Many papers on retinoblastoma open with a statement that it is the most common primary intraocular cancer in children and a rare disease. True enough. After reflecting a little longer on the epidemiology of eye cancer, however, this statement probably should be revised to read: “retinoblastoma is the most frequent primary intraocular cancer and is gaining importance rapidly.” Let us substantiate this claim and evaluate its consequences in some detail.

First, just how frequent is retinoblastoma? Not an easy question to answer because investigators are estimating its incidence in many ways. A recent issue featured two papers on the subject from the Western hemisphere. MacCarthy and coworkers,1 who analysed data from 1601 children with retinoblastoma diagnosed in England, Scotland and Wales between 1963 and 2002, arrive at an annual incidence rate of 3.5 per million children less than 15 years of age. Broaddus and colleagues2 from the USA estimate the rate to be 11.8 per million children below 5 years of age, based on 658 entries in the Surveillance, Epidemiology, and End Results (SEER) programme database from 1975 to 2004.

MacCarthy and coworkers also provide a cumulative incidence estimate of 53 per million for the first 14 years of life, which is useful because by that age, essentially all retinoblastomas have been detected, and they report that 95% of tumours were diagnosed before the age of 5 years.1 Presuming that this percentage is constant, one can calculate the corresponding estimate to be 62 per million for the SEER study. Although they are of the same order of magnitude, the estimates, or any estimate based on the number of children of particular age, will be influenced by differences in child and adolescent mortality between countries, the more so the wider is the age cohort considered.

Because retinoblastoma is a developmental cancer, it is not obvious that the population at risk are the children who are living at any given moment.3 Unnecessary variation likely is introduced by different growth rates of tumours that potentially have initiated even before birth. An informative Northern European study from Sweden and Finland, which have comprehensive national cancer registries, explicitly shows that retinoblastoma incidence rates based on age cohorts are the most volatile; the most stable ones cumulate children to their birth cohorts.3 This method may minimise the effect of variable age at diagnosis of developmental cancers. The data showed that apparent trends in incidence rates disappeared in the birth cohort analysis.3 Cancer databases do not yet always allow direct birth cohort analysis of incidence rates.2

The Nordic study also cautioned against drawing conclusions from analyses based on a short time span and a small population.3 As an example, the high Malawian incidence rate of 21 retinoblastomas per million children younger than 5 years of age4 was based on data collected over just 1 year from a population of 4.9 million people. Because of chance, the annual incidence rate exceeded 20 per million also four times in four decades in Finland, which has a population of 5.2 million.3 Indeed, epidemiological data suggest that the incidence of retinoblastoma is uniform across populations, ranging from one in 16 000 to 18 000 live births.2 3 If we accept this, retinoblastoma can be estimated to develop in 7202 to 8102 children in the world annually (table 1).

It is less straightforward to estimate the annual number of uveal melanomas, because the risk heavily depends on ethnicity5 6 and, apparently, latitude.7 8 Uncertainty is introduced by extensive population migrations and intermarriage, and great ambiguity regarding the percentage of different ethnic groups in the world population. A tentative estimate can be based on two studies which estimated the incidence rates of uveal melanoma in Europe, Asia and Africa5 and in citizens of these origins in the USA,8 and on rough percentages of major ethnic population groups. The number of uveal melanomas is thus estimated to range from 6679 to 7095 yearly (table 1). This makes it as common as retinoblastoma at most and, more likely, puts it on the second tier. The future definitely is in the latter direction. Although the white population in Europe and Northern America is ageing, which will increase the number with uveal melanoma, this population is stable or decreases, whereas African and Asian populations are growing rapidly.

Second, how often do these cancers kill? Long-term mortality from uveal melanoma slightly exceeds 50%,9 and because it appears to be both early metastasising and resistant to available treatments once disseminated,10 11 it is unlikely that we can materially reduce mortality in the near future. Given recent progress in targeted therapy of hepatic metastases of uveal melanoma, one might roughly estimate that 50% or 3339 to 3544 people die of this cancer annually (table 1). The number dying will increase moderately because the world population is growing and, as mentioned, also getting older in developed countries.

Hopes of reducing deaths from eye cancer definitely focus on retinoblastoma. MacCarthy and coworkers,12 in their second paper in the same recent issue, report truly excellent 5-year survival rates of 97% for 996 English, Scottish and Welsh children with unilateral and 100% for 580 children with bilateral retinoblastoma between 1998 and 2002. Given that 63% of children had unilateral retinoblastoma,12 the overall 5-year survival rate is calculated as 98%. Broaddus and colleagues13 in their companion paper estimate the overall 5-year survival rate of 992 children with retinoblastoma in the USA to have been 96% between 1995 and 2004.

The survival issue is also addressed in a concurrently published study by Masood and coworkers,14 who report an overall 5-year survival rate of 83% among 105 children with retinoblastoma treated from 2000 to 2007 in Iran. Corresponding figures for children with unilateral and bilateral tumours were 94% and 85%, respectively. The figures parallel overall survival rates of 80% to 89% in more developed Latin America.15 16 17 They document that establishment of tertiary referral centres for eye cancer can improve survival rates, compared with statistics that range from 48% to 81% survival in India and China18 19 20 and from less than 20% to 46% in Africa.21 22

Based on these mortality data and birth rates in corresponding continents, one can estimate that 3001 to 3376 children die of retinoblastoma annually; most deaths are fairly equally divided between Africa and Asia (table 1). The number dying is rather similar to that estimated for uveal melanoma. If we were able to refer all children with retinoblastoma to tertiary care centres similar to those in Iran,14 mortality would be reduced by 62%, and the number dying would drop to approximately 1200 per year. In the event that every continent could reach survival rates that are achieved today in Europe,12 Northern America13 and Japan,23 mortality from retinoblastoma would decrease by a huge 88% and level at around 400 deaths per year.

The study by Masood and colleagues14 suggests that the former goal is realistic. The challenge lies in improving awareness and knowledge of eye cancer and in enhancing compliance. Several projects have recently been initiated toward this direction: a retinoblastoma education programme linked to a national vaccination campaign in Honduras reduced the incidence of extraocular retinoblastoma from 73% to 35%.24 However, it was not successful in improving compliance: more than one-third of families refused therapy or abandoned treatment before and after implementation of the programme.24 Similar projects are under way in South American, Africa and Asia,25 many in partnership with European and Northern American centres, institutions and charities such as the Daisy’s Eye Cancer Fund in UK, which collaborates with Kenya.

The estimates tabulated do not take into account deaths from second cancer in patients with retinoblastoma, which are a problem in those with hereditary disease. MacCarthy and colleagues26 in their third paper in the present issue (see page 1159) estimate the 25- and 50-year risk of developing such a cancer to be 10% and 48%, respectively. However, they did not adjust for deaths from competing causes, making these overestimates.26 In case of uveal melanoma, taking competing risks into account reduces the 25-year estimate for death from second cancer from 12% to 6%, and the corresponding 35-year estimate from 23% to 7%.9 Fewer competing events affect younger patients, but one should nevertheless read with caution this part of the study by MacCarthy and coworkers.

Admittedly, the obstacles to reduce overall mortality from retinoblastoma are huge, but the rewards will also be tremendous. We are genuinely grateful for investigators who continue to provide up-to-date epidemiological data on retinoblastoma from different parts of the world. They are a key force driving progress to meet the challenge of conquering death and blindness from this cancer, which already is the single most frequent intraocular malignancy and the one which grows most rapidly in number.