A Systematic Review of Intravaginal Testosterone for the Treatment of Vulvovaginal Atrophy

Robin J. Bell, MBBS, PhD, MPH, FAFPHM; Farwa Rizvi, MBBS, MPH; Rakabul M. Islam, MPH, PhD; Susan R. Davis, MBBS, PhD, FRACP


Menopause. 2018;25(6):704-709. 

In This Article


Despite the carrying out of six separate clinical trials of intravaginal testosterone in women to date, the efficacy and safety of this therapy remains uncertain. This is primarily because the only study that provided clinical outcomes did not involve random allocation, was not blinded, and had no placebo arm,[20] and the studies with a placebo treatment were not blinded.[16,18]

The study by Fernandes et al[17] suggests that intravaginal testosterone may lower vaginal pH, increase the proportion of vaginal lactobacilli, and the vaginal health score compared with the lubricant gel. However, the findings from this study for sexual function outcomes have to be considered inconclusive, because of failure to adjust for baseline differences between the treatment groups.[16] As Raghunandan et al included CEE in both active treatment arms, the effects of testosterone on vaginal health cannot definitely be separated from those of CEE. Although there was a suggestion of a greater effect of combined CEE and testosterone, than CEE alone, on sexual function, in the Raghunandan et al[18] study, this was not formally tested. The three studies that recruited women with breast cancer were all open-label studies, limiting interpretation of the findings for the reported outcomes, particularly for self-reported symptom outcomes, which may be placebo effects.[20–22]

With respect to safety, three studies provided full details of serum levels of estradiol and testosterone.[18,19,21] Serum estradiol alone underestimates total estrogen exposure, especially when conjugated estrogen cream, which includes a variety of estrogens, is administered.

As AIs act by inhibiting estrogen biosynthesis throughout the body, baseline serum levels of estradiol should be very low in all women taking an AI. This was the case in the study by Witherby et al.[20] However, in the study by Melisko et al,[21] baseline serum estradiol levels were considered elevated in over one-third of the participants at baseline, indicating variable compliance with AI therapy. The supraphysiological testosterone levels seen in the studies by Melisko et al[21] and Apperloo et al[19] reflect the high doses of testosterone used. The dose of testosterone administered in the study by Raghunandan et al was greater than in the other studies; yet the reported serum testosterone levels were not supraphysiological. In this study, the testosterone was self-administered twice a week, but the timing of the blood draws in relation to dosing was unspecified. If the blood was not drawn close to dosing, then it is not surprising that serum levels of testosterone were not elevated.

The risk of bias in the reviewed studies would be considered high. Four of the six studies were open-label. One was blinded[19] and one in which it is likely that the two treatments could not be distinguished by the patients, stated that the outcome assessors were not blinded to treatment allocation.[20] Of the six studies, four were randomized controlled trials, one was a single group before and after study,[22] and one stated that of the 20 participants, the first 10 received one dose and the second 10 received the alternative dose. Of the randomized trials, three were parallel-group and one was a crossover design.[19] Of the randomized trials, only one specified the method of randomization and specified that allocation concealment was maintained.[16,17]

The strength of this review is inclusion of all published clinical trials of intravaginal testosterone therapy that reported urogenital or sexual symptom outcomes. The findings of the review are limited by the small sizes of the included studies, inconsistent and incomplete outcome reporting, lack of statistical reporting of between-group differences, no adjustment for variations in baseline values in some studies, and little placebo-controlled data.