Semiology: reading a corneal ulcer

Location

• Central or paracentral ulcer → first consider a trophic cause, an iatrogenic cause, an infection, or inflammation.
• Peripheral ulcer → first consider an inflammatory cause (peripheral ulcerative keratitis), more rarely infectious.

This dichotomy relates to vascularization and innervation: the periphery is close to the limbal vessels and immune cells; the center is the most densely innervated territory and the most dependent on the tear film.

Shape of the floor and edges

• Gently sloping, harmonious ulcer, with smooth rounded edges and an edema halo → neurotrophic profile. Tissue loss is progressive and regular.
• Ulcer that is deep from the outset, punched-out, excavated → keratolysis: iatrogenic causes (aminoglycosides, fluoroquinolones, NSAIDs), immunologic cause, lytic infection. The stroma is digested faster than it is denuded.

The sign to actively look for: edge hyperplasia

A chronic neurotrophic ulcer typically shows hyperplastic epithelial edges — raised, gray, detached, overhanging a bare floor — to be distinguished from the normal migrating front, which is thin and translucent.

The mechanism is instructive: in a setting of inflammation, iatrogenic injury or neurotrophic keratitis, the epithelium undergoes premature stratification — the balance shifts toward proliferation/stratification > migration, driven by IL-1, MMP-9 and TGF-β. Cells stack up at the edge instead of advancing. The permissive role of MMP-9 is well established (Mohan et al., JBC 2001: MMP-9-deficient mice develop epithelial hyperplasia; Gabison et al., Ophthalmology 2005).

Two other useful semiological clues: fluorescein seeping under the edges points toward active herpetic keratitis; whorl (vortex) keratopathy signals disorganized epithelial hyperplasia, frequent after keratoplasty for herpes under corticosteroids and aminoglycosides.

Etiologies: look for one, and often find several

Any lesion of V1, from the trigeminal nucleus to the corneal nerve endings, can produce NK. In the series (354 eyes), the cause was multifactorial in 34.2% of cases and idiopathic in 8.2% — finding one cause therefore never exempts from looking for a second.

Ocular causes

• Herpetic disease — 114 eyes, 32.2% (HSV 91, i.e. 25.7%; VZV 23, i.e. 6.5%): the leading etiology, combining direct nerve destruction and recurrent inflammation. Classic pitfall: "herpetic keratitis that fails to heal despite well-conducted antiviral treatment" has very often become a post-herpetic NK, perpetuated by the topical antiviral itself — hence the value of a therapeutic window, spectacular when indicated (absence of stromal inflammation, marked hypoesthesia).
• Iatrogenic causes — 113 eyes, 31.9%, broken down into three families:
  • Surgical: vitrectomy with endolaser 36 eyes (10.2%), multiple ocular surgeries 18 (5.1%), penetrating or lamellar keratoplasty 17 (4.8%) — which severs the stromal nerves —, sequelae of radiotherapy 16 (4.5%), refractive surgery 6 (1.7%), panretinal photocoagulation 4 (1.1%). After retinal detachment surgery, contributing factors add up: limbal peritomy, scleral buckle, intraoperative deepithelialization, postoperative iatrogenic injury, associated diabetes. After LASIK, one must distinguish dryness extending beyond the flap from SPK strictly confined to the flap, which is neurotrophic.
  • Local drug-induced — 39 eyes (11.0%): anti-glaucoma drugs 14 (4%), antibiotics 14 (4%), NSAIDs 9 (2.5%), topical anesthetics 2 (0.6%); with preservatives, topical antivirals and prolonged fortified eye drops adding to the burden.
  • Systemic drug-induced: neurotoxic chemotherapy 11 eyes (3.1%); also worth knowing, EGFR inhibitors (erlotinib, panitumumab), responsible for multiple epithelial defects, melting and perforations (Saint-Jean et al., Ophthalmology 2012).
• Chronic ocular surface disease — 62 eyes, 17.5%: severe blepharitis and rosacea 38 (10.7%), entropion 9 (2.5%), trauma and chemical burns 9 (2.5%), GVHD 4 (1.1%), Sjögren's syndrome 3 (0.8%), trachoma 3 (0.8%), cicatricial pemphigoid 2 (0.6%).

Systemic causes

• Central nervous system disease — 98 eyes, 27.7%: intracerebral tumors and/or their surgeries 62 (17.5%), stroke 13 (3.7%), hemorrhage or arteriovenous malformation 9 (2.5%), V1 injury during brain surgery 6 (1.7%), trigeminal thermocoagulation 3 (0.8%), carotid-cavernous fistula 3 (0.8%), multiple sclerosis 2, leukodystrophy 2, brain infection 2, degenerative or infiltrative disease 2, neurofibromatoses, tuberous sclerosis, MEN IIB, Guillain-Barré syndrome. This high proportion partly reflects the referral pattern of a center that is also a reference site for neurology and neuroradiology.
• Diabetes — 37 eyes, 10.5% (including 25, i.e. 7.1%, without vitrectomy or panretinal photocoagulation): diabetic corneal neuropathy, largely under-diagnosed.
• Rare — 9 eyes (2.5%): severe alcoholism or B12 deficiency (3), Acanthamoeba sequelae (2), leprosy (2), vitamin A deficiency (1), Goldenhar syndrome (1).

Bilateral forms (19 patients, 5.7%): multifactorial 7, idiopathic 3; dominant etiologies iatrogenic (36.8%), chronic surface disease (26.3%), herpes (21.1%), central origin or diabetes (15.8% each).

(Para)central aseptic keratolysis without hypoesthesia should prompt a search for severe dry eye disease: Sjögren's syndrome, GVHD, ongoing chemotherapy.

Mackie classification, epidemiology and prognosis

The three stages

• Stage I — superficial epithelial alterations: superficial punctate keratitis, dull and irregular epithelium, epithelial edema, sometimes hyperplasia and superficial neovascularization.
• Stage II — epithelial erosion without stromal thinning: oval or circular defect, recurrent or persistent, smooth edges often hyperplastic, stromal edema halo.
• Stage III — ulceration with stromal thinning: melting, descemetocele, risk of perforation.

(Mackie IA. Neuroparalytic keratitis. WB Saunders, 1995.)

The reference series

Single-center retrospective study, Fondation Adolphe de Rothschild, November 2009 – October 2017: 305,351 consultations screened, 392 records identified, 335 patients / 354 eyes included (57 patients excluded, 14.5%).

The observed frequency is roughly seven times higher than the classic estimate of 1.6/10,000, which had only been extrapolated from a few associated conditions (6% of herpetic keratitis, 12.8% of ophthalmic zoster, 2.8% of trigeminal neuralgia surgeries).

Visual outcomes and prognostic factors

Mean CDVA improved significantly from 0.61 ± 0.49 to 0.38 ± 0.49 LogMAR, a mean gain of 0.19 ± 0.34 — almost two ETDRS lines.

Reading this table: the greatest visual gain concerns the advanced stages — because they start from a lower baseline — but their final acuity remains significantly worse. Treating a stage III aggressively recovers a lot; treating a stage I early avoids having to do so.

Overall, CDVA improved in 104 eyes (45%), remained stable in 75 (33%) and worsened in 50 (22%). Eyes at "counting fingers or worse" went from 121 (34.2%) to 97 (27.4%).

Three factors are significantly correlated with final CDVA (Spearman): initial acuity (r = 0.66; p < 0.001), initial Mackie stage (r = 0.24; p = 0.002), age (r = 0.24; p = 0.003).

Not correlated: etiology (herpes, VZV, iatrogenic, central causes, surface disease, diabetes, unknown cause), lagophthalmos, and — notably — diagnostic delay (neither in days nor in number of visits). This last absence of correlation does not mean that delay is unimportant: the authors explain it by the rapid initiation of appropriate treatment despite missed initial diagnoses (most often stage I), by short times to presentation and close monitoring, and by the early discontinuation of wrongly prescribed antibiotics.

The stage I paradox

A counter-intuitive and pedagogically valuable finding:

Overall figures: healing rate 79.5%, mean delay 44.8 days; 87.9% healing at 5 weeks for stages II-III — to be compared with the 13.7% at 4 weeks and 33.3% at 8 weeks of the untreated stages II-III of Bonini et al.'s vehicle arm. Mean follow-up 21.4 months (CI 13.7-29.1), 15.0 consultations per patient on average, 73 patients hospitalized (21.8%) for a mean duration of 11.5 days (median 7). Recurrences: 70 eyes, on average 1.5 recurrence per eye.