Statistics from Altmetric.com
We read with interest the report by Radford et al on the incidence of acanthamoeba keratitis and the suggested relation with hardness of water.1 However, we disagree with their findings and believe their study is flawed for two reasons.
Firstly, the incidence of acanthamoeba keratitis has been underestimated by approximately 33% since their monthly questionnaire reporting system yielded an average return of only 67%. A high non-response rate over 30% is considered unreliable for making inferences for the whole population.2 Furthermore, the authors estimated the “incidence” rate using the “complete” cases, which is a biased sample. Even then, if the non-response rates were below 30% in this study, the authors did not provide the precision of the “incident” rate and attempt to apply a statistical weighting approach to remove or reduce non-response bias.
Of 179 positive reports collected, only 106 (59%) met their case definition criteria. This further reduced their number of cases but will have contributed to their diagnostic accuracy. A further five cases were excluded as non-UK residents but where did they acquire their infection—in the United Kingdom or overseas?
The incidence figures calculated for England and Wales for 1997-8 and 1998-9 of 0.21 and 0.18 per 10 000 contact lens wearers (CLW) are thus an underestimate of the expected true infection rate. The expected true figure needs to be increased by a factor of 1.493 (100/67) to account for the missed cases with the low reporting rate used of 67%. The English figures then become 0.31 (0.21 × 1.493) and 0.27 (0.18 × 1.493) per 10 000 CLW respectively, of whom the majority (88%) wore soft hydrogel lenses.
These latter (corrected) figures for England and Wales are remarkably similar to the true figure estimated at the same time in a 17 month prospective cohort study in Hong Kong3 of 0.33 per 10 000 “soft” hydrogel CLW. Reporting “questionnaire” studies of infectious disease, albeit prospective and of new cases only, usually underestimate the true incidence figure by at least twofold and that used by the authors (BOSU) is no exception. Incidence figures for rare diseases are best estimated by performing a prospective cohort study, in which all cases occurring within a community of two million to three million people over a defined period (often a year or more) are examined and tested in several referral hospitals according to defined criteria in advance.
Secondly, the authors suggest that there is a relation between water hardness and acquisition of acanthamoeba keratitis, on the basis of correlation of their questionnaire reports with water quality. They suggest a higher risk in areas of hard water but such findings are not supported by our results. In the west of Scotland, where there is very soft water (mean value for calcium carbonate 15.22 mg/l), we recorded the highest incidence ever of acanthamoeba keratitis of 1.49 per 10 000 “soft” hydrogel CLW in 1994-5 in a prospective cohort study.4 In Hong Kong,3 with an incidence of 0.33 per 10 000 CLW, the water is also soft (mean value for calcium carbonate 40 mg/l).
We believe that the suggested correlation by the authors of acanthamoeba keratitis with water hardness is spurious and demonstrates the danger of projecting “correlation” to “causation” within their study. The authors do not have definitive evidence to support their case because of sampling bias that results in part from evaluating associations between risk factors and a disease well after the risk exposure first occurs. During that time interval, cases of short duration or cases mild in severity or asymptomatic, and cases in which the presence of disease alters or entirely removes the exposure, are missed.5
Water “hardness” remains unproved as a risk factor for acanthamoeba keratitis being no more than an observational quirk. While we have had good molecular based evidence for the association of acanthamoeba keratitis with home tap water,6 recent work using 18S rDNA typing and tracking has shown that there can be different subtypes of keratitis causing strains of Acanthamoeba spp (T4) between the tap water in the home and the cornea, making the epidemiology more involved.7,8
In conclusion, the incidence of acanthamoeba keratitis reported for England and Wales by the authors is an underestimate of the true figure by at least 33%. The finding that the reporting of cases was “static” is of concern as it means that there is still need to reinforce ways of preventing this infection.9–11 While considerably less frequent since the introduction of multipurpose solutions 7 years ago,12 the infection remains a problem to be avoided by CL wearers as predicted by one of us 8 years ago.13 If good hygiene is practised, however, with monthly changing of storage cases, avoidance of tap water, and use only of sterile multipurpose solutions or overnight storage in hydrogen peroxide, then the risk of infection becomes negligible. Use of true daily disposable lens wear has also been virtually free of acanthamoeba infection.