Unless diagnosed and excised early, malignant melanoma is a cancer which carries a poor prognosis and its incidence is rising world-wide, particularly among fair-skinned people. There is growing evidence from many sources that immune responses are elicited against cancers and may contribute towards prevention of disease and limitation of progression of established cancers. Since the mid 19th century there have been many reports of regressions and remissions of cancers following serious infections such as erysipelas, and the causative organisms have been used therapeutically with some claimed success. A study group of the European Organization for Research and Treatment of Cancer (EORTC) found that a history of vaccinations with BCG and/or vaccinia virus in childhood and/or a history of certain uncommon but serious infections are associated with a reduced risk of melanoma . It has been reported that about 95% of malignant transformed melanocytes express an antigen, HERV-K-MEL, encoded by a human endogenous retrovirus K (HERV-K). This antigen is also expressed in a minority of carcinomas and sarcomas but, with the exceptions of the testis and some naevi, not in normal tissues. In further studies it was established that BCG vaccine, the vaccinia virus and pathogens - that appeared to protect against melanoma - had antigenic determinants showing homology in the sequence of amino acids with the melanoma antigen HERV-K-MEL. Diseases which showed no association with reduced melanoma risk were predominantly caused by pathogens without this homology.
Additional information on the impact of vaccination with vaccinia virus and/or BCG on the risk of melanoma has been derived from epidemiological studies. Vaccinations given neonatally or during the first two years of life was associated with a life-long reduced risk of melanoma, although the protective effect waned after the age of 50 years. Among those who had been vaccinated but nevertheless developed melanoma, their risk of dying of the disease within an 8-year follow-up period after surgical resection of the primary tumour was significantly reduced, although use of vaccinia and/or BCG as therapeutic agents for established melanoma has generally been unsuccessful. This suggests that, for protection, the immune responses induced by vaccination need to be established long before the tumour is clinically evident. The temporal pattern of association seems to depict a "crucial point" in the natural history of melanoma. Yellow fever vaccine (YFV) contains an antigenic determinant with a close homology to HERV-K-MEL antigen and to the homologous sequences in BCG and vaccinia virus. However, to date, there is no evidence as whether vaccination with YFV has a protective effect against melanoma. Thus, an epidemiological study was undertaken in order to investigate whether there is such an association and, if there is, to determine the temporal pattern of the association, as any such protection might be optimal at a very early stage of malignant transformation, perhaps preceding the presentation of overt melanoma by many years.
The study was carried out in Veneto in the North-Eastern part of Italy, where a large part of the region was covered by the Veneto Tumor Registry (VTR). According to Italian law, YFV must only be administered in an authorized health service (AHS) and personal data of vaccinated subjects must be recorded. The study protocol was presented at a meeting to most physicians working in these AHSs. Among eligible AHSs (those corresponding to areas covered by VTR), seven agreed to collaborate: Padova, Verona, Vicenza, Treviso, Bassano, Montebelluna, Ufficio di Sanità Marittima ed Aerea. After approval for the study had been granted by the Ethics Committee of the University of Padova, all the records of subjects vaccinated against yellow fever in the collaborating vaccination centers were collected. Information on the incidence of cancer in this group of subjects was obtained through a record-linkage with cancer data of VTR. The age at vaccination and at the diagnosis of cancer, and the years elapsed from the date of vaccination to the date of diagnosis of cancer (latency) were calculated. In the statistical analysis, latency was divided into four classes: <0; 0> to 5; >5 to 10; and >10 years. The first class (<0) included subjects vaccinated after the date of cancer diagnosis considerable as non exposed. In order to determine the time-effect relationship between the risk of melanoma and YFV administration we used a classical cohort study with Veneto general population as reference and an internal cohort analysis with Poisson regression where the references were subjects with latency <0 years.
Furthermore we carried out a cohort nested case-control study. The cases were 60 patients with melanoma and the controls were 178 patients affected by tumours of brain, cervix uteri, colon and rectum, kidney, as well as with mesothelioma and lymphoid leukaemia. The HERV-K-MEL gene has not been found to be expressed in these tumours, and therefore, for control purposes, they may be regarded as "normal" with respect to such gene expression. The TSV class ">0 to 5" was taken as the reference subgroup with an assumed risk of 1.0. The odds ratio (OR) for melanoma, 95% confidence interval (CI), error probability for a two-tail test (p-value) were estimated through logistic regression analysis, stratified by vaccination centres. The dependent variable was dichotomous (1 for cases and 0 for controls); the independent variables were age at incidence, age at vaccination, gender and the polytomous variable TSV at multivariate analysis, and TSV only at univariate analysis
The standardized incidence ratio (SIR) was 1,25 (IC 0,90-1,42) in the overall cohort. Latency analysis showed a SIR=1,29 (IC 0,81-2,06), 1,52 (IC 0,93-2,50) and 0,56 (IC 0,48-1,74), respectively in the latency categories 0-5, 5-10, and ?10 years.
In order to compare the risk estimates within the cohort, we made an internal analysis using Poisson regression. After more than 10 years elapsed from the vaccination, the risk of melanoma reduced to 0.38. The risk estimates increased with age and calendar periods and were higher in women than in men.
In the nested case-control study, subjects vaccinated against YF from 0 to 4.99 were the reference group in the logistic regression analysis. Melamona cases were compared with 178 tumors not expressing HERV-K-MEL gene. OR significantly (p=0.01) reduced to 0.19 in the class >10 years of latency.
The conclusion of this study is that yellow fever 17D vaccine appears to protect against melanoma if subjects were vaccinated 10 or more years before the clinical presentation of the disease, but ineffective later on. Our findings indicate that a simple preventive strategy against melanoma is possible although confirmatory studies are required before administration of yellow fever 17D vaccine can be advocated for the prophylaxis of this aggressive form of cancer.