A Meta-analysis of the Efcacy of Ocular Prophylactic Agents Used for the Prevention of Gonococcal and Chlamydial Ophthalmia Neonatorum

Elizabeth K. Darling, RM, MSc, and Helen McDonald, RM, MHSc

Introduction: Neonatal eye prophylaxis has been routine in North America for more than a century. Contextual changes justify reexamining this practice, and prompted a systematic review of the efcacy of prophylactic agents. Methods: We searched MEDLINE (19662008), EMBASE (19802008), CINAHL (19822008), and the Cochrane library (the rst quarter of 2008) for relevant clinical trials and hand-searched the resulting reference lists. We independently evaluated eligibility and study quality. Meta-analyses were performed using a random effects model. Results: Each of the eight included studies had substantial methodologic weaknesses. Data to estimate the efcacy of prophylaxis in the prevention of gonococcal ophthalmia neonatorum (GON) were not available. One study found no differences in rates of chlamydial ophthalmia neonatorum (CON) when three agents were compared to no prophylaxis: silver nitrate (relative risk [RR] = 1.06; 95% condence interval [CI], 0.552.02; 2225 newborns), erythromycin (RR = 0.93; 95% CI, 0.481.79; 2306 newborns), and tetracycline (RR = 0.82; 95% CI, 0.421.63; 2299 newborns). No statistically signicant differences were found between agents in the prevention of GON. Erythromycin and povidone-iodine both decrease the risk of CON when compared to silver nitrate (RR = 0.71; 95% CI, 0.520.97; 4514 newborns, and RR = 0.52; 95% CI, 0.380.71; 2005 newborns, respectively). Discussion: Failure rates of universal eye prophylaxis support reexamination of this policy where the prevalence of maternal infection is low. J Midwifery Womens Health 2010;55:319327 2010 by the American College of Nurse-Midwives. keywords: ophthalmia neonatorum, conjunctivitis, bacterial, infant, newborn, prophylaxis, systematic review, antibacterial agents, chlamydia, gonorrhea

INTRODUCTION When neonatal eye prophylaxis was introduced in the late 1800s, it led to a dramatic reduction in gonococcal ophthalmia neonatorum (GON) and prevented many cases of childhood blindness.1 Mandatory neonatal eye prophylaxis was soon legislated in much of North America,2 and it remains a routine part of care to this day in most areas of the United States and Canada. Over the years, the development of antibiotics and other advances in health care have signicantly altered the context for this intervention. Most women who receive prenatal care are screened for chlamydia and gonorrhea and are successfully treated with antibiotics before giving birth if they are infected. Likewise, if a newborn develops ophthalmia neonatorum (ON) in a setting with adequate postpartum care, access to antibiotic therapy makes blindness extremely unlikely. Another change was the identication of Chlamydia trachomatis in 1907 and subsequent recognition in the 1930s and 1940s that it can also cause ON.3,4 This infection is much more prevalent in North American

Address correspondence to Elizabeth K. Darling, RM, MSc, Midwifery Group of Ottawa, 700-265 Carling Ave., Ottawa, ON K1S 2E1, Canada. E-mail: ldarling@laurentian.ca

women than gonorrhea (e.g., 543.6 per 100,000 vs. 123.5 per 100,000 in the United States in 2007).5 Chlamydial ON (CON) progresses more slowly than GON and is less likely to cause blindness. In wealthy countries, rates of ON are often extremely low. In the United States in 2002 the rate was 8.5 per 100,000 births.6,7 Several other countries no longer require universal prophylaxis (e.g., Britain, Australia, Sweden, Norway, and Denmark), with some abandoning prophylaxis completely and others offering parental choice.8 In settings where universal prophylaxis has been abandoned, prenatal screening and treatment of sexually transmitted infections along with selective neonatal prophylaxis are used to prevent ON. In Britain, where routine prophylaxis was abandoned in the mid-1950s, no cases of blindness resulting from GON were reported during the rst 25 years after this change.8 A crude comparison of the rates of ON in Canada and Britain suggests that routine eye prophylaxis in Canada may have limited impact on the rate of ON.9 Rates of GON and CON in the United Kingdom in 2003 were 3.7 per 100,000 and 6.9 per 100,000, respectively,10,11 while in Ontario, Canada, the combined rate of GON and CON was 4.5 per 100,000 in 2004.12,13 Overall rates of chlamydial and gonorrheal infections were 161.5 per 100,000 and 41.9 per 100,000, respectively, in the United Kingdom in 2003,14
319
1526-9523/$36.00  doi:10.1016/j.jmwh.2009.09.003

Journal of Midwifery & Womens Health  www.jmwh.org

2010 by the American College of Nurse-Midwives Issued by Elsevier Inc.

and 197.1 per 100,000 and 28.9 per 100,000, respectively, in Ontario in 2004.15 Given the contextual changes since the introduction of neonatal eye prophylaxis, and the evidence from settings where this policy has been abandoned, it seems reasonable to reevaluate the utility of this practice. A systematic review of the efcacy and safety of agents used to prevent ON was conducted to facilitate one aspect of such an evaluation. The objective of this review is to determine the efcacy of prophylactic agents used to prevent GON and CON and to compare the efcacy of different agents in the prevention of these infections. METHODS Studies were eligible for inclusion if they were randomized or quasirandomized controlled trials. Study participants were newborns. Eligible interventions included one or more of the following comparisons: one prophylactic agent applied to the eyes of a newborn at or shortly after birth versus placebo, no prophylactic agent, or a different prophylactic agent. Studies were eligible for inclusion if GON or CON was an outcome. ON was dened as conjunctivitis occurring within the rst 28 days of life, and was determined clinically by the presence of symptoms such as redness, swelling, and discharge. GON and CON were dened as the presence of ON symptoms plus positive culture. MEDLINE (1966January 2008) was searched using the following MeSH terms and textwords: conjunctivitis, inclusion; conjunctivitis, bacterial; ophthalmia neonatorum; trachoma; infant, newborn; prophylaxis.mp; prevention.mp; and clinical trial (publication type). Searches were also conducted in the Cochrane Central Register of Controlled Trials (CCRCT), EMBASE, and CINAHL using a similar strategy. Reference lists of published systematic and narrative reviews and all included studies were reviewed to identify additional studies. Unpublished material was not solicited, but clarication or additional information was requested from published authors. Abstracts were eligible for inclusion if a full publication could not be identied through a MEDLINE search or contacting the authors. Language restrictions were not applied to the search, but translation was only available for studies published in French, German, and Swedish. Both authors independently selected articles for retrieval from a literature search conducted by the primary

author (E.K.D). Articles were independently assessed for inclusion eligibility. Where multiple publications of the same trial were identied, the single most comprehensively reported publication was included. Each author independently assessed methodologic quality using standardized criteria to evaluate randomization, allocation concealment, blinding of interventions and outcomes, and completeness of follow-up data. Attempts to resolve uncertainty about methodology were made by contacting the corresponding authors. Responses were obtained from three authors (Fischer and Reta,16 Isenberg et al.,17 and Ramirez-Ortiz et al.18) and independently considered before comparison of the assessments of study quality. The reviewers then met to compare assessments of study eligibility and methodologic quality, and differences were resolved by consensus. Statistical analyses were conducted using Review Manager software (v. 4.2.9; The Nordic Cochrane Center, Copenhagen, Denmark). Outcomes are reported using relative risk (RR) for categorical data, with 95% condence intervals (CIs). Double data entry of outcome data was performed by the primary author and veried for accuracy by the second author. Meta-analysis was conducted where possible. Tests for heterogeneity were performed. A random effects model was used because of its tendency to yield more conservative estimations of treatment effects and wider CIs. It was decided a priori that if there were serious concerns about the validity of the results of any of the studies included in the metaanalysis because of other methodologic weaknesses (e.g., a high proportion of participants lost for follow up), further secondary analyses would be conducted excluding the studies of concern. Examination of funnel plots to assess for evidence of publication bias was planned. Agreement between the authors on evaluation of methodologic quality was assessed using calculations of weighted Kappa for three ordered categories (yes, cannot tell, and no). RESULTS Description of Studies The nal literature searches were conducted on January 31, 2008. Twenty-three studies were identied in the searches in MEDLINE, the CCRCT, and EMBASE. Two of these studies were excluded based on their abstracts, and 21 were retrieved for assessment.8,1635 Four additional studies were identied from the reference lists of retrieved studies,3639 and one additional study was identied from the reference list of an evidence-based guideline on prophylaxis for ON.40 All together, a total of 26 studies were retrieved for assessment. Of the 26 publications retrieved, eight studies were eligible for inclusion in this review. Many of the ineligible
Volume 55, No. 4, July/August 2010

Liz Darling is a registered midwife who practices in Ottawa, Ontario, Canada, and is an Assistant Professor in the Midwifery Education Program at Laurentian University. She has a Master of Science degree in Health Research Methodology from McMaster University in Hamilton, Ontario, Canada. Helen McDonald is a registered midwife who practices in Hamilton, Ontario, Canada. She is an Associate Professor in the Midwifery Education Program at McMaster University in Hamilton, Ontario, Canada.

320

publications were excluded because they did not report GON or CON as an outcome.8,21,22,27,31,33,36,38,39 Two publications were review articles32,34 one was a commentary,29 one study did not investigate an eligible intervention,19 translation was not available for two publications (so eligibility could not be fully determined),28,35 one study was an observational design,40 and the remaining two studies were secondary analyses of included studies.20,23 One of the included studies was published in French,24 and the rest were published in English. Together, the eight included trials report the outcomes of more than 26,000 newborns. A summary of the study methods, participants, interventions, outcomes, and results of each trial is provided in Table 1. Methodologic Quality of Included Studies The included studies all had at least one area of substantial methodologic weakness, as discussed below. Agreement between the reviewers in the assessment of study quality was good, with the mean weighted Kappa for the ve different questions being 0.854. Randomization Hammerschlag et al.30 reported that participants were randomized but did not specify the randomization method. In the remaining studies, alternate treatment allocation methods were used (i.e., quasi-random treatment assignment).1618,2426,37 Rotation or randomization occurred on a daily,24 weekly,1618,25,26 or monthly37 basis. Chen,37 who used a monthly rotation of treatment, was most vulnerable to bias that might be introduced by seasonal variation in the prevalence of maternal infections.41 Allocation Concealment Hammerschlag et al.30 did not report if allocation concealment was achieved. Concealment of treatment allocation was not possible in any of the other trials, which all used quasirandom allocation methods. Blinding of the Interventions Blinding of the interventions was not described in any trials. In at least some of the trials, it is probable that parents were unaware of the agent received by their newborn, because prophylaxis is often administered outside of their presence, but povidone-iodine drops will stain the skin and sclera. Blinding of Outcome Measurement Isenberg et al.17 and Brussieux et al.24 described some blinding of the assessment of outcomes. Blinding of some outcome assessment (e.g., laboratory analysis) may have occurred in the remaining trials but was not reported.
Journal of Midwifery & Womens Health  www.jmwh.org

Follow Up Incomplete follow up was a signicant issue in all trials. Losses to follow up tended to occur after newborns were discharged from hospital. The usual duration of hospital stay was not reported in most studies. Hammerschlag et al.25and Chen37 did not report the number of participants that returned for scheduled follow-up appointments. In Ramirez-Ortiz et al.,18 Brussieux et al.,24 and Laga et al.,26 the majority of newborns did not complete scheduled follow-up appointments. Isenberg et al.17 did not schedule follow-up appointments but instead relied on parents to return for assessment if they observed signs of infection. Fischer and Reta16 did not report follow-up details. Only Hammerschlag et al.,30 in their small study of 60 newborns born to mothers who had chlamydia, reported complete follow up after exclusions. Efcacy of Neonatal Eye Prophylaxis Prevention of Gonococcal Ophthalmia Neonatorum No trials provided estimations of the efcacy of prophylactic interventions in the prevention of GON based on comparisons with participants assigned to a no prophylaxis group. Chen37 included a no prophylaxis group but found no cases of GON (4544 newborns). In Fischer and Reta,16 a large portion of newborns did not receive their allocated treatment because those administering it disliked how messy it was; the only observed cases of GON in this study occurred in these untreated newborns. Analysis based on actual treatment rather than intention to treat found no signicant difference in the rate of GON between silver nitrate and no prophylaxis (RR = 0.06; 95% CI, 0.001.06; 327 newborns) and between tetracycline and no prophylaxis (RR = 0.11; 95% CI, 0.012.03; 214 newborns).16 However, this study was not adequately powered to detect statistically signicant differences in this outcome and provides poor quality evidence for such an estimate. Comparison of Efcacy: Gonococcal Ophthalmia Neonatorum Table 2 presents comparisons of prophylactic agents with respect to the prevention of GON. Erythromycin, tetracycline, and povidone-iodine are each compared to silver nitrate, and povidone-iodine is then compared to erythromycin. No statistically signicant differences were found between these prophylactic agents in the prevention of GON. Secondary analyses excluding trials without true allocation concealment were not possible because all contributing studies were quasirandomized. Prevention of Chlamydial Ophthalmia Neonatorum Estimates of the efcacy of silver nitrate, erythromycin, and tetracycline in the prevention of CON are only
321

Chloramphenicol

Povidone-Iodine

Author and Date Ramirez-Ortiz et al., 2007 Isenberg et al.,17 1995

18

Methods Quasi-RCT; interventions randomly assigned weekly Quasi-RCT; interventions rotated weekly

Participants and Setting 2004 newborns from three rural hospitals in Southern Mexico 3117 newborns in a hospital setting in Kikuyu, Kenya

Gonococcal ON U U U U U U U

No Prophylaxis

Chlamydial ON

Silver Nitrate

Erythromycin

Tetracycline

322 Volume 55, No. 4, July/August 2010

Table 1. Characteristics of Included Studies

Interventions Outcomes

Summary of Findings, RR (95% CI) Outcome: CONchloramphenicol vs. povidone-iodine, 0.57 (0.311.03) Outcome: CONpovidone-iodine vs. silver nitrate, 0.51 (0.380.71); erythromycin vs. silver nitrate, 0.70 (0.530.93) Outcome: GONpovidone-iodine vs. silver nitrate, 1.94 (0.606.29); erythromycin vs. silver nitrate, 2.29 (0.737.19) Outcome: CONsilver nitrate vs. no prophylaxis, 1.06 (0.552.02); tetracycline vs. no prophylaxis, 0.82 (0.421.63); erythromycin vs. no prophylaxis, 0.93 (0.481.79) Outcome: GONRR not possible to calculate (no cases in any group) Outcome: CONtetracycline vs. silver nitrate, RR = N (NaNN)a Outcome: GONRR not possible to calculate (no cases in either group) Outcome: CONerythromycin vs. silver nitrate, 0.72 (0.361.41); tetracycline vs. silver nitrate, 0.57 (0.251.31) Outcome: GONerythromycin vs. silver nitrate, 3.66 (0.4132.72); tetracycline vs. silver nitrate, 2.55 (0.2724.55) Outcome: CONtetracycline vs. silver nitrate, 0.66 (0.261.66) Outcome: GONtetracycline vs. silver nitrate, 0.27 (0.051.36) Outcome: GONRR not possible to calculate (no cases in either group) Outcome: CONerythromycin vs. silver nitrate, 0 (0NaN)a Outcome: GONRR not possible to calculate (no cases in either group)

U U U

U U

U U

Chen,37 1992

Quasi-RCT; interventions rotated monthly

4544 newborns in a hospital setting in Taichung, Taiwan, China

Brussieux et al.,24 1991

Quasi-RCT; interventions randomly assigned daily Quasi-RCT; interventions alternated weekly

900 newborns in a hospital setting in Saint-Germain-en-Laye, France 230 newborns in a hospital setting in Brooklyn, NY; all newborns born to mothers with chlamydia at the time of birth; secondary analysis reported GON in all 12431 newborns born during the study period (alternating treatments were given to all) 2732 newborns in a hospital setting in Nairobi, Kenya 450 newborns in a hospital setting in northeastern Zaire 60 newborns in a hospital setting in Seattle, WA; all newborns born to mothers with chlamydia at the time of birth

Hammerschlag et al.,25 1989

Laga et al.,26 1988

Quasi-RCT; interventions rotated weekly Quasi-RCT; interventions rotated weekly RCT; randomization method not specied

Fischer and Reta,16 1988 Hammerschlag et al.,30 1980

U U U

U U

N = innity; CI = condence interval; CON = chlamydial ophthalmia neonatorum; GON = gonococcal ophthalmia neonatorum; NaN = not a number; RCT = randomized controlled trial; RR = relative risk. a In the comparisons with these results, there were no cases in one of the intervention groups.

Table 2. Comparison of Efcacy of Various Agents in the Prevention of Gonococcal Ophthalmia Neonatorum
Comparison Erythromycin vs. silver nitrate Tetracycline vs. silver nitrate Povidone-iodine vs. silver nitrate Povidone-iodine vs. erythromycin
CI = condence interval; RR = relative risk. a Random effects model.

Studies Hammerschlag et al.,25 Isenberg et al.17 Hammerschlag et al.,25 Laga et al.26 Isenberg et al.17 Isenberg et al.17

No. of Participants 10,004 11,004 2005 2188

Effect Size,a RR (95% CI) 2.54 (0.926.98) 0.72 (0.086.30) 1.94 (0.606.29) 0.85 (0.352.03)

Table 3. Comparison of Efcacy of Various Agents in the Prevention of Chlamydial Ophthalmia Neonatorum
Comparison Erythromycin vs. silver nitrate Studies No. of Participants Effect Size,a RR (95% CI) 4514 6008 2005 2188 2004 0.71 (0.520.97) 0.70 (0.451.10) 0.52 (0.380.71) 0.74 (0.541.03) 1.77 (0.973.22) Chen,37 Hammerschlag et al.,25 Hammerschlag et al.,30 Isenberg et al.17 Tetracycline vs. silver nitrate Brussieux et al.,24 Chen,37 Hammerschlag et al.,25 Isenberg et al.17 Povidone-iodine vs. silver nitrate Isenberg et al.17 Povidone-iodine vs. erythromycin Isenberg et al.17 Povidone-iodine vs. chloramphenicol Ramirez-Ortiz et al.18

CI = condence interval; RR = relative risk. a Random effects model.

available from Chen,37 who included a no prophylaxis group. None of the interventions made a signicant difference in the rate of CON when compared to no prophylaxis: silver nitrate (RR = 1.06; 95% CI, 0.552.02; 2225 newborns), erythromycin (RR = 0.93; 95% CI, 0.481.79; 2306 newborns), and tetracycline (RR = 0.82; 95% CI, 0.421.63; 2299 newborns).37 Comparison of Efcacy: Chlamydial Ophthalmia Neonatorum Table 3 presents comparisons of prophylactic agents used to prevent CON. Erythromycin and povidone-iodine both decreased the risk of CON when compared to silver nitrate (RR = 0.71; 95% CI, 0.520.97; 4514 newborns17,25,30,37 and RR = 0.52; 95% CI, 0.380.71; 2005 newborns,17 respectively). Compared to erythromycin, povidone-iodine was associated with a nonsignicant trend towards a reduced risk of CON (RR = 0.74; 95% CI, 0.541.03; 2005 newborns),17 but there was a nonsignicant trend toward a higher risk of CON when povidone-iodine was compared to chloramphenicol (RR = 1.77; 95% CI, 0.973.22; 2004 newborns).18 Secondary analyses were not conducted because allocation concealment was not reported in the one randomized study contributing data to these analyses.30 DISCUSSION Gonococcal Ophthalmia Neonatorum There is little doubt that silver nitrate signicantly reduced the incidence of GON when it was introduced. However, the trials included in this review provide no high-quality evidence regarding the effect size of prophylactic agents
Journal of Midwifery & Womens Health  www.jmwh.org

in the prevention of GON nor do they provide adequate data to compare the efcacy of the various agents in the prevention of GON. In both cases this is related to a lack of adequate power. When the prevalence of maternal gonorrhea is low, very large numbers of participants are required to show a statistically signicant reduction in the incidence of GON. Studies of this size are difcult to conduct, and in most settings it would now be considered unethical to randomize newborns at risk of acquiring GON to receive no prophylaxis because this intervention is generally accepted as highly effective. Estimations of the efcacy of various agents in the prevention of GON therefore cannot be determined solely on the data available from randomized participants. Laga et al.26 compared rates of infection to historical controls. In this study, vaginal swabs were collected at delivery from historical controls and from mothers of newborns enrolled in the trial. The authors compared rates of infection in exposed newborns in each treatment group with exposed newborns in the historical control group; only newborns that were seen for follow-up visits were included. When compared to no prophylaxis, both silver nitrate and tetracycline led to a signicant reduction in GON (RR = 0.17; 95% CI, 0.070.41; 138 newborns, and RR = 0.07; 95% CI, 0.020.29; 133 newborns, respectively).26 These ndings are in line with the results of a very large prospective before/after observational study conducted in Cape Town, South Africa.40 During the pre-trial period, no prophylaxis was being used. Prophylaxis (either silver nitrate or erythromycin) was then introduced at three hospitals. When prophylaxis was compared to no prophylaxis, the risk of GON was signicantly reduced (RR = 0.19;
323

95% CI, 0.070.50; 30530 newborns). There was no signicant change in the rate of GON during this time at three midwifery obstetric units where prophylaxis was not introduced (17928 newborns), suggesting no natural variations in prevalence during the study.40 Another way of framing these RRs is to say that prophylaxis fails to prevent GON approximately 7% to 19% of the time. Estimations of efcacy are impacted by compliance with treatment protocols, which may vary from one setting to another. Laga et al.26 noted that three of the cases of GON in newborns who received prophylaxis occurred early in the study when the research nurses were not yet familiar with the technique for applying prophylaxis. During the before/after study, four of the ve newborns that presented with GON during the trial period had inadvertently not received prophylaxis.40 It is possible that with better adherence to treatment protocols, the effectiveness of prophylactic agents will be higher than the estimates provided by the studies above. Another possible reason for the failure of prophylaxis is that there are other portals of entry for infections during birth, including the oropharynx. Despite ocular prophylaxis, a newborn might self-inoculate the eye via their hands if the oropharynx is colonized. Cases of GON acquired through contact spread (e.g., via contaminated ngers or towels) have been documented.8 Chlamydial Ophthalmia Neonatorum Only one of the studies included in this review contributed data to examine how effectively prophylactic agents reduce the risk of CON.37 This study found that none of the agents signicantly reduced the risk of CON. However, the results of this study must be interpreted with caution given the methodologic weakness of the trial. While the sample size of this study had a power of 0.80 to detect a 75% reduction in the risk of CON compared to the risk of 1.6% in the group without prophylaxis, the study was quasirandomized, blinding was not reported, and it is not clear how many newborns were seen for complete follow up. While it is plausible that eye prophylaxis is not as effective in the prevention of CON as it is in the prevention of GON, the nding of no signicant reduction in risk does not t with other available evidence. Several types of evidence suggest that prophylactic agents do have some effect on reducing the risk of CON. First, the results of this review show statistically signicant differences between different agents in the prevention of CON, which logically would not exist if all the agents have no effect. Second, in Hammerschlag et al.25in which all included newborns were born to mothers with known chlamydial infection and received prophylaxis the overall rates of CON were lower than the rates of infection found in a cohort of newborns whose mothers had chlamydia and did not receive prophylaxis (1120% vs. 33%). Finally, Laga et al.26 compared trial ndings with
324

data from historical controls as described above, and demonstrated some effect in reducing the risk of CON: when compared to historical controls who received no prophylaxis, both silver nitrate and tetracycline led to a signicant reduction in CON (RR = 0.32; 95% CI, 0.170.60; 300 newborns, and RR = 0.23; 95% CI, 0.110.46; 312 newborns, respectively). In other words, these data suggest that prophylaxis fails to prevent CON 23% to 32% of the time. The accuracy of this estimation of effect size is limited by possible bias inherent in a nonrandomly allocated control group. Overall, these data suggest that prophylactic agents lead to some reduction in the risk of CON, but the reduction is not as substantial as the reduction in the risk of GON. This review provides fair evidence that both erythromycin and povidone-iodine are signicantly more effective than silver nitrate in the prevention of CON. Incomplete follow up in the studies contributing to these ndings limits the accuracy of estimations of effect size through possible underestimation of the frequency of outcomes, and through potential bias in comparison between intervention groups if there were differences between groups in follow-up rates. Number Needed to Treat In settings with low rates of maternal gonorrheal and chlamydial infections at birth, very large numbers of newborns will need to be given prophylaxis to prevent a single case of ON. For example, in the United States in 2002, the rate of GON was 1.1 per 100,000, and the rate of CON was 7.4 per 100,000.6,7 One might generously estimate the effectiveness of prophylaxis to be slightly better than the estimations from historical control data discussed above, and arbitrarily assume that the RR of CON with prophylaxis is 0.20, and that the RR of GON is 0.05 with prophylaxis. Based on these conditions, 3378 newborns need to receive prophylaxis to prevent one case of CON and 4785 newborns need to receive prophylaxis to prevent one case of GON. Note that these assumptions are conservativethey are likely to lead to lower estimations of the number needed to treat when compared to the estimations of the efcacy of prophylaxis from clinical studies. Implications for Practice Parents should be informed of the limitations of eye prophylaxis and should be instructed to report any signs of eye infection. The risks of long-term sequelae from GON or CON are minimal when there is adequate postpartum follow up and prompt access to antibiotic therapy upon diagnosis.8 Systemic rather than topical antibiotic therapy is recommended for treatment of GON and CON.42 Given the failure rates of prophylaxis, prenatal detection and treatment of maternal chlamydia and gonorrhea is ideal, and may also prevent other adverse outcomes for mother
Volume 55, No. 4, July/August 2010

and newborn. When there is known maternal gonorrheal infection at the time of birth, systemic antibiotics are recommended for the newborn.42 While it may seem reasonable based on the clinical evidence to offer parents an informed choice regarding prophylaxis, midwives should familiarize themselves with the legal requirements for prophylaxis in their jurisdiction (e.g., the state in which they practice), because parents have no legal right to refuse prophylaxis in most North American jurisdictions. Where maternal infection rates are high, and particularly where health care resources are limited, selection of the most effective method of prophylaxis will be of benet. Tetracycline-resistant gonococci have led to a shift away from use of tetracycline for prophylaxis. Presently, erythromycin is primarily the prophylaxis of choice in North America, but povidone-iodine is increasingly used elsewhere.19 Povidone-iodine has proven antibacterial properties, is used for other ocular purposes, is inexpensive, and drug-resistance is not a concern. Proponents argue that it is the most effective agent in the prevention of all causes of infectious ON.34 This claim is plausiblea systematic review by the primary author found povidone-iodine to signicantly reduce the risk of infectious ON (all causes) when compared to silver nitrate, but there was inadequate power to conrm a difference when povidone-iodine was compared to erythromycin (RR = 0.86; 95% CI, 0.70 1.06; 2188 newborns).9 Povidone-iodine may be less acceptable than antibiotic ointment to some parents because it causes temporary staining of the skin and sclera. Povidone-iodine ophthalmic solutions designed for neonatal use and approved by the US Food and Drug Administration are not yet available in the United States. Limitations This analysis was limited to an examination of efcacy and did not explore the safety or cost-effectiveness of prophylactic agents, which are also key considerations in reevaluating the utility of neonatal eye prophylaxis. CONCLUSION The evidence from randomized and quasirandomized trials regarding the efcacy of prophylactic agents used to prevent GON and CON is not of high quality. When additional evidence is also considered, it appears that prophylaxis does reduce the risks of both GON and CON; however, it is apparent that all prophylactic agents have clinically signicant failure rates. Estimates of failure rates are likely inaccurate, with various CIs ranging from 2% to 50% for GON and from 11% to 60% for CON. While there is not sufcient evidence to detect statistically signicant differences in the efcacy of various agents in the prevention of GON, both erythromycin and povidone-iodine are more effective than silver nitrate in the prevention of CON.
Journal of Midwifery & Womens Health  www.jmwh.org

In settings where the prevalence of maternal gonorrheal or chlamydial infection at birth is high, the evidence supports the universal use of eye prophylaxis. Taking cost, adverse effects, and efcacy against both GON and CON into consideration, povidone-iodine appears to be the best agent for neonatal eye prophylaxis in these settings. It is important to understand that neonatal eye prophylaxis has a signicant failure rate. Universal prophylaxis will not prevent all cases of ON, and early identication and treatment of newborns infected with either GON or CON will also be necessary to prevent adverse outcomes caused by these infections. Despite weaknesses in the available evidence, there is sufcient evidence available such that further trials involving a no prophylaxis group are not warranted. Although more accurate estimations of effect size (particularly of erythromycin and povidone-iodine) would be helpful in considering the utility of universal prophylaxis in low-risk settings, the benets of this information do not warrant the costs of conducting the large high-quality trial that would be required. Further research to evaluate the safety, efcacy, and most appropriate concentration of povidone-iodine is a priority. Erythromycin is more appropriate than silver nitrate for comparisons with povidone-iodine because erythromycin is more effective than silver nitrate in the prevention of CON, and maternal chlamydial infections are much more prevalent globally than gonorrheal infections. Future trials should ensure adequately powered sample sizes and prioritize the complete follow up of participants. Settings with high maternal infection rates at birth would be most appropriate. Interpretation of the ndings would be strengthened if swabs to detect chlamydia and gonorrhea are collected in labor. Finally, the evidence suggests that current North American laws mandating universal neonatal eye prophylaxis have limited benet, and a reexamination of this policy is warranted. Evidence regarding efcacy from this review might inform both economic analyses and riskbenet analyses that could be used to determine the best approach within a given context. Such analyses should compare the implications of universal prophylaxis with those of alternative strategies for the prevention of GON and CON.

REFERENCES
1. Schaller UC, Klauss V. Is Crede s prophylaxis for ophthalmia neonatorum still valid? Bull World Health Organ 2001;79:2626. 2. Schneider G. Silver nitrate prophylaxis. Can Med Assoc J 1984;131:1936. 3. Budai I. Chlamydia trachomatis: Milestones in clinical and microbiological diagnostics in the last hundred years: A review. Acta Microbiol Immunol Hung 2007;54:522. 4. Taylor-Robinson D. The history of nongonococcal urethritis. Thomas Parran Award Lecture. Sex Transm Dis 1996;23:8691. 5. Centers for Disease Control and Prevention. Sexually transmitted disease surveillance 2007. Atlanta, GA: US Department of Health and Human Services, 2008. 325

6. Centers for Disease Control and Prevention. Sexually transmitted disease surveillance 2002. Atlanta, GA: US Department of Health and Human Services, 2003. 7. Hamilton BE, Martin JA, Sutton PD. US Department of Health and Human Services Centers for Disease Control and Prevention, Births: Preliminary data for 2002. Hyattsville, MD: National Center for Health Statistics, 2003. 8. Wahlberg V. Reconsideration of Crede prophylaxis. A study of maternity and neonatal care. Acta Paediatr Scand Suppl 1982; 295:173. 9. Darling E. Mandatory prophylaxis for ophthalmia neonatorum: Does the evidence support the policy? [Masters thesis]. Hamilton, Ontario, Canada: McMaster University, 2007. 10. HIV and Sexually Transmitted Infections Department. All new episodes seen at GUM clinics: 19992003, Country specic tables. London: Health Protection Agency Centre for Infections, 2004. 11. Ofce for National Statistics. Birth statistics: Review of the Registrar General on births and patterns of family building in England and Wales, 2003. London: Ofce for National Statistics, 2004. 12. Public Health Information and Information Technology (PHIIT) Ofce Web site. Health surveillance: Infectious/communicable disease surveillance: 2004 reports. Available from: www. publichealthontario.ca/portal/server.pt/gateway/PTARGS_0_2_274_ 209_0_43/http%3B/cmsv1.srv.ehealthontario.ca/API/eHealth/Public+ Health/Programs/Health+Surveillance/_default.htm?lcache=en# DiseaseReport [Accessed January 28, 2007]. 13. Statistics Canada Web site. Births and birth rate, by province and territory. Available from: www40.statcan.ca/l01/cst01/ demo04a.htm [Accessed September 27, 2006]. 14. HIV and Sexually Transmitted Infections Department. Selected STI diagnoses and diagnosis rates from GUM clinics in the UK: 20032007. London: Health Protection Agency, Centre for Infections, 2008. 15. 2004 Canadian sexually transmitted infections surveillance report. Can Commun Dis Rep 2007;33(Suppl 1):169. 16. Fischer PR, Reta BB. Prevention of neonatal conjunctivitis in Zaire. Ann Trop Paediatr 1988;8:856. 17. Isenberg SJ, Apt L, Wood M. A controlled trial of povidoneiodine as prophylaxis against ophthalmia neonatorum. N Engl J Med 1995;332:5626. 18. Ramirez-Ortiz MA, Rodriguez-Almaraz M, Ochoa-Diaz Lopez H, Diaz-Prieto P, Rodriguez-Suarez R. Randomised equivalency trial comparing 2.5% povidone-iodine eye drops and ophthalmic chlorampenicol for preventing neonatal conjunctivitis in a trachoma endemic area in southern Mexico. Br J Ophthalmol 2007;91:14304. 19. Isenberg SJ, Apt L, Del Signore M, Gichuhi S, Berman NG. A double application approach to ophthalmia neonatorum prophylaxis. Br J Ophthalmol 2003;87:144952. 20. Isenberg SJ, Apt L, Wood M. The inuence of perinatal infective factors on ophthalmia neonatorum. J Pediatr Ophthalmol Strabismus 1996;33:1858. 21. Graf H, Retzke U, Schilling C, Schmidt M. Reaction of the anterior eye segment to preventive Crede treatment [in German]. Zentralbl Gynakol 1994;116:63942.

22. Isenberg SJ, Apt L, Yoshimori R, Leake RD, Rich R. Povidone-iodine for ophthalmia neonatorum prophylaxis. Am J Ophthalmol 1994;118:7016. 23. Krohn MA, Hillier SL, Bell TA, Kronmal RA, Grayston JT. The bacterial etiology of conjunctivitis in early infancy. Eye Prophylaxis Study Group. Am J Epidemiol 1993;138:32632. 24. Brussieux J, Boisivon A, Theron HP, Faidherbe C, Machado N, Michelon B. Prevention of neonatal conjunctivitis. A comparative clinical and bacteriologic study of 2 eyedrops: Silver nitrate and oxytetracycline chlorhydrate [in French]. Ann Pediatr (Paris) 1991;38:63741. 25. Hammerschlag MR, Cummings C, Roblin PM, Williams TH, Delke I. Efcacy of neonatal ocular prophylaxis for the prevention of chlamydial and gonococcal conjunctivitis. N Engl J Med 1989; 320:76972. 26. Laga M, Plummer FA, Piot P, Datta P, Namaara W, NdinyaAchola JO, et al. Prophylaxis of gonococcal and chlamydial ophthalmia neonatorum. A comparison of silver nitrate and tetracycline. N Engl J Med 1988;318:6537. 27. Hick JF, Block DJ, Ilstrup DM. A controlled study of silver nitrate prophylaxis and the incidence of nasolacrimal duct obstruction. J Pediatr Ophthalmol Strabismus 1985;22:923. 28. Rajantie J, Osterlund K. Effect of Crede prophylaxis on the eyelid edema of newborn infants [in Finnish]. Duodecim 1982;98:18069. 29. Wahlberg V. Crede s prevention method has lost its justication today [in Swedish]. Nord Med 1982;97:367. 30. Hammerschlag MR, Chandler JW, Alexander ER, English M, Chiang WT, Koutsky L, et al. Erythromycin ointment for ocular prophylaxis of neonatal chlamydial infection. JAMA 1980;244:22913. 31. Matinzadeh ZK, Beiragdar F, Kavemanesh Z, Abolgasemi H, Amirsalari S. Efcacy of topical ophthalmic prophylaxis in prevention of ophthalmia neonatorum. Trop Doct 2007;37:479. 32. Zar HJ. Neonatal chlamydial infections: Prevention and treatment. Paediatr Drugs 2005;7:10310. 33. Richter R, Below H, Kadow I, Kramer A, Mu ller C, Fusch C. Effect of topical 1.25% povidone-iodine eyedrops used for prophylaxis of ophthalmia neonatorum on renal iodine excretion and thyroid-stimulating hormone level. J Pediatr 2006;148:4013. 34. Isenberg SJ, Apt L, Campeas D. Ocular applications of povidone-iodine. Dermatology 2002;204(Suppl 1):925. 35. Ozkan H, Abacioglu H, Duman N, Celikkol B, Ozkutuk A. A controlled trial of efcacy and safety of povidone-iodine as prophylaxis against ophthalmia neonatorum [in Turkish]. C ocuk Sag lg ve Hastalklar Dergisi 1999;42:45967. 36. Bell TA, Grayston JT, Krohn MA, Kronmal RA. Randomized trial of silver nitrate, erythromycin, and no eye prophylaxis for the prevention of conjunctivitis among newborns not at risk for gonococcal ophthalmitis. Eye Prophylaxis Study Group. Pediatrics 1993;92:75560. 37. Chen JY. Prophylaxis of ophthalmia neonatorum: Comparison of silver nitrate, tetracycline, erythromycin and no prophylaxis. Pediatr Infect Dis J 1992;11:102630. 38. Buttereld PM, Emde RN, Svedja MJ. Does the early application of silver nitrate impair maternal attachment? Pediatrics 1981;67:7378.

326

Volume 55, No. 4, July/August 2010

39. Buttereld P, Emde R, Platt B. Effects of silver nitrate on initial visual behavior. Am J Dis Child 1978;132:426. 40. Lund RJ, Kibel MA, Knight GJ, van der Elst C. Prophylaxis against gonococcal ophthalmia neonatorum. A prospective study. S Afr Med J 1987;72:6202.

41. Schroeder B, Tetlow P, Sanlippo JS, Hertweck SP. Is there a seasonal variation in gonorrhea and chlamydia in adolescents? J Pediatr Adolesc Gynecol 2001;14:257. 42. Centers for Disease Control and Prevention. Workowski KA, Berman SM. Sexually transmitted diseases treatment guidelines, 2006. MMWR Recomm Rep 2006;55:194.

Journal of Midwifery & Womens Health  www.jmwh.org

327