Quantitative transvaginal two- and three-dimensional sonography of the ovaries: reproducibility of antral follicle counts

Objective To assess the intra- and interobserver reproducibilities in addition to the between-method reliability of antral follicle counts using two (2D)- and three (3D)-dimensional transvaginal sonography (TVS). Methods Two groups of women with regular menstrual cycles were studied. One group consisted of healthy volunteers with proven fertility and the other group consisted of patients visiting the general infertility clinic. In each woman, 2D or 3D TVS was performed in the early follicular phase (day 2, 3 or 4) of the menstrual cycle to measure the number of antral follicles (2-10 mm). Results Intraobserver reproducibility was calculated from follicle counts using 3D TVS in 41 women. The intraclass correlation coefficient was 0.99 and the 95% coverage interval of the difference (CID) was -3.2 to +3.2. Interobserver reproducibility was assessed from both 2D (n = 37) and 3D (n = 49) TVS-based follicle counts. An interclass correlation coefficient of 0.98 was found for both methods. The 95% CID was -5.0 to +4.1 for 2D and -5.6 to +5.7 for 3D measurements. The latter CID appeared to increase in the higher range of counts. Finally, the degree of agreement between 2D and 3D TVS counts (n = 76) was characterized by a 95% CID of -5.3 to +8.3. This coverage interval widened when higher numbers of follicles were counted. With the exception of the between-method analysis, kappa values indicated overall that follicle counts will hardly change from one category to another when repeatedly carried out. Conclusions Determination of antral follicle numbers by both 2D and 3D TVS is adequate with regard to the intra- and interobserver reproducibility. The between-method reproducibility of follicle counts measured both by 2D and 3D ultrasound is moderate. When higher follicle counts are observed both interobserver and between-method reproducibilities tend to decline. If used in categorical classifications, ultrasound-based follicle counts appear to have a high level of agreement between and within observers.