LETTERS

Retinal Detachment and Fluoroquinolones

To the Editor: Dr Etminan and colleagues1 demonstratedthat current users of oral fluoroquinolones were at higherrisk to be diagnosed with retinal detachment than nonus-ers. The authors discussed the limitations of the study, but2 other issues may affect interpretation of the results.

First, the authors stated that “The risk of retinal detach-ment from our study was only elevated among currentusers but not among recent or past users, indicating anacute adverse event.”1 Table 3 of the article showed thatthe numbers of patients with retinal detachment and fluo-roquinolone use were 145, 12, and 288 in current, recent,and past users, respectively. Although the sample sizes inthe current and past users subgroups seem adequate,enrollment of only 12 patients in the recent use subgroupraises concern about the power of the study to detect anassociation. Providing information about the power of thestudy to detect statistically significant associations wouldbe beneficial.

Another important issue is the potential association be-tween the cumulative dose (or duration) of fluoroquinolo-nes and the risk of retinal detachment. In Table 3 of the ar-ticle, respiratory tract infections comprised 77% and 40%and genitourinary tract infections comprised 9% and 27%of primary indications for fluoroquinolone use in currentand past users, respectively. Common respiratory indica-tions for fluoroquinolone use such as acute sinusitis or com-munity-acquired pneumonia typically require a 7- to 14-day regimen of the drug, while acute uncomplicated urinarytract infection requires only a 3-day regimen.2-4 Therefore,it is likely that the current subgroup of users had indica-tions that required a larger cumulative dose of the medica-tion that may consequently confound the outcomes. Be-cause the authors indicated that drug strength, quantity, andnumber of days of supply were captured in the comprehen-sive prescription drug database, I suggest that they providean analysis with cumulative dose or duration of treatment,or both, included as an additional covariate in their condi-tional logistic regression model.

Mohammad H. Nowroozzadeh, MD

Author Affiliation: Poostchi Ophthalmology Research Center, Shiraz Universityof Medical Sciences, Shiraz, Iran (norozzadeh@gmail.com).Conflict of Interest Disclosures: The author has completed and submitted the ICMJEForm for Disclosure of Potential Conflicts of Interest and none were reported.

1. Etminan M, Forooghian F, Brophy JM, Bird ST, Maberley D. Oral fluoroquino-lones and the risk of retinal detachment. JAMA. 2012;307(13):1414-1419.2. Schaeffer AJ. The expanding role of fluoroquinolones. Am J Med. 2002;113(suppl 1A):45S-54S.3. Taur Y, Smith MA. Adherence to the Infectious Diseases Society of Americaguidelines in the treatment of uncomplicated urinary tract infection. Clin InfectDis. 2007;44(6):769-774.4. Croom KF, Goa KL. Levofloxacin: a review of its use in the treatment of bac-terial infections in the United States. Drugs. 2003;63(24):2769-2802.

To the Editor: Dr Etminan and colleagues1 conducted a case-control analysis of the association of oral fluoroquinoloneuse with retinal detachment. Understanding the mecha-nism by which fluoroquinolones may increase retinal de-tachment risk is important, especially if topical fluoroqui-nolones, the most common class used during theperioperative period of cataract surgery2 and for contact lens–related infections,3 carry a similar risk.

Elucidating the mechanism by which oral fluoroquino-lone use increases retinal detachment risk may be helpedby the determination of retinal detachment type: rheg-matogenous, serous/exudative, tractional, or a combina-tion. Each retinal detachment type typically has a distinctetiology. Fluoroquinolone use may favor a specific type ofretinal detachment, similar to corticosteroid-associatedserous retinal detachment in central serous retinopathy.4

Depending on the type of retinal detachment, distinct diag-nostic tests as well as varied approaches to surgical repairare undertaken. However, the authors stated they could not“distinguish the type of retinal detachment based on proce-dure codes because the surgical procedure codes for alltypes of retinal detachment are the same.”1 Yet in the Meth-ods section, they described the use of International Classifi-cation of Diseases, Ninth Revision (ICD-9) diagnostic codesto classify the incident cases from the cohort. The ICD-9codes do distinguish retinal detachment types based on thepresence of a retinal defect (rhegmatogenous, 361.00-361.07; serous/exudative, 361.2; tractional, 361.81).Although the authors hypothesized that rhegmatogenousretinal detachment is most commonly associated with oralfluoroquinolone use, they did not analyze this. The authorscould have used ICD-9 codes to classify retinal detachmenttype associated with fluoroquinolone use.

The authors theorized that the mechanism of oral fluo-roquinolone–associated retinal detachment was vitreous col-lagen breakdown, which promotes development of poste-

GUIDELINES FOR LETTERS. Letters discussing a recent JAMA article shouldbe submitted within 4 weeks of the article’s publication in print. Letters receivedafter 4 weeks will rarely be considered. Letters should not exceed 400 words oftext and 5 references and may have no more than 3 authors. Letters reportingoriginal research should not exceed 600 words of text and 6 references and mayhave no more than 5 authors. They may include up to 2 tables or figures but on-line supplementary material is not allowed. All letters should include a word count.Letters must not duplicate other material published or submitted for publication.Letters not meeting these specifications are generally not considered. Letters willbe published at the discretion of the editors and are subject to abridgement andediting. Further instructions can be found at http://jama.com/instructions. A signedstatement for authorship criteria and responsibility, financial disclosure, copyrighttransfer, and acknowledgment and the ICMJE Form for Disclosure of Potential Con-flicts of Interest are required before publication. Letters should be submitted viathe JAMA online submission and review system at http://manuscripts.jama.com(note: do not include “www” before the URL). For technical assistance, pleasecontact jama-letters@jamanetwork.org.

Letters Section Editor: Jody W. Zylke, MD, Senior Editor.

©2012 American Medical Association. All rights reserved. JAMA, July 18, 2012—Vol 308, No. 3 233

Downloaded From: http://jama.jamanetwork.com/ by a University of Virginia User on 01/05/2014

rior vitreous detachment, a known risk factor forrhegmatogenous retinal detachment. A possible relation-ship between oral fluoroquinolone use and posterior vitre-ous detachment could have been examined with the use ofICD-9 codes 379.21 and 379.22. They described current—but not recent—use of fluoroquinolones to be related to reti-nal detachment. However, the natural history of posteriorvitreous detachments is progressive, and one would expectthat fluoroquinolone use would initiate an irreversible pro-cess of vitreous collagen degeneration.5 Thus recent, as wellas current, oral fluoroquinolone use should expose the pa-tient to an elevated retinal detachment risk.

We suggest further analysis of ICD-9 codes to examineretinal detachment type and posterior vitreous detachmentstatus to advance understanding of the pathogenesis of oralfluoroquinolone–associated retinal detachment.

Rajesh C. Rao, MDBrian J. Dlouhy, MD

Author Affiliations: Washington University School of Medicine, St Louis, Mis-souri (Dr Rao; raor@vision.wustl.edu); and University of Iowa Hospitals and Clin-ics, Iowa City (Dr Dlouhy).Conflict of Interest Disclosures: The authors have completed and submitted theICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Rao reported re-ceiving a merit award from the Heed Ophthalmic Foundation. Dr Dlouhy re-ported no disclosures.

1. Etminan M, Forooghian F, Brophy JM, Bird ST, Maberley D. Oral fluoroquino-lones and the risk of retinal detachment. JAMA. 2012;307(13):1414-1419.2. Hammoudi DS, Abdolell M, Wong DT. Patterns of perioperative prophylaxisfor cataract surgery in Canada. Can J Ophthalmol. 2007;42(5):681-688.3. Alexandrakis G, Alfonso EC, Miller D. Shifting trends in bacterial keratitis in southFlorida and emerging resistance to fluoroquinolones. Ophthalmology. 2000;107(8):1497-1502.4. Carvalho-Recchia CA, Yannuzzi LA, Negrao S, et al. Corticosteroids and cen-tral serous chorioretinopathy. Ophthalmology. 2002;109(10):1834-1837.5. Uchino E, Uemura A, Ohba N. Initial stages of posterior vitreous detachmentin healthy eyes of older persons evaluated by optical coherence tomography. ArchOphthalmol. 2001;119(10):1475-1479.

To the Editor: The nested case-control study by Dr Etminanand colleagues1 demonstrated an association between sys-temic fluoroquinolone use and episodes of treatment for reti-nal detachment. As the authors stated, association does notimply causation. Although the adjusted rate ratios and con-fidence intervals were noteworthy, several important con-founding factors need to be considered.

First, the timing of fluoroquinolone use and the treat-ment for a retinal detachment were very short. The asso-ciation was only present for current fluoroquinolone us-ers, not for those with recent or past use. If there was a changein the vitreous that could precipitate a retinal detachment,a retinal detachment would be expected to take some timeto develop. There is usually a time lag of several days or weeksbetween the development of a posterior vitreous detach-ment and the formation of retinal breaks that may lead to aretinal detachment.2 For this reason, the association be-tween retinal detachment and fluoroquinolone use wouldhave been expected to hold for those with recent fluoro-quinolone use as well.

Second, the timing of retinal detachment repair dependson the extent of retinal detachment. If the macula has de-

tached, surgery can take place within a week of the diag-nosis with no effect on the final visual outcome.3 For thisreason, patients with retinal detachment and an identifi-able risk factor for endophthalmitis, such as a mild respi-ratory infection (that would otherwise have received no an-tibiotics) may have had fluoroquinolone antibioticsprescribed to reduce any bacterial load before surgery andthe risk of postoperative endophthalmitis.4

Third, the vast majority of retinal detachment caseswere prescribed fluoroquinolones for a respiratory infec-tion. Coughing could precipitate an impending posteriorvitreous detachment and a retinal detachment could fol-low. �-Lactam antibiotics were also investigated as a con-trol group and no association with retinal detachmentwas found. However, no detail was provided on the indi-cations for treatment with �-lactam antibiotics, which isimportant because fluoroquinolones have a much greaterrole in the management of community-acquired pneumo-nia in the United States.5 The �-lactam cohort is likely tohave included a greater proportion of patients with non-respiratory infections, which are not associated withcoughing and possible precipitation of a posterior vitre-ous detachment.

Petros Aristodemou, FRCOphth(Lond)Andreas C. Stylianides, MD, FRCP(Lond)

Author Affiliations: Academic Unit of Ophthalmology, University of Bristol, Bris-tol, United Kingdom (Dr Aristodemou; topetros@yahoo.com); and Hygeia Poly-clinic, Limassol, Cyprus (Dr Stylianides).Conflict of Interest Disclosures: The authors have completed and submitted theICMJE Form for Disclosure of Potential Conflicts of Interest and none were re-ported.

1. Etminan M, Forooghian F, Brophy JM, Bird ST, Maberley D. Oral fluoroquino-lones and the risk of retinal detachment. JAMA. 2012;307(13):1414-1419.2. Schweitzer KD, Eneh AA, Hurst J, Bona MD, Rahim KJ, Sharma S. Predictingretinal tears in posterior vitreous detachment. Can J Ophthalmol. 2011;46(6):481-485.3. Diederen RM, La Heij EC, Kessels AG, Goezinne F, Liem AT, Hendrikse F. Scleralbuckling surgery after macula-off retinal detachment: worse visual outcome aftermore than 6 days. Ophthalmology. 2007;114(4):705-709.4. Bacon AS, Davison CR, Patel BC, Frazer DG, Ficker LA, Dart JK. Infective en-dophthalmitis following vitreoretinal surgery. Eye (Lond). 1993;7(pt 4):529-534.5. Niederman MS. Community-acquired pneumonia: the US perspective. SeminRespir Crit Care Med. 2009;30(2):179-188.

To the Editor: A robust association between current fluo-roquinolone use and retinal detachment was reported by DrEtminan and colleagues.1 They stated that a group of re-sidual confounders would need to be excessively large tochange the results of the study. However, the cumulativeeffect of several important factors may account for, at leastin part, the reported association.

Cataract surgery was considered in the regression model;however, there was no measure of several known risk fac-tors after cataract surgery that have been significantly re-lated to the rate of retinal detachment. In particular, com-plicated cataract surgery and a history of retinal detachmentin the fellow eye were associated with odds ratios of ap-proximately 20 and 12, respectively, for subsequent retinaldetachment in a study by Tuft et al.2 The ethnicity of the

LETTERS

234 JAMA, July 18, 2012—Vol 308, No. 3 ©2012 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ by a University of Virginia User on 01/05/2014

study population has been demonstrated to contribute to a3-fold variation in the reported incidence of retinal detach-ment in a large population-based study,3 while socioeco-nomic status of the study population and the proportion ofocular trauma have both been demonstrated to have inde-pendent associations with the incidence of retinal detach-ment.4

Using administrative data to estimate disease occur-rence often limits the ability of the investigator to examinein detail other influences on disease. Although the poten-tial influence of these factors may not have been measuredin the present study, we suggest that their cumulative ef-fect may not be negligible and they should be taken into con-sideration in future pharmacoepidemiological studies seek-ing to characterize an association between fluoroquinolonesand retinal detachment.

Aoife C. Molloy, MD, DTMH, MRCPUKDanny Mitry, MD, MRCSI(Ophth)Author Affiliations: Department of Microbiology, Royal Free Hospital NHS Foun-dation Trust, London, England (Dr Molloy); and Department of Public Health Sci-ences, University of Edinburgh, Edinburgh, Scotland (Dr Mitry; mitryd@gmail.com).Conflict of Interest Disclosures: The authors have completed and submitted theICMJE Form for Disclosure of Potential Conflicts of Interest and none were re-ported.

1. Etminan M, Forooghian F, Brophy JM, Bird ST, Maberley D. Oral fluoroquino-lones and the risk of retinal detachment. JAMA. 2012;307(13):1414-1419.2. Tuft SJ, Minassian D, Sullivan P. Risk factors for retinal detachment after cata-ract surgery: a case-control study. Ophthalmology. 2006;113(4):650-656.3. Wong TY, Tielsch JM, Schein OD. Racial difference in the incidence of retinaldetachment in Singapore. Arch Ophthalmol. 1999;117(3):379-383.4. Mitry D, Charteris DG, Yorston D, et al; Scottish RD Study Group. The epide-miology and socioeconomic associations of retinal detachment in Scotland: a two-year prospective population-based study. Invest Ophthalmol Vis Sci. 2010;51(10):4963-4968.

In Reply: Dr Nowroozzadeh requests information on thestudy power. For past use, the relatively narrow confi-dence interval (RR, 1.03; 95% CI, 0.89-1.19) indicates suf-ficient power; however, the wider confidence interval withrecent use (RR, 0.92; 95% CI, 0.45-1.87) reveals less powerto reliably estimate an association with retinal detachment.This is not surprising given the short exposure window forrecent use (7 days) compared with that for past use (8 to365 days prior to index date). Nowroozzadeh also suggestsexamining the effect of cumulative dose. The small num-ber of exposed cases did not allow us to adequately exam-ine the effect of dose. We agree that future studies shouldexamine this question.

Drs Rao and Dlouhy and Drs Aristodemou and Stylianidesspeculate that retinal breaks take time to develop and a sig-nal would be expected with recent use. The small samplesize for recent use did not allow us to further examine thisassociation. Given that a single fluoroquinolone dose hasprecipitated detachment of the Achilles tendon, a shorteronset to a retinal detachment is not improbable.1 The au-thors state that there are ICD-9 codes for posterior vitreousdetachment. Ophthalmologists in British Columbia are onlyrequired to list the first 3 digits of an ICD-9 code. Because

379 is a nonspecific code and posterior vitreous detach-ment does not involve a surgical procedure, we could notassess the association with fluoroquinolone use.

Aristodemou and Stylianides wonder if antibiotics couldhave been prescribed after retinal detachment but before sur-gical intervention. We avoided this potential reverse cau-sality bias by indexing cases at the first retinal detachmentdiagnosis and measuring exposure antecedent to this date.They also ask if the association could result from proto-pathic bias, whereby fluoroquinolones are prescribed for re-spiratory tract infections and coughing could induce reti-nal detachment. They suggest that an analysis of those with�-lactam antibiotic use may not control for this bias be-cause these antibiotics are not as frequently prescribed forrespiratory conditions in the United States. In British Co-lumbia, �-lactam antibiotics alone or in combination withmacrolides are commonly prescribed for the treatment ofcommunity-acquired pneumonia. If confounding fromcoughing was present, we would expect a signal with oursecond control group, short-acting �-agonists, which alsoshowed no association with retinal detachment.

Rao and Dlouhy proposed further analysis by subtype ofretinal detachment. Retinal surgeons in British Columbiaare required only to submit a general retinal detachment bill-ing code (361), which makes examination of specific sub-types through ICD-9 coding unreliable and inaccurate. Be-cause serous/exudative subtypes usually do not requiresurgery, we suspect the majority of the cases in our studywere of the rhegmatogenous type.

Drs Molloy and Mitry suggest that previous history of reti-nal detachment or complications from cataract surgery mayhave contributed to the positive association. In our study,we excluded patients with a prior retinal detachment andstratified the analysis by previous history of cataract sur-gery, finding patients with no cataract history to have a sig-nificantly higher risk of developing retinal detachment (RR,3.34; 95% CI, 2.06-5.40). They comment that we did notcontrol for socioeconomic status, ethnicity, and oculartrauma. None of these factors are true confounders, whichby definition have to be associated with both retinal detach-ment and fluoroquinolone use. Socioeconomic status is un-likely to act as a confounder because all residents of BritishColumbia have access to the same level of ophthalmologiccare and receive coverage for fluoroquinolones through auniversal prescription drug plan. Similarly, it is unlikely thatethnicity or ocular trauma act as confounders because thereis no reason to believe that fluoroquinolones may be pre-scribed differentially to patients based on ethnicity or his-tory of ocular trauma.

Mahyar Etminan, PharmD, MScJames M. Brophy, MD, PhDDavid Maberley, MD, MScAuthor Affiliations: Therapeutic Evaluation Unit, Child and Family Research In-stitute of British Columbia, Vancouver, Canada (Dr Etminan; metminan@popi.ubc.ca); Department of Epidemiology, Biostatistics and Medicine, McGill University,

LETTERS

©2012 American Medical Association. All rights reserved. JAMA, July 18, 2012—Vol 308, No. 3 235

Downloaded From: http://jama.jamanetwork.com/ by a University of Virginia User on 01/05/2014

Montreal, Canada (Dr Brophy); and Department of Ophthalmology and Visual Sci-ences, University of British Columbia, Vancouver, Canada (Dr Maberley).Conflict of Interest Disclosures: The authors have completed and submitted theICMJE Form for Disclosure of Potential Conflicts of Interest and none were re-ported.

1. Khaliq Y, Zhanel GG. Fluoroquinolone-associated tendinopathy: a critical re-view of the literature. Clin Infect Dis. 2003;36(11):1404-1410.

Supplementary Imaging for Breast CancerScreening in High-Risk Women

To the Editor: In a clinical trial by Dr Berg and colleagues1

evaluating the benefits of supplementary ultrasound in 3rounds of annual breast cancer screening and a single mag-netic resonance imaging (MRI) scan after completion of the3 screenings, the Breast Imaging-Reporting and Data Sys-tem (BI-RADS) was used to categorize and report imagingfeatures of breast lesions. Scores of 3, 4a, 4b, 4c, or 5 wereconsidered test positive. In 7473 screenings with mammog-raphy and ultrasound and 612 MRIs, the authors found thatadding ultrasound or MRI to mammography increased thedetection of cancer but also increased false-positive re-sults. Supplementary ultrasound increased the recall rate from10.2% with mammography alone to 20.3%, which is higherthan the benchmark rate of 10% recommended by the USscreening guidelines.2

These findings also raise questions about the recommen-dation for more than routine screening in women with aBI-RADS score of 3 because most of these lesions will notprove malignant and such a strategy is likely to lead to in-creased recall rates and short-term follow-up. In a studyevaluating the variability in BI-RADS scores for ultrasoundamong breast radiologists, the percentage of category 3 le-sions was 11.6% in the assessment of 113 benign and 154malignant breast masses, with moderate agreement be-tween observers.3 In the breast cancer screening programin the Netherlands, category 3 was excluded from the na-tionwide introduction of BI-RADS because of the need forshort interval follow-up despite its benign features, whichwas not available in the Dutch screening setting.4 It wouldbe interesting to know if Berg et al1 analyzed their protocolfor BI-RADS 3 lesions separately.

Tetsuji Fujita, MD

Author Affiliation: Jikei University of School of Medicine, Tokyo, Japan (tetsu@jg8.so-net.ne.jp).Conflict of Interest Disclosures: The author has completed and submitted the ICMJEForm for Disclosure of Potential Conflicts of Interest and none were reported.

1. Berg WA, Zhang Z, Lehrer D, et al; ACRIN 6666 Investigators. Detection ofbreast cancer with addition of annual screening ultrasound or a single screeningMRI to mammography in women with elevated breast cancer risk. JAMA. 2012;307(13):1394-1404.2. Bassett LW, Hendrick RE, Bassford TL, et al. Quality determinants ofmammography. In: Clinical Practice Guideline No. 13: AHCPR Publication No.95-0632. Rockville, MD: Agency for Healthcare Policy and Research; 1994.3. Abdullah N, Mesurolle B, El-Khoury M, Kao E. Breast imaging reporting anddata system lexicon for US: interobserver agreement for assessment of breast masses.Radiology. 2009;252(3):665-672.4. Timmers JMH, van Doorne-Nagtegaal HJ, Zonderland HM, et al. The BreastImaging Reporting and Data System (BI-RADS) in the Dutch breast cancer screen-ing programme: its role as an assessment and stratification tool [published onlineMarch 14, 2012]. Eur Radiol. doi:10.1007/s00330-012-2409-2.

To the Editor: One objective of the recent study by Dr Bergand colleagues1 was to evaluate the potential for screeningusing MRI to identify breast cancers that were missed by acombination of ultrasound and mammography. After 3rounds of screening by ultrasound and mammography, 612women went on to have an MRI. In the 2662 women whoreceived a mammogram and ultrasound, there were 111breast cancers detected (4.2%), 91 by mammogram, ultra-sound, or both. In the 612 women who later had an MRI,16 cancers (2.6%) were detected.

The report appears to be misleading and underestimatesthe potential benefits of MRI. The women who had an MRIwere heavily prescreened. The MRI took place a mean of25 days after the last mammogram, not 1 year after (eTable1 in article). The ultrasound/mammography cohort in-cluded both prevalent and incident cancers, whereas the MRIcohort included only incident cancers. If only incidencescreenings in the second and third years were considered,there were 75 women diagnosed with cancer. Nine intervalcancers were detected. Based on data in Table 2 of the ar-ticle, the median sizes of the cancers detected by mammog-raphy and/or ultrasound were 10 mm to 16 mm and 23%were node-positive. The median size of the cancers de-tected by MRI was 8.5 mm and none were node-positive.But as far as sensitivity is concerned, all cancers are weightedthe same.

This obscures the real advantage of MRI screening overconventional screening—a downstaging of breast cancers.In a recent study2 of MRI screening in BRCA1 and BRCA2carriers, the adjusted hazard ratio for the development ofstages II to IV breast cancer associated with MRI screeningcompared with conventional screening was 0.30 (95% CI,0.12-0.72; P=.008). It is possible that if all the women inthe study by Berg et al1 had received an MRI on the firstround, many of the incident cancers would have been de-tected, thereby reducing the sensitivity of mammographyand ultrasound for rounds 2 and 3. The current study pro-vides evidence that ultrasound is a useful addition to mam-mography for breast cancer screening in women with densebreasts but says very little about MRI.

Steven A. Narod, MD

Author Affiliation: Women’s College Research Institute, Toronto, Ontario, Canada(steven.narod@wchospital.ca).Conflict of Interest Disclosures: The author has completed and submitted the ICMJEForm for Disclosure of Potential Conflicts of Interest and none were reported.

1. Berg WA, Zhang Z, Lehrer D, et al; ACRIN 6666 Investigators. Detection ofbreast cancer with addition of annual screening ultrasound or a single screeningMRI to mammography in women with elevated breast cancer risk. JAMA. 2012;307(13):1394-1404.2. Warner E, Hill K, Causer P, et al. Prospective study of breast cancer incidencein women with a BRCA1 or BRCA2 mutation under surveillance with and withoutmagnetic resonance imaging. J Clin Oncol. 2011;29(13):1664-1669.

In Reply: As Dr Fujita points out, our study showed thatsupplemental ultrasound screening increased the cancer de-tection rate. To achieve this, recall rates increased from 10.2%to 20.3% on average (16.8% with incidence screenings).

LETTERS

236 JAMA, July 18, 2012—Vol 308, No. 3 ©2012 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ by a University of Virginia User on 01/05/2014