Here, we report an investigation of the epidemiology and aetiology of conjunctival squamous cell carcinoma, examining over 50 risk factors for the tumour. Conjunctival cancer affects relatively young individuals of both sexes and is strongly associated with HIV infection and poverty. In addition, the risk of the disease decreases with decreasing age at which an individual leaves home and with increasing time spent cultivating, although both results are of borderline statistical significance. Furthermore, significance tests were conducted on over 50 risk factors, greatly increasing the likelihood of significant results arising by chance alone. The role of HPV-16, -18 and -45 requires further evaluation, but KSHV does not appear to be associated with conjunctival carcinoma.
The study reported here is subject to the potential problems of incomplete diagnostic verification and HIV testing. Laboratory verification of cancer diagnosis (for example, by histology, cytology or blood chemistry) was available on 60% of cases (all of which were invasive tumours) and 63% of controls. Typical of studies in developing countries where laboratory services are limited, this proportion compares favourably with other cancer series reported from Africa (Bassett et al, 1995
; Newton et al, 1996a
; Wabinga et al, 2000
). Another potentially important source of bias is the high proportion of adults who were not tested for HIV in the early stages of the study. However, the age and sex-specific seroprevalence rates of HIV infection among those tested is broadly similar to that found in studies of HIV seroprevalence in Uganda (STD/AIDS Control Programme, Uganda, 1997
). Also, the introduction of a saliva test resulted in a dramatic reduction in refusals, but little change in HIV seroprevalence rates, suggesting that patients who were refusing to have a HIV test were unwilling to have blood taken, rather than unwilling to learn their HIV serostatus. The association of conjunctival cancer with HIV infection has been reported in a general paper on HIV and cancer among residents of Kampala, and the odds ratio was the same as that from the current study of all patients who were seen at hospitals in Kampala (Newton et al, 2001
). A proportion of individuals in the study were also tested for antibodies to three HPV subtypes and to KSHV infection. This additional testing occurred in all those for whom enough stored sera were available and (as people are unaware of their own status with respect to infection), is therefore an unbiased sample.
Another potential source of bias arises from the fact that most (but not all) cases had been identified from out-patient clinics, which tend to draw patients from the local area, whereas controls had been hospital in-patients, who are from a wider area. Despite adjusting for region of residence in the analyses, the possible impact of this on the results, is difficult to assess.
Case reports of squamous cell carcinoma of the conjunctiva in human immunodeficiency virus (HIV) infected men in the USA and France, coupled with a marked increase in the numbers of tumours being seen by ophthalmologists in at least two African centres (which mirrored the increases seen for Kaposi's sarcoma), led to the suggestion of an association with HIV (Winward and Curtin, 1989
; Ateenyi-Agaba, 1995
; Kestelyn et al, 1990
; Kim et al, 1990
; Denis et al, 1994
). Among those with HIV, the lesions often affect young adults, in a fashion reminiscent of Kaposi's sarcoma in HIV-seropositive people (Ateenyi-Agaba, 1995
; Kestelyn et al, 1990
; Waddell et al, 1996
). This study, together with others from Africa and the USA indicate about a 10-fold increased risk of the tumour in HIV infected, compared to HIV uninfected individuals (Ateenyi-Agaba, 1995
; Kestelyn et al, 1990
; Goedert and Coté, 1995
; Newton et al, 1995
; Waddell et al, 1996
). Indeed, the spread of HIV in Uganda probably accounts for much of the approximately eight-fold increase in incidence of conjunctival carcinoma observed there since the 1960s (Wabinga et al, 2000
). Using standard equations for case–control studies, the population attributable fraction is about 60%: almost two-thirds of cases would not occur in the absence of HIV infection (dos Santos Silva, 1998
). However, the frequency of conjunctival carcinoma is not such that it is yet a particularly common manifestation of HIV disease (Piot et al, 1992
The mechanism whereby HIV infection increases the risk of conjunctival cancer is not clear. There is no evidence that it is directly oncogenic and the impact on cancer risk is most likely to be mediated via immunosuppression, as is the case for other HIV-associated cancers (Beral and Newton, 1998
; Newton et al, 1999
). It is not known if the risk of conjunctival squamous cell carcinoma is increased in other immunosuppressed groups, such as transplant recipients, because the tumour is very rare in parts of the world where tissue transplantation occurs. However, there is a case report of a conjunctival cancer in a patient with malignant lymphoma on immunosuppressive chemotherapy (Kushner and Mushen, 1975
). In general, HIV infection is thought to increase cancer risk by facilitating the action of other oncogenic viruses, such as KSHV, the principal cause of Kaposi's sarcoma. In relation to conjunctival cancer however, we found no evidence in these data of an association with KSHV.
Several types of squamous carcinoma are associated with human papillomavirus (HPV) infection, most notably cancer of the uterine cervix, induced primarily by HPV-16, -18 and others. Squamous carcinoma of the skin has also been associated with HPV-5 and -8 in immunosuppressed individuals (Iarc, 1995
). Evidence for an association between human papillomavirus and squamous cell carcinoma of the conjunctiva is conflicting, although bovine papillomaviruses are thought to cause conjunctival carcinomas in cattle (Iarc, 1995
). The presence of human papillomavirus DNA (HPV; predominantly types 16, but also other types) in human ocular surface squamous neoplasias, including invasive carcinomas, has been reported in some studies, but not others (reviewed by Newton, 1996
). This is the first study to look for an association between anti-HPV antibodies (types -16, -8 and -45) and the risk of conjunctival carcinoma. The seroprevalence of antibodies against HPV-18 and -45 was too low to make reliable conclusions. Results for anti-HPV-16 antibodies were suggestive of an association among individuals with high titres, but the data presented here are too few to draw valid conclusions and studies of larger numbers of cases are required.
The HPV assays used in this study are based on the expression of L1 major capsid proteins of HPV-16, -18 and -45 in insect cells by using recombinant baculoviruses (Touzé et al, 1998
). The resulting virus-like particles (VLPs), which appear similar to empty virions, can be used in serological studies to test for type specific immunological responses to viral capsid proteins, although there is evidence that a particular assay may cross react with related HPV subtypes (Combita et al, 2002b
). Presence of anti-VLP antibodies is an indicator of past and current infection (Kirnbauer et al, 1994
; Le Cann et al, 1995
; Wideroff et al, 1995
; Dillner et al, 1996
). The utility of such assays has been demonstrated in previous studies of anti-HPV-16 antibodies in relation to the risk of cancer of the uterine cervix (Lehtinen et al, 1996
; Dillner et al, 1997
; Shah et al, 1997
; Vonka et al, 1999
; Hisada et al, 2001
). Results from this study on the relationship between anti-HPV-16 antibodies and the risk of cancer of the uterine cervix are broadly comparable to those reported before and will be the subject of a separate report.
Although relatively rare everywhere, conjunctival carcinoma is more frequent in sub-Saharan Africa and other tropical areas than in temperate countries, leading to the hypothesis that it may be associated with exposure to solar ultraviolet radiation (Templeton, 1973
; Clear, 1979
). Geographical studies support this premise, in that the incidence of squamous carcinoma of the eye increases as exposure to ambient solar ultraviolet radiation increases. Levels of solar radiation are higher towards the equator and the incidence of conjunctival cancer increases by about 50% for each 10° decline in latitude (Newton et al, 1996b
; Sun et al, 1997
). In addition, a single case–control study reported that the risk of ocular surface epithelial dysplasias is greatest in those reporting a past history of skin cancer, which is known to be caused by exposure to solar ultraviolet radiation (Lee et al, 1994
). Time spent cultivating may indicate the time an individual spends in direct sunlight, and so the increasing risk with increasing time cultivating may reflect exposure to ultraviolet radiation. This variable may also reflect exposure to dust or dirt, and it has been hypothesised that ocular trauma could facilitate development of the tumour (Templeton, 1973
; Margo and Groden, 1986
). The lower risk among people who leave home at an earlier age may reflect migration to towns for work, where exposure to solar ultraviolet radiation could be less. The extent to which the higher risk among those of low income is measuring the previously mentioned exposures is not clear. The fact that cases are less likely than controls to have had a blood transfusion is probably an artefact of the control selection. Cases were generally seen as out-patients, whereas controls with other cancers were generally hospital in-patients. The latter may therefore be more likely to have had a transfusion as part of their medical care.
In summary, the current study gives independent support for the strong epidemiological evidence that solar ultraviolet radiation is an important cause of squamous cell carcinoma of the conjunctiva. Another established risk factor is HIV infection, although the mechanism whereby it increases the risk of conjunctival cancer is not clear. We find little evidence supporting the possible role of sexually transmitted forms of HPV in the aetiology of conjunctival cancer, but larger studies are required.