,000 underwent T&A.1 In many,
if not most, of the children undergoing adenoidectomy and in many of the children
undergoing T&A, the primary or secondary surgical diagnosis has been otitis
media.2 Based on early recommendations3-5 and also on more recent
adenoidectomy has generally been considered the component of tonsil and adenoid
surgery that is efficacious in reducing the occurrence of otitis media. Standard
textbooks of pediatrics10,11 and
otolaryngology12,13 as well as
current guidelines of the American Academy of Otolaryngology–Head and
Neck Surgery14 recommend adenoidectomy, but
not tonsillectomy, for this purpose. Nonetheless, many practicing otolaryngologists
recommend T&A for children with persistent and/or recurrent otitis media,
and among children undergoing surgery in whom the primary surgical diagnosis
was recorded as otitis media, T&A frequently has been the operation performed.2
Notwithstanding the widespread performance of adenoid-tonsil surgery
to reduce the occurrence of otitis media, the efficacy of such surgery under
various clinical circumstances has remained in doubt, largely because most
reported clinical trials have important limitations.15-17
Previously we reported results of clinical trials of adenoidectomy in a group
of children we considered severely affected because they had undergone tympanostomy
tube placement for persistent and/or recurrent otitis media and, after tubal
extrusion, had again developed otitis media.17
However, those trials did not address either the efficacy of adenoidectomy
in children with persistent or recurrent otitis media who had not previously
undergone tube placement or the relative efficacy in such children of adenoidectomy
compared with T&A. We conducted 2 parallel randomized clinical trials
to address these issues.
Two parallel, randomized controlled trials were conducted at the Children's
Hospital of Pittsburgh, Pittsburgh, Pa, from April 1980 to April 1994 as an
extension of a previously instituted study of indications for tonsillectomy
The study was approved annually by the hospital's Human Rights Committee.
Subjects were recruited from among hospital outpatients and children referred
by community practitioners or directly by parents. Details have previously
been reported regarding initial evaluation of subjects, study exclusion criteria
(children with overt or submucous palatal clefts were specifically excluded),
and procedures regarding informed consent and surgery.16,17,19,20
To be eligible for the present trials, children were required to be
age 3 to 15 years, not have undergone previous tympanostomy tube placement,
and have experienced (1) at least 3 episodes of acute otitis media during
the preceding 6 months, or at least 4 episodes during the preceding 12 months
including at least 1 episode during the preceding 6 months, with at least
1 of the episodes having been documented with a recorded description of symptoms
and tympanic membrane findings or confirmed by tympanometry, tympanocentesis,
or myringotomy; or (2) middle-ear effusion in 1 or both ears extending over
at least 180 days during the preceding year and documented by at least 2 clinical
observations at least 6 months apart, the most recent by a study-team clinician
and/or confirmed by tympanometry.
Of 2122 children evaluated, 582 met the eligibility criteria (Figure 1). Of these, 17 had disqualifying
medical conditions and 104 were not enrolled because their parents withheld
consent for them to be assigned randomly. The remaining 461 children were
stratified into 3 age categories (age 3 and 4, 5 and 6, and 7 to 15 years)
and further classified by means of clinical and radiographic criteria21 as to whether they had adenoidal nasal obstruction.
The children were then assigned randomly, within age and nasal obstruction
categories, within 1 of 2 clinical trials. In 1 trial, children without apparent
tonsil-related indications for tonsillectomy were assigned, in balanced blocks
of 6 subjects, to 1 of 3 treatment groups: adenoidectomy, T&A, or control
(the 3-way trial). In the other trial, children whose tonsils appeared potentially
obstructing or who had a history of recurrent throat infection that met or
exceeded entry criteria used in previously reported20
or concurrent22 tonsillectomy trials were assigned,
in balanced blocks of 4 subjects, to 1 of 2 treatment groups: T&A or control
(the 2-way trial). Separate, computer-generated random number lists were used
for the assignments. Parents were advised that children assigned to control
status could subsequently receive surgery at parental request provided that
the children continued to meet trial eligibility criteria.
To maximize equivalency of subjects' baseline status, we required that
all subjects be free of otitis media at their trial starting point (ie, the
time from which outcome events were counted). Subjects who had otitis media
at assignment received standardized antimicrobial treatment, as described
below. Subjects assigned to undergo adenoidectomy or T&A were scheduled
for surgery on the earliest date feasible. In those in whom middle-ear effusion
was present at the time of surgery, myringotomy and aspiration of effusion
also were performed. All operations were performed or supervised by a study-team
otolaryngologist; standard surgical techniques were used.17
For all surgical subjects the trial starting point was the day after surgery.
For control subjects who were effusion-free at assignment, the trial starting
point was the following day. For those with effusion at assigment, the starting
point was the first effusion-free day, computed as described below. Control
subjects in whom effusion had been present for 90 days or longer without improvement
underwent myringotomy and aspiration, and their trial starting point was the
Follow-up procedures have been described previously.17,20
These included biweekly inquiries about day-to-day status and clinical assessments
by study-team pediatric nurse practitioners and/or pediatricians using standardized
procedures and algorithms at 6-week intervals; at the time of acute illnesses
(a study-team member was on call at all hours); and for episodes of otitis
media, at 1- to 4-week intervals until resolution. Pneumatic otoscopes with
airtight lens assemblies were used for examining the tympanic membrane. When
the membrane was intact, the diagnosis of otitis media and its classification
as either acute or otitis media with effusion were based on criteria reported
previously.23 In brief, otitis media was classified
as acute when, in addition to middle-ear effusion, a child gave evidence of
recent ear pain and/or showed marked redness or distinct fullness or bulging
of the tympanic membrane. When a tympanostomy tube or preexisting perforation
was present, otitis media evidenced by otorrhea through the tube or the perforation
was classified as acute if accompanied by pain or fever, and otherwise as
"secondary otorrhea." Doubtful otoscopic diagnoses were decided by a study-team
pediatrician or, if still doubtful, by a study-team otolaryngologist on the
basis of otomicroscopic examination. The accuracy of the study team's otoscopic
assessments has been reported.23
Tympanometry was performed at most visits. Tympanograms were classified
and interpreted according to criteria reported previously.24,25
When otoscopic and tympanometric findings appeared discrepant, examinations
were repeated and the final otoscopic (or rarely, otomicroscopic) diagnosis
was used to classify middle-ear status. Audiometry was performed by a certified
audiologist at the time of trial entry, at the first postsurgical visit, during
and following episodes of otitis media as deemed clinically advisable, and
at maximum intervals of 6 months.
For each subject we estimated the cumulative proportions of days on
which unilateral and bilateral otitis media, respectively, were present, based
on diagnoses at individual visits and interpolations for intervals between
visits provided that the intervals did not exceed 90 days when the otitis
status on the 2 visits was the same or 60 days when the status differed. All
days during intervals that exceeded those limits were classified as "otitis-unknown."
In all subjects, each new episode of otitis media of any type was treated
with an antimicrobial in conventional dosage for 10 days to 6 weeks, depending
on recent clinical course and response to treatment. Amoxicillin was used
whenever feasible; second-line drugs mainly used were erythromycin-sulfisoxazole
and amoxicillin-clavulanate potassium. When middle-ear effusion persisted
for 90 days without improvement, myringotomy with aspiration was performed.
When effusion recurred within 6 months after myringotomy and persisted for
60 days without improvement, myringotomy with tube placement was performed.
Secondary otorrhea was treated with an antimicrobial orally for up to 2 weeks,
and if persistent thereafter, with polymyxin B-neomycin-hydrocortisone ototopical
The primary outcome measure was the the number of episodes of acute
otitis media within a follow-up year. Secondary measures were the estimated
proportion of time with otitis media, the numbers of myringotomies and tube
procedures, and the numbers of days, respectively, on which ear pain occurred
and antimicrobial treatment was received.
Sample-size calculation was deferred until experience could be gained
concerning the occurrence rate of episodes of acute otitis media. An interim
analysis after 3 years indicated that 15% of control subjects had remained
free of such episodes during their first follow-up year. On an assumption
that a corresponding value of 40% in surgical subjects (ie, a difference of
25%) would constitute a clinically important difference and setting 2-tailed α
at .05 and β at .20, we estimated that 57 subjects per treatment group
in each of the 2 trials would be required.26
For this degree of difference to justify undertaking surgery in comparably
affected children, we set as a further requirement that the difference be
sustained for 2 years. Outcome data for control subjects withdrawn from control
status to receive surgery were analyzed in 2 modes: in one mode, events that
followed surgery were included and charged to subjects' original status as
controls; in the other mode, such events were excluded.
All statistical tests were 2-tailed. Mean rates of occurrence of episodes
of acute otitis media were compared by assuming a Poisson probability distribution27 and applying a generalized linear model.28 A weighted regression model, with weights equal to
the lengths of observed time, was used to compare mean proportions of time
with otitis media. Regression models also were used in comparisons between
treatment groups to adjust for between-group imbalances in subject characteristics.
Times to occurrence of first, second, and third episodes of acute otitis media
and first accumulation of 30, 60, and 90 days of otitis media were summarized
in life-table analyses and compared using the log-rank test29,30
and an extension of the proportional hazards model for multiple recurrences.31 Analysis of hearing data obtained after subjects'
trial starting points was limited to each subject's first audiogram when middle-ear
effusion was present and first audiogram when middle-ear effusion was absent.
Of the 461 enrolled children, 304 were eligible for the 3-way trial
and 157 for the 2-way trial (Figure 1).
A total of 47 subjects assigned to surgical treatment groups were withdrawn
from the trials without undergoing surgery, in some instances because their
parents lacked insurance coverage and were unwilling to incur the resultant
costs (later, arrangements were made to waive costs for such subjects). Four
subjects assigned to control status were also withdrawn. To conserve resources,
these 51 withdrawn subjects were discharged from the study. Accordingly, follow-up
data were available for 410 subjects: 266 in the 3-way trial and 144 in the
2-way trial. Of the 410 subjects, 354 (86.3%) were followed up for at least
1 year, 308 (75.1%) for at least 2 years, and 250 (61.0%) for 3 years. The
main reason for failure to complete the trial, apart from loss to follow-up,
was termination of the study.
Selected demographic and clinical characteristics of subjects in the
2 trials are summarized in Table 1.
Of the 410 children, 374 (91.2%) were eligible on the basis of recurrent acute
otitis media, 22 (5.4%) on the basis of persistent middle-ear effusion, and
14 (3.4%) on the basis of both conditions. However, of the 374 subjects enrolled
on the basis of recurrent acute otitis media only, 119 (31.8%) had prior histories
of middle-ear effusion of at least 2 months' duration and 189 (50.5%) had
middle-ear effusion at entry that, in many instances, had not been previously
diagnosed. The only significant between-group difference in characteristics
occurred in the 2-way trial, in which the T&A group contained proportionately
more girls and fewer boys than the control group (40% vs 58%; P=.04). In the 3-way trial, primarily in the T&A group, socioeconomic
status was higher among subjects who completed 3 years of follow-up than among
subjects who did not. Otherwise there were no significant differences in characteristics
in either trial between subjects who completed and subjects who did not complete
3 years of follow-up, or between subjects whose parents accepted initial assignment
and subjects discharged without follow-up because of noncompliance with initial
The median and mean intervals from assignment to starting point in the
3-way trial were 73 and 84.0 days, respectively, among adenoidectomy subjects;
60 and 66.4 days, respectively, among T&A subjects; and 8 and 22.6 days,
respectively, among control subjects. Corresponding values in the 2-way trial
were 60 and 63.3 days, respectively, among T&A subjects; and 0 and 19.0
days, respectively, among control subjects.
In analyses for the outcomes presented here, illnesses and treatments
that followed surgery in subjects withdrawn from control status to receive
surgery were charged to subjects' original status as controls. Analyses that
excluded such events gave similar results. Similar results also were obtained
in analyses that excluded the 22 subjects eligible for the trials only on
the basis of persistent otitis media with effusion.
Differences in outcome between the groups in the 3-way trial were generally
small. As shown in Table 2, the
mean rate of episodes of acute otitis media in adenoidectomy subjects was
actually higher than in control subjects in the second follow-up year. In
T&A subjects the rate was lower than in control subjects in the first
follow-up year and for the 3 follow-up years combined (P=.009, data not shown). The mean rate also was lower than in adenoidectomy
subjects in the first follow-up year (P=.03, not
shown) and for the 3 follow-up years combined (P=.04,
data not shown). As shown in Table 3,
both adenoidectomy subjects and T&A subjects had, on average, less estimated
time with otitis media than control subjects in the first follow-up year but
not thereafter. The first-year difference between T&A and control subjects
also was responsible for a difference between these 2 groups for the 3-year
follow-up period as a whole (P=.04, data not shown).
The results of life-table analyses were consistent with these results. Data
concerning the remaining outcome measures are summarized in Table 4. During the first follow-up year T&A subjects received,
on average, less antimicrobial treatment than control subjects; no other between-group
differences during the 3-year follow-up period were significant.
Differences in outcome in the 2-way trial again were generally small.
As shown in Table 2, differences
favoring T&A subjects over control subjects in the mean rate of episodes
of acute otitis media were not significant in any individual follow-up year,
but for the 3 follow-up years combined the difference was significant (P=.005, data not shown). Table 3 shows that T&A subjects had, on average, less estimated
time with otitis media than control subjects during the first follow-up year;
this resulted in a difference between the 2 groups for the 3 follow-up years
combined (P=.02, data not shown). The results of
life-table analyses were consistent with these results. Data for the remaining
outcome measures are summarized in Table
4. During the first and second follow-up years, T&A subjects
received less antimicrobial treatment than control subjects, but no other
differences during the 3-year follow-up period were significant.
In both the 3-way and 2-way trials, among control subjects who eventually
underwent surgery, mean rates of acute otitis media and estimated proportions
of time with otitis media during the segments of follow-up years that preceded
the surgery were modestly higher than the corresponding whole-year values
among control subjects who remained under surveillance without change in status.
Accordingly, had the surgery not been performed and had these higher values
been maintained for the remainders of the follow-up years in question, the
resulting calculated whole-year values in control subjects would have been
marginally higher than those shown in Table
2 and Table 3. Among
surgical and control subjects in both trials who were lost to follow-up, rates
of acute otitis media and estimated proportions of time with otitis media
during the segments of follow-up years that preceded the loss to follow-up
differed relatively little, and in no consistent direction, from the corresponding
whole-year values among subjects who remained under surveillance.
Outcomes were generally less favorable in younger than in older subjects
and in subjects with bilateral effusion at the time of assignment than in
subjects with unilateral or no effusion. However, surgical-vs-control outcomes
did not change substantially after adjusting individually for age group, sex,
middle-ear effusion status at assignment, baseline presence of nasal obstruction
attributable to large adenoids, probable presence of upper respiratory allergy,
and parents' socioeconomic status.
Analysis indicated that hearing acuity was related consistently only
to whether otitis media was present and not to subjects' treatment groups.
When middle-ear effusion was absent, group mean pure-tone averages ranged
from 3.3 to 5.7 dB hearing level, and when effusion was present, from 7.5
dB to 17.4 dB hearing level.
Of the 261 subjects who underwent surgery, 30 (11.5%) developed perioperative
or postoperative complications (Table 5). Complications were more common with T&A than with adenoidectomy.
The mean reported duration of postoperative sore throat in adenoidectomy subjects
was 1.1 days (range, 0-7 days), and in T&A subjects, 5.8 days (range,
0-21 days). In the course of antimicrobial treatment erythematous rashes occurred
in 6 adenoidectomy subjects, 4 T&A subjects, and 7 control subjects; serum
sickness occurred in 1 T&A subject and 1 control subject.
Most previously reported randomized trials of adenoid-tonsil surgery
for otitis media had as their main entry criterion persistent otitis media
with effusion, and most tested the efficacy of adenoidectomy and not of T&A.
Only the trials reported by Rynnel-Dagöö et al32
and by ourselves17 included recurrent acute
otitis media as an entry criterion, and only the trials reported by McKee,7 Maw,8 and Maw and Herod9 compared the efficacy of adenoidectomy with that of
T&A. Rynnel-Dagöö et al32 found
adenoidectomy inefficacious. McKee,7 whose
trial excluded children with "recurrent or chronic otitis media" because those
conditions were considered absolute indications for adenoidectomy, found the
incidence of otitis media (type not specified) during the 2-year follow-up
period lower in the adenoidectomy group than in the T&A group but not
significantly so. The trials reported by Maw8
and Maw and Herod9 found adenoidectomy and
T&A equally efficacious in clearance of middle-ear effusion; neither trial
described the occurrence of acute otitis media. Limitations of these trials,
as discussed previously,17 included nonstringent
eligibility criteria, limited methods for detection and follow-up of otitis
media, failure to specify use or nonuse of antimicrobial treatment for recurrences
of otitis media, and failure to consider the possibility of a type II error.
Our previous trials,17 involving children
who had undergone tympanostomy tube placement and who, after tube extrusion,
had again developed otitis media, showed adenoidectomy to have been effective
in these children for 2 years in reducing the occurrence of otitis media.
The effect was greater with regard to time with middle-ear effusion than with
regard to the occurrence rate of acute otitis media and seemed large enough
overall to be considered clinically important.
The present trials were designed with 3 considerations in mind. First,
to test the efficacy of adenoid-tonsil surgery in children less severely affected
with otitis media than those in our earlier trials and to avoid the potentially
confounding influence of tubes, we designed the trials to be "tubeless" to
the extent feasible by enrolling only children who had not previously undergone
tube placement and, after enrollment, by limiting first surgical interventions
for persistent middle-ear effusion to myringotomy and aspiration alone. Second,
we wished to test the efficacy of adenoid-tonsil surgery in children who had
troublesome histories specifically of recurrent acute otitis media, whether
or not they also had had persistent middle-ear effusion. Finally, in children
with either recurrent acute otitis media, or persistent otitis media with
effusion, or both, we wished to compare the efficacy of adenoidectomy with
that of T&A. To meet these objectives and at the same time avoid mixing
children who might differ prognostically, separate 3-way and 2-way trials
In both of these trials, the efficacy of adenoid-tonsil surgery proved
to be quite modest with respect to both the occurrence of acute otitis media
and the proportion of time spent with otitis media. However, within those
bounds, in the 3-way trial T&A appeared narrowly more efficacious than
adenoidectomy. In neither trial did efficacy reach the level we initially
posed as necessary to justify surgery in comparably affected children. The
rate of surgical complications in the 2 trials was not inconsiderable.
In both our trials, failure to comply with treatment assignment occurred
more often among subjects assigned to undergo surgery than among those assigned
to control status (Figure 1). If
such failure to comply involved less severely affected subjects disproportionately,
the resulting bias might have led to an understatement of the efficacy of
surgery by trial findings. Conversely, assumptions by parents that surgery
was efficacious might have led to less vigilance and less detection of illness
among surgical subjects than among control subjects, which would have resulted
in an overstatement of the efficacy of surgery. The possibility of observer
bias was minimized by the adjunctive use of tympanometry. We think that the
potential bias introduced by differences between surgical and control subjects
in times to starting points was limited in effect and less consequential than
the bias that might have resulted had we not ensured initial effusion-free
status in all subjects. Finally, as noted previously, our calculations suggest
that bias resulting from the withdrawal of certain subjects from control status
in order to undergo surgery or from the loss of subjects to follow-up did
not affect trial outcomes substantially.
In the present trials, subjects' illnesses might have been minimized
as a consequence of the trials' unusually close monitoring and prompt and
sustained treatment. Accordingly, it seems possible that in nonresearch settings
the degree of benefit afforded by surgery might differ from that found in
How should these findings be applied? Our study sample seemed similar
to children generally who are considered candidates for adenoidectomy or T&A
because of histories of recurrent acute otitis media. Given that we found
both operations to have limited efficacy and in view of their not inconsiderable
risks, morbidity, and costs, we believe that neither operation ordinarily
should be considered as an initial intervention in such children. Instead
we believe that nonsurgical management33 should
be attempted first, with tube placement to be considered later as the preferred
first surgical recourse if the burden of illness becomes intolerable,34 and with adenoidectomy reserved for those who develop
recurrent otitis media after tubal extrusion.17
Our finding that T&A appeared narrowly more efficacious than adenoidectomy
prompts 2 further questions: whether the occurrence of otitis media is related
to infection not only in the nasopharynx but also in the oropharynx; and relatedly,
whether children, such as those in our earlier trials,17
who have undergone tube placement and who continue to develop otitis media
after tube extrusion might benefit more from T&A than from adenoidectomy
alone. The latter question could be answered only by a randomized clinical
trial in which such children undergo either adenoidectomy or T&A, but
the additional risks posed by the addition of tonsillectomy to the procedure
would, in our judgment, outweigh any potential benefit.
Create a personal account or sign in to: