While I was going by analogy, here is one poster. A similar study points out the low risk in MSMs. Case reports of HIV from eating out do exist, and their accuracy can be questioned.
http://bmj.bmjjournals.com/cgi/eletters/326/7392/730
This study was presented at the Barcelona AIDS Conf., and published last year in AIDS:
I) del Romero J, Marinocovich B, Castilla J et al. Evaluating the risk of HIV transmission through unprotected orogential sex. AIDS 2002;16:1296-97.
Between 1990 and June 2000, in a STD and HIV testing clinic in Madrid, the authors followed an open cohort of individuals who were initially seronegative for HIV and whose heterosexual steady partner had had a diagnosis of HIV infection confirmed. Each member of each couple was followed in the same clinic with regular six monthly examinations. Subjects were surveyed about each type of contact, whether protected or unprotected, as well as about ejaculations and accidents with condoms. Determination of antibodies to HIV was carried out for the member of the couple who was not initially infected. Individuals who showed any risk exposure to HIV other than to sex with the mentioned partner were excluded from the study.
A total of 135 seronegative individuals (110 women and 25 men) whose only risk exposure to HIV was unprotected orogenital sex with their infected partner, participated in the study. For the 110 HIV-seronegative women whose partner was an infected man, 179 person-years of follow-up were taken into account. Of these women, 96 performed fellatio without a condom, which gave an estimated total of 8,965 unprotected fellatios, of which 3,060 (34 percent) were with ejaculation in the oral cavity. Ninety- eight infected men carried out unprotected cunnilingus on the uninfected woman with an estimate of 8,656 practices of this type. Among the HIV- infected members of the couples, 8.1 percent had been diagnosed with AIDS and 15.6 percent had a CD4 cell count lower than 200 x 106/l. For 60 individuals an HIV-1-RNA quantification was available, and six of them presented with greater than 10,000 copies/ml. Of the 135 initially infected partners, 39 percent received antiretroviral therapy during the follow-up. The authors reported registering 210 person- years of follow-up with participants.
While a number of women presented with vaginal infections, the 135 individuals, “who had had over 19,000 unprotected orogential contacts with their HIV-infected partner,” presented without a single case of seroconversion to HIV. According to the authors, “this seems to point to a very low probability of HIV transmission related to this practice, when other risk exposures are excluded.”