Article Text

Download PDFPDF

A systematic review of drug induced ocular reactions in diabetes
  1. J P Hampson,
  2. J N Harvey
  1. University of Wales College of Medicine, Gwenfro Academic Unit, Wrexham Technology Park Centre, Croesnewydd Road, Wrexham LL13 7YP
  1. J P Hampson


AIMS To conduct a systematic review of drug induced adverse ocular effects in diabetes to determine if this approach identified any previously unrecognised adverse drug effects; to make a preliminary assessment of the feasibility of this approach in identifying adverse drug reactions; and to assess the current accessibility of this information to prescribing physicians.

METHODS Literature search of online biomedical databases. The search strategy linked eye disorders with adverse drug reactions and diabetes. Source journals were classified as medical, pharmaceutical, diabetes related, or ophthalmological. It was determined whether the reactions identified were recorded in drug datasheets and theBritish National Formulary.

RESULTS 63 references fulfilled the selection criteria, of which 45 were considered to be relevant to the study. The majority of these were case reports but cross sectional surveys, case-control and cohort studies, and review articles were also identified. 61% of the reactions were not recorded in the British National Formulary and 41% were not recorded in the datasheets. 55% appeared in specialist ophthalmology journals.

CONCLUSIONS This is a feasible approach to the identification of adverse drug reactions. Adverse reactions not listed in the most commonly used reference sources were found. The majority were published in specialist ophthalmology journals which might not be seen by prescribing physicians.

  • adverse drug reactions
  • eye
  • diabetes

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Diabetes is associated with a variety of ocular manifestations and is a major cause of visual impairment.1 2 Thus, the recognition or anticipation of adverse drug reactions involving the eye is important. However, drug induced ocular side effects are uncommon. Measurements of prevalence are difficult to do and precise estimates are generally unavailable. The data which do exist rely on the astute physician making a connection between an eye condition and the patient's systemic therapy. In the USA, data on all drug induced ocular side effects can be obtained from the National Registry of Drug Induced Ocular Side Effects.3 It is postulated that, as a result of this registry, ocular adverse effects are identified earlier and patients protected. In the UK, to the authors' knowledge, there is no such equivalent. The Committee on Safety of Medicines (CSM) holds data on all adverse drug reactions, but not specifically on those related to the eye. General physicians must rely on their personal knowledge, official warnings, and the range of journals which they read regularly to help them identify and manage suspected reactions. Two of the most readily accessed information sources are theBritish National Formulary(BNF) and the Summary of Product Characteristics (SPCs), formerly known as datasheets.4 5However, the possibility that the majority of ocular reactions are reported only in specialist ophthalmology journals reduces the chance of general physicians being aware of some important adverse drug effects.

The purpose of this study was to systematically review the world literature on drug induced ocular side effects in diabetes. Ultimately, a list of drugs and side effects would be produced to aid general physicians in identification of such reactions. This systematic review strategy has not previously been applied to adverse drug reactions and thus a secondary aim was to assess the feasibility of this approach to drug reactions in a clinical area of manageable size. We also wanted to determine whether these adverse reactions were to be found in the standard reference sources—that is, the BNFand SPC, and to look at the type of journals where they were reported in order to assess accessibility to this kind of information.



An online literature search of the databases Medline, Embase, Biosis, Toxline, Pharmline, IOWA, and International Pharmaceutical Abstracts was performed towards the end of June 1998. The search strategy (available from the authors) principally linked eye disorders with adverse drug reactions and was narrowed down by adding the term “diabetes”. Criteria for selection of titles produced by the search included the presence of a commonly prescribed drug name and ocular effect in the title. Animal studies were excluded. Titles fulfilling these criteria were downloaded and the full reference obtained where possible. In a small number of cases, for foreign journals, the English abstract was used. If no English translation was available the article was excluded. On scanning the full reference the corresponding author of any article thought to be of value published in the past 10 years was contacted to determine whether they had additional data on similar or other work (published or unpublished). A search of the Cochrane Database was also performed.6 Data from every relevant article were extracted in a systematic format. Data extracted included study type, drug prescribed, details of adverse reaction and population, randomisation, study blindness, inclusion/exclusion criteria, statistical significance and power, follow up, and generalisability. Using this information, a comprehensive list of drugs and their suspected effects on the eye was produced.


The journal type from all of the articles used in the review were categorised as medical (general or specialist), pharmaceutical, diabetes related, or ophthalmological. The proportion of each type was calculated.


Using the list above, the side effects were crosschecked in theBNF and SPC. Each adverse reaction was classified as (A) present in both sources, (B) partial agreement, or (C) present in review list only. Partial agreement was defined as when the BNF or SPC contained a phrase which did not precisely define the reaction in question—for example, visual disturbance was in partial agreement with short sightedness.


The electronic search produced a total of 528 titles for further assessment. Sixty three references were identified using the selection criteria, and 45 full references were eventually selected as being relevant to the study. These included four articles available in abstract only and six review articles.7-12 The majority of studies were case reports (Table 1).13-26 The other articles consisted of cross sectional, cohort, and case-control studies (Table 2).27-43

Table 1

Case reports of drug induced ocular toxicity

Table 2

Drug induced ocular side effects: case-control, cohort, and cross sectional studies


References were included in the review when the subjects were diabetic or where diabetes was considered to be one of the confounding factors. Despite “diabetes” being one of the search terms, it was sometimes not clear whether the patient groups did include diabetes. For the purposes of the review, this small number of studies was included where the drug reaction was also mentioned by at least one other study which did include diabetes as above. Table 3 summarises the reactions obtained, listed according to drug or drug group. In addition, the review articles mentioned a number of side effects not specifically occurring in diabetes and included reduced visual acuity with cisplatin and cytarabine, papilloedema with ketoconazole, and retinopathy with chloroquine.8 10

Table 3

Summary of reported drug induced ocular reactions


Of the 45 relevant journals, seven (15.6%) were specialist medical, 11 (24.4%) were general medical, 25 (55.6%) were ophthalmological, with one (2.2%) pharmaceutical and one (2.2%) diabetes related journals. Thus, over half of the journals would, most probably, be read by ophthalmologists only.


Table 4 displays the results of the cross check between the drug list and the BNF/SPC. Fewer side effects were reported for the datasheet because certain products were not listed in the SPC. OKT3 was not listed in either theBNF or SPC.

Table 4

Comparison of side effects obtained from literature search with those found in BNF and datasheet


This literature search identified studies of all types. These are discussed in relation to ocular structure affected.


Cross sectional studies suggested that oral contraceptive pill use does not have any effect on diabetic retinopathy.28 29Interferon treatment for hepatitis C, however, does cause retinopathy. Several case reports were identified documenting this. The funduscopic appearances have some features in common with diabetic retinopathy suggesting that interferon retinopathy is also the result of a micro- angiopathy.32 The suggestion was made that the retinopathy was worse in patients with diabetes. This was found to be statistically significant in another study of 63 patients with hepatitis treated with interferon where 11/12 (92%) of patients with diabetes developed evidence of retinopathy although it was asymptomatic in the majority.33 Treatment with intravenous cidofovir for cytomegalovirus retinitis caused iritis in 26% of 43 patients.39 The risk of iritis appeared to be increased in patients with diabetes.

Lakowski and Morton describe colour vision changes that occur with diabetes and also with oral oestrogen usage.27 44Steinberg et al described visual hallucinations in dialysis patients after erythropoietin.31 Risk factors for hallucinations on this treatment included diabetic retinopathy or cataracts. Inhibition of oxidative metabolism in the retina was thought to be responsible for blindness which occurred in a diabetic patient with lactic acidosis following phenformin treatment.46


The effects of glucocorticoids on the eye were examined in a number of studies. Risk of cataract was examined in a large matched cohort study by Isaac et al who found that risk was increased with use of systemic steroids, phenothiazines, antidiabetic agents, and benzo- diazepines.35 It has recently been claimed that use of inhaled steroids is associated with the development of cataracts.12 41 Topical dexamethasone eye drops appear to cause cataracts.43 The relation between cataract and allopurinol use has also been examined. Studies by Liuet al and Leske et al did find a relation while in the study by Clairet al no significant increase in odds ratio was observed.36-38 The presence of diabetes appears to increase the risk.

Lightman et al suggested that the modern, more potent sulphonylureas cause changes within the lens altering refraction.17 This phenomenon has previously been described by Keller.45 Hyperglycaemia in diabetes results in sorbitol accumulation in the lens, along with other diabetes specific metabolic changes at the cellular level. D'Arcy reported a single case of glibenclamide induced accommodation paralysis and cited older references to sulphonylurea induced myopia.16 It is suggested that there is swelling of the lens and ciliary body with forward displacement of the lens-iris diaphragm. Topical ophthalmic and oral glucocorticoids may cause glaucoma.42

The effect of pupillary dilatation on visual acuity is important in relation to driving. It is now policy in most diabetes units that diabetic patients should undergo mydriasis before funduscopy. Tropicamide is generally used for its short duration of action, but despite this a recent study has shown that a minority of patients have binocular visual acuity insufficient to meet the legal requirement for driving 1 hour after mydriasis.47 Thus patients should now be advised (usually at the previous visit) not to drive themselves after mydriasis. The risk of precipitating acute glaucoma is considered small enough to be acceptable. Of the various possible adverse outcomes of pupil dilatation, recent experience is that the alternative—missing sight threatening retinopathy—carries the greatest medicolegal risk.


Wymore and Carter described a case of optic neuropathy with chlorpropamide which recovered with withdrawal of the drug and cited two other examples from the literature.15 This adverse effect is not mentioned in the BNF. Elsewhere, Sedwick discusses the possible relation between amiodarone induced and ischaemic optic neuropathy in a diabetic patient.19 Optic neuropathy is well recognised with ethambutol. The search identified a diabetic patient with tuberculosis in whom isoniazid was thought to be the predominant factor.25 These authors cite 13 other cases of optic neuropathy due to isoniazid.


Several reports described vitreous haemorrhage in patients given warfarin, streptokinase, or recombinant tPA.23 24 Since these treatments are commonly used, generally without detailed retinal assessment, this adverse effect is likely to be substantially more common than generally realised. Physicians perhaps do not always appreciate that haemorrhage can occur despite laser treatment if the new vessels have not fully regressed. One report described bilateral vitreous haemorrhage requiring vitrectomy in a non-diabetic patient following tPA.48 There is evidence that aspirin is not contraindicated in diabetic patients with proliferative retinopathy.7


Greven et al described retinal artery occlusions in 21 patients under the age of 40.30 The majority had one or more risk factors, two had diabetes, and five were taking oral contraceptive agents.

Diabetes is a risk factor for retinal vein occlusion (RVO).49 Kirwan et al found an associaton between oral contraceptive use and RVO in women aged under 35.50 Hormone replacement therapy with lower doses of oestrogens did not appear to be a risk factor. They concluded that RVO is a contraindication to use of oral contraception.


Of particular interest was whether our approach would identify new adverse reactions. This was likely to come from individual case reports. Single examples must be considered unconfirmed. For example, the potential for thiazide treatment to cause oculomotor palsy and cataracts has not been reported elsewhere.13 14 However, chlorpropamide optic neuropathy and lens changes due to second generation sulphonylureas both appeared in more than one report but do not seem to have made it into the reference works. Such reports are by their nature anecdotal but the prevalence of such reactions is likely to be higher than realised. Thus a system of prospective surveillance is needed.


To estimate how much of this information is readily available to physicians we determined how many of them were listed in theBNF and the data sheets. From Table 4, the data sheets were slightly more accurate than theBNF with 45% and 26% respectively in complete agreement with the literature review list. This is not surprising because the datasheet, which outlines the licensed indications for a specific drug, is generally much more detailed than the equivalent BNF entry. The most striking result is that between 41% and 61% of the side effects from the review were not present in the BNF or datasheet. It is uncertain whether this reflects lack of awareness of the report or delay before new side effects are included, or an editorial opinion that the associations between drug and effect was not strong enough to merit inclusion.

Approximately one quarter of all reactions were reported in general medical journals. This means that 75% of the reactions would be unlikely to be read by physicians who are responsible for prescribing the majority of these drugs. Although the reactions cited in Table 3are mostly rare, unless the general population of prescribers are made aware of these suspicions, their true incidence may never be known. One possible solution is to ensure that the BNFshould include more of these reactions. However, as discussed above, this is not without its problems. There is a need, therefore, for an early warning system where prescribers can share their experiences of potential ocular reactions. Perhaps the first stage of this process should involve the setting up of a UK registry of drug induced ocular side effects.


Financial support for this study was provided by the NHS Wales Office of Research and Development.