Authors' reply

EDITOR, - Louis Clearkin doubts the value of treatment for glaucoma. In line with most ophthalmologists we favour it: plenty of good quality research has proved that treatment of chronic glaucoma slows or halts its natural progression over many years to blindness.1,2 Clearkin supports his argument with reference to two studies. The first found that only one of 200 patients who underwent surgery for glaucoma became blind before death or during the 20 years of follow up3; we interpret this as successful preservation of vision. The second was a randomised controlled trial comparing medical treatment with no treatment in 15 patients followed up for between 12 and 36 months. In relation to the natural course of untreated chronic glaucoma1 the size and follow up in this trial were balefully inadequate; additionally, the trial's statistical ability to detect a difference (power) was vanishingly small. Anything other than a negative result was therefore unlikely.

John M Gardner, Graham P Kirkby, and K D Phillips each suggest that the referral data for 1988 should have been omitted from our analysis. In doing so they make a fundamental mistake of scientific logic. Any time series is bound to fluctuate around the underlying trend, and omitting a year from such a series cannot be justified unless that year has first been shown to have been exceptional; the authors did not test this hypothesis.

We did test this hypothesis and showed that, in terms of the numbers of referrals to Bristol Eye Hospital, 1988 was not an exceptional year, regardless of the number of sight tests being performed in the community. Having established this, we included 1988 in our predictions. Excluding the data for 1988 roughly doubles the confidence intervals for the predicted number of referrals from 1989 onwards, which artificially and unjustifiably leads to a fallacious interpretation of the available data as not significant. The same arguments apply to the omission of the data for 1987 and 1988 from trend predictions, which Janet Pooley and colleagues proposed.

Peter J Gray and K W Pullum conjecture about the possible aetiologies of the proportionally equal reduction that we observed in both the total number of referrals and the number of true positive glaucoma referrals to Bristol Eye Hospital from 1989 onwards. Any appreciable improvement in optometric screening should have maintained the number of true positives while allowing the total number of referrals to fall. We clearly showed that both total referrals and the number of true positives fell in equal proportion, thereby disproving these authors' hypothesis. Gray also suggests that general practitioners may have referred patients elsewhere from 1989 onwards, but in Bristol there are few practical alternatives to the eye hospital.

Finally, Gray contends that screening for glaucoma is required only in the population aged over 40 with a family history of the disease. This is both incorrect and misleading; if implemented such a policy would miss two thirds of people with glaucoma.4