Dear Sir,

Given the intensive efforts to treat melanoma by various forms of immunotherapy utilizing mostly Bacille Calmette-Guérin (BCG) vaccine or, to a lesser extent, vaccinia vaccine during the past decades, it is surprising that no epidemiological study so far has examined the potential relationship between vaccination history and melanoma risk.1 Although there is quite a difference between immunotherapy of an established lesion and immunological prevention of melanoma, we found it tempting to analyze in an epidemiological study whether the application of BCG vaccination, commonly administered to the newborn to provide protection against tuberculosis, and vaccinia vaccination, also commonly administered twice to children to provide protection against smallpox until the late 1970s, influences melanoma risk. Experimental evidence2 and clinical trials3, 4, 5, 6 on the therapeutical benefits of these vaccinations for melanoma patients have shown promising results in the 1970s that could not be confirmed in later studies on large numbers of patients.7, 8 Substantial heterogeneity among melanoma patients with respect to their response to such therapeutic regimes was, however, identified by Tan and Ho9 in a reanalysis of 20 randomized controlled clinical trials. The development of new strategies for melanoma therapy resulted in a decline of interest in this form of therapy, and further research in this area nearly stopped during recent years.

We conducted a case-control study on incident melanoma cases and population controls in Sofia, Bulgaria, to address specifically the relationship between melanoma risk and vaccination history. From 1994 until 1997 we recruited 115 incident cases with histopathologically verified diagnosis of malignant melanoma of the skin and 115 population controls sampled from the same geographical area as the cases and frequency-matched to them with respect to gender and age. The participation rates among eligible cases and controls were 92% and 78%, respectively. In a standardized personal interview conducted at the study subject's home the same interviewer ascertained detailed information on socio-economic and life style variables as well as on a variety of established risk factors for melanoma such as, e.g., number of nevi, skin type, freckling, and sun exposure. In addition, study subjects were asked for their personal history of BCG vaccination against tuberculosis and vaccinia vaccination against smallpox during childhood. Answers were verified by the interviewer by inspection of the subjects' vaccination cards.

We employed multiple logistic regression analysis to analyze the effect of the 2 different vaccinations on melanoma risk. Confounding variables included in all models were gender, age, skin type (according to Fitzpatrick's classification),10 freckling index (according to the charts described by Gallagher et al.),11 number of nevi, and number of sunburns. First, we analyzed the effect of the different vaccinations in 2 separate logistic models, each of which comprised a dichotomous exposure variable for the specific vaccination. Then, to address the combined effect of both vaccinations we fitted a joint logistic model in which both exposure variables together with their interaction and all the confounders were incorporated. In this joint approach, results for the effect of vaccination history can be given for the following 4 disjoint vaccination patterns: neither vaccinated with BCG nor with vaccinia (reference category), only vaccinated with BCG but not with vaccinia, only vaccinated with vaccinia but not with BCG, and vaccinated with both BCG and vaccinia. Alternatively, in the joint approach the vaccination history has also been modelled as a dichotomous variable (being vaccinated with BCG or vaccinia vs. neither vaccinated with BCG nor with vaccinia). Results of the logistic regression analyses are given by the estimates of the adjusted odds ratios (OR) for the corresponding exposure variables related to vaccination status and their accompanied 95% confidence intervals (CI) obtained by the profile likelihood method.12 All statistical analyses were carried out by means of validated programs within the SAS software (SAS Institute Inc., Cary, NC).

The study sample consisted of 118 males (59 cases, 59 controls) and 112 females (56 cases, 56 controls). The median age (in years) among cases and controls was 61 and 59 years, respectively, accompanied by a wide range in the age of cases (24–90 years) as well as controls (24–84 years). In the case group, the melanomas were rather advanced: one-third had a tumour thickness (according to Breslow's method)13 above 4 mm, only one-sixth were thin melanomas less than 0.75 mm in thickness. The tumour was primarily located on the trunk (39%), followed by legs and feet (33%), head and neck (19%), and finally, arms and hands (9%).

Whereas BCG and vaccinia vaccination were reported by 100 and 111 controls, respectively, only 85 and 104 cases did so. In the separate logistic analyses, both vaccinations were inversely associated with melanoma risk after control for confounding (OR = 0.49, 95% CI = 0.20–1.17 for BCG and OR = 0.34, 95% CI = 0.08–1.33 for vaccinia, respectively). The joint analysis further corroborated this inverse association. Compared to those subjects without either vaccination, we observed drastically lower melanoma risks for those having only been vaccinated against tuberculosis (OR = 0.13, 95% CI = 0.004–2.23), having only been vaccinated against smallpox (OR = 0.17, 95% CI = 0.007–1.52), and having been vaccinated against both tuberculosis and smallpox (OR = 0.11, 95% CI = 0.005–0.85). Thus, the combination of both vaccinations shows the most pronounced effect on melanoma risk reaching statistical significance at the conventional significance level of 5%. The apparent discrepancy between the results of the separate and joint analysis concerning the effect of the 2 vaccinations on melanoma risk points to the fact that the effect of each single vaccination on melanoma risk is not independent from the other. We found lower ORs in the joint analysis that can methodologically be explained by the different definition of the reference category in the two analyses. Because the group of subjects without either vaccination carries the highest melanoma risk (OR = 8.40, 95% CI = 1.10–166.67 in the logistic model incorporating vaccination status as a dichotomous variable), the joint analysis shows the risk-lowering effect of the 2 vaccinations much clearer than the separate analyses where the reference category for 1 vaccination always is contaminated with subjects with a positive history of the other vaccination.

We considered carefully potential sources of bias for this new epidemiological finding. Selection bias seems unlikely to us because: 1) cases in the study were representative of melanoma cases in Sofia, where such a distribution of tumour thickness with a high proportion of advanced melanomas is normal, 2) controls were recruited from the same population that gave rise to the cases without imposing any restrictions on them other than those given by the frequency-matching procedure and 3) the response rates for both groups were sufficiently high. Exposure misclassification is a severe problem in all epidemiological studies addressing retrospectively events early in life and will have affected our data to some degree. There are, however, no reasons to assume that this form of information bias will have operated differentially between cases and controls. It is therefore reasonable to assume that we have to face nondifferential exposure misclassification that leads to some bias toward the null. Confounding by other risk factors for melanoma as a possibility for producing spurious associations can never be ruled out completely. In our study, we have carefully ascertained information on other risk factors such as skin type, freckling index, number of nevi and sun exposure. For these variables, we could confirm their known relationship to melanoma risk. In logistic regression models the confounding effect of these variables and the factors utilized in the frequency-matching (gender and age) has been controlled for when estimating the impact of the vaccinations on melanoma risk.

The immunologic aspects of the underlying mechanisms of melanoma initiation and progression are currently a rapidly evolving area of research,14 their discussion is beyond the scope of this letter. Current knowledge about the immune reaction of BCG-vaccinated subjects suggests, however, that an early timing of the vaccination seems to be important with respect to its potential tumour preventive effect.15 This might explain why therapeutical attempts to treat melanoma with BCG and vaccinia vaccinations in adults were not that successful, whereas the protective potential of an early vaccination observed in our study seems to be much higher.

In conclusion, our results suggest a rather strong inverse association between melanoma risk and previous vaccinations against tuberculosis and smallpox. Due to the limited sample size the findings should interpreted with caution. Larger studies are needed to confirm this observation that could then have public health implications.

Yours sincerely,


We are indebted to Dr. J.M. Grange for his help in guiding us when attempting to understand the immunological plausibility of our findings. The study was financially supported by the Cancer Research Institute, New York. The grant was given to Prof. Kölmel, who inaugurated the study.


  1. Top of page
  • 1
    Berwick M. Epidemiology: current trends, risk factors, and environmental concerns. In: BalchCM, HoughtonAN, SoberAJ, SoongS, eds. Cutaneous melanoma, 3rd ed. St. Louis: Quality Medical Publishing, Inc., 1998. 55171.
  • 2
    Zbar B, Bernstein ID, Bartlett GL, Hanna MG Jr, Rapp HJ. Immunotherapy of cancer: regression of intradermal tumors and prevention of growth of lymph node metastases after intralesional injection of living Mycobacterium bovis. J Natl Cancer Inst 1972; 49: 11930.
  • 3
    Seiger HF, Shingleton WW, Metzgar RS, Buckley CE 3rd. Immunotherapy in patients with melanoma. Ann Surg 1973; 178: 3529.
  • 4
    Pinsky CM, Hirshaut Y, Oettgen HF. Treatment of malignant melanoma by intratumoral injection of BCG. Natl Cancer Inst Monogr 1973; 39: 2258.
  • 5
    Everall JD, O'Doherty CJ, Wand J, Dowd PM. Treatment of primary melanoma by intralesional vaccinia before excision. Lancet 1975; 2: 5836.
  • 6
    Eilber FR, Morton DL, Holmes EC, Sparks FC, Ramming KP. Adjuvant immunotherapy with BCG in treatment of regional-lymph-node metastases from malignant melanoma. N Engl J Med 1976; 294: 23740.
  • 7
    Veronesi U, Adamus J, Aubert C, Bajetta E, Beretta G, Bonadonna G, Bufalino R, Cascinelli N, Cocconi G, Durand J, De Marsillac J, Ikonopisov RL, Kiss B, Lejeune F, MacKie R, Madej G, Mulder H, Mechl Z, Milton GW, Morabito A, Peter H, Priario J, Paul E, Rumke P, Sertoli R, Tomin R. A randomized trial of adjuvant chemotherapy and immunotherapy in cutaneous melanoma. N Engl J Med 1982; 307: 9136.
  • 8
    Paterson AH, Willans DJ, Jerry LM, Hanson J, McPherson TA. Adjuvant BCG immunotherapy for malignant melanoma. Can Med Assoc J 1984; 131: 7448.
  • 9
    Tan JK, Ho VC. Pooled analysis of the efficacy of bacille Calmette-Guérin (BCG) immunotherapy in malignant melanoma. J Dermatol Surg Oncol 1993; 19: 98590.
  • 10
    Fitzpatrick TB. The validity and practicality of sun-reactive skin types I through VI. Arch Dermatol 1988; 124: 86971.
  • 11
    Gallagher RP, McLean DI, Yang CP, Coldman AJ, Silver HK, Spinelli JJ, Beagrie M. Suntan, sunburn, and pigmentation factors and the frequency of acquired melanocytic nevi in children. Similarities to melanoma: the Vancouver Mole Study. Arch Dermatol 1990; 126: 7706.
  • 12
    Venzon DJ, Moolgavkar SH. A method for computing profile-likelihood based confidence intervals. Appl Stat 1988; 37: 8794.
  • 13
    Breslow A. Thickness, cross-sectional areas and depth of invasion in the prognosis of cutaneous melanoma. Ann Surg 1970; 172: 9028.
  • 14
    Nesbit M, Setaluri V, Herlyn M. Biology of melanocytes and melanoma. In: BalchCM, HoughtonAN, SoberAJ, SoongS, eds. Cutaneous melanoma, 3rd ed. St. Louis: Quality Medical Publishing, Inc., 1998. 46373.
  • 15
    Grange JM, Stanford JL, Rook GA. Tuberculosis and cancer: parallels in host responses and therapeutic approaches? Lancet 1995; 345: 13502.

Annette Pfahlberg*, Ivan N. Botev†, Klaus F. Kölmel‡, Olaf Gefeller*, * Department of Medical Informatics, Biometry and Epidemiology, University of Erlangen-Nuremberg, Germany, † Department of Dermatology and Venerology, Alexander's University Hospital, Sofia, Bulgaria, ‡ Department of Dermatology and Venerology, University Hospital Göttingen, Germany.