A commonly held belief within the community is that undetectable HIV viral load in the blood means that the person cannot transmit the virus to others.

While that is true much of the time, it isn't true all of the time. And the risk may be higher than previously thought.

Prameet Sheth and colleagues at the University of Toronto closely followed 25 men who were about to initiate highly active antiretroviral therapy. The researchers took blood and seminal fluid samples at the start of the study, week two, week four, and then every month for six months.

All of the men responded well to treatment; blood levels of HIV became undetectable in four to 16 weeks. But the same wasn't true in seminal fluid; at week 16 nearly half (48 percent) still had detectable levels of virus in that compartment, which sits behind a blood barrier that some drugs do not penetrate very well.

The level of detectable virus in semen was low in some patients and probably posed little risk of transmission during sex. But four of the 25 (16 percent) had a viral load in their seminal fluid that was above 5,000 copies of HIV RNA, and one was above 16,000 copies.

Sheth took a sample from the patient with the highest viral load and found that it could infect CD4 cells in the lab.

"In this individual, the virus was infectious," Sheth said.

Another patient's blood viral load became undetectable within four weeks of starting therapy but he "was a sustained semen shedder," his virus was still detectable in seminal fluid, for the entire six months of the study.

Sheth looked at a second group of 13 men who had been on HAART with an undetectable viral load in blood for at least four years, four had isolated HIV shedding episodes in their semen.

"We did see quite a differential in the penetration of drugs; 3TC was present in seminal plasma 100-fold higher than it was in blood, but we were still able to get isolated HIV shedding in this individual," said Sheth. "There was no evidence of drug resistance in the virus in the compartments."

"We could not find that [shedding] was related to the concentration of antiretroviral drugs, or the drug regimen that people were on," he said. But he also noted that the study probably was too small to detect differences between drugs.

Nor was there any association with CD4 count or herpes serostatus. "The only predictor that we were able to find is higher semen shedding prior to initiation of therapy. Obviously that is not readily available" to doctors or patients, Sheth said.