Dear Editor,

We are pleased at the interest shown by Dr. Stapleton and colleagues in our work [1] and we note their endorsement of our approach in adopting a clinical severity classification system for the assessment of contact lens associated keratitis.

Dr. Stapleton and colleagues have expressed concern at the high incidence of severe keratitis that we reported for hydrogel extended wear lenses. They suggested that our claim that this result could be attributed to a higher capture rate of cases (due to our single-centre methodology) is flawed because, if this were true, the incidence rate we reported for other lens types should also be higher than those reported previously. In support of their argument, Dr. Stapleton and co-workers presented data from only a selected range of previous studies. However, our reported incidence rates are typically higher than those reported previously for all categories of lens types. For example, the incidence values reported in our study for daily wear rigid, daily wear hydrogel and extended wear hydrogel were 2.6, 1.8 and 4.8 times greater than those reported by Cheng et al. [2], respectively.

In relation to the specific case of data for hydrogel extended wear lenses (Figure 1) it is clear that the incidence figure we report is of the same order of magnitude as that reported by Poggio and Abelson [3] and Holden et al. [4]. It is noteworthy that the higher values reported by Poggio and Abelson [3], Holden et al. [4], and ourselves are derived from studies in which case capture was the responsibility of a small number of centres (eight centres, two centres and one centre, respectively), whereas the lower values reported by other authors [2,5-7] were derived from studies that typically involved hundreds of reporting centres. This observation lends support to our supposition that the use of a limited number of reporting centres will result in higher, and necessarily more accurate, estimates of incidence, albeit with larger statistical confidence intervals. We would suggest that the recently- published estimate of Holden et al. [4] (57.9 cases per 10,000 wearers per year, 95% confidence interval 21.3 – 126) is likely to be the most accurate incidence value for severe keratitis with hydrogel extended wear contact lenses, as this is based on the largest prospective cohort study yet published; an approach which virtually ensures total case capture.

Figure 1. Published annualised incidence rates for presumed microbial keratitis in extended wear hydrogel lens use from population based studies using more than 200 report centres [2,5-7] (black circles) and less than 10 report centres [1,3,4] (white circles). Vertical lines show 95% confidence intervals.

Notwithstanding the arguments presented above, we agree with the general sentiment expressed by Dr. Stapleton and colleagues of the desire to reconcile the results of various studies addressing the question of contact lens safety. A key factor in our own study was the criteria adopted to differentiate between non-severe and severe keratitis. The use of different criteria is likely to result in different incidence values between studies. Indeed, this is the outcome of a separate analysis which we have conducted and are in the process of preparing for publication.

Attention has been drawn to the ‘indirect’ methodology for determining the incidence denominator in our study. It is of course the case that all methodologies that have been used to determine information about the control population in contact lens incidence studies (with the exception of Holden et al. [4]) have been indirect. For example, Cheng et al. [2] and Poggio et al. [5] estimated the denominator using telephone surveys. Such techniques can also be problematic [8].

Dr. Stapleton and colleagues criticised our methodology as being likely to over-estimate the penetrance of new modalities such as silicone hydrogel lens use and under-estimate the penetrance of modalities such as rigid gas permeable lenses. They appear to have assumed that our estimate of the number of wearers of each lens type was determined from prescribing data.

While others, such as Radford, Minassian and Dart [9] have previously adopted this approach, we agree that this would result in over- and under- estimations of lens usage. Contrary to the assertion of Dr. Stapleton and co- workers, we used lens sales data for 2003 (the year our survey was conducted) which gives a precise estimate of the number of lens wearers. The model which underpins these estimates has arisen from a three year collaboration between one of us (PBM) and the UK Association of Contact Lens Manufacturers and this approach has been previously validated as an accurate method for predicting the number of wearers of various contact lens type [10]. The use of prescribing data in our study was confined to establishing the ratio of lenses worn on a daily vs. extended wear basis. This information has been collected on an annual basis over the past nine years and in that way we gained an accurate estimate of lens wear modalities in the UK during the time of the survey.

Dr. Stapleton and colleagues challenge our assumption that the prevalence of contact lens usage in the catchment population of the Royal Eye Hospital in Manchester is essentially the same as that throughout the UK. The appropriateness of their suggestion of extrapolating the data from Market Opinion Research International (MORI) on overall lens penetrance (as cited by Radford, Minassian and Dart [9]) to our results is unclear; however, assuming that it is sound, we agree that this could impact on our data by uniformly raising or lowering our estimates of incidence to some extent. Nevertheless, we are unaware of any evidence of regional differences of usage between different lens types.

We reject the assertion of Dr. Stapleton and co-workers that the use of the chi-square test for comparing estimates of incidence for lens types is unjustified, whereas citing 95% confidence intervals is preferred. As the underlying assumption of both of these approaches is that one has accurate numbers of cases and controls, it is contradictory to advocate only one of these approaches. We are confident that our estimates of cases and controls are sufficiently robust to be able to apply the chi-square test. This chi- square approach was adopted by Radford, Minassian and Dart [9] in coming to their conclusion that the incidence of Acanthamoeba keratitis is lower in users of daily disposable lenses versus users of other lens types.

Dr. Stapleton and colleagues cited the overlap of 95% confidence intervals for the incidence of severe keratitis in extended wear hydrogel lenses and extended wear silicone hydrogel lenses as undermining our assertion that these incidence values are significantly different. However, the correct application of 95% confidence intervals for comparative purposes is to determine whether the estimate of incidence for one lens types falls within the 95% confidence interval for the other lens type, and vice versa. In our study, the incidence of severe keratitis for extended wear hydrogel lenses (96.4 cases per 10,000 wearers per year) falls outside the 95% confidence interval for the incidence of severe keratitis for extended wear silicone hydrogel lenses (6.7 – 58.0); conversely, the incidence of severe keratitis for extended wear silicone hydrogel lenses (19.8) falls outside the 95% confidence interval for the incidence of severe keratitis for extended wear hydrogel lenses (37.5 – 245.2). This confirms the chi-square result reported in our paper of a significant difference between these two groups, a finding which is consistent with predictions of reduced rates of severe keratitis with silicone hydrogels lenses based on reports of worldwide cases of lens-related infections [11] and the general clinical performance with this lens type [12].

In conclusion, we feel that our paper presents a novel, robust and clinically- meaningful portrayal of the risks associated with contemporary contact lenses. We are confident that our methodological and statistical approaches are sound, and we believe that the criticisms of Dr. Stapleton and colleagues are largely unfounded. We are aware that other studies of the incidence of contact lens related keratitis are currently underway, and we look forward with interest to the outcome of these studies inasmuch as they will assist the contact lens community in arriving at a more comprehensive picture of the relative safety of different contact lens types.

Yours faithfully,

Philip B Morgan¹
Nathan Efron¹
Elizabeth A Hill¹
Mathew K Raynor²
Mark A Whiting²
Andrew B Tullo²

1 Eurolens Research, The University of Manchester, Manchester, UK
2 Royal Eye Hospital, Manchester, UK

References

1. Morgan PB, Efron N, Hill EA,at al. Incidence of keratitis of varying severity among contact lens wearers. Br J Ophthalmol 2005; 89:430-436.

2. Cheng KH, Leung SL, Hoekman HW, et al. Incidence of contact-lens- associated microbial keratitis and its related morbidity. Lancet 1999;354: 181-185.

3. Poggio EC and Abelson M. Complications and symptoms in disposable extended wear lenses compared with conventional soft daily wear and soft extended wear lenses. Contact Lens Assoc Ophthalmol J 1993; 19: 31-39.

4. Holden BA, Sankaridurg PR, Sweeney DF, et al. Microbial keratitis in prospective studies of extended wear with disposable hydrogel contact lenses. Cornea 2005; 24: 156-161.

5. Poggio EC, Glynn RJ, Schein OD, et al. The incidence of ulcerative keratitis among users of daily-wear and extended-wear soft contact lenses. N Engl J Med 1989;321: 779-783.

6. Nilsson SE and Montan PG. The hospitalized cases of contact lens induced keratitis in Sweden and their relation to lens type and wear schedule: results of a three-year retrospective study. Contact Lens Assoc Ophthalmol J 1994; 20: 97-101.

7. Lam DS, Houang E, Fan DS, et al. Incidence and risk factors for microbial keratitis in Hong Kong: comparison with Europe and North America. Eye 2002; 16: 608-618.

8. Aquilino WS. Telephone versus face-to-face interviewing for household drug use surveys. Int J Addict 1992; 27: 71-91.

9. Radford CF, Minassian DC, and Dart JK. Acanthamoeba keratitis in England and Wales: incidence, outcome, and risk factors. Br J Ophthalmol 2002; 86: 536-542.

10. Morgan PB. A healthcheck on the UK contact lens market. Optician 2002; 223 (5854): 14-16.<

11. Holden BA, Sweeney DF, Sankaridurg PR, Carnt N, Edwards K, Stretton S, Stapleton F. Microbial keratitis and vision loss with contact lenses. Eye Contact Lens 2003; 29: S131-4.

12. Covey M, Sweeney DF, Terry R, Sankaridurg PR, Holden BA. Hypoxic effects on the anterior eye of high-Dk soft contact lens wearers are negligible. Optom Vis Sci 2001; 78: 95-99.

Dear Editor,

We would like to congratulate the authors on their recent paper ‘Incidence of keratitis of varying severity amongst contact lens wearers’ [1]. The use of a clinical severity matrix as described by Aasuri et al. [2] is a useful tool in the classification of contact lens related infiltrates and provides insight to the spectrum of acute lesions presenting to an ophthalmic casualty unit. The proportion of cases with severe (38/118, 32%) and non-severe keratitis (80/118, 68%) is very similar to an earlier ophthalmic casualty study where the spectrum of acute contact lens related suppurative keratitis was separated into ‘presumed microbial’ (60/207, 29%) and ‘presumed sterile’ lesions (147/207, 71%) [3]. We agree that the historical definition of presumed microbial keratitis in earlier studies [4-6] is likely to be consistent with the authors’ classification of severe keratitis.

However, we are concerned that the data presented is misleading. The authors present an incidence of ‘severe’ keratitis with extended wear soft contact lenses of 96.4 (37.5-245.2) per 10,000 wearers per year. This is five times higher than previously published data (Figure 1) and is central to the author’s conclusion that silicone hydrogels “have a five times decreased risk of severe keratitis for extended wear compared with hydrogel lenses”.

We believe the most likely explanation for the difference between the present data and previous studies relates to the selection of controls for the denominator and not, as the authors claim, a higher capture rate of cases. Indeed a higher capture rate of cases might be expected to affect the incidence rate for all lens types equally and this appears not to be the case.

The determination of the denominator in the present study is indirect and is carried out as follows:

1. Estimation of the size of the hospital casualty population.

2. Determination of the proportions of lens use in the UK population using prescribing trends data collected over a nine year period. Information on new lens fittings and re-fitting of existing wearers is provided by practitioners and the proportion of wearers using different lens types and modalities is estimated.

3. Estimation of the penetration of contact lens wear was based on sales data provided by the Association of Contact Lens Manufacturers in 2003.

4. Estimation of the numbers of wearers using different lens types in the catchment population was therefore estimated from a combination of the above three calculations.

One difficulty with this complex and indirect approach is the use of fitting data to estimate lens use in the community. This approach is likely to overestimate the penetrance of new modalities such as silicone hydrogel lens use and underestimate the penetrance of modalities, such as rigid gas permeable lenses where individuals are less likely to present for follow up and extended wear hydrogel lenses, which is a modality rarely recommended by contact lens practitioners. The precision of estimates of infrequently used lens modalities such as extended wear of hydrogel lenses is likely to be poor. Sales data usually represents units of lenses sold and not wearers. Additionally, sales data and fitting study data response rates of around 10% are unlikely to provide accurate estimates of actual wear practice in the community, which is important in precisely estimating the denominator. This also limits the appropriateness of the control group [4]. Lastly, there may be a threefold difference in CL’s prevalence for different UK regions [9], making extrapolation of national figures to a region likely to lead to inaccuracies.

Furthermore, we believe that the statistical analysis for the comparison of incidence between extended wear hydrogel and extended wear silicone hydrogel lenses is inappropriate. The authors have used a chi- square test to determine significant difference in incidence rate of severe keratitis between extended wear hydrogel and extended wear silicone hydrogel. The data and methodology presented in the paper do not justify the use of chi-square test. The chi-square test is a test of association using dichotomous / multichotomous data. The test assumes that the outcome variable is recorded for every individual in the analysis. However, when the outcome is not known amongst the controls, the observed and expected counts in the 2x2 table are invalid. Moreover, tests of significance such as this make no inference about population parameters, which is the aim of studies reporting incidence or prevalence.

A more appropriate method of assessing differences in incidence rates is the use of 95% confidence intervals (CI). The CI set the upper and lower bounds of the population parameter (incidence rate) for a given level of significance. The CI is useful as it provides the relative meaningfulness of the given estimate. The authors have reported the 95% CI for severe keratitis in extended wear hydrogel lenses and extended wear silicone hydrogel lenses and have shown that these intervals overlap. The overlap indicates that there is less than 95% confidence that the true population incidence rates of severe keratitis with these two lens types are different.

In summary, the strength of this paper lies in its investigation of the relative severity of corneal infiltrative events associated with contact lenses presenting to an eye casualty department. The low number of cases of serious keratitis (38) and methodological choices in estimating contact lens use in the community has implications for the strength of this data. The statement of a five times decreased risk of severe keratitis with silicone hydrogel lenses on an extended wear basis is not supported statistically or methodologically in this study. Reporting of such a conclusion may mislead practitioners about the true risks of serious keratitis and choice of lens type.

Yours faithfully

Fiona Stapleton ^1,2,3
Lisa Keay ^1,3
Katie Edwards ^1,3
Thomas Navidulath ^1,2,3
Cherry Radford ⁴
Brien Holden ^1,2,3
John Dart ^4,5

1 School of Optometry and Vision Science, University of New South Wales, Sydney, Australia
2 Institute for Eye Research, Sydney, Australia
3 Vision CRC, Sydney, Australia
4 Moorfields Eye Hospital, London, England
5 Institute of Ophthalmology, University College London, England.

References

1. Morgan PB, Efron N, Hill EA, et al. Incidence of keratitis of varying severity amongst contact lens wearers. Br J Ophthalmol 2005;89:430 -436.

2. Aasuri MK, Venkata N, Kumar VM. Clinical (differential) diagnosis of microbial keratitis (MK) and contact lens induced peripheral ulceration. Eye Contact Lens 2003;29:S60-S62.

3. Stapleton F, Dart JKG, Minassian D. Risk factors in contact lens related suppurative keratitis. CLAO Journal 1993;19:204-210.

4. Poggio EC, Glynn RJ, Schein OD, et al. The incidence of ulcerative keratitis among users of daily-wear and extended-wear soft contact lenses. N Engl J Med 1989;321:779-783.

5. Dart JKG, Stapleton F, Minassian D. Contact lenses and keratitis. Lancet 1991;338:1146-1147.

6. Cheng K, Leung D, Hoekman H, et al. Incidence of contact-lens- associated microbial keratitis and its related morbidity. Lancet 1999;354:181-184.

7. Nilsson SEG, Montan PG. The annualized incidence of contact lens induced keratitis in Sweden and its relation to lens type and wear schedule: results of a 3-month prospective study. CLAO Journal 1994;20:225 -230.

8. Lam DSC, Houang E, Fan DSP, et al. Incidence and risk factors for microbial keratitis in Hong Kong: comparison with Europe and North America. Eye 2002;16:608-618.

9. Radford C, Minassian D, JKG D. Acanthamoeba keratitis in England and Wales: incidence, outcome and risk factors. Br J Ophthalmol 2002;86:536-542.

Figure legend

Figure 1. Published annualised incidence rates for presumed microbial keratitis in extended wear hydrogel lens use from population based studies [1, 4, 6-8]. Vertical lines show 95% confidence intervals.