As we report in
this issue, the 16th Retrovirus (CROI) Conference in Montreal
featured the first ever non-negative result in a trial of a microbicide.
‘non-negative’ rather than ‘positive’ because the 30% reduction in HIV
infections seen in women who used the microbicide PRO2000 wasn’t ‘statistically
significant’. It could have been a chance finding.
the first thing to do is to celebrate. The idea for a microbicide has been
around since 1990 when the South African epidemiologist Zena Stein published a
piece called "HIV Prevention: The Need for Methods Women Can Use", pointing
out that whether to use a condom was essentially the man’s choice and women
were often in a position where it wasn’t possible to “choose safer sex”.
Nearly 20 years
later we at last have a hint that microbicides could work. This is a testament
to the scientific belief and persistence that has got us this far – often
driven by feisty female and gay male advocates - and the success of some of the
most difficult research studies ever attempted in the history of medicine.
But what about
statistical significance? This is a complex area and misunderstanding it was
the reason why the headlines about the trial ranged from the confident “Vaginal
gel effective in preventing HIV infection” (Hindu
Times) to the disappointed “Microbicide Gel Falls Short of Showing
Significant Efficacy” (Doctor’s Guide).
significance is an arbitrary limit that scientists impose on the data they get
from experiments, in order to decide what means something and what doesn’t. Odd
things can happen that are purely due to chance: if you toss a coin for long
enough, eventually you’ll get 20 heads turning up in a row.
out the probability that the findings from their study are due to chance. If
there’s a less than one-in-20 chance that the result is just ‘noise’, then that
is ‘statistically significant’; if there’s a more than one-in-20 chance, it is ‘not
significant’. In the microbicide study, the probability that the 30% reduction
in HIV infection seen was not real was one-in-ten: so it was ‘not significant’…
…which, to put
it another way, means that there was a 90% chance that it was real.
Still, you may
say, 30% isn’t very impressive. Would you trust a condom that was 30%
effective? PRO2000 is, nonetheless, promising rather than disappointing because:
some people, 30% may be better than nothing. As Professor Abdool Karim, the
principal investigator in this study said, “This may be a niche product for
women with no other choices”.
the still ongoing three-times-bigger UK-supported study produces similar
results at the end of this year, we’ll really be on to something.
more gel women used, the more protected they were. In women with above-average
adherence to PRO2000, HIV infections were reduced by 44%.
It may work even better than this. You can’t ethically test a
prevention product without offering women the best in existing methods, and
condoms and safer sex advice were freely available. A woman who is a regular condom
user has already protected herself against nearly nine in ten possible HIV
infections. Adding a microbicide won’t reduce her risk much further.
But for a woman who doesn’t use condoms at all, a microbicide may add a
considerable amount of protection. In the group of women who used condoms
rarely but PRO2000 frequently, it stopped 68% of infections. This could be
closer to its ‘real’ efficacy, but we don’t as yet have the figures to support
that conclusion with confidence.
Right now the result of the trial poses more questions than answers.
Do we start
using PRO2000 as the comparator drug in forthcoming microbicide trials because
it will be unethical to use an inert placebo? Or can we say “It wasn’t proved
to work” with a clear conscience? If we use it as the comparator drug, it means
new studies will have to be larger to prove an effect.
efficacy is confirmed to be in the 30 to 40% region, is it worth developing and
how do we ethically market it? Will there be pressure to market it anyway?
What about gay
men and others who have anal sex? A rectal safety study is starting this year,
but isn’t that a bit late if they decide to license it?
Would you use it…?