Epidemiology of ocular surface squamous neoplasia in A frica

1Ocular surface squamous neoplasia (OSSN) is the most common ocular surface tumour (Grossniklaus et al. ). Other synonymous terms include ‘conjunctival epithelial neoplasia’, ‘ocular surface epithelial dysplasia’ and ‘conjunctival squamous cell neoplasia’ (Lee & Hirst; McDonnell et al.; Tulvatana ). OSSN covers a spectrum of disease ranging from non-invasive intra-epithelial dysplasia of the conjunctiva and cornea (CCIN) to invasive squamous cell carcinoma (Lee & Hirst ).

2The disease may present with irritation, red eye, raised gelatinous mass and leucoplakia (Tunc et al. ). In Africans, it is often pigmented brown (Figure1). OSSN is usually unilateral (Chisi et al. ) and arises at the limbus – the junction between the cornea and conjunctiva (Lee & Hirst ). Most lesions occur within the exposed part of the eyeball between the lids (Ateenyi-Agaba; McKelvie; Waddell et al. ). Up to 31.2% of cases seen are recurrent lesions (Chisi et al. ). Late stages present with a large fungating oculo-orbital mass (Ogun et al. ). Early lesions resemble benign growths such as pterygia and pingueculae. OSSN can be the first manifestation of HIV infection in about 50% of cases in HIV-endemic settings (Porges & Groisman; Spitzer et al. ).

3Histologically, OSSN may be classified into 3 forms: benign, pre-invasive and invasive (Table 1; Basti & Macsai ). The term OSSN usually excludes the benign forms. The term ‘invasive’ indicates infiltration through the basement membrane of the conjunctival epithelium into the underlying stroma (Basti & Macsai; Shields & Shields ).

4Two disease patterns of OSSN are recognised: older, predominantly male in temperate climates, not associated with HIV or human papilloma virus (HPV); and younger men and women, in tropical climates, associated with HIV and HPV. The latter represents a public health challenge in Africa in relation to the HIV pandemic and late presentation of large tumours (Ukponmwan et al.; Chisi et al.; Ogun et al. ), diagnostic difficulties (Furahini & Lewallen ), malignant transformation and high recurrence rates after treatment (1-year recurrence of 16.6% reported in Tanzania; Makupa et al. ). Experienced surgeons report lower recurrences (3.2%) after excision (Waddell et al. ). Trial data to guide management in this context are lacking (Gichuhi & Irlam ). For the temperate pattern of disease, one randomised controlled crossover trial in Australia compared mitomycin-C with placebo in participants whose average age was 67 years (Hirst ). There was a significant treatment effect on clinically assessed complete resolution of lesions (P = 0.0005), but no effect on histologically assessed complete resolution (P = 0.49).

5Incidence estimates for OSSN are difficult to ascertain and vary regionally (Table 2). The first paper to examine this used cancer registry data from International Agency for Research on Cancer (IARC; Newton et al. ). A subset of these data were used in a subsequent publication looking at variation in incidence across the USA (Emmanuel et al. ). However, published results need to be interpreted with caution – firstly, all eye cancers are classified together by the International Classification of Diseases for Oncology (ICD-O-3 C.69) while other databases classify squamous cell carcinoma of the conjunctiva (SCCC) with head and neck cancers (Lee et al.; Curado et al.; Parkin et al. ). OSSN is not recognised as a separate entity. Squamous cell carcinomas that are site-coded for the eye (C69) probably include some cancers that originate in the eyelid skin (WHO,; Curado et al. ). Secondly, the availability of histopathology services to confirm OSSN diagnosis is often limited in low- and middle-income countries (Furahini & Lewallen ). Thirdly, health information systems tend to capture invasive squamous cell carcinoma (SCC) but not earlier stages. Countries reporting higher rates of SCC (mostly in Africa) only started sending cancer registry data to IARC in the mid-1980s (Curado et al. ). Completeness of the current IARC database is hampered in that only data from 80 countries were submitted, of which 75% was of acceptable quality, and not all countries had data on squamous cell carcinoma in the eye under code C69. Africa had the lowest level of acceptable quality of data (36%). Fourthly, crude incidence rates can be influenced by population structure, a problem often addressed by reporting age-standardised incidence rates. Finally, in areas with limited health facilities for cancer treatment where a large number of patients are treated outside the reference area, incidence may be underestimated. Moreover, in defining incidence from different sources, it may be difficult to distinguish between recurrence or extension of an existing cancer on one hand and the development of a new primary on the other. Analysis of incidence time trends is also difficult if geographical coverage, ICD revisions and disease definitions in a registry change.

6Systematic and non-systematic review methods were used. No a priori systematic review protocol had been published. Incidence data were obtained from the current IARC report (9th Volume) covering the period 1998–2002. The IARC collates data from cancer registries worldwide. The report uses ICD codes to show the age-standardised incidence per 100 000 population stratified by sex and histological type. Under code C.69 where eye cancers are reported, the four main groups are retinoblastoma, malignant melanoma, carcinomas (11.4% of all eye cancers), sarcoma and other unspecified tumours. Under carcinomas, there are three subgroups – SCC (principally tumours of the conjunctiva and cornea, comprising 70% of the carcinoma subgroup), other specified carcinoma (adenocarcinomas of the lacrimal gland and lacrimal duct) and unspecified carcinomas. We extracted data from the SCC subgroup. The coordinates locating each registry were obtained from http://itouchmap.com/latlong.html.

7We searched PubMed, EMBASE and Web of Science for systematic reviews, meta-analysis and case–control studies using ‘OSSN’, ‘conjunctival squamous cell carcinoma’, ‘risk factors’ and their synonyms as key words with no language restrictions. Abstracts were assessed and studies were selected if they reported analysis of known or suspected risk factors. The search was conducted on 2 January 2013 and updated on 31 May 2013. Data were extracted from the full texts of articles and additional articles obtained from their reference lists. Meta-analyses were conducted where appropriate. A fixed-effects model was used for HIV and cigarette smoking. A random-effects model was chosen for HPV after investigation of heterogeneity.

8Africa has the highest age-standardised incidence rate of ocular SCC followed by Central and South America then Oceania (Australia, New Zealand and Hawaii), respectively (Table 2 and Figure 2). The rate in Africa is about 9–10 times higher than in Europe and North America. The highest incidence rate is 3.4 cases/year/100 000 among males and 3.0 cases/year/100 000 among females in Zimbabwe (Curado et al. ). Uganda follows with 1.6 cases/year/100 000 for males and females. Australia comes third with 0.3–0.5 cases/year/100 000 in parts of that country. Other countries have rates between 0 and 0.1 cases/year/100 000. The rates have a right-skewed bell-shaped distribution peaking at latitude 16° South (Figure3). Incidence rates are higher in the Southern Hemisphere than the Northern Hemisphere, with male ASR = 0.61 cases/year/100 000 (95% CI: 0.14–1.09) and female ASR = 0.33 (95% CI: −0.12 to 0.78) in the Southern Hemisphere, compared with male ASR = 0.10 (95% CI: 0.06–0.14) and female ASR = 0.05 (95% CI: 0.00–0.08) in the Northern Hemisphere.

9The high rates in Africa are consistent with other estimates from the region. A Tanzanian study estimated the incidence of suspected OSSN from 2006 to 2008 using operating theatre records across the country. Although there was no histological confirmation in all cases, the incidence was found to be 2.2 cases/year/100 000 (Furahini & Lewallen ). Uganda reported a peak incidence of 3.5 cases/year/100 000 in 1992 (Ateenyi-Agaba ). More recent data from the Kampala Cancer Registry also show a marked increase, although it is reported as ocular cancer, rather than specifically as OSSN (Wabinga et al. ).

10Cancer registry data in two African countries show that OSSN has become more prevalent with time. In Zimbabwe, the age-adjusted annual incidence rates of SCCC underwent a more than 10-fold increase from 0.17 to 1.8/100 000 between 1990 and 1999 (Masanganise et al. ) while the prevalence of OSSN among ocular surface tumour biopsy specimens increased from 33% in 1996 to 58% by 2000 (Pola et al. ).

11OSSN is the most common indication for orbital exenteration performed in adults in Africa (Table 3). This surgical procedure to excise all the orbital tissue including stripping the periosteum from the orbital walls is performed in cases with advanced disease. More than half (≥57%) the exenterations performed in Africa are for OSSN compared with 32% in Australia and 9–15% in Europe and India. Although available data does not clearly distinguish those performed for primary eyelid disease from conjunctival disease, SCC still emerges as an important cause in Africa. Eyelid SCC is uncommon in Africa (Templeton, ).

12In temperate countries, OSSN remains a rare, slow-growing tumour of elderly males (70–80% are males with a mean age of about 60 years; Lee & Hirst; Tunc et al. ). In contrast, in tropical countries, particularly in Eastern and Southern Africa, the prevalence is highest among young people in their 30s and among women (50–70%; Table 4; Poole; Pola et al.; Chisi et al.; Furahini & Lewallen ). Within East Africa, the pattern of SCCC in the 1960s differed to that seen today. In 1967, the average age of affected patients was 48 years, and males were four times more frequently affected than females (Templeton ).

13Worldwide, IARC data show that the overall incidence is higher in males than females but the difference is not statistically significant (Figure3 and Table 2). The mean male ASR worldwide is 0.18 cases/year/100 000 (95% CI: 0.09–0.26) and 0.08 (95% CI: 0.01–0.15) among females (P = 0.09). Incidence is significantly higher in males than females except in Africa and Asia where both sexes are equally affected (Table 2). Prevalence in Africa is higher in females than males (Table 4). This may be related to Africa having the highest prevalence of both HIV and HPV, which may increase the risk of OSSN in women and gender differences in mortality of HIV-infected adults. In South Africa, HIV-infected females have a longer life expectancy than HIV-infected males (Cornell et al.; Johnson et al.; Maskew et al. ). Men present in later stages of HIV/AIDS for antiretroviral therapy (ART) and possibly have poorer adherence to ART (Taylor-Smith et al. ). This has also been observed in Latin America, China and Lao (Dou et al.; Gonzalez et al.; Bastard et al. ). In Europe, the response to ART and mortality is similar for both sexes (Perez-Molina et al.; Thorsteinsson et al. ).

14There may be variation in disease stage at presentation, with more advanced disease present at time of surgery in East Africa, compared with other regions (Table 5; Chisi et al.; Waddell et al.; Kao et al.; Makupa et al. ). This may reflect delayed presentation to ophthalmic services in this region, leading to more advanced pathology by the time of surgery. Histopathological reporting is also subjective, and pathologists may not always grade tumours the same way (Margo et al. ). Alternatively, the disease may be intrinsically more aggressive in the East African region or HIV worsens disease progression.

15Various factors are thought to influence the causation of OSSN, but it is not clear how they interact or which is the most potent. The rising incidence of OSSN in recent decades may be driven by increased prevalence of these factors. We found no systematic reviews of risk factors for OSSN after the literature search. Of the case–control studies found, two in Uganda and Australia examined the association with solar exposure; six in Africa examined the association with HIV; sixteen examined the association with HPV; seven in Africa, five in Asia, one in Brazil, two in USA and one in Australia. Two studies examined cigarette smoking in Uganda.

16Several cutaneous malignancies, including melanoma and SCC, have a strong association with solar radiation. It was first noted in the 1960s that SCCC was relatively common in East Africa, and this apparent excess risk was attributed to higher exposure to sunlight (Templeton ). There is a strong relationship between the incidence of SCCC and increasing Ultraviolet (UV) levels (Newton et al. ). Using IARC data and published measurements of ambient solar ultraviolet light, the incidence of SCCC was found to reduce by 49% for every 10° increase in latitude from 1.2 cases/year/100 000 (Table 7) in Uganda (latitude 0.3°) to <0.02/year/100 000 in the UK (latitude > 50°). More recently, the National Institutes of Health/American Association of Retired Persons (NIH-AARP) Diet and Health Study in the USA found a slightly lower risk of SCCC in those who lived >35° compared with ≤35° from the equator, although this was not statistically significant (adjusted Hazard Ratio = 0.92, 95% CI: 0.49–1.71; Emmanuel et al. ). The USA has comparatively lower HIV prevalence, solar irradiance and incidence of OSSN than Africa, which is bisected by the equator. The high incidence of ocular SCC near the equator may be related to high solar irradiance (the amount of solar radiant energy incident on a surface per unit area and per unit time) in the world (World Energy Council ).

17A case–control study in Uganda adjusted for age, sex, residential district, and HIV serostatus demonstrated that the risk of OSSN was higher with increasing time spent in daylight (Waddell et al. ). Compared with those who reported spending up to 1 h a day in direct sunlight, the odds ratio (OR) for those who spent 2–4 h was 1.7 (95% CI: 1.2–2.4), and for those who spent 5 or more hours a day, it was 1.8 (95% CI: 1.1–3.1). A case–control study in Australia reported that the strongest risk factor was a past history of skin cancer (OR = 15, 95% CI: 2.0–113.6), although other factors, including outdoor activity, pale skin and irides and propensity to burn, were also important (Lee et al. ).

18More direct evidence for UV radiation induced damage in the pathophysiology of SCCC was described in another case–control study in Uganda in which 52% of the cases had mutations in the tumour suppressor gene TP53 compared with 14% of controls (Ateenyi-Agaba et al. ). The mutations were mainly of the CC TT type, consistent with UV-induced mutagenesis. This gene also downregulates the replication of HPV type 16 via the viral E2 protein, suggesting that its mutation may allow replication of HPV particles (Brown et al. ). Further, exposure to UV radiation is associated with altered expression of matrix metalloproteinases (MMPs) and the tissue inhibitors of these metalloproteinases (TIMPs), molecules that may be responsible for tissue invasion and metastasis of tumours (Ng et al. ).

19In addition, OSSN lesions occur more often at the limbus. A study in Uganda demonstrated that tumours almost always occur in sun-exposed areas of the eye (Waddell et al. ). It is thought that the human eye is more exposed laterally, making this a large collecting zone of peripheral sunlight, which, depending on the incident angle and radius of curvature of the cornea, is focused on the limbus, lens and lid margin, which are the main foci of sun-related eye diseases such as pterygium, OSSN, cataract and lid malignancies (Maloof et al. ). Low doses of ambient sunlight received on every day exposure inhibit immunity in the skin and internal organs (Halliday et al. ).

20There is strong evidence that HIV is a major risk factor for OSSN. Uganda, which had a cancer registry since 1951, was the first country to report a dramatic increase in the annual incidence of SCCC shortly after the outbreak of HIV/AIDS. There was a sixfold increase from 0.6 cases/year/100 000 between 1970 and 1988 to 3.5/year/100 000 by 1992 (Figure4; Ateenyi-Agaba ). A marked rise was also observed in the USA with the onset of the HIV pandemic (Guech-Ongey et al. ). At the same time, a US study observed a strong association in an HIV-infected cohort (OR = 13.0, 95% CI: 4–34; Goedert & Cote ). In Tanzania, regional incidence rates were significantly correlated with regional HIV prevalence (Pearson's r = 0.53, P = 0.03; Furahini & Lewallen ). The majority of patients (60–77%) with OSSN seen in Africa are HIV-infected (Table 6). A meta-analysis of 6 case–control studies (Table 7) in Uganda, Rwanda and Zimbabwe shows a strong association with HIV infection (pooled OR = 6.17, 95% CI: 4.83–7.89; Figure5).

21The association with HIV suggests that immunosuppression plays a role in OSSN; however, a linear association between the CD4 lymphocyte count and OSSN has not been confirmed. A cross-sectional study conducted in Tanzania found a median CD4 cell count of 71 cells/μl among HIV-infected individuals with OSSN (Makupa et al. ). HIV-infected cases tended to have larger lesions: 71% had lesions >5 mm in diameter vs. 27% among HIV-negative individuals (OR = 3.13, 95% CI: 1.5–6.5). HIV-infected cases were also more likely to develop recurrent tumours within a year of excision (82% vs. 18%; OR = 3.54, 95% CI: 1.12–11.2). However, there was no significant trend found between CD4 count and the grade of OSSN (P = 0.94). In a Ugandan study, among 112 HIV-infected cases of CIN and invasive SCC, the median CD4 count at diagnosis was 111 cells/μL (IQR; 62–221; Waddell et al. ). Excess risks standardised incidence ratio (SIR = 19.5, 95% CI: 6.3–45.5) have also been observed among a cohort of kidney transplant recipients in Australia suggesting that immune suppression from other causes may play a role (Vajdic et al. ).

22HAART does not reduce the incidence of SCCC according to data from the US HIV/AIDS Cancer Match (HACM) Study (Guech-Ongey et al. ) which compared SIRs in the pre-HAART and HAART eras among 491 048 adults aged ≥15 years with HIV/AIDS diagnosed from 1980 to 2004. The SIRs here estimate the excess risk of SCCC attributable to HIV/AIDS compared with a population with negligible HIV/AIDS prevalence and were similar at 12.0 (95% CI: 5.5–22.8) and 12.6 (95% CI: 4.6–27.4) in the pre- and post-HAART eras, respectively (P = 0.79). There is, however, a case report of ART causing tumour regression in an otherwise inoperable case (Holkar et al. ).

23The relationship between human papilloma virus (HPV) and OSSN is rather controversial with variable results. (Tulvatana; Moubayed et al.; Sen et al.; de Koning et al.; Guthoff et al.; Simbiri et al.; Yu et al. ). A review of 12 case series and 17 case–control studies concluded that there was no causal association between mucosal HPV types and OSSN while the role of cutaneous types was uncertain (de Koning et al. ). The studies included used different methods for testing of HPV (including PCR and serology), and different HPV types were examined. Conversely, a random-effects meta-analysis of various case–control studies shows that OSSN is associated with HPV infection in sub-Saharan Africa (pooled OR = 2.64, 95% CI: 1.27–5.49) and worldwide (pooled OR = 4.00, 95% CI: 2.11–7.57; Figure6). The prevalence of HPV in OSSN ranges from 0% to 100% depending on geographical region with subtypes HPV18 and HPV16 being the most common (Table 8; di Girolamo ). Most African studies report prevalence of 75–85% (Ateenyi-Agaba et al.; Simbiri et al.; Yu et al. ). HPV is more commonly isolated in OSSN than pterygium – on average, considering studies from different regions of the world, 33.8% of OSSN lesions and 18.6% of pterygia are HPV positive (di Girolamo ). There may be a true geographical variation in the prevalence of HPV in OSSN.

24Differences in HPV prevalence in OSSN may be influenced by patient selection, sample handling in the operating theatre, preparation, storage, overseas shipping and the detection method. Variations may also be due to different testing methodology and the specific HPV types tested for. Most existing molecular diagnostic tests applied in OSSN testing for HPV were developed for cervical tissue testing. The sensitivity and specificity of various polymerase chain reaction (PCR) tests varies and may be influenced by various factors including the PCR design (nested, broad spectrum or type-specific), size of amplified product and choice of polymerase used (Munoz et al.; Mesher et al. ). Detection of E6/E7 mRNA transcripts by quantitative reverse transcriptase–PCR (qRT-PCR) has been proposed as the gold standard for HPV testing (Smeets et al. ). However, RNA is unstable limiting this test to fresh frozen tissue (Kim et al. ). Testing for HPV DNA by PCR from paraffin-embedded archived tumour blocks may be complicated by contamination between samples at the time of initial tissue sectioning for DNA harvest (Boyd et al.; Iftner & Villa ).

25Generally, only a limited subset of HPV types has been investigated among OSSN cases. There are 170 genotypes of HPV described to date, which are broadly subdivided into cutaneous and mucosal types (de Villiers ). There are conflicting reports on which of these two are more commonly associated with OSSN. One study conducted in Uganda reported that among OSSN cases, the prevalence of mucosal types was higher than cutaneous types (38% vs. 22%) while from another study in the same population, the prevalence of cutaneous types was higher than mucosal types (43.6% vs. 6.8%; Table 8; de Koning et al.; Ateenyi-Agaba et al. ). Multiple HPV types have been found in individual patients with OSSN tumours. One Ugandan study reported multiple HPV types in 57.1% of SCCC and 75% of dysplasia cases by PCR (Ateenyi-Agaba et al. ). In Botswana, multiple HPV types were identified in all OSSN and all pterygium specimens by DNA sequencing (Simbiri et al. ). The HPV types found by sequencing ranged from 4 to 21 types per sample. The same study also described co-infection with multiple other viral types per individual in 17 of 18 (94%) histologically proven OSSN specimens by PCR; 83% were positive for Epstein–Barr virus (EBV), 72% were HPV positive, 67% were Kaposi's sarcoma-associated herpesvirus (KSHV) positive, 67% were herpes simplex virus (HSV-1/2) positive and 56% were cytomegalovirus (CMV) positive. All the pterygium specimens from that study similarly had multiple viruses; 75% were positive for each of EBV, KSHV, CMV and HSV while 50% were HPV positive. The proportion of HPV infection in this series was much higher than any other studies in the region have reported raising the question whether this could be due to the methodology used.

26The mechanism by which HPV is associated with OSSN is unknown. HPV is associated with causation of metaplasia in squamocolumnar epithelial transition zones such as the corneoscleral limbus and eyelid skin of the eye, the cervix and anus where there is active cell turnover and continuous cell division to replace desquamated cells (Chow et al. ). HPV also promotes degradation of the p53 gene (Scheffner et al. ).

27The epidemiology of OSSN is closely related to that of cervical cancer with respect to high incidence in Africa and the association with HIV and HPV mainly types 18 and 16 (Sun et al.; Clifford et al.; Stanley ). A meta-analysis of HPV prevalence reports worldwide shows that Africa has the highest adjusted prevalence (22.1%; 95% CI: 20.9–23.4%) among women with cytologically normal cervical pap smears using PCR-based or high-risk Hybrid Capture 2 (HC-2) technology to detect HPV DNA (de Sanjose et al. ). Whether vaccination against HPV may help to reduce the incidence of OSSN remains to be seen (Hughes et al. ).

28Outdoor occupations have been associated with OSSN, probably related to UV solar radiation exposure. In Uganda, those with outdoor occupations had an OR of 1.7 (95% CI: 1.1–2.6) compared to those with indoor occupations (Waddell et al. ). Another in Uganda reported that 74% of 133 patients with SCCC or dysplasia had outdoor occupations (Ateenyi-Agaba et al. ). In Japan, exposure to petroleum products was also described as a risk factor for conjunctival intraepithelial neoplasia (synonym of OSSN) in a small age–sex-matched case–control study (Napora et al. ). Exposure to smoke from burning wood in the kitchen was described as a risk factor for cervical cancer among HPV-infected women in Honduras (Velema et al. ).

29Cigarette smoking is implicated in other squamous cell cancers (Haverkos ). There is, however, evidence of no effect from smoking on OSSN in Africa. In Uganda, two case–control studies showed that current smokers were not at a significantly higher risk for OSSN than non-smokers (Waddell et al.; Ateenyi-Agaba et al.; pooled OR = 1.40; 95% CI: 0.94–2.09; Figure7). In a Nigerian series of 37 SCCC cases, only two patients (5.4%) had a history of cigarette smoking (Ogun et al. ) while in a series from Australia, 5 of 11 cases of SCCC (45%) were smokers (McKelvie ).

30There is little evidence that allergic conjunctivitis is a risk factor. Among 215 SCCC cases in Tanzania, 1.9% had allergic conjunctivitis (Poole ). In Rwanda, allergic conjunctivitis was found in 4% of children and was responsible for 3–6% of hospital visits of all ages (de Smedt et al. ). In a case–control study in Uganda, none of the cases of OSSN had a history of allergic eye disease (Waddell et al. ). However, a case series of SCCC from Germany reported that 6/10 cases had atopic eczema, so this may be of more importance in temperate climates (Heinz et al. ).

31Xeroderma pigmentosum (XP), a rare, inherited skin disease characterised by high sensitivity to UV damage is associated with a high prevalence (40%) of specific mutations of the TP53 tumour suppressor gene (Dumaz et al. ). Over a 25-year period in Zimbabwe, in a series of 12 cases, 2 had SCCC while the rest had SCC of the skin, lip or tongue (Chidzonga et al. ). From a series of 7 XP cases in India, 6 of the 14 eyes (42.9%) had invasive SCC and eight eyes (57.1%) had CIN (Gupta et al. ). A larger series of 32 cases in France found that 59% of them had ocular and periocular malignancies (Touzri et al. ).

32The importance of vitamin A in maintaining the health of the ocular surface is well known, but the role of vitamin A deficiency in OSSN has not been established. Deficiency of vitamin A induces keratinisation of the ocular surface (Beitch; Pfister & Renner ). Keratinisation is commonly observed as leucoplakia in OSSN lesions (Figure1). There is a synergistic interaction between vitamin A and zinc in maintenance of the corneal and conjunctival epithelium (Kanazawa et al. ). In South Africa, it was shown that 54% of HIV-infected adults are deficient in vitamin A (plasma retinol <1.05 μm) and 33% deficient in zinc (<10.7 μm; Visser et al. ). In Ethiopia, 53% of HIV-infected adults were deficient in vitamin A (Fufa et al. ). As most patients with OSSN are also HIV-infected, it is plausible that vitamin A deficiency contributes to the aetiology.

33There is limited evidence of a role for exposure to dust, ocular trauma and pre-existing benign conjunctival lesions such as pterygia and pingueculae (Templeton; Margo & Groden; Waddell et al. ).

34One of the Ugandan case–control studies found that some factors are associated with a lower risk for SCCC such as higher personal income (adjusted OR = 0.4, 95% CI: 0.3–0.7) and decreasing age at leaving home (P = 0.05), perhaps reflecting less exposure to sunlight consequent to rural-to-urban migration (Newton et al. ).

35Various models have been proposed to simultaneously address the role of two or more risk factors in cancer causation within hierarchical levels (Victora et al. ). Most such models focus on social and environmental hypothesis but do not incorporate biological factors. A recently proposed framework called Multi-level Biological And Social Integrative Construct (MBASIC) includes biological factors together with macro-environmental and individual level factors (Lynch & Rebbeck ). Using the existing evidence reviewed in this article, we propose an aetiological model that might explain how the risk factors discussed may be involved development of OSSN (Figure8).

36OSSN is a disease of increasing importance in Africa. A triad of ultraviolet solar radiation, HIV and HPV form the major risk factors and this may explain the high incidence rates in Africa. There is evidence from case–control studies that exposure to UV radiation, outdoor occupations – perhaps due to exposure to sunlight, HIV and HPV infection are associated with a higher risk for OSSN. These studies also show no evidence of effect of cigarette smoking. Dust, ocular trauma and pre-existing benign conjunctival tumours may play a role. Although mentioned in the literature, the effect of atopy and xeroderma pigmentosa is unclear. The effect of vitamin A deficiency has not been examined in case–control studies.

37The highest incidence of OSSN is found in Africa, where males and females are equally affected, unlike other continents where male disease predominates. This probably reflects that African women have increased risk due to their higher prevalence of HIV and HPV infections. As people with HIV are living longer, and given no evidence that ART reduces risk of OSSN, one could expect incidence of OSSN to increase in Africa in coming years.

38Currently, the best available options for OSSN control remain early detection and effective treatment. However, there are no early non-invasive diagnostic methods in use and no trial evidence to guide treatment. OSSN is currently largely neglected by both eye and HIV care programmes. Eye care programmes prioritise preventable blindness while OSSN often in early stages does not affect vision. OSSN may, however, lead to facial disfigurement and death in late stages. In Africa, a key research question is whether scale-up of ART and HPV vaccination will impact on OSSN.