British Journal of Ophthalmology 1996;80:962-968

Non-tuberculous mycobacterial keratitis: a studyof 22 cases

Samuel C M Huang, H Kaz Soong, Jen-Shiang Chang, Yu-Sung Liang

AbstractAim-To investigate causes and clinicalfindings of non-tuberculous mycobacte-rial keratitis, and to study its response totopical antibiotic therapy and surgicalextirpative keratectomy.Method-A single centre, retrospectivereview of22 patients with non-tuberculousmycobacterial keratitis seen in a 3 yearperiod. Laboratory diagnoses were estab-lished with Ziehl-Nielsen acid fast stain-ing and Lowenstein-Jensen cultures.Results-In 20 patients (91%), there wasan antecedent history of foreign body eyetrauma (18 patients) or elective surgery(two patients). There were 19 cases ofMycobacterium chelonei, and three ofMfortuitum. Clinical signs included epithe-lial defects, satellite or ring stromalinfiltrates, crystalline keratopathy, andhypopyon. For topical antibiotic therapy,20 patients received amikacin, while onepatient received rifampin and anotherreceived ciprofloxacin, each in accord-ance with the results of the in vitro drugsensitivities. An exdrpative keratectomywas performed in 15 cases; four of thesecases additionally required a temporaryconjunctival flap in order to finally eradi-cate the infection. At the end of the followup period (median 18 months; range 3months to 3 years) all eyes were stable andfree of infection, with 19 (86%) havingfinal visual acuities of 201200 or better.Conclusion-Early clinical recognitionand prompt laboratory diagnosis, to-gether with aggressive topical antibiotictherapy and early keratectomy, mayshorten morbidity and improve the clini-cal outcome ofnon-tuberculous mycobac-terial keratitis.(BrJ Ophthalmol 1996;80:962-968)

Non-tuberculous or atypical mycobacteria arewidespread in the environment' and were oncethought to be non-pathogenic. Non-tuberculous mycobacteria are now known tocause a variety of human diseases, includinginfections of the cornea. In 1965, Turner andStinson' described the first case of non-tuberculous mycobacterial keratitis in a cor-neal ulcer that developed 4 months afterremoval of a superficial foreign body; Mycobac-terium fortuitum was identified as the aetiologi-cal agent of the ulcer. The first case of cornealinfection withM chelonei was reported by Gan-gadharam and coworkers' in 1978. Severalsubsequent reports of M fortuitum and M

chelonei keratitis have been recorded in theliterature.'4 Corneal infections with Mflaves-cens,M avium-intracellulare, M gordonae, andMmarinum have also been reported, but appearto be quite infrequent in comparison with Mfortuitum and M chelonei.3 1" Most reportedcorneal infections by non-tuberculous myco-bacteria are associated with trauma or surgery.

Topical fortified amikacin (14-100 mg/ml) isthe drug of choice in the treatment of mostnon-tuberculous mycobacterial cornealinfections.4 1`1 The non-tuberculous mycobac-teria also show varying degrees of susceptibilityto other drugs, including the fluoroquinolones,aminoglycosides, tetracycline family, erythro-mycin, cephalosporins, sulphonamides, neo-mycin, clarithromycin, and vancomycin.9 182"'We report herein 22 culture proved cases of

infectious keratitis caused by non-tuberculousmycobacteria. The purpose of this study was toinvestigate the causes and clinical findings innon-tuberculous mycobacterial corneal infec-tion, and to study its response to topicaltherapy and surgical keratectomy.

Materials and methodsBetween January 1991 and June 1994, weinvestigated 22 consecutive cases of non-tuberculous mycobacterial keratitis. The studyincluded reviews of the patient charts at theChang Gung Memorial Hospital and at thereferring ophthalmologists' offices, interviewswith the patients and the referring ophthal-mologists, microbiological cultures of theinfected corneas, drug sensitivity tests for com-monly used antibiotics and for antituberculousagents, and pathological examination of extir-pative lamellar keratectomy specimens.

DIAGNOSTIC CRITERIAPositive non-tuberculous mycobacterial cul-tures on L6wenstein-Jensen agar medium con-stituted a definitive diagnosis of non-tuberculous mycobacterial keratitis. Althoughthese organisms may often grow on standardmedia, Lowenstein-Jensen medium providesmuch more optimal and specific growth condi-tions for mycobacteria. Whether or not thesmears for acid fast organisms are positive, thematerial should be cultured for several reasons:(1) cultures can detect lower quantities of theorganism, and (2) tests for drug sensitivitieswill be useful in directing medical treatment.All 22 patients had positive cultures fornon-tuberculous mycobacteria. Despite posi-tive L6wenstein-Jensen cultures, only 10 ofthese specimens showed acid fast bacteria withthe Ziehl-Nielsen stain. An antecedent history

Chang Gung MemorialHospital, Chang GungMedical and TechnicalCollege, Taipei, TaiwanS C M HuangJ-S ChangY-S Liang

WK Kellogg EyeCenter, University ofMichigan MedicalSchool, Ann Arbor,Michigan, USAH K Soong

Correspondence to:Samuel C M Huang, MD,Department ofOphthalmology, ChangGung Memorial Hospital,199, Tung Hwa North Road,Taipei, Taiwan 10591.

Accepted for publication23 August 1996

962

on Decem

ber 27, 2019 by guest. Protected by copyright.

http://bjo.bmj.com

/B

r J Ophthalm

ol: first published as 10.1136/bjo.80.11.962 on 1 Novem

ber 1996. Dow

nloaded from

Non-tuberculous mycobacterial keratitis: a study of22 cases

oftrauma, especially superficial corneal foreignbody injuries, greatly increased the index ofsuspicion of infection with this micro-organism. Negative routine cultures for bacte-ria, herpes simplex virus, fungus, and Acan-thamoeba, together with lack of clinicalresponse to routine antibiotic (and occasion-ally corticosteroid) therapy, usually helpedconfirm our suspicions. Routine culture meth-ods included the use of the following media:blood agar, chocolate agar, thioglycollatebroth, and Sabouraud dextrose agar. Whenevernecessary, herpes simplex cultures were trans-ported in Hanks's balanced salt medium andinoculated on rabbit kidney cells, and Acan-thamoeba was cultured on non-nutritive agarwith Escherichia coli confluent overlay. Directsmears for Gram and Ziehl-Nielsen stainswere prepared in each case. In two cases, a'cracked windshield' corneal infiltrate patternwas virtually diagnostic in itself. As weencountered more cases of non-tuberculousmycobacterial corneal infections in our prac-tice, not only did our diagnostic acumenbecome honed, but we also began to realisethat this organism was not really as rare as wehad once assumed.

LABORATORY DIAGNOSTIC PROCEDURECorneal specimens for culture were obtainedby superficial scraping of stromal infiltrateswith a Kimura spatula. These were inoculatedonto Lowenstein-Jensen medium and weremaintained for over 4 weeks in some cases.

In the cases which required extirpativecorneal surgery, part of the keratectomy speci-men was placed in formalin for histopathologi-cal examination with haematoxylin and eosinand Ziehl-Nielsen acid fast stains and part wassent to the microbiological laboratory forL6wenstein-Jensen cultures and for antibioticsensitivity assays. The broth microdilutionmethod was used for the antibiotic susceptibil-ity testing.

TREATMENTNon-tuberculous mycobacteria are generallyresistant to conventional antibiotics, but theydo respond favourably to amikacin. Initial con-ventional antibiotics used routinely by us con-sisted of fortified topical gentamicin andcefazolin. Varying degrees of in vitro suscepti-bility have been reported to a broad range ofother agents including fluoroquinolones,aminoglycosides, tetracycline derivatives,erythromycin, vancomycin, cephalosporins,and sulphonamides.Once the diagnosis was established by

culture, our mainstay of treatment was topicalamikacin 50 mg/ml hourly around the clock(20 patients). In one patient (case 5), onlytopical rifampin (30 mg/ml) was used, as drugsensitivity tests indicated a susceptibility torifampin only. In another patient (case 11),only topical ciprofloxacin was used (once againin accordance with the drug sensitivity assay).Once the diagnosis of non-tuberculous myco-bacterial keratitis was established by laboratorymethods, antimicrobial therapy was alteredappropriately.

Extirpative keratectomy was required insevere or recalcitrant cases in order to reduceor eliminate the numbers of infectious organ-isms in the cornea. This also ostensiblyenhanced the antimicrobial drug penetrationand improved patient comfort. Furthermore,the surgical excision provided an excellentbiopsy source. Keratectomies were performedunder retrobulbar anaesthesia with the aid ofan operating microscope. A slightly oversized,round zone marker was used to delineate theinfected area to be excised. A No 64 or No 69Beaver blade was used to make the initial verti-cal incision up to about two thirds of thecorneal depth. Either a No 57 or No 66 Beaverblade was then used for the intralamellardissection of the infected and/or necroticstroma.

If the infection failed to respond to extirpa-tive surgery and intensive antimicrobialtherapy, a Gundersen conjunctival flap wasperformed (four patients). In all cases, the con-junctival flaps were successful in controllingthe infection and stabilising the eye. Theconjunctival flaps were taken down after 6months, leaving corneas with varying degreesof scarring. One case eventually underwent asuccessful penetrating keratoplasty followingtakedown of the conjunctival flap (case 17).

ResultsDEMOGRAPHICS AND CLINICAL HISTORYThe patients ranged in age from 20 to 66 years(median 38.5 years). The median follow upperiod was 18 months (range 3 months to 3.5years). The majority (82%) of the patients weremale. In 20 of the 22 patients (91%), there wasan antecedent traumatic or surgical history.This subgroup of 20 patients was more specifi-cally divided into 18 who had trauma and twowho had surgery. All 18 trauma cases involvedsuperficial corneal foreign body injuries, themajority with metallic agents (61%) and theremainder with mineral or vegetable matter.The two postsurgical corneal infections oc-curred within 1 year of elective eye surgery-one was associated with pterygium excision(bare sclera technique) (case 17) and the otherwas associated with a failed corneal graft(originally done for contact lens relatedPseudomonas keratitis) (case 4).

Eighteen of 22 patients (82%) were male,paralleling the predominance of traumaticaetiologies in most cases. Twelve cases wereassociated with foreign body injuries incurredat work, thus illustrating the importance of thisproblem among the blue collar workforce andthe need for enforcing adequate eye protectionin the workplace. In the non-traumatic cases,one had chronic ocular surface disease associ-ated with cicatricial pemphigoid (case 2) andthe other had neurotrophic keratopathy afterexcision of a cerebropontine angle tumour(case 5).

Typically, the clinical symptoms and findingsof corneal infection did not appear immedi-ately after corneal injury. The time intervalbetween onset oftrauma and appearance of thecorneal infection ranged from 3 days to 3weeks. Time of diagnosis from the onset of

963

on Decem

ber 27, 2019 by guest. Protected by copyright.

http://bjo.bmj.com

/B

r J Ophthalm

ol: first published as 10.1136/bjo.80.11.962 on 1 Novem

ber 1996. Dow

nloaded from

Huang, Soong, Chang, Liang

keratitis ranged from 1 week to 2 months.Generally, as we gained more experience in thediagnosis and management ofnon-tuberculousmycobacterial keratitis, this latter time intervalbecame shorter.

MICROBIOLOGYNineteen of 22 cases were culture positive forM chelonei, while three were culture positive forMfortuitum. Typical histopathological findingsare shown in Figure 1.

In vitro minimal inhibitory concentrations(MIC) for amikacin in the 20 susceptible casesranged from 9 to 15 gg/ml. In cases 5 and 1 1,the MIC indicated resistance to amikacin, butgood sensitivity to either rifampin (case 5) orciprofloxacin (case 11). Antibiotic resistancewas indicated by the following criteria: (1)amikacin > 64 jg/ml, (2) rifampin 1 ,ug/ml, and(3) ciprofloxacin 0.8 ,ug/ml.

In total, keratectomies were performed in 15out of 22 patients (68%). This included sevenpatients in whom the organisms showed goodsensitivity in vitro to the antibiotics employed.In these patients, the infection was so severeand deep that drug therapy alone was not suf-ficient to eradicate the disease.

In some cases, the aetiology of keratitis wasinitially mistaken for fungus, herpes simplex, orAcanthamoeba, as the clinical findings wereoften masquerades for these agents. Poor clini-cal responses to therapy, together with negativefungal, viral, and parasitic laboratory studies,were helpful (albeit belatedly) in establishingthe final correct diagnosis. No cases of mixedinfections involving non-tuberculous mycobac-teria and other organisms were encountered.

Surprisingly, only 11 (50%) of the scrapingsshowed positive Ziehl-Nielsen acid fast stain-ing. The Lowenstein-Jensen cultures, on the

other hand, were positive for all 22 patients.This may be due to: (1) low numbers of organ-isms not adequately detected by the smear, butdetected after 'amplification' provided by culti-vation; (2) tendency by the clinician toprioritise the specimen for culture, rather thanfor simple smearing, when the availablecorneal inoculum is sparse; and/or (3) acidwash being too potent in the Ziehl-Nielsenmethod, leading to overdecolorisation of theorganisms (see Discussion for alternative stain-ing method, the Fite-Ferraco stain). Thisexperience re-emphasised the importance ofobtaining culture identification, rather thandepending solely on the acid fastness criterion.Definitive mycobacteria were not seen onretrospective examination of the Gram stainspecimens.

CLINICAL FINDINGSIn all cases, except for the one with underlyingneurotrophic keratopathy, the patients com-plained of moderate pain and photophobia.Excruciating pain, characteristic of Acan-thamoeba keratitis, was not encountered. Vary-ing degrees of blurring were noted by thepatients.

Clinical signs are listed in Table 1. Allpatients showed persistent corneal epithelialdefects and underlying inflammatory stromalinfiltrates at various depths. Eight patientsmanifested a main stromal infiltrate sur-rounded either by discrete satellite infiltrates orfluffy extensions of the parent infiltrate. Sixpatients showed either a complete or partialring infiltrate. Three patients manifested crys-talline keratopathy, including two who had acracked windshield stromal lesion (Fig 2), con-sidered by some to be diagnostic of non-tuberculous mycobacterial keratitis.""1723 Vary-

Figure I Histopathological appearance ofa colony ofM chelonei in an anterior lamellar keratectomy specimen, showingacidfast mycobacteria. (Ziehl-Nielsen stain; original magnification x 1584).

964

40AW

06-Ar

ow

on Decem

ber 27, 2019 by guest. Protected by copyright.

http://bjo.bmj.com

/B

r J Ophthalm

ol: first published as 10.1136/bjo.80.11.962 on 1 Novem

ber 1996. Dow

nloaded from

Non-tuberculous mycobacterial keratitis: a study of22 cases

Figure 2 Slit-lamp biomicroscopic appearance of cracked windshield corneal infiltratepattern (case 20).

ing degrees of anterior chamber inflammatorreactions were seen in eight patients, of whic]five had hypopyon. No cases of infectiouendophthalmitis or scleritis were seen.

CLINICAL COURSEThe final best corrected visual acuity afteeradication of the infection and after all thassociated surgeries ranged from hand move

ments to 20/20 (Table 2). In 19 patient(86%), the best corrected visual acuity wa

20/200 or better. The remaining three patientwith poorer visual acuities have not undergonvision restoring penetrating keratoplasties fotwo reasons: (1) their contralateral eyes weri

good enough to allow adequate performance odaily visual tasks, and (2) good donor corneatissue is relatively difficult to obtain in Taiwan

REPORTS OF CASESCase 5A 47-year-old woman underwent craniotom'on 8 October 1993 for excision of a benigicerebellopontine angle tumour. Subsequentl3she developed right hemifacial palsy wit]lagophthalmos and severe right neurotrophikeratopathy. Two weeks later, she was sent tus for evaluation and treatment of a corneaulcer in her right eye. The visual acuities wer

hand movement in the right eye and 20/20 ilthe left. Significant lagophthalmos and corneaexposure were noted in the right eye. The con

junctiva in the right eye was very injected. Slitlamp examination of the right cornea showed

large epithelial defect centrally with a large,underlying disciform, intrastromal white cellinfiltrate. A 10% layered hypopyon was presentin the anterior chamber. The lens appearedclear, but the fundus could not be adequatelyvisualised because of the corneal opacity.Intraocular pressures by applanation tono-metry were normal. The corneal sensation wasabsent in the right eye.

Bacterial and fungal studies were obtainedon corneal scrapings from the right eye, andthe patient was started on hourly doses of topi-cal cefazolin 50 mg/ml. Routine Gram stainsand cultures were negative, and Ziehl-Nielsenstaining was negative for acid fast organisms.The corneal ulcer responded poorly to therapy.The L6wenstein-Jensen culture grew M che-lonei 3 weeks after inoculation. The organismwas sensitive in vitro only to rifampin. Thecefazolin eyedrops were discontinued and thepatient was started on hourly doses of topicalrifampin 30 mg/ml. The corneal infiltratesresponded gradually to this regimen andmedication was tapered over a 1 month period.In the setting of neurotrophic and exposurekeratopathy, the epithelial defect persisted andthe hypopyon never completely disappeared.One month after cessation of the eyedrops, anew stromal infiltrate appeared in the centralcornea in the right eye. Reinstitution of hourlytopical rifampin resulted in complete resolu-tion of the infection, leaving a dense centralcorneal stromal scar. Six months later, thepatient underwent penetrating keratoplasty inthe right eye, but owing to a persistent epithe-lial defect, the graft eventually failed.

'yh Case 7Ls A 62-year-old mechanic sustained a superficial

metallic foreign body injury to his left eye whileworking. The foreign body was removed by thefactory medical staff; however, the patient con-

r tinued to feel intermittent foreign body sensa-e tion in this eye for about 2 weeks. When the

pain persisted, he was referred to our clinic fors further evaluation. In the right eye, the visualIs acuity was 20/400 and the conjunctiva wass severely inflamed. A 3 x 3 mm peripheral cor-e neal epithelial defect, with an underlying, ring-or shaped, anterior stromal infiltrate, was presente near the limbus in the 1 o'clock meridian. Thef ring infiltrate was surrounded by several smallil satellite infiltrates. A provisional diagnosis ofi. herpetic keratitis was made and the patient was

started on acyclovir ophthalmic ointment four

Table 1 Clinical signs of non-tuberculous mycobacterialY keratitis

.c

0

II-e

n

a

No ofClinical sign cases

Corneal epithelial defect 22Corneal stromal infiltrates 22Disciform infiltrates 8Main infiltrate surrounded by satellites 8Partial or full ring infiltrates 6Crystalline keratopathy (including crackedwindshield infiltration) 3

Iritis 8Hypopyon 5Infectious endophthalmitis 0Scleritis 0

965

on Decem

ber 27, 2019 by guest. Protected by copyright.

http://bjo.bmj.com

/B

r J Ophthalm

ol: first published as 10.1136/bjo.80.11.962 on 1 Novem

ber 1996. Dow

nloaded from

Huang, Soong, Chang, Liang

Table 2 Case summaries of non-tuberculous mycobacterial keratitis

No Agelsex Nature ofcorneal insult Organism Topical treatment Surgery Final VA Acidfastness

1 39/M Metallic FB M chelonei Amik EK 20/25 +2 61/M Pemphigoid M chelonei Amik EK/F CF 60 cm3 47/M Metallic FB M chelonei Amik EK 20/50 +4 62/M Failed graft M chelonei Amik None HM 60 cm5 47/F Neurotrophic/exposure M chelonei Rifampin None HM 60 cm6 20/M Unknown FB M chelonei Amik EK 20/40 +7 62/M Metallic FB Mfortuitum Amik EK 20/100 -

8 49/F Unknown FB M chelonei Amik None 20/40 -

9 38/M Unknown FB M chelonei Amik EK/F 20/200 -

10 31/M Metallic FB M chelonei Amik EK 20/20 -

11 36/M Clay FB Mfortuitum Cipro EK 20/40 +12 25/M Metallic FB M chelonei Amik EK twice 20/30 +13 33/M Unknown FB Mfortuitum Amik None 20/200 +14 45/M Metallic FB M chelonei Amik None 20/5015 66/M Wood FB M chelonei Amik EK/F 20/200 +16 30/M Metallic FB M chelonei Amik None 20/2017 55/F Pterygium excision M chelonei Amik EK/F 20/60 +18 30/M Metallic FB M chelonei Amik EK 20/25 +19 55/M Metallic FB M chelonei Amik EK twice 20/100 +20 34/F Metallic FB M chelonei Amik EK 20/4021 26/M Clay FB M chelonei Amik None 20/25 +22 35/M Metallic FB M chelonei Amik EK 20/30

FB = foreign body; EK = extirpative keratectomy; F = conjunctival flap; VA = visual acuity; Amik = amikacin; Cipro = ciprofloxacin.*By direct microscopy, Ziehl-Nielsen stain.

times daily and cefazolin 50 mg/ml eyedropsevery hour around the clock. The keratitis con-tinued to worsen on this regimen. Viral, bacte-rial, and fungal cultures were negative; Ziehl-Nielsen stained specimens were also negative.The L6wenstein-Jensen medium grew Mfortuitum after 7 days. The patient was startedon topical amikacin 50 mg/ml hourly aroundthe clock, as an addition to the cefazolineyedrops already in use. Ten days later, thepatient underwent anterior lamellar keratec-tomy in the right eye because of what weperceived to be a poor clinical response (that is,continued stromal inflammation, infiltration,and poor wound healing) of the infection to themedical therapy. Amikacin and cefazolin werecontinued after the keratectomy. He respondedwell to the extirpative therapy and his corneaepithelialised completely within 2 weeks, leav-ing only a faint anterior scar. Histopathologicalexamination and Lowenstein-Jensen culturesof the surgical specimen did not show anyresidual organisms within the stromal lamellae.The final visual acuity at the latest examinationa year later was 20/100 in the right eye.

Case 20While riding her motorcycle, a 34-year-oldwoman sustained a small, superficial, metallicforeign body injury to her left cornea. This wasremoved that same day by a general medicaldoctor. Five days later, the patient developedpain, photophobia, and tearing in the injuredeye. Her doctor noted a corneal ulcer andreferred the woman to us for evaluation andtreatment.On examination, her visual acuity was 20/20

in the right eye and 20/200 in the left. Moder-ate conjunctival injection of the left eye wasnoted and slit-lamp biomicroscopy showed a 3x 3 mm central corneal epithelial defect with a'cracked windshield' anterior stromal infiltrate(Fig 2).We suspected non-tuberculous mycobacte-

rial keratitis and immediately started thepatient on hourly doses of topical amikacin 50

mg/ml. Our suspicion was confirmed 1 weeklater with the growth of M chelonei in theL6wenstein-Jensen medium. The organismwas sensitive in vitro to amikacin. Resolution ofthe infiltrate appeared to come to a standstillafter initial improvement the first 5 days. Avery superficial lamellar keratectomy (essen-tially subepithelial) was performed 4 days laterin order to debulk the infiltrate and to improvetopical drug penetration. The cornea re-epithelialised within 2 weeks and the patientwas left with good uncorrected visual acuity of20/40 in the left eye. A faint, mid-stromal scarremained in the cornea.

DiscussionNon-tuberculous mycobacteria are aerobic,non-spore forming, non-motile bacilli. Owingto the many differences in characteristics fromthe 'typical' M tuberculosis, non-tuberculousmycobacteria have been called atypical, anony-mous, or unclassified.24 In the past, most non-tuberculous mycobacteria were thought to beenvironmental saprophytes, widely distributedin soil and water, and incapable of producingclinical infections.

In 1959, Runyon' proposed a classificationofnon-tuberculous mycobacteria, under whichgroup IV is the Mfortuitum complex, consist-ing of M fortuitum, M chelonei, M smegmatis,and M vaccae.'4 Non-tuberculous mycobacte-ria have been isolated from the normal flora ofhuman sputum, gastric contents, and ocularsurfaces.'8 Both Mfortuitum and M chelonei areresponsible for a growing number of skin andsoft tissue diseases. The rise in the number ofreported cases is probably due to: (1) betterunderstanding of the clinical features andenhanced awareness of diseases caused by thisorganism, (2) better diagnostic methods andcriteria, and (3) increased iatrogenic or noso-comial infections associated with surgery,renal dialysis, and immunosuppression. Surgi-cal debridement is now widely advocated forthe treatment of postoperative infections andother outbreaks of nosocomial infections.25 Inthe eye, non-tuberculous mycobacterial

966

on Decem

ber 27, 2019 by guest. Protected by copyright.

http://bjo.bmj.com

/B

r J Ophthalm

ol: first published as 10.1136/bjo.80.11.962 on 1 Novem

ber 1996. Dow

nloaded from

Non-tuberculous mcobacterial keratitis: a study of 22 cases

keratitis occurs most often after cornealtrauma, especially those involving foreignbodies2"7 15-17 23 26-28; however, it also occursafter anterior segment surgery, including radialkeratotomy,' penetrating keratoplasty," ' ' in-cisional posterior lens capsulotomy,9 and incorneal suture abscesses.9'1029 Contact lenseshave also been causally associated with non-tuberculous mycobacterial keratitis.31830Several investigators have advocated surgicalextirpation therapy in runaway cornealinfections that respond poorly to medicaltherapy.3 69 12 18 19 27 29 31-33

In our present experience with non-tuberculous mycobacterial corneal infections,we have learned about the important role ofextirpative lamellar keratectomy in the treat-ment of keratitis recalcitrant to antimicrobialmedical therapy alone. This procedure debulksthe microbial inoculum, facilitates drug pen-etration, removes optical opacities, providestissue for cultures and histopathologicalexamination, and allows the eye to stabilise andbecome quiescent before it undergoes a moreextensive operation such as corneal transplan-tation. In our experience, it sometimes mayeven obviate the need for lamellar or penetrat-ing keratoplasty altogether. Our findings paral-lel those of recent reports.3233

If anterior keratectomy fails, a conjunctivalflap appears to be very effective in stopping theinfection. The flap could be taken down forrestoration ofvision in the future, either with orwithout the use of corneal transplantation. Webelieve the overlying conjunctiva allows a moredirect contiguity of the intrastromal mycobac-teria to the cell mediated immune mecha-nisms.We were initially surprised by the high

incidence (50%) of negative acid fast stainingwith the Ziehl-Nielsen stain. At the Universityof Michigan, we have begun using the Fite-Ferraco stain in delineating non-tuberculousmycobacteria. This method uses a gentler acidwash than the Ziehl-Nielsen technique, thusallowing the bacilli to better retain their stain.34For future cases of non-tuberculous mycobac-terial keratitis, we plan to rely predominantlyon the Fite-Ferraco stain and L6wenstein-Jensen cultures.

Both Mfortuitum andM chelonei are resistantto most antibiotics, including the standardantituberculous drugs (for example, isoniazid,streptomycin, rifampin, and ethambutol). Formore than 20 years, amikacin has been the sin-gle most frequently used topical antimicrobialdrug in the treatment of non-tuberculousmycobacterial keratitis.45 9101418-21 Dalovisioand associates4 reported highly successfultreatment of infections with amikacin, espe-cially when used in conjunction with a secondantibiotic, such as doxycycline. Thus, it may beadvantageous to treat the keratitis with topicalamikacin and add a second agent, such as kan-amycin, tetracycline or its relatives, fluoroqui-nolones, erythromycin, vancomycin, cepha-losporins, sulphonamides, neomycin, orclarithromycin.9182022 The benefits of two drugsynergy may not be obvious on routine in vitrotesting.

Keratitis caused by non-tuberculous myco-bacteria often manifests a relatively indolent,recalcitrant stromal infection which developswith a 2 to 3 week delay after trauma orsurgery. The disease may wax and wane over aperiod of several months.62328 Corticosteroidsare likely to aggravate the infectious process.Pain is variable, but is certainly never excruci-ating as in Acanthamoeba keratitis.The corneal stromal infiltrates may show

feathered edges, satellite lesions, crystallinekeratopathy, or a ring pattern. Satellite lesionsand feathery edges of infiltrates may causeconfusion with a fungal process. The crystal-line keratopathic changes are due to colonisa-tion of the organisms along discrete lamellarplanes in the stroma-that is, the organismsspread along the paths of least resistance.35This peculiar intrastromal distribution isthought to be associated with a combination oflow virulence of the involved micro-organismand suppression of the 'white cell response' bycorticosteroids.35 Three of our patients mani-fested crystalline keratopathy. Two such pa-tients had a cracked windshield infiltrative pat-tern, which in actuality is a main infiltratesurrounded by a radiating pattern of crystallinekeratopathy. The cracked windshield pattern isthought by some to be diagnostic of non-tuberculous mycobacterial keratitis.1223 Partialor complete ring infiltrates are common withthis infection and may cause confusion withherpetic, fungal, or Acanthamoeba keratitis. It isunclear whether this ring is the result of animmune deposition or due to other causes. Thecorneal epithelium has been reported to beintact over stromal infiltrates in a number ofcases reported elsewhere6; however, none ofour patients had intact epithelium.

Iritis with or without hypopyon was com-monly observed in our patients. The degree ofanterior chamber inflammation appears to par-allel the gravity of the events in the cornea, andis probably a reactive phenomenon rather thaninfectious per se.We did not encounter any cases of infectious

scleritis in our series. The only reported case ofnon-tuberculous mycobacterial scleritis in theliterature is a sclerokeratitis occurring contigu-ous to a limbal corneal infection.'7 None of thecorneal infiltrates in our patients were periph-eral enough to straddle the limbus.

In summary, non-tuberculous mycobacterialkeratitis should be suspected in corneal ulcersdeveloping after foreign body injuries (espe-cially metallic ones), particularly if the ulcer isindolent and fails to respond adequately tocommon antibacterial, antifungal, antiviral,and antiparasitic drugs. Surgically acquirednon-tuberculous mycobacterial infections, al-though less common, have the potential tooccur in epidemic clusters in surgical facili-ties.89 Better awareness of the disease andimproved familiarity with diagnostic and thera-peutic methods are important factors in reduc-ing the incidence of debilitating ocular damagefrom this infection. Misdiagnoses are commonand may lead to crucial delays in therapy. TheFite-Ferraco stain should perhaps be usedinstead of the Ziehi-Nielsen stain.

967

on Decem

ber 27, 2019 by guest. Protected by copyright.

http://bjo.bmj.com

/B

r J Ophthalm

ol: first published as 10.1136/bjo.80.11.962 on 1 Novem

ber 1996. Dow

nloaded from

Huang, Soong, Chang, Liang

L6wenstein-Jensen cultures should not be dis-carded after only 1 week, but should beretained for 4 weeks if necessary to allow colo-nies adequate time to grow. Although Mchelonei andfortuitum are rapid growers on sub-culture, the initial isolation can be slow as seenin the three cases (5, 7, and 20) detailed in theResults section. Despite excellent in vitroactivity of amikacin against non-tuberculousmycobacteria, the clinical response is oftenunpredictable"2 and does not correlate wellwith the minimum inhibitory concentration(MIC). Hu" postulated that poor drug pen-etration into the cornea, emergence of resistantstrains, immune ulceration, and stromal necro-sis impeding epithelialisation and wound heal-ing may all contribute to this discrepancy. Inrefractory cases showing no visible improve-ment after 1 week of conservative therapy, sur-gical extirpative therapy should be stronglyconsidered. Conjunctival flap surgery appearsefficacious if extirpative surgery fails to controlthe infection. Penetrating corneal transplanta-tion a' chaud may be an additional mode oftherapy for the very deep infections close toDescemet's membrane; however, if possible, itis best to delay transplantation until the infec-tion is eradicated and the eye is quiescent.

The authors have no proprietary interest in any of the drugs ormaterials used in this study.

1 Runyon EH. Anonymous mycobacteria in pulmonarydisease. Med Clin North Am 1959;43:273-8.

2 Turner L, Stinson I. Mycobacterium fortuitum as a cause ofcorneal ulcer. Am J Ophthalmol 1965;60:329-3 1.

3 Gangadharam PRJ, Lanier JD, Jones DB. Keratitis due toMycobacterium chelonei. Tubercle 1978;59:55-60.

4 Dalovisio JR, Pankey GA, Wallace RJ Jr, Jones DB. Clinicalusefulness of amikacin and doxycycline in the treatment ofinfection due to Mycobacterium fortuitum and Mycobac-terium chelonei. Rev Infect Dis 1981;3:1068-74.

5 Wallace RJ Jr, Jones DB, Wiss K. Sulfonamide activityagainst Mycobacterium fortuitum and Mycobacteriumchelonei. Rev Infect Dis 1981;3:898-904.

6 Turner L. Atypical mycobacteria infections in ophthalmol-ogy. Trans Am Ophthalmol Soc 1970;68:667-729.

7 Meisler DM, Friedlaender MH, Okumoto M. Mycobacteriumchelonei keratitis.Am J Ophthalmol 1982;94:398-401.

8 Robin JB, Beatty RF, Dunn S, Trousdale MD, RiffenburghR, Rao N, et al. Mycobacterium chelonei keratitis afterradial keratotomy. Am J Ophthalmol 1986;102:72-9.

9 Newman PE, Goodman RA, Waring GO III, Finton RJ,Wilson LA, Wright J. A cluster of cases of Mycobacteriumchelonei keratitis associated with outpatient office proce-dures. Am J Ophthalmol 1984;97:344-8.

10 Laflamme MY, Poisson M, Chehade N. Mycobacteriumchelonei keratitis following penetrating keratoplasty. Can J7Ophthalmol 1987;22:178-80.

11 Aylward GW, Stacey AR, Marsh RJ. Mycobacteriumchelonei infection of a corneal graft. BrJ Ophthalmol 1987;71:690-3.

12 Mirate DJ, Hull DS, Steel JH Jr, Carter MJ. Mycobacteriumchelonei keratitis: a case report. BrJ Ophthalmol 1983;67:324-6.

13 Zabel RW, Mintsioulis G, MacDonald I. Mycobacteriumchelonei keratitis in a soft-contact lens wearer. Can J Oph-thalmol 1988;23:315-7.

14 Bullington RH Jr, Lanier JD, Font RL. Non-tuberculousmycobacterial keratitis. Arch Ophthalmol 1992;110:519-24.

15 Knapp A, Stern GA, Hood C I. Mycobacterium avium-intracellulare corneal ulcer. Cornea 1987;6:175-80.

16 Moore MB, Newton C, Kaufman HE. Chronic keratitiscaused by Mycobacterium gordonae. Am J Ophthalmol1986;102:516-21.

17 Schonherr U, Naumann GO, Lang GK, Bialasiewicz AA.Sclerokeratitis associated with Mycobacterium marinum.Am J Ophthalmol 1989;108:607-8.

18 Dugel PU, Holland GN, Brown HH, Pettit TH, HofbauerJD, Simons KB, et al. Mycobacterium fortuitum keratitis.Am J Ophthalmol 1988;105:661-9.

19 Matoba A. Mycobacterium chelonei keratitis. Am JOphthalmol 1987;103:595-6.

20 Helm CJ, Holland GN, Lin R, Berlin OG, Bruckner DA.Comparison of topical antibiotics for treating Mycobacte-rium fortuitum keratitis in an animal model. Am 7 Ophthal-mol 1993;116:700-7.

21 Stevens RK, Holland GN, Paschal JF, Phinney RB, BerlinOG, Bruckner DA. Mycobacterium fortuitum keratitis. Acomparison of topical ciprofloxacin and amikacin in ananimal model. Cornea 1992;11:500-7.

22 Lin R, Holland GN, Helm CJ, Elias SJ, Berlin OG, Bruck-ner DA, et al. Comparative efficacy of topical ciprofloxacinfor treating Mycobacterium fortuitum and Mycobacteriumchelonei keratitis in an animal model. Am J Ophthalmol1994;117:657-62.

23 Lazar M, Nemet P, Bracha R, Campus A. Mycobacteriumfortuitum keratitis. Am J' Ophthalmol 1974;78:530-2.

24 Sommers HM, Good RC. Mycobacterium. In: LennetteEH, Balows A, Hausler WJ Jr, eds. Manual of clinical micro-biology, 4th ed. Washington, DC: American Society ofMicrobiology, 1985;216-48.

25 Wallace RJ Jr, Swenson JM, Silcox VA, Good RC, TschenJA, Stone MS. Spectrum of disease due to rapidly growingmycobacteria. Rev Infect Dis 1983;5:657-9.

26 Lauring LM, Wergeland FL, Sack GE. Anonymousmycobacterium keratitis. Am J Ophthalmol 1969;67: 130-3.

27 Levenson DS, Harrison CH. Mycobacterium fortuitumcorneal ulcer. Arch Ophthalmol 1966;75: 189-91.

28 Willis WE, Laibson PR. Intractable Mycobacterium fortui-tum corneal ulcer in man. Am J Ophthalmol 1971;71:500-4.

29 Wunch SE, Boyle GL, Leopold IH, Littman ML. Mycobac-terium fortuitum infection of corneal graft. Arch Ophthal-mol 1969;82:602-7.

30 Kent HD, Cohen EJ, Laibson PR, Arentsen JJ. Microbialkeratitis and corneal ulceration associated with therapeuticsoft contact lenses. CLAOJ 1990;16:49-52.

31 Hu FR. Infectious crystalline keratopathy caused by Myco-bacterium fortuitum and Pseudomonas aeruginosa. Am JOphthalmol 1990;109:738-9.

32 Hu FR. Extensive lamellar keratectomy for treatment ofnontuberculous mycobacterial keratitis. Am J Ophthalmol1995;120:47-54.

33 Tseng SH, Hsiao WC. Therapeutic lamellar keratectomy inthe management ofnontuberculous mycobacterium kerati-tis refractory to medical treatment. Cornea 1995;14:161-6.

34 Fite GL, Honolulu TH. The fuchsin-formaldehyde methodof staining acid-fast bacilli in paraffin sections. J Lab ClinMed 1940;25:743-4.

35 Gorovoy MS, Stern GA, Hood CI, Allen C. Intrastromalnoninflammatory bacteria colonization of a corneal graft.Arch Ophthalmol 1983;101:1749-52.

968

on Decem

ber 27, 2019 by guest. Protected by copyright.

http://bjo.bmj.com

/B

r J Ophthalm

ol: first published as 10.1136/bjo.80.11.962 on 1 Novem

ber 1996. Dow

nloaded from