Topical pharmacologic interventions versus placebo for epidemic keratoconjunctivitis

Summary of findings 1. Any intervention compared with artificial tears or saline

Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Studies contributing to the pooled effects	Certainty of the evidence (GRADE)	Comments
Any intervention compared with artificial tears (or saline) for epidemic keratoconjunctivitis
Patient or population: epidemic keratoconjunctivitis Setting: eye hospital Intervention: any intervention (PVP‐I, ganciclovir, trifluridine, dexamethasone phosphate plus neomycin) Comparison: artificial tears or saline
Outcomes	Assumed risk with artificial tears or saline	Corresponding risk with any intervention	Relative effect (95% CI)	№ of participants (studies)	Studies contributing to the pooled effects	Certainty of the evidence (GRADE)	Comments
Mean number of days from treatment initiation to resolution of symptoms or signs (shorter is favored)	We chose to compare times to resolution of selected ocular signs and symptoms, including conjunctivitis (red eye or excessive tearing) and severe eye discomfort (see comments).		‐	‐	‐	⊕⊝⊝⊝ Very low^1,2,3	Ward 1993 compared trifluridine 1% with artificial tears (49 participants) and found no differences (MD 0 days, 95% CI 3 days shorter to 3 days longer), whereas Tabbara 2001 reported a significant reduction of days with signs of conjunctivitis (MD 11 days shorter, 95% CI 8 to 13 days shorter) comparing ganciclovir to artificial tears.
Proportion of participants with resolution of symptoms within 10 days of treatment (higher is favored)	829 per 1000	953 per 1000 (887 to 1028)	RR 1.15 (1.07 to 1.24)	409 (2 RCTs)	Elwan 2020; Ricciardelli 2021	⊕⊕⊝⊝ Low^1,2	A third trial (Kovalyuk 2017, 52 eyes in 45 participants) compared PVP‐I plus dexamethasone with artificial tears and reported an RR of 9.00 (95% CI 1.23 to 66.05) at the eye level.
Proportion of participants with resolution of signs within 10 days of treatment (higher is favored)	177 per 1000	565 per 1000 (405 to 788)	RR3.19 (2.29 to 4.45)	409 (2 RCTs)	Elwan 2020; Ricciardelli 2021	⊕⊕⊝⊝ Low^1,2
Proportion of participants in whose eyes SEIs had developed within 30 days of treatment initiation (lower is favored)	529 per 1000	127 per 1000 (53 to 296)	RR 0.24 (0.10 to 0.56)	77 (2 RCTs)	Ricciardelli 2021; Tabbara 2001	⊕⊝⊝⊝ Very low^1,2,4	Kovalyuk 2017 and Trauzettel‐Klosinski 1980 also reported this outcome within 4 weeks of treatment but at the eye level, with an RR of 0.57 (95% CI 0.07 to 4.81; 139 eyes). Elwan 2020 reported an RR of 0.04 (95% CI 0.02 to 0.09; 350 participants) after 3 months of treatment.
Proportion of participants in whose eyes SEI had disappeared within 21 days of treatment (higher is favored)	298 per 1000 eyes	399 per 1000 eyes (262 to 608)	RR 1.34 (0.88 to 2.04)	156 eyes (2 RCTs)	Ricciardelli 2021; Trauzettel‐Klosinski 1980	⊕⊝⊝⊝ Very low^1,2,4	Ricciardelli 2021 reported at the participant level (59 participants); Trauzettel‐Klosinski 1980 reported at the eye level (97 eyes).
Proportion of participants who discontinued medication and did not develop rebound SEI within 21 days (higher is favored)	Not measured by any included study		‐	‐	‐	‐
Proportion of participants who had evidence of adenoviral eradication within 14 days of treatment (higher is favored)	See comments		‐	‐	‐	⊕⊝⊝⊝ Very low^1,2,5	Kovalyuk 2017 (42 eyes) reported on percentages of reduction in mean viral titers, with an MD of −23% (95% CI −40.84% to ‐5.16%) on day 7 of treatment.
Proportion of initially unaffected fellow eyes that had signs/symptoms of infection within 7 days of onset of signs/symptoms in the first eye (lower is favored)	Not measured by any included study		‐	‐	‐	‐
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; MD: mean difference; PVP‐I: povidone‐iodine; RCT: randomized controlled trial; RR: risk ratio; SEI: subepithelial infiltrates
GRADE Working Group grades of evidence High certainty: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
¹Downgraded for risk of bias (−1). ²Downgraded either for small sample size OR the confidence interval crossing or equaling no differences (−1). ³Downgraded for inconsistency (−1). ⁴Downgraded for inconsistency (details in Effects of interventions) and indirectness (−1). ⁵Downgraded for indirectness (−1).

Summary of findings 2. Any intervention compared with steroid alone or plus levofloxacin

Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Studies contributing to the pooled effects	Certainty of the evidence (GRADE)	Comments
Any intervention compared with steroid alone or plus levofloxacin for epidemic keratoconjunctivitis
Patient or population: epidemic keratoconjunctivitis Setting: eye hospital Intervention: any intervention (tacrolimus, CsA, trifluridine, PVP‐I, FML plus PVA‐I) Comparison: steroid (dexamethasone, loteprednol, FML) or FML plus levofloxacin
Outcomes	Assumed risk with steroid	Corresponding risk with any intervention	Relative effect (95% CI)	№ of participants (studies)	Studies contributing to the pooled effects	Certainty of the evidence (GRADE)	Comments
Mean number of days from treatment initiation to resolution of symptoms or signs (shorter is favored)	We chose to compare times to resolution of selected ocular signs and symptoms, including conjunctivitis (red eye or excessive tearing) and severe eye discomfort (see comments).	‐	‐	‐	‐	⊕⊝⊝⊝ Very low^1,2,3	Ward 1993 (N = 49) found 1.0 day longer (95% CI 2.3 shorter to 4.3 longer) in mean duration with symptoms/signs in the trifluridine group as compared with the dexamethasone group.
Proportion of participants with resolution of symptoms within 7 days of treatment (higher is favored)	See comments	‐	‐	‐	‐	⊕⊝⊝⊝ Very low^1,2,3	Kovalyuk 2017 (52 eyes) reported this outcome at the eye level, showing an 8‐fold increase in the likelihood of symptom resolution (RR 9.00, 95% CI 1.23 to 66.05) in the PVP‐I group as compared with the dexamethasone group.
Proportion of participants with resolution of signs within 7 days of treatment (higher is favored)	See comments	‐	‐	‐	‐	⊕⊝⊝⊝ Very low^1,2,3	Matsuura 2021 compared PVA‐I plus FML with levofloxacin plus FML at the eye level, which was analyzed as a control treatment, hence the indirectness of this evidence; the result was inconclusive (RR 1.62, 95% CI 0.32 to 8.18).
Proportion of eyes in which SEI had developed within 21 days of treatment (lower is favored)	410 per 1000 eyes	33 per 1000 eyes (4 to 226)	RR 0.08 (0.01 to 0.55)	69 eyes (2 RCTs)	Kovalyuk 2017; Matsuura 2021	⊕⊝⊝⊝ Very low^1,2,3	The combined RR was 0.12 (95% CI 0.02 to 0.65) within 30 days of treatment (69 eyes, 2 RCTs).
Proportion of participants in whose eyes SEI had disappeared within 4 weeks of treatment initiation (higher is favored)	See comments	‐	‐	‐	‐	⊕⊝⊝⊝ Very low^1,2,4	Rafe 2020 reported this outcome at the participant level, showing an inconclusive effect of CsA when compared with steroid (RR 0.84, 95% CI 0.67 to 1.06; 88 participants). Gouider 2021 also reported this outcome at the eye level after 3 months of CsA as compared with steroid (RR 0.32, 95% CI 0.12 to 0.88; 70 eyes), and after 6 to 7 months of treatment (RR 0.67, 95% CI 0.43 to 1.04; 63 eyes).
Proportion of participants who discontinued medication and did not develop rebound SEI within 21 days (higher is favored)	872 per 1000 eyes (see comments)	907 per 1000 (820 to 994)	RR 1.04 (0.94 to 1.14)	231 eyes (3 RCTs)	Bhargava 2019; Gouider 2021; Rafe 2020	⊕⊝⊝⊝ Very low^1,2,3	Rafe 2020 and Bhargava 2019 reported this outcome at the participant level (168 participants, assuming 1 eye per participant) at week 12 and month 6, separately. Gouider 2021 reported at the eye level (63 eyes) after 6 to 7 months of treatment.
Proportion of participants who had evidence of adenoviral eradication by day 14 after initiation of treatment (higher is favored)	See comments		‐	‐	‐	⊕⊝⊝⊝ Very low^1,2,3	Kovalyuk 2017 (42 eyes) reported on percentages of reduction in mean viral titers, with an MD of −35% (95% CI −57% to −13%), in favor of the intervention.
Proportion of initially unaffected fellow eyes that had signs/symptoms of infection within 7 days of onset of signs/symptoms in the first eye (lower is favored)	Not measured by any included study		‐	‐	‐	‐
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; CsA: cyclosporin A; FML: fluorometholone; MD: mean difference; PVA‐I: polyvinyl alcohol iodine; PVP‐I: povidone‐iodine; RCT: randomized controlled trial; RR: risk ratio; SEI: subepithelial infiltrates
GRADE Working Group grades of evidence High certainty: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
¹Downgraded for risk of bias (−1). ²Downgraded either for small sample size OR the confidence interval crosses or equals no difference (−1). ³Downgraded for indirectness of evidence (−1). ⁴Downgraded for inconsistency and indirectness of evidence (−1).

Summary of findings 3. Topical steroid compared with artificial tears

Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Topical steroid compared with artificial tears for epidemic keratoconjunctivitis
Patient or population: epidemic keratoconjunctivitis Setting: eye hospital Intervention: dexamethasone Comparison: artificial tears
Outcomes	Assumed risk with artificial tears	Corresponding risk with dexamethasone	Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Mean number of days from treatment initiation to resolution of symptoms or signs (shorter is favored)	The mean number of days from initiation of treatment until signs were resolved was 11 days (SD 5.96). (see comment)	MD 1.0 day shorter (4.3 shorter to 2.3 longer)	‐	50 (1 RCT)	⊕⊝⊝⊝ Very low^1,2,3	Ward 1993 reported time to resolution of signs at the participant level.
Proportion of participants with resolution of symptoms within 7 days of treatment (higher is favored)	38 per 1000 eyes	38 per 1000 eyes (3 to 576)	RR 1.00 (0.07 to 15.15)	52 eyes (1 RCT)	⊕⊝⊝⊝ Very low^1,2,3	Kovalyuk 2017 reported this outcome at the eye level; verification of symptom resolution was by telephone interview.
Proportion of participants with resolution of signs within 7 days of treatment (higher is favored)	Not measured by either included study		‐	‐	‐
Proportion of participants in whose eyes SEI had developed within 21 days of treatment (lower is favored)	192 per 1000 eyes (see comment)	423 per 1000 eyes (171 to 1046)	RR 2.20 (0.89 to 5.41)	50 eyes (1 RCT)	⊕⊝⊝⊝ Very low^1,2,3	Kovalyuk 2017 reported this outcome at the eye level at day 7.
Proportion of participants in whose eyes SEI had disappeared within 21 days of treatment initiation (higher is favored)	Not measured by either included study		‐	‐	‐
Proportion of participants who discontinued medication and did not develop rebound SEI within 21 days (higher is favored)	Not measured by either included study		‐	‐	‐
Proportion of participants who had evidence of adenoviral eradication within 14 days of treatment (higher is favored)	See comment		‐	50 eyes (1 RCT)	⊕⊝⊝⊝ Very low^1,2,3	Kovalyuk 2017 reported on percentages of reduction in mean viral titers at the eye level, with an MD of 12.00% (95% CI −34.88% to 58.88%) comparing dexamethasone to artificial tears.
Proportion of initially unaffected fellow eyes that had signs/symptoms of infection within 7 days of onset of signs/symptoms in the first eye (lower is favored)	Not measured by either included study		‐	‐	‐
The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI:* confidence interval; MD: mean difference; RCT: randomized controlled trial; RR: risk ratio; SD: standard deviation; SEI: subepithelial infiltrates
GRADE Working Group grades of evidence High certainty: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
¹Downgraded for risk of bias (−1). ²Downgraded either for small sample size OR the confidence interval crosses or equals no difference (−1). ³Downgraded for indirectness of evidence (−1).

Background

Description of the condition

Viruses cause about 80% of all cases of acute conjunctivitis (Fitch 1989; Matsui 2008; Ronnerstam 1985; Stenson 1982; Uchio 2000; Woodland 1992). Human adenoviruses (HAdVs) are believed to account for 65% to 90% of cases of viral conjunctivitis (O'Brien 2009), or 20% to 75% of all causes of infectious keratoconjunctivitis worldwide, making them the most common cause of viral conjunctivitis (Gigliotti 1981; Ishii 1987; Woodland 1992).

Epidemic keratoconjunctivitis (EKC) is a highly infectious subset of conjunctivitis caused by HAdVs. To a much lesser extent, ribonucleic acid (RNA) viruses such as Coxsackievirus and Enterovirus can also cause EKC. In a six‐month period in Chicago, 401 individuals (patients, staff, physicians) developed adenoviral EKC at the Illinois Eye and Ear Infirmary, University of Illinois, Chicago, USA (Warren 1989). EKC is accompanied by severe conjunctival inflammation, watery discharge, and light sensitivity, and can lead to chronic complications such as corneal and conjunctival scarring (Dawson 1972). These clinical findings are accompanied by discomfort and vision of poor quality.

There are seven species of HAdVs, of which HAdV types in the B and D species can cause ocular infections. In addition to HAdV‐D8, D19 (now retyped as D64; Zhou 2012) and D37 cause EKC. In the last two decades new viral genomes that can cause EKC have been identified. Some B‐species types (e.g. type 3) cause less severe conjunctivitis (Meyer‐Rusenberg 2011), but can cause fatal pneumonia, especially in individuals with immature or compromised immune systems, such as neonates (Sammons 2019). Other HAdV types can cause pneumonia, encephalitis, sepsis, and death in healthy people, hence the development and use of vaccines for types 4 and 7 have been developed and used in military recruits. It is important to note that the earlier nomenclature is undergoing revision as a result of whole genome sequencing, which is also leading to the identification of new pathogens that cause EKC. In addition, recombination events, to which HAdV‐D types are susceptible, are leading to the evolution of new pathogen genomes.

The incubation period for adenoviral ocular infections is 2 to 14 days. EKC begins unilaterally, but often spreads to the unaffected eye. Patients with EKC present with severe inflammation of the cornea and conjunctiva accompanied by clear discharge, lid swelling, light sensitivity, and foreign body sensation. This acute phase can last two weeks. In a subset of patients with EKC, complications such as corneal subepithelial infiltrates (SEI) develop. SEI can affect vision for months to a year (Butt 2006), and result in permanent corneal scarring and decreased vision. Other complications include symblepharon, conjunctival scarring, lacrimal drainage abnormalities, dry eye, and symblepharon formation (Hammer 1990; Hyde 1988).

Epidemiology

Most cases of EKC affect adults aged 20 to 40 years old, with men and women affected equally (Ford 1987). Precise incidence and prevalence rates of EKC are unknown because most cases of 'red eye' (including EKC) are first seen by non‐ophthalmologists. In one study of patients in a national managed care network, over 80% of cases of acute conjunctivitis had been treated by non‐ophthalmologists (Shekhawat 2017).

Adenoviral infections (including conjunctivitis) are reportable to national registries in Japan and Germany for national surveillance (Hiroi 2013; Meyer‐Rusenberg 2011; Tabbara 2010), but are not reportable in most countries (CDC 2013; Ford 1987). The US Centers for Disease Control and Prevention (CDC) established a voluntary registry for adenoviral infections in 2014 (CDC 2017).

Transmission of adenoviral infection is predominantly through contact with ocular secretions via contaminated surfaces, instruments, eyedroppers, bottle tips, or hands. A person is thought to be infectious from a few days before symptoms develop to approximately 14 days after symptom onset. Infection rates vary, but have been reported to be as high as 50% in a long‐term care facility in France (Piednoir 2002). Investigators of another study based on a simultaneous outbreak of EKC in a large ophthalmologic institute in Chicago and in the community estimated the secondary household attack rates to be 20% (Vastine 1976).

Presentation

Symptoms of EKC usually appear within 14 days after exposure and typically last 7 to 21 days (Meyer‐Rusenberg 2011). Corneal involvement (keratitis and SEI; hence, the term 'keratoconjunctivitis') can occur in 50% of patients with EKC (Tabbara 2010). Epithelial keratitis occurs around day 7; in the corresponding subepithelial locations, SEI can develop as early as the second week of presentation, but usually in the third or fourth week after the onset of acute infection (Dosso 2008). SEI are considered pathognomonic for previous EKC (Jones 1958).

Histopathological investigation of SEI reveals lymphocytes, histiocytes, and fibroblasts that are accompanied by disruption of the collagen fibers of the Bowman layer (Lund 1978). Pseudomembranes may develop in a high proportion of patients with EKC (Fukumi 1958; Laibson 1968). SEI have been observed in more than 40% of cases of EKC in an outbreak. In a six‐year study at a tertiary ophthalmology clinic, 25% of patients had symptomatic infiltrates that persisted for 45 days after presentation (Butt 2006). These infiltrates can persist for months to years and cause disabling visual symptoms such as blurriness, glare, haloes, and light sensitivity. In some cases, SEI are permanent. Pseudomembranes are sheets of fibrin‐rich exudate lacking blood or lymphatic vessels that adhere to the upper and lower tarsal conjunctiva. In the same study (Butt 2006), 24% of patients had pseudomembranes or membranes; 15% of this subset developed symblepharon, which can cause dry eye symptoms, diplopia from restriction of ocular motility, mucus discharge, and redness.

Outbreaks of EKC typically occur in healthcare environments such as hospitals and eye clinics (Hamada 2008; Montessori 1998; Warren 1989), but also occur in the community (Darougar 1983; Dawson 1963; Jawetz 1959; Kuo 2018). EKC is seasonal (Lee 2018). The global health impacts of EKC outside high‐income countries are not widely known. It is known, however, that seasonal and spontaneous outbreaks of adenoviral infections, ocular or not, occur in Africa and Asia that are unrelated to exposure in healthcare environments. In fact, studies indicate that adenoviral infections are endemic in the community with shifting prevalence of causative types (Barnadas 2018; Hiroi 2013; Kuo 2018; Lin 2019).

Diagnosis

Lack of a rapid, accessible diagnostic test makes diagnosis of adenoviral conjunctivitis in 'red eyes' difficult for non‐ophthalmologists. Diagnosis is primarily clinical because laboratory testing, using the gold standards of viral culture or polymerase chain reaction (PCR) assays to detect HAdV in conjunctival specimens, is neither widely available nor routinely performed. Whole genome sequencing soon may become the next gold standard, but this testing is not yet widely available (Singh 2015).

Rapid point‐of‐care testing for adenoviral conjunctivitis (based on virus hexon antigen detection) has variable sensitivity, from 39.5% to 90% against the gold standards of PCR or culture (Kam 2015; Sachdev 2018).

In an unpublished study, clinical diagnosis of adenoviral conjunctivitis by corneal specialists had a clinical accuracy of 48% (unpublished data alluded to in O'Brien 2009). Of all presentations of ocular adenoviral infection, EKC is the most striking. The accuracy of clinical diagnosis is therefore likely higher for EKC than for adenoviral conjunctivitis in general. As the signs and symptoms of EKC can be severe, most affected patients will probably be seen by an ophthalmologist during their disease course.

Treatment options

No antiviral drug is approved for the treatment of EKC. Antibiotics are not effective against adenovirus or EKC. Shekhawat 2017 found that non‐ophthalmologists prescribed antibiotics for patients with acute conjunctivitis—many with presumed viral conjunctivitis—at a much higher rate than ophthalmologists. This practice increases antibiotic resistance, particularly for fluoroquinolones, which are often the first‐line treatment prescribed by many non‐ophthalmologists for red eye; they are used more judiciously by ophthalmologists (Asbell 2015; McDonald 2010). Surprisingly, many non‐ophthalmologists also prescribe steroid‐antibiotic combination drops for acute conjunctivitis (Shekhawat 2017), despite the inherent risks of topical steroid use. Treatment preferences for EKC are wide‐ranging, reflecting lack of accessible, accurate diagnostic testing; the wide range of providers who treat such patients (nurses, optometrists, urgent care doctors, internists, pediatricians, and ophthalmologists); and the lack of consensus regarding effective treatment.

Description of the intervention

The goals of pharmacotherapy are to lessen the severity of symptoms; shorten the duration of signs and symptoms from infection and inflammation; restore comfort and visual function; and decrease long‐term morbidity and complications (SEI, pseudomembranes, and membranes), without long‐term dependence on or adverse effects from pharmacotherapy. Although not commonly reported in studies of EKC, an important goal of therapy is the prevention of transmission to the fellow eye. This outcome is difficult to ascertain. Because incubation time in the fellow eye may precede initiation of pharmacotherapy for the first eye, many therapies may be ineffective against fellow eye transmission.

Although the sequelae of EKC can be long‐lasting, EKC is usually self‐limited. Supportive care in the form of lubrication and cold compresses can provide symptom relief. Other treatments include or have included topical corticosteroids (Laibson 1970), virustatic agents such as cidofovir and cyclosporin A (Hillenkamp 2001; Hillenkamp 2002), interferon B (Negoro 1980; Romano 1980; Sundmacher 1982; Uchio 2011), cyclosporin A (CsA) (Hillenkamp 2001; Hillenkamp 2002; Reinhard 2000), and tacrolimus (Berisa 2017).

The participants treated in several other studies appear to have had adenoviral conjunctivitis, not EKC; treatments included interferon (Adams 1984; Romano 1980; Wilhelmus 1987), non‐steroidal anti‐inflammatory drugs (NSAIDs) (Shiuey 2000), antiseptic povidone‐iodine (PVP‐I) (Isenberg 2002; Pelletier 2009), and combination PVP‐I and corticosteroid (Pepose 2018).

How the intervention might work

It is believed that an immune response to viral antigens occurs on the corneal epithelium, which then diffuses into the cornea (Dawson 1972; Jones 1958), sinking to the epithelium basement membrane where these chemokines induce leukocyte migration, forming SEI. Corticosteroids suppress the inflammatory response and delay SEI formation, thus relieving signs and symptoms. However, when given early in the course of infection, corticosteroids may increase viral replication and prolong viral shedding (Romanowski 1996; Romanowski 2001). Consequently, when corticosteroid use is discontinued, the SEI return (Laibson 1970). Another risk of topical corticosteroid use is elevation of intraocular pressure and thus increased risk of glaucoma and cataract formation.

Topical CsA, tacrolimus, and interferon may both modulate and suppress inflammatory responses. A study of the effects of topical CsA yielded the same results as corticosteroids with improved symptoms during medication use, but recurrence of SEI when the medication was discontinued (Jeng 2011). Topical tacrolimus, a calcineurin inhibitor similar to CsA, has also been used to treat (not prevent) SEI (Berisa 2017). During a median follow‐up time of about one year, a retrospective study comparing tacrolimus 0.03% drops with 0.02% ointment found reduced numbers and sizes of SEIs in about 60% of eyes as well as elimination of SEIs in 32% of eyes after six months of treatment. The improvements from baseline visual acuity after treatment with each formulation of tacrolimus were both clinically and statistically significant, corresponding to about two lines of improvement on a Snellen acuity chart with use of the ointment, and almost one line of improvement with use of the drop form of tacrolimus. At a mean of seven months after discontinuation, about 19% of eyes (not reported separately for ointment and drop forms) had a recurrence of SEI.

Interferon‐b, an immunomodulator secreted by virally infected cells, may act by inducing an antiviral state or modulating the immune response (Adams 1984; Negoro 1980; Romano 1980; Sundmacher 1982; Uchio 2011; Wilhelmus 1987).

Topical virustatic agents, such as cidofovir, may reduce viral replication and possibly prevent some sequelae of infection. Cidofovir (a cytosine analog) inhibits DNA polymerase. Topical cidofovir has been shown to be effective against adenovirus in cases of pediatric bone marrow transplantation, and may be effective in EKC (Hillenkamp 2002). However, it has also been shown to cause ocular surface toxicity. When tested in humans, with and without cyclosporin, cidofovir appeared to have no effect on the disease course or the symptoms of SEI (Hillenkamp 2002).

Topical antiseptic agents such as PVP‐I may have the potential to reduce the viral load in patients with adenoviral conjunctivitis by killing the infectious agent on contact (Isenberg 2002). Recent phase 2 clinical trials of PVP‐I 0.6% and combination PVP‐I 0.6%/dexamethasone 0.1% in 144 patients in India showed promise in improving clinical symptoms and shortening recovery time compared with vehicle (Pepose 2018). Combining a topical antiseptic such as PVP‐I with corticosteroids may mitigate viral replication while suppressing the formation of SEI.

PVP‐I 0.6% alone was no different from PVP‐I 0.6% plus dexamethasone 0.1% in viral eradication, which was measured by negative cell culture immunofluorescence assay [CC‐IFA]; there were no differences in length of time to clinical resolution (Pepose 2018) in participants with adenoviral conjunctivitis, but not EKC. Phase 3 trials of the combination of PVP‐I 0.6%/dexamethasone 0.1% in populations that include pediatric patients are under way.

Endogenous antimicrobial N‐chlorotaurine has been tried for treatment of adenoviral conjunctivitis (Romanowski 2006; Teuchner 2005; Uchio 2010). Topical NSAIDs, such as ketorolac, are sometimes used to reduce inflammation without the risk of side effects associated with topical corticosteroids, but have not been shown to reduce the signs or symptoms of adenoviral conjunctivitis, Shiuey 2000, or EKC.

Why it is important to do this review

A clear standard of care for EKC is lacking because there is no consensus on the efficacy of any pharmacotherapy to alter the clinical course of EKC. Supportive care in the case of EKC usually entails artificial tears, an over‐the‐counter formulation. We sought to compare the effects of virustatic agents, antiseptic agents with or without steroids, and immune‐modulating topical therapies against placebo (e.g. artificial tears), vehicle, or steroids as an active control. We focused on these therapies because we believe them to be the most promising. Some previously studied treatments have been shown to be ineffective or to cause toxicity to the ocular surface and are therefore rarely used today. Most published studies have compared active medications against artificial tears or vehicle drops. A three‐arm retrospective study compared non‐preserved artificial tears versus prednisolone acetate 1% and non‐preserved artificial tears versus cyclosporin 2% and non‐preserved artificial tears in a population with EKC (Asena 2017). A three‐way comparison of PVP‐I 0.6%/dexamethasone versus PVP‐I 0.6% versus vehicle (1:1:1) was also performed in a population of patients with adenoviral conjunctivitis, not EKC (Pepose 2018).

It is important to examine the benefits and harms of pharmacologic therapies currently prescribed for EKC because it is unclear whether any of these therapies offer more rapid resolution of signs and symptoms or long‐term sequelae such as SEI than no treatment or artificial tears. Rapid resolution allows better quality of life and ability to return to work. There are also unintended effects of treatment as well as costs associated with treatment. Therapies may induce drug allergy, sensitivity, or toxicity to the patient's eye or be associated with prolonged viral shedding, and may cost significantly more than vehicle or no treatment.

Objectives

To assess the efficacy and safety of topical pharmacological therapies versus placebo, an active control, or no treatment for adults with EKC.

Methods

Criteria for considering studies for this review

Types of studies

We planned to include both randomized controlled trials (RCTs) and quasi‐RCTs (comparative studies that use methods of treatment allocation such as alternation, date of birth, or case record number) had we not found sufficient numbers of RCTs in the search. However, we were able to identify enough RCTs and chose to include only RCTs in the review.

Types of participants

We planned to include studies that enrolled adults with clinically suspected or laboratory‐confirmed EKC; however, as we identified several studies that enrolled both children and adults, we expanded the inclusion criteria of the review (see Differences between protocol and review). We excluded studies of acute hemorrhagic conjunctivitis (which can be caused by viruses other than adenovirus) and studies of pharyngoconjunctival fever (which is caused by adenovirus but has less severe signs and symptoms than EKC).

Types of interventions

Originally, interventions of interest were antiseptic agents (PVP‐I or equivalent), and immune‐modulating therapy (corticosteroids and the calcineurin inhibitor CsA), irrespective of dosage or duration of treatment. Control treatments specified in the protocol were placebo, vehicle, no treatment, or another intervention of interest (Kuo 2020). We excluded other treatment comparisons that were not of interest in the current review. However, we added virustatic agents, ganciclovir and trifluridine, and the calcineurin inhibitor tacrolimus to the list of intervention treatments. We also considered studies that examined steroids or artificial tears (and saline) as control treatment. We describe these and other deviations from the review protocol in the Differences between protocol and review section.

Types of outcome measures

Primary outcomes

Symptom resolution
- Mean time (days) from initiation of treatment until symptoms of tearing and discomfort had resolved. Symptoms of discomfort include pain, itching, foreign body sensation, and photophobia.
- Proportion of participants with subjective resolution of symptoms by day 7 after initiation of treatment.
Sign resolution
- Mean time (days) from initiation of treatment until clinical signs had resolved. Examples of clinical signs are chemosis, hyperemia, bulbar conjunctival injection (redness), and subepithelial infiltrates (SEI).
- Proportion of participants with resolved signs by day 7 after initiation of treatment.

Secondary outcomes

RCTs of treatment of EKC report on a wide variety of outcomes. In the absence of an accepted standard set of outcomes, we included as many outcomes that we believed to be relevant, while recognizing that not all outcomes were evaluated in any single study.

Proportion of participants in whose eyes SEI had developed by day 21 after initiation of treatment.
Proportion of participants in whose eyes SEI had disappeared by day 21 after initiation of treatment.
Proportion of participants who discontinued medication and did not develop rebound signs or symptoms within 21 days.
Proportion of participants with evidence of adenoviral eradication (i.e. negative cell culture immunofluorescence assay [CC‐IFA] or other test like polymerase chain reaction) by day 14 after initiation of treatment.
Proportion of initially unaffected fellow eyes with signs/symptoms of infection within seven days of onset of signs/symptoms in first eye.

We extracted the available data for each outcome closest to the time points specified.

Adverse effects

We have reported the proportion of participants who had experienced any of the following adverse outcomes between treatment initiation following diagnosis of EKC and the longest follow‐up time point of each study.

Severe discomfort in the eye.
Severe conjunctival reaction such as redness or excessive tearing or other discharge (possible indication of toxicity).
Rebound of SEI after cessation of topical medication (and length of time between cessation and recurrence when reported).
Intraocular pressure ≥ 22 mmHg.
Bacterial superinfection.

Search methods for identification of studies

Electronic searches

The Cochrane Eyes and Vision (CEV) Information Specialist searched the following electronic databases for RCTs and controlled clinical trials. There was no restriction regarding language or year of publication. The date of the last database searches was 27 April 2021.

Cochrane Central Register of Controlled Trials (CENTRAL, which contains the CEV Trials Register; 2021, Issue 4) in the Cochrane Library (searched on 27 April 2021) (Appendix 1).
MEDLINE Ovid (1946 to 27 April 2021) (Appendix 2).
Embase Ovid (1947 to 27 April 2021) (Appendix 3).
PubMed (1948 to 27 April 2021) (Appendix 4).
LILACS (Latin American and Caribbean Health Science Information database) (1982 to 27 April 2021) (Appendix 5).
US National Institutes of Health Ongoing Trials Register ClinicalTrials.gov (www.clinicaltrials.gov) (Appendix 6).
World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) (www.who.int/ictrp) (Appendix 7).

Searching other resources

We searched the reference lists of included studies to identify any additional eligible studies. We contacted study investigators for trial registration records without publications or studies for which inclusion criteria were unclear. If the investigator's contact information was missing or incorrect, we iteratively searched Google and PubMed by investigator and study identifiers. For unpublished trials, we searched regulatory and pharmaceutical databases using intervention therapeutics as search terms (Appendix 8). We did not search conference proceedings or journals specifically for this review; abstracts from many conferences have been identified by Cochrane collaborators and included in CENTRAL and Embase, which were searched as part of the electronic database searches.

Data collection and analysis

Selection of studies

After removal of duplicates from the search results, two review authors independently screened the titles and abstracts of all records identified by the search using the internet‐based review management software Covidence (Covidence 2021). The review authors classified each record as either relevant ('yes' or 'maybe') or irrelevant ('no') for full‐text review. Pairs of review authors independently assessed the full‐text copies of all reports from each study identified as relevant by any review author during title and abstract screening to determine whether the study met the review inclusion criteria. We contacted the authors of the study reports to obtain clarification needed to make a complete assessment of study eligibility. Whenever authors did not respond within 14 days, we used information available from publications and trial registers. We documented the reasons for exclusion for each study assessed as ineligible after full‐text review. Any discrepancies between review authors were resolved by discussion at each stage of the screening process. For reports published in languages other than English, we first applied Google translation, and then consulted colleagues who were fluent in the language to determine eligibility; we also asked for their assistance with data extraction whenever the study was deemed eligible for inclusion.

Data extraction and management

Two review authors independently extracted data using Covidence for each included trial (Covidence 2021). We extracted the following information: study setting, countries where participant recruitment and treatment took place, sample size, study duration, planned and actual participant follow‐up time, study design features (unit of randomization; one or two study eyes per participant, paired‐eyes design), method of analysis (e.g. intention‐to‐treat, completers only), source(s) of funding and potential conflicts of interest, characteristics of the participants, underlying diseases or conditions, symptoms at time of diagnosis (pain, tearing/watery discharge, itching, foreign body sensation, and photophobia), clinical findings at diagnosis (chemosis, hyperemia, bulbar conjunctival injection, epithelial keratitis, SEI), medical history (clinically suspected or laboratory‐confirmed EKC), type of topical interventions (e.g. virustatic agents, antiseptic agents, immune modulators and suppressors such as corticosteroid, tacrolimus), alternative medication, or placebo (no treatment), outcomes (e.g. domain, specific measurement, specific metric, method of aggregation, and time frame), and quantitative results.

Two review authors compared the extracted data, resolving any discrepancies by discussion. One review author entered data into RevMan Web (RevMan Web 2020), and a second review author verified the data entry.

Assessment of risk of bias in included studies

Two review authors independently assessed the risk of bias in included studies, following the guidance in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Specific domains for consideration included random sequence generation and allocation concealment (selection bias), masking of participants and study personnel (performance bias), masking of outcome assessors (detection bias), missing data and failure to use an appropriate method of data analysis (e.g. intention‐to‐treat analysis) (attrition bias), and selective outcome reporting (reporting bias). We judged each domain for each study as having low risk, high risk, or, when the information provided was insufficient to make an assessment, unclear risk of bias. We documented reasons for our assessments, resolving any discrepancies through discussion. Our overall assessments are presented in a risk of bias summary figure (Higgins 2011).

Measures of treatment effect

We used mean difference values for comparison of continuous outcomes, including mean time (measured in days) from initiation of treatment until clinically assessed signs and symptoms had resolved. We used risk ratios for comparison of dichotomous or categorical outcomes, including the proportion of participants with resolved clinical signs and subjective symptoms, proportion of participants in whom SEI had developed, who experienced resolution of SEI, and who experienced any adverse effects.

Unit of analysis issues

EKC is often bilateral, thus all the review outcomes of interest were at the participant level. An outcome of interest was the proportion of participants in whose fellow eye EKC had developed within seven days of signs and symptoms in the first eye. When both eyes were deemed 'study eyes' and reported at the eye level, we examined the analyses to determine whether a method to account for potential between‐eyes correlation of outcomes had been used. In three of the included studies (Gouider 2021; Kovalyuk 2017; Matsuura 2021), both eyes of all or some participants were included. However, none of these studies considered intraperson correlation in the analysis of outcomes. In the absence of such an analysis or of sufficient information to calculate the correlation, we analyzed the data as reported rather than omitting the study from the review. As a result, confidence intervals were wider than they would have been if the potential within‐person correlation had been accounted for.

Dealing with missing data

Where outcome data from the included studies were unclearly reported, inadequate, or missing, we contacted the authors of study reports via email. We received responses that included data clarification, corrections, and additional information from several study authors (Gouider 2021; Matsuura 2021; Rafe 2020; Tabbara 2001). We documented these communications, along with descriptions of each study, in the Characteristics of included studies section (see also Included studies). When no response was received within two weeks, we analyzed the data using the best information available. We did not impute any missing data. We also attempted to locate complete reports of trials on pharmaceutical websites. In the event that the quality of the available data prevented any meaningful analysis, we omitted the study from quantitative analyses and reported the data in a narrative format when appropriate.

Assessment of heterogeneity

Heterogeneity of individual effect estimates among studies may affect the overall strength of the evidence to support an overall effect estimate. To assess clinical and methodological heterogeneity, we compared participant characteristics, pharmacotherapies, and outcomes by carefully reviewing the study reports and taking into consideration potential risk of bias. We planned to summarize findings in forest plots, and reported I² statistics to convey heterogeneity of effect estimates among studies included in a given analysis. We also examined confidence intervals on effect estimates for overlap among studies. We considered an I² value greater than 50% as indicative of substantial statistical heterogeneity, suggesting that a meta‐analysis of the heterogeneous group of studies may not be appropriate. We also considered the consistency of the effect estimates across studies. When all estimates were in the same direction, we frequently conducted a meta‐analysis even in the presence of substantial statistical heterogeneity, and sought an explanation for the heterogeneity.

Assessment of reporting biases

We did not use funnel plots to assess small‐study effects, which could be due to publication bias, given that fewer than 10 trials—typically only two trials—contributed data to any meta‐analysis. We assessed selective outcome reporting as part of the risk of bias assessment.

Data synthesis

We followed the recommendations in Chapter 9 of the Cochrane Handbook for Systematic Reviews of Interventions for data synthesis and analysis (Higgins 2011). We used a fixed‐effect model for quantitative syntheses when fewer than three studies reported data for the same outcome. Otherwise, we provided a descriptive qualitative synthesis of all the included studies and their results.

We compared virustatic agents (ganciclovir, trifluridine); antiseptic agent (PVP‐I, polyvinyl alcohol‐iodine [PVA‐I]); and immune modulators including calcineurin inhibitors (CsA, tacrolimus) versus artificial tears of various formulations, saline, or dexamethasone of different strengths with or without antimicrobial agent.

Subgroup analysis and investigation of heterogeneity

We planned to perform subgroup analysis according to the type of pharmacological intervention. Because only a few studies using artificial tears or steroids as a comparator reported the same outcome metrics, we performed subgroup analysis by therapeutic intervention only on the following prespecified outcomes:

proportion of participants reporting symptom resolution within seven days of treatment;
proportion of participants in whose eyes SEI had developed by day 21 after initiation of treatment;
proportion of participants in whose eyes SEI had disappeared by day 21 after initiation of treatment.

Sensitivity analysis

We planned to conduct sensitivity analyses to determine the impact on the effect estimates of restricting the analysis to studies that were not industry funded, studies of high methodological quality as reflected in low overall risk of bias across domains for which each study was assessed, and studies that evaluated one eye per participant versus used a paired‐eyes design and accounted for between‐eyes correlation. However, insufficient data precluded these analyses. Most studies analyzed and reported data at the participant level, with a few exceptions in which authors reported outcomes at the individual, eye level, or at both participant and eye levels (Gouider 2021; Matsuura 2021; Ricciardelli 2021; Trauzettel‐Klosinski 1980). None of the included studies was funded by industry, although in one trial the active intervention was supplied by a pharmaceutical company (Matsuura 2021).

Summary of findings and assessment of the certainty of the evidence

We planned to present a summary of findings table that included estimates for seven outcomes, that is all the primary and secondary outcomes specified in this review.

Mean time from initiation of treatment until resolution of symptoms, signs, or both.
Proportion of participants with resolved symptoms, signs, or both by day 7 after initiation of treatment.
Proportion of participants in whose eyes SEI had developed by day 21 after initiation of treatment.
Proportion of participants in whose eyes SEI had disappeared by day 21 after initiation of treatment.
Proportion of participants who discontinued medication and did not develop rebound SEI within 21 days
Proportion of participants with evidence of adenoviral eradication (by CC‐IFA or other test) by day 14 after initiation of treatment.
Proportion of initially unaffected fellow eyes with signs/symptoms of infection within seven days of onset of signs/symptoms in the first eye.

We summarized these outcomes for each comparison of interest whenever data were available and indicated the strength and limitation of the evidence for these and adverse outcomes. We used the GRADE approach to assess the certainty of evidence (effect estimates) for each outcome (Schünemann 2021). We began our GRADE assessment by judging the randomized design of each RCT to confer a high certainty of evidence for each outcome. We downgraded certainty to moderate, low, or very low when there was evidence of high risk of bias in reporting, inconsistency, indirectness, imprecision, or publication bias, following the guidance in Chapter 14 of the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2021).

Results

Description of studies

Results of the search

Our last search of the electronic databases was in April 2021, which yielded 4978 records. We also added three records through handsearching other sources. Overall, we screened 4166 titles and abstracts after removal of duplicates, excluding 4065 records. We retrieved 101 full‐text reports for further screening; we excluded 54 trials (81 records) and documented reasons for exclusion (see Characteristics of excluded studies table). We assessed one study (one record) as ongoing (CTRI/2020/06/025666), three studies (five records) as awaiting classification (Lerche 2001; Shabazi 2018; Validad 2017), and included 10 studies (14 records) in the review (see Included studies) (Figure 1).

Figure 1

Study flow diagram.

Included studies

Type of studies

All 10 included studies were RCTs with a parallel‐group design conducted among participants diagnosed with EKC. Each participant was assigned randomly to the intervention or comparator group, except in one study where the test intervention (ChibroCadron, containing 0.1% dexamethasone phosphate and neomycin) was initiated for 17 patients before the etiology of infection was identified (Trauzettel‐Klosinski 1980). The 36 remaining participants were assigned randomly to the test or control intervention; the 17 participants already treated with the intervention were retained in the ChibroCadron arm (Trauzettel‐Klosinski 1980).

Two studies had three study arms (Kovalyuk 2017; Ward 1993), while the remaining studies were two‐arm studies. The included studies were conducted at a single center in one of the following countries: Asia (India, Japan, Pakistan, the Philippines), Europe (Germany, Italy), the Middle East (Egypt, Israel, Saudi Arabia), and Africa (Tunisia). Studies were published between 1980 and 2021. The duration of symptom onset and participant follow‐up varied among studies; however, all participants had at least seven days of follow‐up after treatment initiation, with the longest follow‐up of six months (Bhargava 2019). In Trauzettel‐Klosinski 1980, a subset of participants returned for a follow‐up examination 9 to 18 months after study entry. No study reported having industry funding. Three studies reported information on trial registration, yet there was no study protocol publicly available (Bhargava 2019; Elwan 2020; Gouider 2021). For details on the included studies, see Characteristics of included studies.

Types of participants

The 10 trials enrolled a total of 892 participants (834 with reported outcomes, 93.5%), with 18 to 350 participants enrolled per study. More than 50% of participants were men (range: 44% to 90%); the age of participants (when reported) ranged from 9 to 82 years. Seven trials enrolled patients with acute findings of EKC. Three studies enrolled only patients who had SEI (Bhargava 2019; Gouider 2021; Rafe 2020), which is a subacute to chronic finding of EKC. At study entry, participants with SEI had a varying duration of the associated signs or symptoms of SEI (four weeks for Bhargava 2019, two weeks for Gouider 2021, and "fresh" for Rafe 2020). Symptoms such as redness or discomfort occur with acute conjunctivitis, not with SEI. Some outcome measures of interest therefore pertained only to patients with acute EKC and others to those with SEI.

Types of interventions

Not only the interventions under evaluation, but also the control treatments varied among the included trials (Table 1). Control treatments fell into two pharmaceutical categories: steroid or steroid‐containing agents and artificial tears or saline. The former included mostly dexamethasone in concentrations of 0.01%, 0.05%, or 0.5%; loteprednol; and fluoromethalone (FML) 0.1%. Various formulations of artificial tears were used (hypromellose, polyvinyl alcohol, or unspecified), with or without preservatives.

See: Summary of findings 1 Any intervention compared with artificial tears or saline; Summary of findings 2 Any intervention compared with steroid alone or plus levofloxacin; Summary of findings 3 Topical steroid compared with artificial tears

Table 1. Summary of included studies

Study ID	Interventions (No. of participants/eyes randomized)
	Test intervention	Comparator(s)
	Test intervention	Steroid	Artificial tears
Elwan 2020	PVP‐I 5%	NA	Saline
Elwan 2020	(200/200)	NA	(150/150)
Ricciardelli 2021	PVP‐I 0.6%	NA	Hyaluronate tears
Ricciardelli 2021	(34/34 analyzed)	NA	(25/25 analyzed)
Tabbara 2001	Ganciclovir ointment 0.15%	NA	Artificial tears
Tabbara 2001	(9/9)	NA	(9/9)
Trauzettel‐Klosinski 1980	Dexamethasone 0.1% + neomycin	NA	Artificial tears
Trauzettel‐Klosinski 1980	(18/18) (35/35 analyzed)	NA	(18/18)
Kovalyuk 2017	PVP‐I 1.0% + dexamethasone 0.1%	Dexamethasone 0.1%	Artificial tears
Kovalyuk 2017	(23/26)	(26/26)	(25/26)
Ward 1993	Trifluridine 1%	Dexamethasone 0.5%	Artificial tears
Ward 1993	(25/25)	(25/25)	(25/25)
Bhargava 2019 (SEI at baseline)	Tacrolimus 0.03%	Dexamethasone 0.05%	NA
Bhargava 2019 (SEI at baseline)	(45/45)	(45/45)	NA
Gouider 2021 (SEI at baseline)	CsA 0.5%	FML 0.1%	NA
Gouider 2021 (SEI at baseline)	(23/34)	(28/38)	NA
Rafe 2020 (SEI at baseline)	CsA 0.05%	Loteprednol 0.5%	NA
Rafe 2020 (SEI at baseline)	(44/44)	(44/44)	NA
Matsuura 2021	FML 0.1% + PVA‐I (eq. PVP‐I 0.055%), switched to: FML 0.1% + levofloxacin 0.5% at day 15	FML 0.1% + levofloxacin 0.5%	NA
Matsuura 2021	(10/14)	(9/13)	NA

CsA: cyclosporin A, eq.: equivalent, FML: fluorometholone, NA: not applicable, PVA‐I: polyvinyl alcohol iodine; PVP‐I: povidone‐iodine, SEI: subepithelial infiltrates

We examined the pharmaceutical effects by categories of control treatments and stratified the analysis by control treatment. Comparison 1 (any treatment versus artificial tears or saline) included data from Elwan 2020 (PVP‐I 5% versus saline); Ricciardelli 2021 (PVP‐I 0.6% versus hyaluronate tears); Tabbara 2001 (ganciclovir 0.15% ointment versus non‐preserved artificial tears); and Trauzettel‐Klosinski 1980 (ChibroCadron, a combination of topical steroid and neomycin, versus Vidisept, polyvinylpyrrolidone, a tear substitute). Comparison 2 (any treatment versus steroid) included data from Bhargava 2019 (tacrolimus 0.03% ointment versus dexamethasone 0.05% ointment); Gouider 2021 (CsA 0.5% versus FML 0.1%); and Rafe 2020 (CsA 0.05% versus loteprednol 0.5%).

In one of the three‐arm trials (Kovalyuk 2017), the investigators compared PVP‐I 1.0% plus dexamethasone 0.1% drops versus dexamethasone 0.1% drops versus artificial tears (hypromellose 0.3%), all administered at a frequency of four times per day. In the other three‐arm study (Ward 1993), the comparisons consisted of trifluridine 1% versus dexamethasone 0.5% versus Liquifilm tears (artificial tears; polyvinyl alcohol 1.4% weight per volume). We included paired comparisons from these two trials in both Comparisons 1 and 2 by the corresponding control treatment. In brief, we included six trials in Comparison 1 (summary of findings Table 1) and five trials in Comparison 2 (summary of findings Table 2). We did not include Matsuura 2021 in Comparison 2 because both arms contained FML, and the control treatment also contained an antimicrobial agent (levofloxacin).

Types of outcomes

Most investigators randomized at the participant level or reported their results at the individual level; the authors of two trials did not provide relevant information (Tabbara 2001; Ward 1993). Kovalyuk 2017 and Trauzettel‐Klosinski 1980 included both eyes of some participants as study eyes but did not consider intraperson correlation when reporting the outcomes. No included study had a paired‐eyes design. We analyzed the data as reported in the study publication(s) and by any study author via personal communication. Outcome measures of interest did not pertain to all trials as seven trials enrolled patients with acute EKC, and three trials enrolled patients with existing SEI (Bhargava 2019; Gouider 2021; Rafe 2020), with SEI duration ranging from "fresh" to four weeks.

Symptom/sign resolution by day 7

Four of 10 included studies reported the proportion of participants whose ocular symptoms had resolved within 10 days after treatment initiation (Elwan 2020; Matsuura 2021); Ricciardelli 2021 reported this outcome at 3 weeks after the first study visit. Another study alluded to the proportion of participants with resolution of symptoms by seven days, but some participants did not attend their last follow‐up examination (Kovalyuk 2017). Similarly, three studies reported the proportion of participants whose ocular signs (in particular "red eyes" or "conjunctival injection") had resolved within 10 days after treatment (Elwan 2020; Matsuura 2021; Ricciardelli 2021). Instead of proportion of participants with resolution within a specified time period, Tabbara 2001 and Ward 1993 reported, respectively, mean and median numbers of days from treatment initiation to resolution of cardinal symptoms or signs of conjunctivitis, such as red eyes, tearing, or severe discomfort. Bhargava 2019 reported mean changes in symptom scores at months 3 and 6 using a self‐developed scoring system.

SEI development/resolution/rebound by day 21

The three studies that enrolled participants with SEI at baseline reported the proportion of participants whose SEI disappeared without continued treatment (Bhargava 2019; Gouider 2021; Rafe 2020), from which the proportion of SEI recurrence at a later time in the clinical course could be calculated. Bhargava 2019 reported mean changes in SEI scores using an unvalidated scoring system at months 3 and 6 as a quantitative measure to follow disease severity.

The remaining studies, except for Ward 1993, reported the proportion of participants who developed SEI at various time points after treatment initiation (Elwan 2020; Kovalyuk 2017; Ricciardelli 2021; Tabbara 2001; Trauzettel‐Klosinski 1980).

Adenoviral eradication by day 14

Kovalyuk 2017 reported percentages of viral titer reduction for participants in each trial arm at the end of their study, day 7 after treatment initiation. In Matsuura 2021, human adenoviral DNA copy number was measured at each visit by sampling each participant's conjunctiva.

Signs or symptoms development of initially unaffected fellow eyes

No included studies reported this outcome.

Adverse effects

Five included studies (50%) reported adverse events (Bhargava 2019; Gouider 2021; Matsuura 2021; Trauzettel‐Klosinski 1980; Ward 1993).

Excluded studies

We excluded 54 studies after evaluating the full‐text report and documented the reasons for their exclusion in the Characteristics of excluded studies tables. Briefly, we excluded 27 studies (50%) because they did not evaluate any intervention of interest; 18 studies did not enroll the population of interest (33%); and seven studies were neither RCTs nor quasi‐RCTs (13%). We also excluded two trials that were aborted before the initiation of participant recruitment (EUCTR2017‐004614‐25‐FI; NCT01600365).

Ongoing studies and studies awaiting classification

We identified one ongoing study (CTRI/2020/06/025666). Three eligible studies for which there was insufficient information for evidence synthesis are awaiting classification (Lerche 2001; Shabazi 2018; Validad 2017).

Risk of bias in included studies

We reported our risk of bias judgements and rationales for these judgements in the Characteristics of included studies tables. We assessed most of the included studies as at high or unclear risk of bias for each of the risk of bias domains (Figure 2). Few authors responded to email requests for clarification regarding the process of randomization, allocation concealment, or masking in terms of risk of bias. We summarized the overall degree of risk of bias for the current review with regard to the six domains in Figure 3.

Figure 2

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Figure 3

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Allocation

Random sequence generation

Five studies reported the method of random sequence generation in sufficient detail to permit a judgement on the potential for bias (Bhargava 2019; Gouider 2021; Kovalyuk 2017; Ricciardelli 2021; Ward 1993). Four studies did not describe how their random sequence was generated or how random assignment of participants was performed beyond simply mentioning that randomization had performed. We judged these studies to be at unclear risk of bias. The remaining study reported using "nonprobability lottery method" and did not provide any information about random sequence generation; we therefore also judged it to be at unclear risk of bias (Rafe 2020).

Allocation concealment before assignment

Four studies clearly reported that treatment allocation was concealed prior to assignment (Bhargava 2019; Gouider 2021; Kovalyuk 2017; Ward 1993). The remaining six studies did not provide information about treatment allocation and were therefore judged to be at unclear risk of bias.

Blinding

Masking of participants and personnel

Three studies stated that both participants and study investigators were masked to the assigned intervention and were therefore judged to be at low risk of performance bias (Bhargava 2019; Gouider 2021; Ward 1993). In contrast, Tabbara 2001, Elwan 2020, and Matsuura 2021 had an open‐label design and were therefore judged to be at high risk of performance bias. Similarly, Trauzettel‐Klosinski 1980 was deemed at high risk because a subset of participants prescribed dexamethasone 0.1% prior to randomization were told to continue the same treatment (steroid), and the authors did not report separate outcome data for those participants who were randomized. We also assessed Rafe 2020 to be at high risk of performance bias because "only the doctors know which patient received which medicine". In Kovalyuk 2017, participants were not masked to the treatment to which they had been assigned; the investigators stated it would have been very difficult to conceal the brown color of PVP‐I. In Ricciardelli 2021, there were no descriptions about how masking of study participants were performed. We judged these last two studies to be at unclear risk of performance bias.

Masking of outcome assessors

Authors of reports from three studies stated that outcome assessors were masked (Bhargava 2019; Gouider 2021; Ward 1993); these studies were deemed as at low risk of detection bias. In Kovalyuk 2017, the investigator attempted to mask the assessor(s) by requesting that the participants not instill treatment drops within two hours of assessment in the belief that the color of PVP‐I would dissipate outside this time frame. Although it was unclear how completely the color dissipated, we classified Kovalyuk 2017 as being at low risk of bias by design. Both Matsuura 2021 and Rafe 2020 were open‐label trials, and thus at high risk of detection bias by design. The investigators of the four other trials did not provide information on whether or how outcome assessment was masked; we judged them as having unclear risk of detection bias (Elwan 2020; Ricciardelli 2021; Tabbara 2001; Trauzettel‐Klosinski 1980).

Incomplete outcome data

Three studies reported outcomes for all or nearly all participants randomized (Matsuura 2021; Rafe 2020; Ward 1993), with few participants lost to follow‐up (Ward 1993). In Kovalyuk 2017, nine study participants (35%) and two control participants (4%) "did not attend their last examination (on day 7) due to (self‐reported) recovery following their day 5 visit." Although the investigators confirmed these participants' resolution of symptoms by telephone interview, we judged such differential missed examination rates as placing the trial at high risk of incomplete outcome data. We similarly judged Ricciardelli 2021 as at high risk of attrition bias because the study did not report outcomes for 9 participants who failed to complete follow‐up visits (13% of 68 randomized).

Other studies were at unclear risk of incomplete reporting due to either inconsistent or insufficient information provided (Bhargava 2019; Elwan 2020; Gouider 2021; Tabbara 2001; Trauzettel‐Klosinski 1980). For example, Bhargava 2019 stated the study data analysis to be "intent‐to‐treat," but denominators indicated that results were reported only for treatment completers (compliers). The authors of Gouider 2021 provided additional information about participants who discontinued treatment or who were withdrawn from the study because of treatment intolerance after one month of treatment. The Gouider 2021 authors still included these participants in the analysis for outcomes reported at months 1 to 3 and 6, without detailing how they addressed missing data. In Trauzettel‐Klosinski 1980, the authors did not report the number of participants examined at each follow‐up visit. Insufficient information from Tabbara 2001 prevented our judging the completeness of outcome assessments.

Selective reporting

Although three studies reported trial registration, we found no publicly available study protocol to permit comparisons of planned outcomes with those reported; therefore, risk of selective reporting was unclear (Bhargava 2019; Elwan 2020; Matsuura 2021). Other studies did not report trial registration or cite a publicly available trial protocol that would have allowed us to assess the potential for reporting bias. However, we judged Kovalyuk 2017 to be at high risk of reporting bias because the authors prespecified the outcome measures at the participant level in the methods section of their publication, but reported outcomes at the eye level. Likewise, we assessed Gouider 2021 to be at high risk of reporting bias as the authors appeared to randomize participants at the individual level but reported outcomes inconsistently at the person or eye level.

Other potential sources of bias

Five of the 10 included trials did not report sources of funding. Four studies reported an academic institution, research institution, or government organization as the funding source (Elwan 2020; Tabbara 2001; Trauzettel‐Klosinski 1980; Ward 1993); Bhargava 2019 stated that the study was self‐funded.

Effects of interventions

Comparison 1: any treatment versus artificial tears or saline

Because the included studies reported varying ocular symptoms or signs, we compared intervention effects on resolution of cardinal symptoms (burning or eye discomfort) and signs (conjunctivitis, 'red eye,' or tearing) (summary of findings Table 1).

Primary outcomes

Mean number of days from treatment initiation to resolution of symptoms or signs

Investigators of Tabbara 2001 and Ward 1993 assessed the intervention treatment on this outcome by comparing ganciclovir 0.15% ointment and trifluridine against artificial tears, respectively (Analysis 1.1). As described in a meeting abstract, Tabbara 2001 recruited 18 consecutive patients with EKC into the open‐label trial and examined each participant every three days until complete resolution of conjunctivitis. By the end of the study on day 30, participants treated with ganciclovir had required a mean of 7.7 days of treatment (range: 7 to 12) for resolution of conjunctivitis, whereas participants in the control group had required a mean of 18.5 days (range: 7 to 30). The single‐study estimate for the effect of ganciclovir on reducing time with signs of conjunctivitis was 10.8 days shorter (95% confidence interval (CI) 8.2 to 13.4 days shorter).

In contrast, both the control group and the trifluridine group in Ward 1993 had 11 days of symptoms or signs, but the standard deviation was not reported. Based on information in the study report, we assumed normal distribution of the times from treatment initiation to symptom/sign resolution; we estimated the standard deviation (5.96) and no differences in mean days with EKC symptoms or signs between the two comparison groups. We did not combine the results from these two trials because of substantial between‐study heterogeneity (I² = 96.0%) and the lack of overlapped 95% CIs (Figure 4). Overall, the certainty of evidence for this effect estimate was very low, downgraded for risk of bias (−1), inconsistency (−1), and imprecision due to small sample size (−1).

Figure 4

Forest plot of comparison 1: Any treatment versus artificial tears or saline, outcome: 1.1 Mean number of days from treatment initiation to resolution of symptoms or signs.

Proportion of participants with resolution of symptoms within 10 days of treatment

Three of the six trials comparing povidone‐iodine (PVP‐I), with or without steroid, versus artificial tears or saline evaluated this outcome within 10 days of treatment (Analysis 1.2). The largest study involved 350 adults randomized to PVP‐I 5% (N = 200) or normal saline eye wash (N = 150) (Elwan 2020). After one week to 10 days of treatment, 95% of participants in the PVP‐I group showed complete resolution of "eye discomfort," whereas 80% of those in the saline group did. In Ricciardelli 2021, all 59 participants in the PVP‐I 0.6% group or the tear substitute group who completed the follow‐up were pain‐free on day 6 with treatment. The combined estimated risk ratio (RR) was 1.15 (95% CI 1.07 to 1.24; 2 studies; N = 409) comparing PVP‐I alone with artificial tears.

In Kovalyuk 2017, 9 participants (34% of 26 eyes) in the PVP‐I plus dexamethasone group and 1 participant in the artificial tear group (4% of 26 eyes) did not return to the last study visit on day 7, stating that they were symptom‐free during a follow‐up telephone interview. The single‐study estimated RR was 9.00 (95% CI 1.23 to 66.05; 52 eyes); there was substantial heterogeneity between the two subgroups (I² = 75.4%, P = 0.04) (Figure 5). Sources of the between‐group heterogeneity likely stemmed from the distinct anti‐inflammatory mechanism of steroid and the uncertainty introduced by eye‐level outcomes presented in Kovalyuk 2017, therefore we chose not to combine the subgroups. The evidence for the overall estimate is of low certainty because of reporting bias (−1) and imprecision due to small sample size (−1).

Figure 5

Forest plot of comparison 1: Any treatment versus artificial tears or saline, outcome: 1.2 Proportion of participants with resolution of symptoms within 10 days of treatment.

Proportion of participants with resolution of signs within 10 days of treatment

Elwan 2020 reported that "red eyes" had resolved in 65% of PVP‐I participants versus 20% artificial tears users at day 7 to 10. In Ricciardelli 2021, we compared proportions of participants in whom conjunctival injection resolved at day 6 between PVP‐I and artificial tears groups (Analysis 1.3). Overall, the estimated RR was 3.19 (95% CI 2.29 to 4.45; 2 studies; N = 409), suggesting that PVP‐I may be more effective in resolving selected ocular signs in EKC patients than artificial tears (Figure 6). Risk of bias (−1), and imprecision (−1) of these findings were reasons for downgrading the evidence to low certainty.

Figure 6

Forest plot of comparison 1: Any treatment versus artificial tears or saline, outcome: 1.3 Proportion of participants with resolution of signs within 10 days of treatment.

Secondary outcomes

Proportion of participants in whose eyes SEI had developed within 30 days of treatment initiation

For this outcome, we first stratified the analysis by time points reported by the included studies (Analysis 1.4). Four studies assessed this outcome within one month of treatment with PVP‐I alone (Ricciardelli 2021, 59 participants treated for 20 days), PVP‐I plus dexamethasone (Kovalyuk 2017, 42 eyes), ganciclovir (Tabbara 2001, 18 participants), or dexamethasone plus neomycin (Trauzettel‐Klosinski 1980, 97 eyes in 53 participants). With participant‐level data (Ricciardelli 2021; Tabbara 2001), the combined estimated RR for the treatment effect was 0.24 (95% CI 0.10 to 0.56; N = 77), suggesting beneficial effects of PVP‐I or ganciclovir in preventing SEI development assessed 30 days after the initiation of treatment (Figure 7). However, evidence from eye‐level data showed otherwise (RR 0.57, 95% CI 0.07 to 4.81; 139 eyes). Analysis by treatment duration indicated that findings from a single study also favored PVP‐I (RR 0.04, 95% CI 0.02 to 0.09; N = 350) for reduction of SEI development at three months (Elwan 2020). The subgroup differences were to be expected, presumably because of the longer observation time, the prolonged treatment effect, the unaccounted‐for intraperson correlation, or chance alone (I² = 81.3%, P = 0.005).

Figure 7

Forest plot of comparison 1: Any treatment versus artificial tears or saline, outcome: 1.4 Proportion of participants in whose eyes SEI had developed with treatment.

Additional post hoc analysis by therapeutic interventions (Analysis 1.5) showed that, except for the subgroup of PVP‐I 5% versus artificial tears (Elwan 2020; Ricciardelli 2021), the evidence did not support or refute the use of PVP‐I 1% plus dexamethasone 0.1% (Kovalyuk 2017), topical ganciclovir (Tabbara 2001), or dexamethasone plus neomycin (Trauzettel‐Klosinski 1980) for preventing the development of SEI when compared with artificial tears (Figure 8). We judged the overall evidence for this outcome as of very low certainty, downgrading for reporting bias (−1), imprecision (−1), and inconsistency plus indirectness (−1).

Figure 8

Forest plot of comparison 1: Any treatment versus artificial tears or saline, outcome: 1.5 Proportion of participants in whose eyes SEI had developed with treatment ‐ sensitivity analysis

Proportion of participants in whose eyes SEI had disappeared within 21 days of treatment

Two trials compared PVP‐I (Ricciardelli 2021, 59 participants) or combined steroid and neomycin (Trauzettel‐Klosinski 1980, 97 eyes in 53 participants) versus artificial tears and assessed disappearance of SEI during treatment (Analysis 1.6). The combined estimate of RR was 1.34 (95% CI 0.88 to 2.04; 2 studies; I² = 78%), suggesting that the test treatment was no more effective at resolving SEI than artificial tears within 21 days of treatment (Figure 9). We assessed the certainty of the evidence for this estimate as very low because of unclear risk of bias (−1), indirectness and inconsistency (−1), and imprecision arising from small sample sizes (−1).

Figure 9

Forest plot of comparison 1: Any treatment versus artificial tears or saline, outcome: 1.6 Proportion of participants in whose eyes SEI disappeared within 21 days of treatment.

Proportion of participants who discontinued medication and did not develop rebound SEI within 21 days

None of the included studies measured this outcome.

Proportion of participants with evidence of adenoviral eradication within 14 days of treatment

Kovalyuk 2017 followed participants' viral titer, finding that on the seventh day of treatment, the mean titer in the study group (PVP‐I plus dexamethasone, 17 eyes returning for the last visit) showed 100% reduction and that in the artificial tear group (25 eyes) showed an average of 77% reduction. Assuming normal distribution of the percentages of viral titer reduction, we derived the estimates of standard deviation and calculated the mean difference (MD) estimate as −23% (95% CI −40.84% to −5.16%; 1 study; 42 eyes), suggesting some benefit of PVP‐I plus dexamethasone over artificial tears in reducing viral shedding (Analysis 1.7). However, the certainty of the evidence for this outcome was very low, downgraded for risk of bias (−1), imprecision (−1), and indirectness (−1).

Proportion of initially unaffected fellow eyes that had signs/symptoms of infection within seven days of onset of signs/symptoms in the first eye

None of the included studies measured this outcome.

Adverse effects

Proportion of participants with severe discomfort upon medication distillations

In Ward 1993, investigators observed that other than mild stinging sensation upon instillation of the treatment, there was no case of severe conjunctival injection (redness or excessive tearing) or severe eye discomfort. In contrast, Trauzettel‐Klosinski 1980 reported that, during the extended follow‐up visit 9 to 18 months after treatment initiation, 7 of 12 participants in the combination steroid‐antibiotic group experienced "annoying sicca symptoms" and at least one objective sicca finding (rose bengal staining or Schirmer's strip < 5 mm at 5 minutes), while none of the participants in the tear substitute group showed any of these findings. The single‐study estimate of RR was 9.23 (95% CI 0.61 to 140.67; N = 19) (Analysis 1.8).

Participants with elevated intraocular pressure (IOP) (an increase > 6 mmHg or IOP ≥ 22 mmHg)

In Ward 1993, investigators observed no cases of elevated IOP, whereas in Trauzettel‐Klosinski 1980, the authors reported that 7.7% of steroid‐treated eyes experienced IOP up to 28 mmHg. The point estimate for RR was 5.50 (95% CI 0.31 to 96.49; 1 study; 97 eyes), indicating no statistically significant differences in risk of elevated IOP (Analysis 1.9).

Other outcomes reported in the included trials

We noted additional outcomes reported in the six trials included in this comparison that were not specified in the protocol for this review (Kuo 2020), as follows.

Mean change in symptom/sign scores: in Kovalyuk 2017, participants were questioned at each visit about five symptoms (eye itching, foreign body sensation, tearing, redness in the eye, and eyelid swelling). The severity of each symptom was scored on a 4‐point scale (0‐none, 1‐mild, 2‐moderate, 3‐severe), and scores were summed. The authors reported that the intervention group consistently showed a "superior" reduction in all scores (all P < 0.001 for comparing median estimates) but did not provide denominators (the number of participants or eyes) being evaluated at each time point.
Other ocular complications: Trauzettel‐Klosinski 1980 reported that 44% of eyes in the tear group versus 12% of steroid‐treated eyes had iritis and "deeper corneal involvement."

Comparison 2: any treatment versus steroid alone or with polyvinyl alcohol iodine (PVA‐I)

We included four studies that compared topical treatment with steroid‐containing control treatment in this comparison (summary of findings Table 2).

Primary outcomes

Mean number of days from treatment initiation to resolution of symptoms or signs

Ward 1993 identified and enrolled 75 participants in a three‐arm study that compared polyvinyl alcohol, trifluridine, and dexamethasone sodium phosphate. This was the only study that compared mean times of symptom or sign resolution between an antiviral agent and steroid. The average duration of symptoms or signs for the trifluridine and the steroid group was 11 and 10 days, respectively. Assuming a normal distribution, we derived the pooled standard deviation (5.69) and estimated MD of 1.0 day (95% CI −2.34 to 4.34; N = 49), indicating little to no difference in treatment effect on shortening the mean duration of common symptoms/signs in EKC patients (Analysis 2.1). Because of the derived standard deviation, we judged the evidence as indirect. The certainty of evidence for this effect estimate was very low, downgraded for risk of bias (−1), imprecision due to small sample size (−1), and indirectness of evidence (−1).

Proportion of participants reporting symptom resolution within seven days of treatment

Kovalyuk 2017 was the only study that reported proportions of eyes with symptom recovery within seven days of treatment. The authors reported that nine participants did not attend the last exam visit on day 7 because of self‐reported symptom resolution, which was later confirmed by telephone interview. Based on the reported numbers of eyes randomized at baseline (26 in each group), the estimated RR was 9.00 (95% CI 1.23 to 66.05; 52 eyes) (Analysis 2.2), suggesting that PVP‐I 1.0% plus dexamethasone 0.1% might hasten symptom recovery when compared with dexamethasone 0.1% alone. However, the certainty of evidence for this estimate was very low because of risk of bias (−1), small sample size (−1), and indirectness of evidence (−1).

Proportion of participants reporting sign resolution within seven days of treatment

None of the studies that compared topical treatment with steroid reported this outcome. However, in Matsuura 2021, the investigators enrolled 19 participants (27 eyes) with symptoms of acute conjunctivitis and randomly assigned these participants into either levofloxacin 1.5% plus fluorometholone (FML) 0.1% or PVA‐I plus FML 0.1% for 14 days of treatment. Beginning on day 15, treatment of levofloxacin plus FML was gradually tapered, whereas participants treated with PVA‐I plus FML were "switched to levofloxacin" (for an unstated period) before tapering. The authors reported that the proportions of eyes with resolution of signs at day 8 (RR 1.62, 95% CI 0.32 to 8.18) and at the other two study visits were similar between groups (Analysis 2.3). We judged the certainty of evidence as very low because of risk of bias (−1), imprecision (−1), and indirectness (−1).

Secondary outcomes

Proportion of eyes in which SEI had developed within 21 days of treatment

Only two studies comparing test treatment with steroid‐based control treatment reported this outcome at the eye level (Analysis 2.4) (Kovalyuk 2017; Matsuura 2021). In Kovalyuk 2017, the authors reported proportions of eyes that developed SEIs within 7 days of treatment; in Matsuura 2021, the authors presented the data at days 8, 15, and 30. The combined estimated RR for treatment effect within 21 days of treatment was 0.08 (95% CI 0.01 to 0.55; 69 eyes), indicating that eyes that received PVP‐I or PVA‐I plus steroid experienced SEIs much less often than those treated with dexamethasone alone or FML plus levofloxacin (Figure 10). Results of sensitivity analysis that included Kovalyuk 2017 and outcomes reported at day 30 by Matsuura 2021 were similar (RR 0.12, 95% CI 0.02 to 0.65). We assessed the overall certainty of the evidence for the effect of PVA‐I or PVP‐I plus steroid as very low, downgrading for risk of bias (−1), imprecision associated with small sample sizes (−1), and indirectness of evidence (−1).

Figure 10

Forest plot of comparison 2: Any treatment versus steroid alone or with levofloxacin, outcome: 2.4 Proportion of eyes in which SEI had developed within 21 days of treatment initiation.

Proportion of participants in whose eyes SEI had disappeared within four weeks of treatment initiation

The investigators of three trials purposefully enrolled EKC patients with SEI at baseline and evaluated the effects of the intervention on SEI resolution within three to four weeks of treatment (Bhargava 2019; Gouider 2021; Rafe 2020). In the latter two trials, cyclosporin A (CsA) was compared with steroid (Analysis 2.5). Rafe and colleagues enrolled 88 "fresh cases of SEIs": 44 participants received either topical loteprednol or CsA 0.05% eye drop via non‐probability lottery method (Rafe 2020). Resolution of SEIs was defined as "complete disappearance of opacities leaving behind clear cornea." The investigators reported proportions of SEI resolution at weeks 2, 4, 8, and 12; all participants in each treatment group showed resolution of SEI by the end of week 12. The single‐study estimated RR for treatment effect at week 4 was 0.84 (95% CI 0.67 to 1.06; N = 88), suggesting no differences between CsA and steroid in amelioration of SEIs with four weeks of treatment.

The 51 EKC patients (72 eyes) enrolled in Gouider 2021 had SEI of at least 14 days duration at the time of study entry. Participants were randomly assigned to receive either topical FML 0.1% (Flucon, N = 28) or CsA 0.5% (N = 24) four times a day for a month, three times a day for another month, and then twice a day for four additional months. The authors reported changes in SEI numbers (documented using slit lamp photography), best corrected visual acuity (BCVA), and tear break‐up time (TBUT) at months 1, 3, and 6 after treatment initiation. The authors noted that by month 3, only 13% of 33 eyes in the CsA group showed complete disappearance of SEI, whereas 39% of 37 eyes in the FML group did (P = 0.02). At month 6, the proportions were 47% of 30 eyes versus 70% of 33 eyes, respectively. The single‐study estimate of RR was 0.32 (95% CI 0.12 to 0.88; 70 eyes) within three months of treatment, but 0.67 (95% CI 0.43 to 1.04; 63 eyes) with up to six months of treatment. We assessed the certainty of the evidence for this outcome, either at week 4 or overall, as very low, downgrading for risk of bias (−1), small sample size (−1), and inconsistency and indirectness (−1).

Proportion of participants who discontinued medication and did not develop rebound SEI within 21 days

The three trials that enrolled only participants with SEI present at study entry reported this important outcome, but at different time points after treatment initiation (Analysis 2.6). For Rafe 2020, we derived the proportions by reverse calculation from the reported numbers of participants with recurrent SEI at week 12. The other two studies presented data at month 6 or 7 (Bhargava 2019; Gouider 2021), but Gouider 2021 reported data at the eye level (63 eyes). Given the discrepancies in treatment duration, we estimated the combined RR at month 6 to 7 to be 1.04 (95% CI 0.94 to 1.14; 3 studies, 231 eyes), assuming that each participant contributed data from one eye in Bhargava 2019 and Rafe 2020. Results were similar when only considering participant‐level data from Bhargava 2019 and Rafe 2020 (RR: 1.10; 95% CI: 0.99 to 1.22; 168 participants). Either the eye‐level or the participant‐level estimates indicated that neither tacrolimus nor CsA was more effective than steroid for preventing SEI rebound after discontinuation of treatment (Figure 11). We assessed the evidence for this estimate as of very low level of certainty because of risk of bias (−1), imprecision secondary to small sample size (−1), and indirectness of evidence (−1).

Figure 11

Forest plot of comparison 2: Any treatment versus steroid alone or with levofloxacin, outcome: 2.6 Proportion of participants who discontinued medication and did not develop rebound SEI within 21 days.

Proportion of participants with evidence of adenoviral eradication by day 14 after initiation of treatment

None of the included studies reported this outcome. Instead, Kovalyuk 2017 reported that the mean viral titer in the PVP‐I plus dexamethasone group (likely 17 eyes) had 100% reduction, whereas that in the dexamethasone group (25 eyes) had 65% reduction on the seventh day of treatment. By assuming normal distribution, we estimated the standard deviation and calculated the MD estimate to be −35% (95% CI −57% to −13%; 42 eyes), suggesting that PVP‐I plus steroid had a superior viral suppression after seven days of treatment when compared with steroid alone (Analysis 2.7). We assessed the evidence for this estimate as of very low certainty because of risk of bias (−1), imprecision (−1), and indirectness (−1).

Proportion of initially unaffected fellow eyes with signs/symptoms of infection within seven days of onset of signs/symptoms in the first eye

None of the included studies measured this outcome.

Adverse effects

Proportion of participants with severe eye discomfort, conjunctival injections, or excessive tearing when instilling intervention eye drops

Both Bhargava 2019 and Gouider 2021 reported this outcome or its equivalence (Analysis 2.8). Gouider 2021 reported proportions of (eyes) having burning sensation upon eye drops instillation at month 1 (15% versus 5%), month 3 (7% versus 6%), and month 6 (11% versus 9%) for the CsA and FML group, respectively. Between months 3 and 6, two participants in the CsA group (9%, one with painful ulcer and the other without) and two in the FML group (7%, without painful ulcer) reported severe intolerance resulting in discontinuation of treatment. In Bhargava 2019, 8 participants (18%) in the tacrolimus group complained about burning, redness, and foreign body sensation in the eyes after 1 month of treatment; 5 of these participants declined to continue the study. The combined estimate for RR of severe treatment intolerance was 4.64 (95% CI 1.15 to 18.71; 2 studies; N = 141) for tacrolimus or CsA as compared with steroid (I² = 64%, Figure 12).

Figure 12

Forest plot of comparison 2: Any treatment versus steroid alone or with levofloxacin, outcome: 2.8 Proportion of participants with severe intolerance to treatment.

Participants with elevated IOP (an increase > 6 mmHg or IOP ≥ 22 mmHg)

In Bhargava 2019, seven participants in the steroid group (15.6%) but none in the tacrolimus group showed an increase in IOP after treatment. In contrast, none of the 25 CsA and 28 steroid participants in Ward 1993 developed high IOP. The overall estimated RR was 0.07 (95% CI 0 to 1.13; 1 study; N = 80), showing no additional risks for tacrolimus as compared to steroid (Analysis 2.9).

Proportion of participants with (bacterial) superinfection

Gouider 2021 was the only study of the 10 included trials that reported a case of bacterial superinfection in the CsA group (4%) at month 3 secondary to self‐medication with topical dexamethasone, which was not the intervention steroid in the trial.

Other outcomes reported in the included trials

The following are additional outcomes not specified in the protocol for this review (Kuo 2020).

Mean change in symptom score: investigators in Bhargava 2019 assessed mean change in symptom score using a "non‐validated" questionnaire that appeared to be the authors' own Dry Eye Scoring System (range: 0 to 18). There was no difference in mean change in scores between tacrolimus and dexamethasone (MD −0.20, 95% CI −1.02 to 0.62) (Analysis 2.11).
Mean change in SEI score: Bhargava 2019 reported mean changes in SEI grading scores from gradings at baseline and six months. This scoring system was not referenced or stated to have been validated. Categories were 0‐no infiltrates, 1‐mild infiltration, 2‐moderate; 3‐severe. Compared to dexamethasone, changes in scores favored tacrolimus (MD −1.20, 95% CI −1.33 to −1.07) (Analysis 2.11).
Mean difference in IOP (mmHg): Bhargava 2019 reported MDs in IOP (mmHg) at six months: the estimated MD was −1.24 mmHg (95% CI −2.24 to −0.24), suggesting a possible effect in favor of tacrolimus versus dexamethasone.
Visual acuity: Bhargava 2019 reported visual acuity on the LogMAR scale. The observed measurements suggested a 3‐letter improvement with tacrolimus compared with dexamethasone (MD ‐0.06, 95% CI ‐0.10 to ‐0.02).

Comparison 3: topical steroid versus artificial tears

Two of the included studies were three‐arm trials; each study compared or reported outcomes associated with topical dexamethasone (0.1% in Kovalyuk 2017, 0.5% in Ward 1993) versus artificial tears. However, none of the prespecified review outcomes were reported by both studies. We assessed the certainty of the evidence for all outcomes as very low (summary of findings Table 3).

Primary outcomes

Mean number of days from treatment initiation to resolution of symptoms or signs

According to Ward 1993, participants receiving dexamethasone had a mean duration of 10 days whereas those receiving artificial tears had a mean duration of 11 days before symptom resolution. The single‐study estimated MD was −1.0 days (95% CI −4.3 to 2.3; N = 50) comparing dexamethasone with artificial tears (Analysis 3.1).

Proportion of participants with resolution of symptoms within seven days of treatment

Kovalyuk 2017 reported this outcome at day 7 of treatment but at the eye level, showing that the likelihood of eyes to resolve symptoms after treatment was comparable between dexamethasone and artificial tears (RR 1.00, 95% CI 0.07 to 15.15; 52 eyes) (Analysis 3.2).

Proportion of participants with resolution of signs within seven days of treatment

None of the included studies measured this outcome.

Secondary outcomes

Proportion of participants in whose eyes SEI had developed within 21 days of treatment

Kovalyuk 2017 measured this outcome but reported it at the eye level at day 7. The authors reported that SEIs were observed in 44% of the dexamethasone group and 20% of the artificial tears group, suggesting that the two treatments had similar effects on SEI prevention (RR 2.20, 95% CI 0.89 to 5.41; 50 eyes) (Analysis 3.3).

Proportion of participants in whose eyes SEI had disappeared within 21 days of treatment initiation

None of the included studies measured this outcome.

Proportion of participants with evidence of adenoviral eradication by day 14 after initiation of treatment

Neither trial measured this outcome. Instead, Kovalyuk 2017 presented percentages of mean viral titer reduction for the two groups: 65% reduction in the dexamethasone‐receiving eyes versus 77% reduction in the artificial tears‐receiving eyes. The estimated MD was 12% more reduction associated with dexamethasone use than with artificial tears, but the difference was not statistically significant (95% CI −34.88% to 58.88%; 50 eyes) (Analysis 3.4).

Proportion of initially unaffected fellow eyes with signs/symptoms of infection within seven days of onset of signs/symptoms in the first eye

None of the included studies measured this outcome.

Adverse effects

Proportion of participants with severe discomfort upon medication distillations

Ward 1993 reported that "no study patients complained of side effects of any drops other than mild stinging."

Participants with elevated IOP (an increase > 6 mmHg or IOP ≥ 22 mmHg)

The authors of Ward 1993 also noted that "no patients showed an increase of intraocular pressure of 6 mmHg or a pressure > 22 mm Hg."

Discussion

Summary of main results

In this review, we reported outcome data from 10 RCTs that compared topical interventions such as antiseptic agent PVP‐I or PVA‐I (with or without steroids), calcineurin inhibitors (CsA or tacrolimus), and antiviral agents (ganciclovir or trifluridine) against artificial tears or steroids (alone or with levofloxacin) in participants with EKC. We reported interventions with steroid‐based control treatment and ones with artificial tears or saline control treatment as separate comparisons. We pooled some studies to estimate the relative effect of these interventions on outcomes of shortening the duration of signs and symptoms, decreasing the proportion of participants (or eyes) with signs or symptoms after treatment, reducing the proportion of participants (or eyes) developing SEI, increasing the proportion with disappearance of SEI, and increasing the proportion without recurrence of SEI after discontinuation of treatment.

Very low certainty evidence from a meeting abstract showed that compared with artificial tears, ganciclovir, but not trifluridine, might shorten the duration of symptoms or signs of EKC by 8 to 13 days; there was no benefit when comparing trifluridine with steroids. Compared with artificial tears, PVP‐I may be associated with an increased proportion of individuals with resolution of signs or symptoms within seven days of treatment. Treatment regimen, duration of treatment, and time of assessment varied widely in the five trials examining prevention of SEI development: very low certainty evidence suggests that PVP‐I or ganciclovir may be associated with a much lower risk of SEI development compared with artificial tears. Very low certainty evidence also suggests that PVP‐I (or PVA‐I) plus steroid was associated with a much lower risk of SEI development compared with steroid use alone or with levofloxacin. Neither PVP‐I nor CsA showed an advantage in clearing SEI after three weeks to six or seven months of treatment. Evidence of any difference between CsA or tacrolimus and steroid in SEI resolution or rebound SEI was of very low certainty. In addition, there was no evidence of any difference in preventing rebound SEI between calcineurin inhibitor and steroid. Moreover, severe eye discomfort, local injection or redness, and excessive tearing were more common adverse effects of calcineurin inhibitors than of steroids. The risk of elevated IOP or any difference in reported outcomes between groups was equivocal for the comparisons between steroids versus artificial tears.

Overall completeness and applicability of evidence

Population representativeness

In all of the included studies, diagnosis was made clinically at study entry based on the striking findings of EKC. One study described diagnosis based on "acute onset of scratchy sensation and conjunctival injection with a watery discharge... acute follicular conjunctivitis, usually with preauricular adenopathy and a fine keratitis on the inferior perilimbal cornea" (Ward 1993), but few studies provided any detailed descriptions of diagnostic criteria beyond "adenoviral keratoconjunctivitis." Most trials from the 1980s to 1990s utilized technology that is still available today but no longer widely used to confirm EKC. In studies that relied on laboratory technologies to exclude ineligible participants after enrollment or randomization, investigators used methods of varying complexities with varying levels of detection sensitivity and specificity, including viral cultures (Ward 1993); serological identification of adenoviral serotype 8 (Trauzettel‐Klosinski 1980); polymerase chain reaction (PCR) but without mention of specific genotypes (Kovalyuk 2017); and a combination of PCR, Giemsa staining, and cytology (Tabbara 2001). As such, whether all participants in the included trials were truly eligible patients with EKC or a mixture with patients with atypical presentation of adenoviral conjunctivitis is questionable. Few studies provided information regarding the time since symptom/sign onset to study entry. In most studies, the time of study enrollment relative to the onset of infection and the disease course was unclear; for example, in three studies, the duration of SEI prior to study enrollment varied by a few weeks. Evidence from studies of patients who were potentially enrolled at different phases of EKC progression could be misleading. Varying proportions of participants in the acute or subacute phase could affect their likely responses to any treatment. In addition, the concentration of intervention medications such as PVP‐I, CsA, and dexamethasone varied widely, which may have also contributed to high heterogeneity in a few of the subgroups studied.

Although studies rarely provided comprehensive baseline information about participant characteristics by age, sex, race or ethnicity, we expanded our eligibility criteria to include studies with children or adolescents (see Differences between protocol and review). Such protocol modification broadened the review scope and made our findings more applicable to the general EKC population.

Pharmacologic interventions

Because of practice patterns, we expanded the list of potential comparators beyond placebo, vehicle, or artificial tears, and permitted the use of steroids as an active control treatment in one set of analyses. In screening and reviewing potentially eligible full‐text publications, we encountered many trials that compared various drug classes or natural products against control treatments. In fact, 27 of the 54 excluded studies examined intervention treatments that were beyond the scope of this review (see Excluded studies); these treatments included interferon, antihistamine, novel antiviral agents, homeopathic products, and others. We did not consider these drug classes mainly because the wide variety of formulations makes characterization of the effects of these drug classes very challenging. Future updates of this review may consider including studies that either examine a single or a few drugs of similar mechanisms, or drugs that are hypothesized to be effective in either controlling acute symptoms/signs or preventing SEI. This more focused approach may result in the analysis of fewer trials, but the evidence would appear less heterogeneous.

Control treatment

By modifying the eligible control treatments (Differences between protocol and review), we were able to include 10 trials in the review. Except for a trial with normal saline wash as the control treatment, which was published as a meeting abstract (Tabbara 2001), there were no placebo‐controlled trials. Even artificial tears can be beneficial to patients with EKC by alleviating irritation associated with discharge and keratitis. While we acknowledge that active treatment‐controlled trials may provide different answers to our review questions than placebo‐controlled trials, steroids or artificial tears are commonly used as 'standard of care.' The evidence generated from this review may therefore be more readily applicable to actual clinical care than that based on placebo‐controlled studies of patients with EKC.

Outcome measurement

As EKC can have major economic repercussions from loss of work or school days and can cause chronic visual disturbance (e.g. from SEI) (Angueyra 2021; Kuo 2018; Kuo 2021; Piednoir 2002), the primary outcomes of reducing the number of days of signs and symptoms or increasing the proportion of participants with resolution within seven days of initiation of treatment are important. We found limited evidence that PVP‐I achieves these outcomes over artificial tears. Other outcomes for which there was little or weak support by few studies were the proportion of participants with development of SEI by day 21 after initiation of treatment, and discontinuation of medication without rebound SEI by day 21 after initiation of treatment. These secondary outcomes are relevant. Patients may not be able to decrease or discontinue treatments without rebound of SEI (Bhargava 2019; Jeng 2011), or they may suffer adverse events from prolonged treatment with medications like topical steroids. In other words, the development and persistence of SEI can affect quality of life and productivity, and treatment complications may do so as well.

In general, our outcomes of interest had weak support from studies with small sample sizes. Several studies implemented scoring systems for general symptoms or signs, Kovalyuk 2017, or for SEI (Bhargava 2019), neither of which has been previously validated. Studies that reported on our outcomes of interest did so at various time points which differed from those specified in our protocol (Kuo 2020). For example, five studies contributed data to the outcome proportion of participants who had developed SEI by 21 days after treatment initiation at varying time points of the treatment course: day 7 (Kovalyuk 2017), day 21 (Ricciardelli 2021), day 30 (Tabbara 2001), week 3 to 4 (Trauzettel‐Klosinski 1980), and month 3 (Elwan 2020). Considering the varied recruitment timelines of the included trials, the specified time frame (21 days in the case of SEI development) could appear arbitrary.

No study reported on two important outcomes that are related to duration of infection (proportion of participants who had evidence of adenoviral eradication by day 14 after initiation of treatment) and level of infectiousness (proportion of initially unaffected fellow eyes that had signs or symptoms of infection within seven days of onset of signs or symptoms in the first eye). Involvement of one eye by EKC, let alone bilateral involvement, can be disabling.

Quality of the evidence

Despite changes from the original protocol to expand the number of qualifying trials by increasing the type of active comparators and including trials with both pediatric and adult patients, many factors led to downgrading of the certainty of the evidence to low or very low. These factors included heterogeneity in participant characteristics, enrollment times with regard to the disease course, pharmacologic agents of different formulations, and outcome measurements (metrics and time points), in addition to small study sizes and high risk of bias across multiple domains for most studies.

Potential biases in the review process

An Information Specialist assisted with the broad electronic database search of multiple databases including trial registries. We applied standard Cochrane methods to conduct this review and avoid potential biases associated with literature search, appraisal, data extraction, analysis, and interpretation. We also performed an extensive handsearch in regulatory databases and made frequent contacts with many trial investigators and study authors to obtain additional details regarding study design, analytic methods, and outcome measures in the case of ambiguities.

Agreements and disagreements with other studies or reviews

To the best of our knowledge, there are no other systematic reviews on the topic of pharmacologic treatment of EKC. Two recently published narrative reviews concluded that artificial tears, ganciclovir, CsA, and tacrolimus were promising interventions for relieving EKC symptoms, treating SEI, or preventing SEI development (Jonas 2020; Labib 2020). These two reviews also covered studies evaluating non‐steroidal anti‐inflammatory agents (NSAIDs), antihistamine, other antiviral agents, and immunoglobulin among other pharmacologic interventions. However, the authors of the two reviews either did not differentiate trials that enrolled only patients with EKC from those with adenoviral conjunctivitis, or also considered observational studies or phase II single‐arm trials to inform their evidence synthesis.

Figure 1

Study flow diagram.

Figure 2

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Figure 3

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Figure 4

Forest plot of comparison 1: Any treatment versus artificial tears or saline, outcome: 1.1 Mean number of days from treatment initiation to resolution of symptoms or signs.

Figure 5

Forest plot of comparison 1: Any treatment versus artificial tears or saline, outcome: 1.2 Proportion of participants with resolution of symptoms within 10 days of treatment.

Figure 6

Forest plot of comparison 1: Any treatment versus artificial tears or saline, outcome: 1.3 Proportion of participants with resolution of signs within 10 days of treatment.

Figure 7

Forest plot of comparison 1: Any treatment versus artificial tears or saline, outcome: 1.4 Proportion of participants in whose eyes SEI had developed with treatment.

Figure 8

Forest plot of comparison 1: Any treatment versus artificial tears or saline, outcome: 1.5 Proportion of participants in whose eyes SEI had developed with treatment ‐ sensitivity analysis

Figure 9

Forest plot of comparison 1: Any treatment versus artificial tears or saline, outcome: 1.6 Proportion of participants in whose eyes SEI disappeared within 21 days of treatment.

Figure 10

Forest plot of comparison 2: Any treatment versus steroid alone or with levofloxacin, outcome: 2.4 Proportion of eyes in which SEI had developed within 21 days of treatment initiation.

Figure 11

Figure 12

Forest plot of comparison 2: Any treatment versus steroid alone or with levofloxacin, outcome: 2.8 Proportion of participants with severe intolerance to treatment.

Analysis 1.1

Comparison 1: Any treatment versus artificial tears or saline, Outcome 1: Mean number of days from treatment initiation to resolution of symptoms or signs

Analysis 1.2

Comparison 1: Any treatment versus artificial tears or saline, Outcome 2: Proportion of participants with resolution of symptoms within 10 days of treatment

Analysis 1.3

Comparison 1: Any treatment versus artificial tears or saline, Outcome 3: Proportion of participants with resolution of signs within 10 days of treatment

Analysis 1.4

Comparison 1: Any treatment versus artificial tears or saline, Outcome 4: Proportion of participants in whose eyes SEI had developed with treatment

Analysis 1.5

Comparison 1: Any treatment versus artificial tears or saline, Outcome 5: Proportion of participants in whose eyes SEI had developed with treatment (sensitivity analysis of 1.4)

Analysis 1.6

Comparison 1: Any treatment versus artificial tears or saline, Outcome 6: Proportion of participants in whose eyes SEI disappeared within 21 days of treatment

Analysis 1.7

Comparison 1: Any treatment versus artificial tears or saline, Outcome 7: Percentage of mean viral titer reduction

Analysis 1.8

Comparison 1: Any treatment versus artificial tears or saline, Outcome 8: Proportion of participants with severe discomfort in the eye upon distillation

Analysis 1.9

Comparison 1: Any treatment versus artificial tears or saline, Outcome 9: Proportion of participants with elevated IOP (an increase > 6 mmHg or IOP ≥ 22 mmHg) during the study

Analysis 2.1

Comparison 2: Any treatment versus steroid alone or with levofloxacin, Outcome 1: Mean number of days from treatment initiation to resolution of symptoms or signs

Analysis 2.2

Comparison 2: Any treatment versus steroid alone or with levofloxacin, Outcome 2: Proportion of participants with resolution of symptoms within 7 days of treatment

Analysis 2.3

Comparison 2: Any treatment versus steroid alone or with levofloxacin, Outcome 3: Proportion of eyes with resolution of signs within 8 days of treatment

Analysis 2.4

Comparison 2: Any treatment versus steroid alone or with levofloxacin, Outcome 4: Proportion of eyes in which SEI had developed within 21 days of treatment initiation

Analysis 2.5

Comparison 2: Any treatment versus steroid alone or with levofloxacin, Outcome 5: Proportion of participants in whose eyes SEI had disappeared with treatment

Analysis 2.6

Comparison 2: Any treatment versus steroid alone or with levofloxacin, Outcome 6: Proportion of participants who discontinued medication and did not develop rebound SEI within 21 days

Analysis 2.7

Comparison 2: Any treatment versus steroid alone or with levofloxacin, Outcome 7: Percentage of reduction in mean viral titer

Analysis 2.8

Comparison 2: Any treatment versus steroid alone or with levofloxacin, Outcome 8: Proportion of participants with severe intolerance to treatment

Analysis 2.9

Comparison 2: Any treatment versus steroid alone or with levofloxacin, Outcome 9: Proportion of participants with elevated IOP (an increase > 6 mmHg or IOP ≥ 22 mmHg) during the study

Analysis 2.10

Comparison 2: Any treatment versus steroid alone or with levofloxacin, Outcome 10: Proportion of participants with superinfection during the study

Analysis 2.11

Comparison 2: Any treatment versus steroid alone or with levofloxacin, Outcome 11: Mean changes in symptom score or SEI score from baseline

Analysis 3.1

Comparison 3: Dexamethasone versus artificial tears, Outcome 1: Mean number of days from treatment initiation to resolution of symptoms or signs

Analysis 3.2

Comparison 3: Dexamethasone versus artificial tears, Outcome 2: Proportion of participants with resolution of symptoms within 7 days of treatment

Analysis 3.3

Comparison 3: Dexamethasone versus artificial tears, Outcome 3: Proportion of participants in whose eyes SEI had developed within 21 days of treatment

Analysis 3.4

Comparison 3: Dexamethasone versus artificial tears, Outcome 4: Percentage of mean viral titer reduction

Summary of findings 1. Any intervention compared with artificial tears or saline

Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Studies contributing to the pooled effects	Certainty of the evidence (GRADE)	Comments
Any intervention compared with artificial tears (or saline) for epidemic keratoconjunctivitis
Patient or population: epidemic keratoconjunctivitis Setting: eye hospital Intervention: any intervention (PVP‐I, ganciclovir, trifluridine, dexamethasone phosphate plus neomycin) Comparison: artificial tears or saline
Outcomes	Assumed risk with artificial tears or saline	Corresponding risk with any intervention	Relative effect (95% CI)	№ of participants (studies)	Studies contributing to the pooled effects	Certainty of the evidence (GRADE)	Comments
Mean number of days from treatment initiation to resolution of symptoms or signs (shorter is favored)	We chose to compare times to resolution of selected ocular signs and symptoms, including conjunctivitis (red eye or excessive tearing) and severe eye discomfort (see comments).		‐	‐	‐	⊕⊝⊝⊝ Very low^1,2,3	Ward 1993 compared trifluridine 1% with artificial tears (49 participants) and found no differences (MD 0 days, 95% CI 3 days shorter to 3 days longer), whereas Tabbara 2001 reported a significant reduction of days with signs of conjunctivitis (MD 11 days shorter, 95% CI 8 to 13 days shorter) comparing ganciclovir to artificial tears.
Proportion of participants with resolution of symptoms within 10 days of treatment (higher is favored)	829 per 1000	953 per 1000 (887 to 1028)	RR 1.15 (1.07 to 1.24)	409 (2 RCTs)	Elwan 2020; Ricciardelli 2021	⊕⊕⊝⊝ Low^1,2	A third trial (Kovalyuk 2017, 52 eyes in 45 participants) compared PVP‐I plus dexamethasone with artificial tears and reported an RR of 9.00 (95% CI 1.23 to 66.05) at the eye level.
Proportion of participants with resolution of signs within 10 days of treatment (higher is favored)	177 per 1000	565 per 1000 (405 to 788)	RR3.19 (2.29 to 4.45)	409 (2 RCTs)	Elwan 2020; Ricciardelli 2021	⊕⊕⊝⊝ Low^1,2
Proportion of participants in whose eyes SEIs had developed within 30 days of treatment initiation (lower is favored)	529 per 1000	127 per 1000 (53 to 296)	RR 0.24 (0.10 to 0.56)	77 (2 RCTs)	Ricciardelli 2021; Tabbara 2001	⊕⊝⊝⊝ Very low^1,2,4	Kovalyuk 2017 and Trauzettel‐Klosinski 1980 also reported this outcome within 4 weeks of treatment but at the eye level, with an RR of 0.57 (95% CI 0.07 to 4.81; 139 eyes). Elwan 2020 reported an RR of 0.04 (95% CI 0.02 to 0.09; 350 participants) after 3 months of treatment.
Proportion of participants in whose eyes SEI had disappeared within 21 days of treatment (higher is favored)	298 per 1000 eyes	399 per 1000 eyes (262 to 608)	RR 1.34 (0.88 to 2.04)	156 eyes (2 RCTs)	Ricciardelli 2021; Trauzettel‐Klosinski 1980	⊕⊝⊝⊝ Very low^1,2,4	Ricciardelli 2021 reported at the participant level (59 participants); Trauzettel‐Klosinski 1980 reported at the eye level (97 eyes).
Proportion of participants who discontinued medication and did not develop rebound SEI within 21 days (higher is favored)	Not measured by any included study		‐	‐	‐	‐
Proportion of participants who had evidence of adenoviral eradication within 14 days of treatment (higher is favored)	See comments		‐	‐	‐	⊕⊝⊝⊝ Very low^1,2,5	Kovalyuk 2017 (42 eyes) reported on percentages of reduction in mean viral titers, with an MD of −23% (95% CI −40.84% to ‐5.16%) on day 7 of treatment.
Proportion of initially unaffected fellow eyes that had signs/symptoms of infection within 7 days of onset of signs/symptoms in the first eye (lower is favored)	Not measured by any included study		‐	‐	‐	‐
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; MD: mean difference; PVP‐I: povidone‐iodine; RCT: randomized controlled trial; RR: risk ratio; SEI: subepithelial infiltrates
GRADE Working Group grades of evidence High certainty: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
¹Downgraded for risk of bias (−1). ²Downgraded either for small sample size OR the confidence interval crossing or equaling no differences (−1). ³Downgraded for inconsistency (−1). ⁴Downgraded for inconsistency (details in Effects of interventions) and indirectness (−1). ⁵Downgraded for indirectness (−1).

Summary of findings 1. Any intervention compared with artificial tears or saline

Summary of findings 2. Any intervention compared with steroid alone or plus levofloxacin

Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Studies contributing to the pooled effects	Certainty of the evidence (GRADE)	Comments
Any intervention compared with steroid alone or plus levofloxacin for epidemic keratoconjunctivitis
Patient or population: epidemic keratoconjunctivitis Setting: eye hospital Intervention: any intervention (tacrolimus, CsA, trifluridine, PVP‐I, FML plus PVA‐I) Comparison: steroid (dexamethasone, loteprednol, FML) or FML plus levofloxacin
Outcomes	Assumed risk with steroid	Corresponding risk with any intervention	Relative effect (95% CI)	№ of participants (studies)	Studies contributing to the pooled effects	Certainty of the evidence (GRADE)	Comments
Mean number of days from treatment initiation to resolution of symptoms or signs (shorter is favored)	We chose to compare times to resolution of selected ocular signs and symptoms, including conjunctivitis (red eye or excessive tearing) and severe eye discomfort (see comments).	‐	‐	‐	‐	⊕⊝⊝⊝ Very low^1,2,3	Ward 1993 (N = 49) found 1.0 day longer (95% CI 2.3 shorter to 4.3 longer) in mean duration with symptoms/signs in the trifluridine group as compared with the dexamethasone group.
Proportion of participants with resolution of symptoms within 7 days of treatment (higher is favored)	See comments	‐	‐	‐	‐	⊕⊝⊝⊝ Very low^1,2,3	Kovalyuk 2017 (52 eyes) reported this outcome at the eye level, showing an 8‐fold increase in the likelihood of symptom resolution (RR 9.00, 95% CI 1.23 to 66.05) in the PVP‐I group as compared with the dexamethasone group.
Proportion of participants with resolution of signs within 7 days of treatment (higher is favored)	See comments	‐	‐	‐	‐	⊕⊝⊝⊝ Very low^1,2,3	Matsuura 2021 compared PVA‐I plus FML with levofloxacin plus FML at the eye level, which was analyzed as a control treatment, hence the indirectness of this evidence; the result was inconclusive (RR 1.62, 95% CI 0.32 to 8.18).
Proportion of eyes in which SEI had developed within 21 days of treatment (lower is favored)	410 per 1000 eyes	33 per 1000 eyes (4 to 226)	RR 0.08 (0.01 to 0.55)	69 eyes (2 RCTs)	Kovalyuk 2017; Matsuura 2021	⊕⊝⊝⊝ Very low^1,2,3	The combined RR was 0.12 (95% CI 0.02 to 0.65) within 30 days of treatment (69 eyes, 2 RCTs).
Proportion of participants in whose eyes SEI had disappeared within 4 weeks of treatment initiation (higher is favored)	See comments	‐	‐	‐	‐	⊕⊝⊝⊝ Very low^1,2,4	Rafe 2020 reported this outcome at the participant level, showing an inconclusive effect of CsA when compared with steroid (RR 0.84, 95% CI 0.67 to 1.06; 88 participants). Gouider 2021 also reported this outcome at the eye level after 3 months of CsA as compared with steroid (RR 0.32, 95% CI 0.12 to 0.88; 70 eyes), and after 6 to 7 months of treatment (RR 0.67, 95% CI 0.43 to 1.04; 63 eyes).
Proportion of participants who discontinued medication and did not develop rebound SEI within 21 days (higher is favored)	872 per 1000 eyes (see comments)	907 per 1000 (820 to 994)	RR 1.04 (0.94 to 1.14)	231 eyes (3 RCTs)	Bhargava 2019; Gouider 2021; Rafe 2020	⊕⊝⊝⊝ Very low^1,2,3	Rafe 2020 and Bhargava 2019 reported this outcome at the participant level (168 participants, assuming 1 eye per participant) at week 12 and month 6, separately. Gouider 2021 reported at the eye level (63 eyes) after 6 to 7 months of treatment.
Proportion of participants who had evidence of adenoviral eradication by day 14 after initiation of treatment (higher is favored)	See comments		‐	‐	‐	⊕⊝⊝⊝ Very low^1,2,3	Kovalyuk 2017 (42 eyes) reported on percentages of reduction in mean viral titers, with an MD of −35% (95% CI −57% to −13%), in favor of the intervention.
Proportion of initially unaffected fellow eyes that had signs/symptoms of infection within 7 days of onset of signs/symptoms in the first eye (lower is favored)	Not measured by any included study		‐	‐	‐	‐
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; CsA: cyclosporin A; FML: fluorometholone; MD: mean difference; PVA‐I: polyvinyl alcohol iodine; PVP‐I: povidone‐iodine; RCT: randomized controlled trial; RR: risk ratio; SEI: subepithelial infiltrates
GRADE Working Group grades of evidence High certainty: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
¹Downgraded for risk of bias (−1). ²Downgraded either for small sample size OR the confidence interval crosses or equals no difference (−1). ³Downgraded for indirectness of evidence (−1). ⁴Downgraded for inconsistency and indirectness of evidence (−1).

Summary of findings 2. Any intervention compared with steroid alone or plus levofloxacin

Summary of findings 3. Topical steroid compared with artificial tears

Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Topical steroid compared with artificial tears for epidemic keratoconjunctivitis
Patient or population: epidemic keratoconjunctivitis Setting: eye hospital Intervention: dexamethasone Comparison: artificial tears
Outcomes	Assumed risk with artificial tears	Corresponding risk with dexamethasone	Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Mean number of days from treatment initiation to resolution of symptoms or signs (shorter is favored)	The mean number of days from initiation of treatment until signs were resolved was 11 days (SD 5.96). (see comment)	MD 1.0 day shorter (4.3 shorter to 2.3 longer)	‐	50 (1 RCT)	⊕⊝⊝⊝ Very low^1,2,3	Ward 1993 reported time to resolution of signs at the participant level.
Proportion of participants with resolution of symptoms within 7 days of treatment (higher is favored)	38 per 1000 eyes	38 per 1000 eyes (3 to 576)	RR 1.00 (0.07 to 15.15)	52 eyes (1 RCT)	⊕⊝⊝⊝ Very low^1,2,3	Kovalyuk 2017 reported this outcome at the eye level; verification of symptom resolution was by telephone interview.
Proportion of participants with resolution of signs within 7 days of treatment (higher is favored)	Not measured by either included study		‐	‐	‐
Proportion of participants in whose eyes SEI had developed within 21 days of treatment (lower is favored)	192 per 1000 eyes (see comment)	423 per 1000 eyes (171 to 1046)	RR 2.20 (0.89 to 5.41)	50 eyes (1 RCT)	⊕⊝⊝⊝ Very low^1,2,3	Kovalyuk 2017 reported this outcome at the eye level at day 7.
Proportion of participants in whose eyes SEI had disappeared within 21 days of treatment initiation (higher is favored)	Not measured by either included study		‐	‐	‐
Proportion of participants who discontinued medication and did not develop rebound SEI within 21 days (higher is favored)	Not measured by either included study		‐	‐	‐
Proportion of participants who had evidence of adenoviral eradication within 14 days of treatment (higher is favored)	See comment		‐	50 eyes (1 RCT)	⊕⊝⊝⊝ Very low^1,2,3	Kovalyuk 2017 reported on percentages of reduction in mean viral titers at the eye level, with an MD of 12.00% (95% CI −34.88% to 58.88%) comparing dexamethasone to artificial tears.
Proportion of initially unaffected fellow eyes that had signs/symptoms of infection within 7 days of onset of signs/symptoms in the first eye (lower is favored)	Not measured by either included study		‐	‐	‐
The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI:* confidence interval; MD: mean difference; RCT: randomized controlled trial; RR: risk ratio; SD: standard deviation; SEI: subepithelial infiltrates
GRADE Working Group grades of evidence High certainty: We are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: We are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: Our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: We have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
¹Downgraded for risk of bias (−1). ²Downgraded either for small sample size OR the confidence interval crosses or equals no difference (−1). ³Downgraded for indirectness of evidence (−1).

Summary of findings 3. Topical steroid compared with artificial tears

Table 1. Summary of included studies

Study ID	Interventions (No. of participants/eyes randomized)
	Test intervention	Comparator(s)
	Test intervention	Steroid	Artificial tears
Elwan 2020	PVP‐I 5%	NA	Saline
Elwan 2020	(200/200)	NA	(150/150)
Ricciardelli 2021	PVP‐I 0.6%	NA	Hyaluronate tears
Ricciardelli 2021	(34/34 analyzed)	NA	(25/25 analyzed)
Tabbara 2001	Ganciclovir ointment 0.15%	NA	Artificial tears
Tabbara 2001	(9/9)	NA	(9/9)
Trauzettel‐Klosinski 1980	Dexamethasone 0.1% + neomycin	NA	Artificial tears
Trauzettel‐Klosinski 1980	(18/18) (35/35 analyzed)	NA	(18/18)
Kovalyuk 2017	PVP‐I 1.0% + dexamethasone 0.1%	Dexamethasone 0.1%	Artificial tears
Kovalyuk 2017	(23/26)	(26/26)	(25/26)
Ward 1993	Trifluridine 1%	Dexamethasone 0.5%	Artificial tears
Ward 1993	(25/25)	(25/25)	(25/25)
Bhargava 2019 (SEI at baseline)	Tacrolimus 0.03%	Dexamethasone 0.05%	NA
Bhargava 2019 (SEI at baseline)	(45/45)	(45/45)	NA
Gouider 2021 (SEI at baseline)	CsA 0.5%	FML 0.1%	NA
Gouider 2021 (SEI at baseline)	(23/34)	(28/38)	NA
Rafe 2020 (SEI at baseline)	CsA 0.05%	Loteprednol 0.5%	NA
Rafe 2020 (SEI at baseline)	(44/44)	(44/44)	NA
Matsuura 2021	FML 0.1% + PVA‐I (eq. PVP‐I 0.055%), switched to: FML 0.1% + levofloxacin 0.5% at day 15	FML 0.1% + levofloxacin 0.5%	NA
Matsuura 2021	(10/14)	(9/13)	NA
CsA: cyclosporin A, eq.: equivalent, FML: fluorometholone, NA: not applicable, PVA‐I: polyvinyl alcohol iodine; PVP‐I: povidone‐iodine, SEI: subepithelial infiltrates

Table 1. Summary of included studies

Comparison 1. Any treatment versus artificial tears or saline

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Mean number of days from treatment initiation to resolution of symptoms or signs Show forest plot	2		Mean Difference (IV, Fixed, 95% CI)	Subtotals only

1.1.1 Ganciclovir 0.15%	1	18	Mean Difference (IV, Fixed, 95% CI)	‐10.80 [‐13.42, ‐8.18]
1.1.2 Trifluridine 1%	1	49	Mean Difference (IV, Fixed, 95% CI)	0.00 [‐3.34, 3.34]
1.2 Proportion of participants with resolution of symptoms within 10 days of treatment Show forest plot	3		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only

1.2.1 PVP‐I	2	409	Risk Ratio (M‐H, Fixed, 95% CI)	1.15 [1.07, 1.24]
1.2.2 PVP‐I + dexamethasone	1	52	Risk Ratio (M‐H, Fixed, 95% CI)	9.00 [1.23, 66.05]
1.3 Proportion of participants with resolution of signs within 10 days of treatment Show forest plot	2	409	Risk Ratio (M‐H, Fixed, 95% CI)	3.19 [2.29, 4.45]

1.4 Proportion of participants in whose eyes SEI had developed with treatment Show forest plot	5		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

1.4.1 Within 30 days, participant‐level	2	77	Risk Ratio (M‐H, Random, 95% CI)	0.24 [0.10, 0.56]
1.4.2 Within 4 weeks, eye‐level	2	139	Risk Ratio (M‐H, Random, 95% CI)	0.57 [0.07, 4.81]
1.4.3 Within 3 months, participant‐level	1	350	Risk Ratio (M‐H, Random, 95% CI)	0.04 [0.02, 0.09]
1.5 Proportion of participants in whose eyes SEI had developed with treatment (sensitivity analysis of 1.4) Show forest plot	5		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only

1.5.1 PVP‐I vs artificial tears	2	409	Risk Ratio (M‐H, Fixed, 95% CI)	0.06 [0.03, 0.11]
1.5.2 PVP‐I + steroid vs artificial tears	1	42	Risk Ratio (M‐H, Fixed, 95% CI)	0.13 [0.01, 2.23]
1.5.3 Ganciclovir vs artificial tears	1	18	Risk Ratio (M‐H, Fixed, 95% CI)	0.29 [0.08, 1.02]
1.5.4 Steroid + neomycin vs artificial tears	1	97	Risk Ratio (M‐H, Fixed, 95% CI)	1.11 [0.80, 1.56]
1.6 Proportion of participants in whose eyes SEI disappeared within 21 days of treatment Show forest plot	2	156	Risk Ratio (M‐H, Fixed, 95% CI)	1.34 [0.88, 2.04]

1.6.1 PVP‐I 0.6% vs artificial tears	1	59	Risk Ratio (M‐H, Fixed, 95% CI)	7.35 [1.01, 53.78]
1.6.2 Dexamethasone 0.1% + neomycin vs artificial tears	1	97	Risk Ratio (M‐H, Fixed, 95% CI)	1.02 [0.67, 1.55]
1.7 Percentage of mean viral titer reduction Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Subtotals only

1.8 Proportion of participants with severe discomfort in the eye upon distillation Show forest plot	2	68	Risk Ratio (M‐H, Fixed, 95% CI)	9.23 [0.61, 140.67]

1.8.1 Trifluridine 1% vs artificial tears	1	49	Risk Ratio (M‐H, Fixed, 95% CI)	Not estimable
1.8.2 Dexamethasone 0.1% + neomycin vs artificial tears	1	19	Risk Ratio (M‐H, Fixed, 95% CI)	9.23 [0.61, 140.67]
1.9 Proportion of participants with elevated IOP (an increase > 6 mmHg or IOP ≥ 22 mmHg) during the study Show forest plot	2	146	Risk Ratio (M‐H, Fixed, 95% CI)	5.50 [0.31, 96.49]

1.9.1 Trifluridine 1% vs artificial tears	1	49	Risk Ratio (M‐H, Fixed, 95% CI)	Not estimable
1.9.2 Dexamethasone 0.1% + neomycin vs artificial tears	1	97	Risk Ratio (M‐H, Fixed, 95% CI)	5.50 [0.31, 96.49]

Comparison 1. Any treatment versus artificial tears or saline

Comparison 2. Any treatment versus steroid alone or with levofloxacin

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Mean number of days from treatment initiation to resolution of symptoms or signs Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Subtotals only

2.2 Proportion of participants with resolution of symptoms within 7 days of treatment Show forest plot	1		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only

2.3 Proportion of eyes with resolution of signs within 8 days of treatment Show forest plot	1		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only

2.3.1 Day 8	1	27	Risk Ratio (M‐H, Fixed, 95% CI)	1.62 [0.32, 8.18]
2.3.2 Day 15	1	27	Risk Ratio (M‐H, Fixed, 95% CI)	1.08 [0.63, 1.83]
2.3.3 Day 30	1	27	Risk Ratio (M‐H, Fixed, 95% CI)	1.07 [0.88, 1.30]
2.4 Proportion of eyes in which SEI had developed within 21 days of treatment initiation Show forest plot	2	96	Risk Ratio (M‐H, Fixed, 95% CI)	0.12 [0.03, 0.48]

2.4.1 PVP‐I 1% + steroid, day 7	1	42	Risk Ratio (M‐H, Fixed, 95% CI)	0.06 [0.00, 1.00]
2.4.2 PVA‐I + steroid, day 15	1	27	Risk Ratio (M‐H, Fixed, 95% CI)	0.10 [0.01, 1.60]
2.4.3 PVA‐I + steroid, day 30	1	27	Risk Ratio (M‐H, Fixed, 95% CI)	0.27 [0.03, 2.11]
2.5 Proportion of participants in whose eyes SEI had disappeared with treatment Show forest plot	2		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only

2.5.1 Within 4 weeks	1	88	Risk Ratio (M‐H, Fixed, 95% CI)	0.84 [0.67, 1.06]
2.5.2 Within 3 months	2	158	Risk Ratio (M‐H, Fixed, 95% CI)	0.84 [0.73, 0.98]
2.5.3 Up to 6 months	1	63	Risk Ratio (M‐H, Fixed, 95% CI)	0.67 [0.43, 1.04]
2.6 Proportion of participants who discontinued medication and did not develop rebound SEI within 21 days Show forest plot	3	231	Risk Ratio (M‐H, Fixed, 95% CI)	1.04 [0.94, 1.14]

2.6.1 At month 3	1	88	Risk Ratio (M‐H, Fixed, 95% CI)	1.08 [0.95, 1.22]
2.6.2 At month 6 to 7	2	143	Risk Ratio (M‐H, Fixed, 95% CI)	1.01 [0.89, 1.15]
2.7 Percentage of reduction in mean viral titer Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Subtotals only

2.8 Proportion of participants with severe intolerance to treatment Show forest plot	2	141	Risk Ratio (M‐H, Fixed, 95% CI)	4.64 [1.15, 18.71]

2.9 Proportion of participants with elevated IOP (an increase > 6 mmHg or IOP ≥ 22 mmHg) during the study Show forest plot	2		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only

2.9.1 Tacrolimus 0.03% vs steroid	1	80	Risk Ratio (M‐H, Fixed, 95% CI)	0.07 [0.00, 1.13]
2.9.2 Trifluridine 1% vs steroid	1	53	Risk Ratio (M‐H, Fixed, 95% CI)	Not estimable
2.10 Proportion of participants with superinfection during the study Show forest plot	1		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only

2.11 Mean changes in symptom score or SEI score from baseline Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Subtotals only

2.11.1 SEI score	1	80	Mean Difference (IV, Fixed, 95% CI)	‐1.20 [‐1.33, ‐1.07]
2.11.2 Symptom score	1	80	Mean Difference (IV, Fixed, 95% CI)	‐0.20 [‐1.02, 0.62]

Comparison 2. Any treatment versus steroid alone or with levofloxacin

Comparison 3. Dexamethasone versus artificial tears

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Mean number of days from treatment initiation to resolution of symptoms or signs Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Subtotals only

3.2 Proportion of participants with resolution of symptoms within 7 days of treatment Show forest plot	1		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only

3.3 Proportion of participants in whose eyes SEI had developed within 21 days of treatment Show forest plot	1		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only

3.4 Percentage of mean viral titer reduction Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Subtotals only

Comparison 3. Dexamethasone versus artificial tears