Oral Erythromycin Prophylaxis vs Watchful Waiting in Caring for Newborns Exposed to Chlamydia trachomatis

Background  Chlamydia trachomatis exposure at birth may cause conjunctivitis or pneumonia. Until recently, a course of oral erythromycin prophylaxis was recommended for C trachomatis– exposed neonates. However, recognition of an association between erythromycin and pyloric stenosis prompted a change to a watchful waiting recommendation under which only infants who develop symptomatic C trachomatis infection are treated with oral erythromycin.

Objective  To compare erythromycin prophylaxis with watchful waiting for a hypothetical cohort of 100 000 neonates exposed to C trachomatis.

Methods  In a decision tree, potential outcomes were C trachomatis conjunctivitis, C trachomatis pneumonia (which could require inpatient or outpatient therapy), no clinical disease, and pyloric stenosis. Published data were reviewed to derive probability point estimates and ranges. Estimated charges served as outcome measures.

Results  Watchful waiting is less expensive than erythromycin prophylaxis ($15.1 million vs $28.3 million); prophylaxis prevents 5986 cases of C trachomatis pneumonia, including 1197 hospital admissions, but causes 3284 pyloric stenosis cases. (For every 30 infants given oral erythromycin prophylaxis, one additional case of pyloric stenosis would be expected to occur, and approximately 1.8 cases of C trachomatis pneumonia would be prevented.) In sensitivity analyses, if more than 3.4% of exposed neonates are hospitalized for C trachomatis pneumonia, prophylaxis becomes favored.

Conclusions  This study supports the watchful waiting recommendation for asymptomatic C trachomatis–exposed neonates. However, there are wide plausible ranges for pyloric stenosis risk after erythromycin administration and for the incidence of C trachomatis pneumonia severe enough to require hospitalization; under some combinations of these rates, prophylaxis could be favored.

GENITAL INFECTION with Chlamydia trachomatis is the most commonly reported sexually transmitted disease in the United States.¹ Because of the high prevalence of C trachomatis infection in women of childbearing age, it is estimated that more than 100 000 US neonates are exposed during the birth process annually.² Many of these infants develop conjunctivitis, pneumonia, or both in the first few months of life.³Chlamydia trachomatis pneumonia, although often mild, may cause severe symptoms that require hospitalization.² Until recently, the Committee on Infectious Diseases (the Red Book Committee) of the American Academy of Pediatrics recommended a prophylactic 14-day course of oral erythromycin for newborns exposed to C trachomatis at delivery.^4,5 However, this recommendation was changed after the recognition,⁶ since confirmed,^7,8 of an association between erythromycin and subsequent pyloric stenosis. The current recommendation is for watchful waiting, treating with erythromycin only those infants who develop symptomatic C trachomatis infection.⁹

We conducted a decision analysis of these 2 strategies—erythromycin prophylaxis vs watchful waiting—for care of asymptomatic C trachomatis–exposed neonates. In this analysis, the benefits of erythromycin prophylaxis (lower rates of conjunctivitis and pneumonia) are weighed against the risk (pyloric stenosis). The recent discovery of the risk of pyloric stenosis following erythromycin makes a decision analysis potentially useful; before studies had shown any serious adverse consequences of neonatal erythromycin therapy, a formal analysis of the risk-benefit tradeoff would not have been required to justify prophylaxis.¹⁰ A decision analysis highlights and weighs the fundamental factors on which clinical decisions can be objectively based and can facilitate individualized recommendations in different clinical situations. The decision analysis method also points out areas where further research might be helpful.

For a hypothetical cohort of 100 000 neonates exposed to C trachomatis, the decision analysis compares a prophylaxis strategy (providing oral erythromycin beginning in the first few days of life to all 100 000 infants) with a watchful waiting strategy (treating with erythromycin only those infants who develop symptomatic C trachomatis infection). Figure 1 shows the decision tree. Potential outcomes for infants without prophylaxis are the development of C trachomatis conjunctivitis or pneumonia (possibly requiring inpatient care) or no clinical disease. With prophylaxis, the probabilities of conjunctivitis and pneumonia are reduced, but the risk of pyloric stenosis is much higher. For simplicity, we assumed that symptomatic C trachomatis infections would be successfully treated with erythromycin and that, since C trachomatis pneumonia usually presents after 2 weeks of age¹¹ (beyond the period during which treatment with erythromycin seems to increase the risk of pyloric stenosis most greatly^6-8), treatment of pneumonia would not increase the subsequent risk of pyloric stenosis. Estimated charges were used as outcomes; the possibility of death from either pyloric stenosis or C trachomatis was not considered, because the risk of death from either cause is very low and difficult to quantify. The analysis was conducted using the software program Data 3.5 (TreeAge Software, Williamstown, Mass).

MEDLINE searches and cross-linking of reference lists using Web of Science were used to examine all published studies of infants exposed to C trachomatis at birth. The terms chlamydia, trachomatis, conjunctivitis, pneumonia, and neonatal, newborn, or infant were used to search worldwide for English-language studies of transmission of C trachomatis from cohorts of infected mothers to their infants. We used data from those studies that reported both the number of infants exposed to C trachomatis and the number who developed confirmed C trachomatis conjunctivitis or pneumonia.^11-23 (Other types of studies, such as those that began with a set of infected infants, were not included because they could not be used to establish the incidence of symptomatic C trachomatis infections among exposed infants.) The data were aggregated, weighting studies in proportion to their sample size, to produce pooled estimates of probability. Studies that did not attempt to identify infants with pneumonia were used only for the probability estimate for conjunctivitis.^12,18 Those that were designed to study conjunctivitis but also reported the number of infants with pneumonia were included in the pneumonia calculation²² unless the minimum follow-up was 1 month or less²¹ (before the peak incidence of C trachomatis pneumonia at 6 to 9 weeks of age).¹¹ We did not include studies of maternal-infant transmission that documented microbiologic or serologic evidence of infection but did not report rates of conjunctivitis or pneumonia.

For the base-case analysis, we used a published estimate of the likelihood that C trachomatis pneumonia will lead to hospitalization.² Only one study,²⁴ to our knowledge, has reported an empirical hospitalization rate. In sensitivity analyses, we varied this rate widely, with 0% as the lower bound.

We found no studies of the efficacy or effectiveness of oral erythromycin in preventing symptomatic C trachomatis infection in exposed infants; we estimated effectiveness from studies of erythromycin treatment of C trachomatis conjunctivitis that reported the success rate in eradicating the organism from the infants' eyes.^11,16,25-30 The data were aggregated, weighting studies in proportion to their sample size, to produce a pooled estimate of probability. Some of these studies compared the effectiveness of oral erythromycin therapy with the effectiveness of topical^25-27 or other oral antibiotic therapy³⁰; others did not have a comparison group.^28,29 We also included data from 2 of the birth cohort studies^11,16 used for estimating the incidence of conjunctivitis and pneumonia; these 2 provided details about infants who were treated with oral erythromycin for conjunctivitis. The probabilities of conjunctivitis and pneumonia under the prophylaxis strategy were calculated by multiplying the corresponding probabilities from the watchful waiting strategy by the likelihood that erythromycin prophylaxis would be ineffective (one minus the estimated effectiveness).

Pyloric stenosis may develop regardless of whether an infant is exposed to erythromycin; its likelihood under each strategy was estimated from the reported absolute risks for exposed infants and unexposed infants in the 2 published cohort studies^6,7 of erythromycin given during the first week of life and subsequent pyloric stenosis.

The Health Care Utilization Project (HCUP) database of charges from children's hospitals³¹ includes information for hospitalizations for pyloric stenosis and C trachomatis pneumonia. We queried the HCUP database for infants younger than 1 year admitted with a principal International Classification of Diseases, Ninth Revision (ICD-9) code of 750.5 (infantile hypertrophic pyloric stenosis) or 483.1 (chlamydial pneumonia). ICD-9 code 483.1 includes both C trachomatis and Chlamydia pneumoniae infections; however, including only infants hospitalized at younger than 1 year should largely exclude C pneumoniae infections.^32,33 Estimated charges for outpatient treatment of C trachomatis pneumonia and conjunctivitis and for a 14-day prophylactic course of erythromycin were based in part on cost estimates published by the Institute of Medicine.² These cost estimates were transformed into charges using a ratio of 1.32, which was derived by dividing the HCUP charge by the Institute of Medicine cost estimate for a hospitalization for chlamydial pneumonia. We assumed that prophylaxis would be prescribed in the normal course of newborn nursery care and would not lead to any additional charges or produce any adverse effects other than pyloric stenosis. Because all modeled events occur within a few months, the modeled outcomes (charges) were not time discounted.

The base-case analysis involved point estimates of probabilities and charges derived from the literature as described herein; for infants with pyloric stenosis or C trachomatis pneumonia that required hospitalization, median inpatient charges were used. In sensitivity analyses, to determine which elements of the model were important determinants of the ultimate decision, we varied the base-case probabilities over ranges that have been reported. We also examined whether the decision was sensitive to changes in the estimated charges for each outcome.

Among infants exposed to C trachomatis at birth, the point estimate of the incidence of conjunctivitis is 15% (Table 1) and of pneumonia is 7% (Table 2). The published estimate of the hospitalization rate among infants with C trachomatis pneumonia is 20%²; thus, among all infants exposed to C trachomatis, 1.4% would be hospitalized. The effectiveness of erythromycin prophylaxis is estimated to be 85% (Table 3). The estimated risk of pyloric stenosis is 3.5% among treated infants compared with 0.26% in untreated infants (Table 4). Median inpatient charges in the HCUP database³¹ for children younger than 1 year are $5807 for chlamydial pneumonia and $5550 for pyloric stenosis; estimated charges for outpatient treatment of C trachomatis pneumonia and conjunctivitis are $462 and $198, respectively, and a course of erythromycin prophylaxis is $66 (Table 5).^2,31

For infants exposed to C trachomatis at birth, published incidence rates vary from 8% to 44% for C trachomatis conjunctivitis (Table 1) and from 0% to 17% for C trachomatis pneumonia (Table 2). We varied the rate of hospitalization among infants with C trachomatis pneumonia from 0% to 39%, the only available empirical figure. (In a study²⁴ at an inner-city teaching hospital, 7 [39%] of 18 infants with C trachomatis pneumonia were hospitalized.) The reported effectiveness of erythromycin ranges from 68% to 93% (Table 3); we used 100% as an upper bound. For infants exposed to erythromycin in the first week of life, the published risk of pyloric stenosis varies from 2.7% to 4.5% (Table 4). For infants not exposed to erythromycin, we used the range of rates in large epidemiologic studies tabulated by Schechter et al³⁴ (0.011% to 0.31%). Although these studies include an unknown number of erythromycin-exposed infants, most cases of pyloric stenosis occur in infants who were not treated with erythromycin.⁷ Therefore, the risk in these general population studies only slightly overestimates the risk among unexposed infants. The risk of pyloric stenosis specifically in unexposed infants is reported in only one large study, to our knowledge⁷; its point estimate falls within the range reported by Schechter et al.³⁴

In the base-case analysis, using the summary rates and charges fromTables 1 through 5, watchful waiting is less expensive than prophylaxis (Table 6). With the watchful waiting strategy, the estimated charge per 100 000 chlamydia-exposed infants is $15 100 000, most of which is for inpatient care for infants with pneumonia ($8 170 000). With the prophylaxis strategy, the estimated charge is $28 300 000, most of which is for care of infants with pyloric stenosis ($19 650 000). In the hypothetical cohort of 100 000 neonates, erythromycin prophylaxis prevents 5986 episodes of C trachomatis pneumonia (1197 of which would have required hospitalization) but increases the number of pyloric stenosis cases by 3284. That is, for every 30 infants given oral erythromycin prophylaxis, 1 additional case of pyloric stenosis would be expected to occur, and approximately 1.8 cases of C trachomatis pneumonia would be prevented.

The watchful waiting recommendation holds across all of the probability ranges, considered one at a time. The estimate of the effectiveness of erythromycin prophylaxis does not affect the choice of strategies. If prophylaxis prevented all cases of C trachomatis pneumonia (7035 cases and 1407 hospitalizations among 100 000 infants), that would not outweigh the 3284 additional cases of pyloric stenosis; watchful waiting would still be strongly favored. If the risk of pyloric stenosis following neonatal erythromycin administration were less than 1.2% (less than the lowest published risk) or if the incidence of C trachomatis pneumonia among exposed infants were greater than 17.1% (slightly higher than the highest published incidence), then the prophylaxis strategy would be preferred in the model. Allowing 2 or more factors to vary reveals plausible circumstances that would favor a decision to offer prophylaxis. The decision is sensitive to simultaneous variation in the incidence of C trachomatis pneumonia and in its hospitalization rate. The product of these 2 variables equals the probability of hospitalization for C trachomatis pneumonia, given exposure at birth. This hospitalization incidence could be as high as 7% (if C trachomatis pneumonia affects 17% of exposed neonates, with a 39% hospitalization rate); if the true hospitalization incidence were greater than 3.4% of exposed neonates, the prophylaxis strategy would be favored.

Figure 2 shows how the recommendation is affected by varying, at the same time, assumptions about the hospitalization incidence (x-axis) and the risk of pyloric stenosis following erythromycin administration (y-axis). The study with the most complete clinical follow-up of exposed infants reported a 16% incidence of C trachomatis pneumonia¹¹; assuming a hospitalization rate of 20%, the probability of hospitalization for C trachomatis pneumonia, given exposure at birth, would be 3.2%. If, at the same time, the risk of pyloric stenosis after neonatal erythromycin exposure were 2.7%,⁷ rather than 3.5% (base case), then prophylaxis would be favored.

Inpatient charges for pneumonia and pyloric stenosis account for most of the hypothetical cohort's charges. Patients with pneumonia and/or conjunctivitis who are treated as outpatients have little effect on the results of sensitivity analyses of incidence or charges. The charge for erythromycin does not influence the choice of strategies. (In the base-case model, watchful waiting would still be favored if erythromycin were free.) Even wide variation in the charges for inpatient care of pneumonia and pyloric stenosis would not alter the base-case preference for watchful waiting (Figure 3). When we replaced the base-case HCUP median inpatient charges (pyloric stenosis, $5500; C trachomatis pneumonia, $5807) with HCUP mean³¹ inpatient charges (pyloric stenosis, $6685; C trachomatis pneumonia, $10 211), watchful waiting still was preferred. However, this change in inpatient charge estimates (a 30% reduction for pyloric stenosis in relation to C trachomatis pneumonia) combined with a greater than 2.5% probability of hospitalization for C trachomatis pneumonia (given exposure at birth) would sway the decision toward prophylaxis.

This analysis supports the 2000 Red Book Committee's recommendation that asymptomatic C trachomatis–exposed newborns not receive oral erythromycin prophylaxis. The decision rests on 2 key probabilities: the risk (in the absence of erythromycin prophylaxis) of developing pneumonia severe enough to require hospitalization, and the risk (given erythromycin prophylaxis) of pyloric stenosis. There is considerable uncertainty about the actual rates of these events, and the sensitivity analysis shows that there is a possibility that the true values, if known, could favor prophylaxis. Therefore, research to better understand the magnitude of these risks would be of particular value.

Uncertainty exists because the C trachomatis pneumonia studies are diverse and the pyloric stenosis studies are few. The 2 largest cohort studies^11,20 of infants exposed to C trachomatis reported pneumonia incidences of 3% and 16%, a broad range. These studies used relatively insensitive diagnostic tests, such as cell culture techniques, which may not have detected C trachomatis in specimens from some infants with symptomatic illness. Nucleic acid amplification tests might be able to detect more infections, but these have not been approved for use in infants and were not available at the time these studies were performed. Only 2 published studies quantify the risk of pyloric stenosis following erythromycin exposure between birth and the seventh day of life. Honein et al⁶ compared 157 newborns treated prophylactically with erythromycin for pertussis exposure with 125 untreated infants; Mahon et al⁷ studied a birth cohort (n = 14 876) in which 182 newborns were prescribed erythromycin in the first week of life (see Table 4 for the reported risks). A recent analysis of claims data reported, in a much larger population, an adjusted incident rate ratio of 7.88 for pyloric stenosis after erythromycin exposure between the 3rd and 13th days of life.⁸ The magnitude of the association in this study is smaller than those in the earlier studies but again demonstrates a strong and statistically significant association, the extent of which lends support to the watchful waiting recommendation. As with the incidence of C trachomatis pneumonia, however, it is impossible from these 3 studies to quantify the absolute risk with much precision.

Apart from the decision analysis results, there are additional reasons why the watchful waiting recommendation is acceptable medical practice. Most clinicians shun iatrogenic harm more than they value preventing equivalent illness-related morbidity.³⁵ In addition, 2 aspects of the model bias the results toward erythromycin prophylaxis and thereby strengthen the finding that watchful waiting still is preferred. First, the model does not consider that C trachomatis conjunctivitis has a peak onset at an earlier age than C trachomatis pneumonia; to the extent that infants who develop C trachomatis conjunctivitis are appropriately treated with a course of oral erythromycin, the watchful waiting strategy could prevent pneumonia in these infants. Second, by leaving out any adverse effects of erythromycin (other than pyloric stenosis), the model makes erythromycin prophylaxis appear somewhat more desirable than it actually would be.

One purpose of decision models is to make explicit the tradeoffs involved in choosing one course of action over another. To measure outcomes, we used hospital charges, since these have the advantages of being readily available and a credible and easily compared measure. Less easily quantified outcomes, such as the burdens (often termed disutilities) that illness imposes on infants and their families, are also relevant and could offer different advantages in decision analysis. However, the brief duration of both pyloric stenosis and C trachomatis pneumonia deters measurement of formal disutility estimates; there is only one published disutility estimate for C trachomatis disease in infants² and none for pyloric stenosis.

Nevertheless, clinicians and parents might compare their own preferences to the relative weights assigned in this model. The tradeoff to be weighed is that a prophylaxis strategy would cause approximately 2.7 additional cases of pyloric stenosis at the same time that it prevented 5 cases (1 hospitalization and 4 outpatient cases) of C trachomatis pneumonia. Using charges as a measure of morbidity, pyloric stenosis ($5550) is about equal to C trachomatis pneumonia that requires hospitalization ($5807) and about 12 times worse than C trachomatis pneumonia that does not require hospitalization ($462). Chlamydia trachomatis pneumonia inpatient and outpatient charges combined, weighted by their respective frequencies in the model (20% and 80%), equal $1531 per case, about one fourth the charge for a case of pyloric stenosis. A decision maker who considers pyloric stenosis to be more serious than its relative weight in our model, in general or in specific circumstances (for instance, a clinician or parent deciding about prophylaxis for an infant with higher-than-average anesthetic risk), would be less inclined to give erythromycin prophylaxis than the model suggests. On the other hand, if future studies confirm long-term consequences of neonatal chlamydia exposure, as have been suggested by studies^36-38 showing that infants with C trachomatis pneumonia may have long-term chronic pulmonary effects and by reports^39-41 of infants with neonatal chlamydia infection and myocarditis or sudden infant death syndrome, then the relative value of preventing C trachomatis infection would increase in comparison with avoiding pyloric stenosis.

Future studies also might establish alternative treatments or strategies that would make this decision moot. Azithromycin or other drugs might be shown to be both effective and safe prophylaxis for infants exposed to C trachomatis. Delaying erythromycin prophylaxis until an infant is at least 2 weeks old, when the risk of pyloric stenosis seems to be lower,^7,8 might also be a reasonable strategy, although it would not prevent all cases of C trachomatis pneumonia.⁴² Ideally, effective public health and obstetric interventions might greatly decrease the number of infants exposed to C trachomatis.

Article

Corresponding author and reprints: Marc B. Rosenman, MD, Regenstrief Institute, 1050 Wishard Blvd, Indiana University School of Medicine, Indianapolis, IN 46202 (e-mail: mrosenma@iupui.edu).

Accepted for publication January 1, 2003.

After the recognition of an association between oral erythromycin and subsequent pyloric stenosis, the American Academy of Pediatrics revised its recommendation for asymptomatic C trachomatis–exposed neonates; the old recommendation of a 14-day course of oral erythromycin prophylaxis was changed to a new recommendation of watchful waiting for conjunctivitis or pneumonia. However, no published studies have examined the risk-benefit tradeoff involved in deciding whether watchful waiting or erythromycin prophylaxis is preferable.

This study synthesizes published estimates: rates of symptomatic C trachomatis illnesses among exposed infants (and the rate of hospitalization for C trachomatis pneumonia), the effectiveness of oral erythromycin, and the risks of pyloric stenosis with and without oral erythromycin exposure. Using decision analysis methods, the 2 strategies—watchful waiting vs erythromycin prophylaxis—are compared. The results support the watchful waiting recommendation. The study also reveals substantial variation in reported rates of C trachomatis pneumonia and of pyloric stenosis after oral erythromycin exposure; there is a possibility that the true values of these rates, if known, could favor a prophylaxis recommendation.