Background: Filamentous fungal infections of the cornea, filamentous fungal keratitis (FK), are challenging to treat. Current topical antifungals are not always effective and are often unavailable. Topical natamycin 5% is usually first-line treatment, however, even when treated intensively, infections may progress to corneal perforation; alternative antifungal medications are needed.Previous pilot studies suggest that topical chlorhexidine 0.2% compares favourably with topical natamycin. Full-scale randomised controlled trials (RCTs) of topical chlorhexidine 0.2% are warranted to answer this question definitively. In addition, outcomes for patients with FK are poor as they often present late, clinical diagnosis is challenging, and investigations are limited. Developing the evidence-base to guide practice and ultimately improve outcomes is therefore required. Methods: All consenting patients with microbial keratitis (MK) attending a tertiary ophthalmic referral hospital in Eastern Nepal were assessed for the presence of FK by in vivo confocal microscopy (IVCM) and/or smear microscopy. Demographic, clinical, journey, and microbiological data were collected from all these patients. These data were analysed in a cohort study to investigate reasons associated with delayed presentation as well as two nested case-controlled studies, investigating indicative clinical features and evaluating investigations. Patients with confirmed FK were enrolled in the RCT and randomly allocated to receive topical chlorhexidine 0.2% or topical natamycin 5%. Primary analysis (intention-to-treat) was by linear regression, using baseline logarithm of the minimum angle of resolution (logMAR) best spectacle-corrected visual acuity (BSCVA) and treatment arm as prespecified covariates. The primary outcome measure was BSCVA at 3 months. Secondary outcome measures included perforation or therapeutic penetrating keratoplasty by 90 days. Results: Between 3 June 2019 and 9 November 2020 we enrolled 643 patients with MK. Of these, 354 were eligible for the RCT and randomly assigned: 178 to chlorhexidine and 176 to natamycin. Excluding mixed infections, primary outcome data were available for 141 and 143 of the chlorhexidine and natamycin groups, respectively. We did not find evidence to suggest chlorhexidine was noninferior to natamycin and in fact found strong evidence to suggest that natamycin-treated participants had significantly better 3-month BSCVA than chlorhexidine-treated participants, after adjusting for baseline BSCVA (regression coefficient, −0.30; 95% confidence interval [CI], −0.42 to −0.18; P < 0.001). There was no difference in re-culture positivity between arms at day 7. The majority of chlorhexidine-treated patients healed (151/175, 86.3%), although this was less than natamycin-treated cases (163/173, 94.2%; P = 0.018). Furthermore, there were more perforations and emergency corneal grafts in the chlorhexidine arm (24/175, 13.7%) than in the natamycin arm (10/173, 5.8%; P = 0.018, mixed infections included). A fungal cause was identified in 482/642 (75.1%) of cases, which increased to 532/642 (82.9%) when including mixed infections. Unusually, dematiaceous fungi accounted for half of the culture-positive cases (50.6%). Serrated infiltrate margins, patent nasolacrimal duct, raised corneal slough, and organic trauma were independently associated with fungal keratitis (p < 0.01). Smear microscopy had the highest sensitivity (90.7% [87.9-93.1%]), followed by IVCM (89.8% [86.9-92.3%]) and culture (75.7% [71.8-79.3%]). Of the three smear microscopy stains, KOH had the highest sensitivity (85.3% [81.9-88.4%]), followed by Gram stain (83.2% [79.7-86.4%]) and calcofluor white (79.1% [75.4-82.5%]). In the cohort study, the majority of patients (96%) self-referred. Over half (328/643) of all cases presented after at least seven days. The total cost of care increased with increasing numbers of facilities visited (P<0.001). Those living furthest away were least likely to present directly (P<0.001). Factors independently associated with delayed presentation included distance >50km from the eye hospital (aOR 5.760 [95% CI 1.829-18.14, p=0.003]), previous antifungal use (aOR 4.706 [95% CI 3.139-5.360]), and two or more previous journeys (aOR 1.442 [95% CI 1.111-3.255]). Conclusion: Treatment with natamycin is associated with significantly better visual acuity, with fewer perforations, compared to treatment with chlorhexidine. However, the proportion of healed chlorhexidine-treated cases is comparable to that of voriconazole reported in earlier trials. Natamycin remains the preferred first-line monotherapy treatment for filamentous fungal keratitis. Chlorhexidine 0.2% may be considered in situations where natamycin is unavailable. Smear microscopy and IVCM were the most sensitive tools for identifying FK in our cohort, whilst certain clinical signs can help direct the clinician to find a presumptive infectious cause, allowing appropriate treatment to be started without delay. Distance to the eye hospital is a significant barrier to prompt, direct presentation.