More women than men take 4AQs because the prevalence of rheumatoid arthritis (RA) and systemic lupus erythematosus (SLE) is higher in women than men [6]. In most series, the proportion of females among cases of retinopathy is 80 % or greater [7–9]. The weighted average percentage of patients who are female is 83 % (Table 5.1). Moreover, the demographic characteristics of patients who take 4AQs are consistent across countries from which case series have been reported [10, 11].

By pooling studies and weighting the reported statistics by sample size, the weighted median age of patients is 54 years. The age distribution of patients taking hydroxychloroquine has been estimated in the literature (Fig. 5.1 and Table 5.2). The ages of treated patients range from 10 to 100 with a peak in the decade 51 to 60.

The list of diseases for which patients take 4AQs is long, but the most common diagnoses are SLE and rheumatoid arthritis (RA). Table 5.3 shows selected data from the literature on this issue. The weighted mean percentages of patients taking 4AQs for RA, SLE, and other autoimmune diseases are 62 %, 29 %, and 9 %, respectively.

The concepts of prevalence and incidence are commonly confused and the terms interchanged [38–40]. The most common error is that authors write about incidence (which implies a number of cases among susceptible individuals over a certain interval of time) when they mean prevalence (a certain number of cases among susceptible individuals at a given time) [39, 40]. The definitions follow.

There are fewer studies of the incidence of 4AQR because an incidence study requires two identical examinations using standardized techniques for detection at two separate times.

There is little controversy that more cases of chloroquine retinopathy than hydroxychloroquine retinopathy have occurred [43, 44]. In 2011, Easterbrook had personally cared for 217 cases of 4AQR, with 200 of them attributable to chloroquine [45]. As a result of the higher prevalence of chloroquine retinopathy and perceived higher risk, prescription of chloroquine has almost vanished in the USA and Japan, although it continues to be widely used in Europe, Mexico, Brazil, Turkey, and China [21, 46–49].

The situation is different with hydroxychloroquine. It has been estimated that more than one million persons have taken hydroxychloroquine, yet the number of cases of retinopathy was reported to be less than 47 in 2006 [50, 51]. There is skepticism that hydroxychloroquine retinopathy occurs often enough to be a public health problem [52–54].

To rationally discuss the issue, good data on prevalence of retinopathy are needed. Yet in 1998, Albert and colleagues wrote “There is no epidemiologically sound study that determines the frequency of ocular toxicity in patients treated with antimalarials; therefore, the incidence, prevalence, and risk of toxicity cannot be determined accurately” [55]. In 2014, this statement remains true. To determine the prevalence and incidence of hydroxychloroquine and chloroquine retinopathy would require a population-based study that could be replicated by others: using standardized examinations, prospective follow-up, and prespecified definitions of hydroxychloroquine and chloroquine retinopathy [41]. It is unlikely that a proper study will ever be done, because the resources required are probably not justified relative to the importance of the information gained, especially in the context of other, more pressing, health care needs. This is part of the reason that screening has intentionally not been recommended in the UK [52, 56–58]. The possible benefit seemed incommensurate with the expense to the panel convened to develop national policy.

In the absence of population-based prevalence estimates, we are left with estimates from observational, retrospective studies and small prospective studies. The prevalences of retinopathy reported from these limited studies for various subsets of patients vary from none to 40 % [3, 13, 14, 16, 17, 20, 35, 43, 47, 50, 58–64]. This uselessly wide range of estimates arises from the different definitions of retinopathy used by different authors (see Chap. 4), different detection methods, differences in risk factor profiles among samples of patients, different stages of retinopathy studied, failure to report cases, publication bias, failure to detect retinopathy by the screening physicians, multiply-counting the same patients with 4AQR over different publications covering the same samples, and failure of patients taking the drugs to comply with screening [21, 43, 58, 65].

Estimates of prevalence depend on sample size. For example, one group reported the prevalence of retinopathy at their center in serial publications over time. As their sample size increased, the reported prevalence decreased from 3.4 to 0.5 %. In this cohort, the sample size at the time of the first report was only 58, but the follow-up report was 400 [14]. Small samples give unreliable point estimates of prevalence.

The risk factors for 4AQR are the drug used (chloroqine or hydroxychloroquine), daily dose adjusted for IBW, cumulative dose (and its surrogate duration of use), age, renal or liver dysfunction, and pre-existing maculopathy (see Chap. 7). Therefore, it is expected that epidemiologic indices of risk would vary in patient samples having different characteristics. The most important variable affecting prevalence is the proportion of patients properly dosed according to IBW. In patients taking doses of hydroxychloroquine less than 6.5 mg/kg/day the prevalence of retinopathy has been estimated to be from 0 to 0.5 % [14, 20, 51, 54, 66–68]. Hydroxychloroquine toxicity at the recommended daily dose adjusted for IBW does occur but is so rare that the only documentation is in the form of isolated case reports [40, 69–72]. At this low rate the concept of prevalence becomes meaningless because the numerator is so small and the denominator so large. For example, Levy and colleagues found no cases of retinopathy in the subgroup of patients taking a nontoxic dose (N not reported) out of a larger sample of 1,207 patients taking hydroxychloroquine within the Kaiser Permanente health care system in California [20]. Mavrikakis found two cases out of 360 patients taking nontoxic daily doses [73]. Perhaps because of this, the index that has been reported more often in the literature has been the total number of cases of hydroxychloroquine retinopathy that have been published. This number was four in 1999 [72] and 47 in 2005 [74]. It follows that a program emphasizing proper dosing of 4AQs based on IBW would be expected to reduce the rate of 4AQR in a cost-effective manner.

The prevalence of chloroquine retinopathy based on daily dosing adjusted for IBW or, if that is not available, for actual body weight (ABW) is less well defined. In one series of patients with chloroquine retinopathy, 7.4 % of those patients with chloroquine retinopathy were dosed at less than 3 mg/kg of ABW/d, but 40 % were taking 3–4 mg/kg (ABW)/d [72]. Different series having different rates of overdosing by weight will be expected to show different prevalences of retinopathy. Although good estimates of prevalence of chloroquine retinopathy are lacking, the available data show that the prevalence is higher than that of hydroxychloroquine retinopathy.

Prevalence of 4AQR would also be expected to depend on cumulative dose or its surrogate, duration of therapy. The prevalence has not been examined in this way. A related epidemiologic index, the hazard function, has been modeled as a function of these variables [16]. The hazard function gives the rate of occurrence of 4AQR at a given value of the cumulative dose or duration of therapy subject to the condition that the 4AQR has not yet occurred. Wolfe and colleagues found that the hazard function for retinopathy increased for cumulative doses of hydroxychloroquine between 750 and 1,500 g or durations of use from 5 to 10 years (Fig. 5.2) [16].

Referral bias is a problem in estimating prevalence of 4AQR. One would expect to see a higher prevalence in patients seen in a retina referral clinic than in a rheumatology clinic, which in turn would be expected to be higher than in a population-based surveillance study of patients taking 4AQs. Table 5.4 lists the common sources of patients studied. Lastly, a problem tending to inflate estimates of prevalence is the occurrence of the same case in multiple publications, which can be difficult to track [37].