Reply

We thank Ludovisi et al. for their interest in our paper and for undertaking a retrospective analysis of data collected from symptomatic women attending their department for clinical evaluation. The authors note visualization rates of over 80% for their postmenopausal patient cohort (n = 6649), compared with 65–67% in the UKCTOCS cohort (n = 43 867). In women aged 70–74 they report a rate of 69.5% compared with 55% in the UKCTOCS (n = 4820). No confidence limits or formal statistical comparisons are presented, making it unclear if these differences are statistically significant. Our data showed that ultrasound visualization of postmenopausal ovaries is affected by a variety of factors (hysterectomy, tubal ligation, age, unilateral oophorectomy, body mass index, infertility and age at menopause) that need to be taken into consideration when comparing cohorts. It is surprising that, given this, there is no description of these factors or even the median age of their patients. Given that their cohort would largely consist of symptomatic women attending an ultrasound service to rule out pelvic abnormalities, the prevalence of these factors is likely to be different from that in an asymptomatic population invited to join a screening program randomly from local health authority registers. Furthermore, it would have been informative if the team had reported on symptoms and pathology detected. In addition, the setting and volume of scans are also different. The ultrasound exams reported by Ludovisi et al. involved 1350 annual postmenopausal scans undertaken at one of the leading academic units in the UK, headed by an international expert in gynecological ultrasound and staffed by closely supervised doctors. While it is not stated explicitly, repeat examinations by more experienced gynecologists were probably the norm in difficult cases. Sonography in the UKCTOCS involved over 60 000 annual scans across 13 NHS Trusts by over 100 sonographers trained in gynecological scanning but with varying expertise/experience. It is unlikely that in the context of population screening, expertise such as that available at the University College Hospital unit would be available for a national screening program. As part of the quality control in the UKCTOCS, the ultrasound subcommittee made a prodigious effort to achieve standards, provide accreditation and audit results on a regular basis. As part of our accreditation process, all sonographers participating in the trial had to complete a protocol questionnaire and submit 12 sets of images for central review. They had to attend at least one UKCTOCS Study Day over 2 years and be reaccredited every 2 years, with submission of more sets of images. One of the problems with the visualization of small, featureless ovaries is verification bias. A strength of our trial is that considerable effort was made to check the validity of ‘ovarian visualization’. A subset of images taken by each sonographer was audited by the UKCTOCS ultrasound subcommittee to determine concordance with an expert. To our knowledge, none of the other ovarian cancer screening studies has been subjected to such intense review. In the UKCTOCS, if a sonographer was unable to demonstrate an ovary with clear and distinct margins, they were instructed to demonstrate a 4-cm length of iliac vein with clearly defined walls in the pelvic sidewall. As Ludovisi et al. suggest, poor visualization of the pelvis may lead to anxiety among patients. Hence a second transabdominal and transvaginal scan was arranged in the absence of good visualization of the pelvis. As the scans analyzed were undertaken between 2001 and 2007, we acknowledge that earlier models of ultrasound machine may have had an impact on ovarian visualization. Our results were averaged across all centers over the duration of the analysis. More detailed longitudinal analysis of visualization rates according to year, center and sonographer across the entire timespan of the trial (2001–2011) is underway. Improvements related to introduction of new machines in 2008 should become evident. On balance, it is unclear whether there are statistically significant differences in visualization rates between the two cohorts. Any differences noted are probably multifactorial, relating to the different populations and indications (symptomatic vs screening), settings (multisite vs single academic center), methods of audit and quality of equipment, and to the inevitably more variable levels of expertise among 100 sonographers compared with the closely monitored scanning standards in a single academic center. The key conclusion is that, unlike a laboratory test, pelvic ultrasound suffers from being subjective and operator-dependent, which makes its use as a primary population screening tool challenging. A. Sharma, A. Gentry-Maharaj, M. Burnell, A. Ryan, N. N. Amso, S. Campbell, I. Jacobs and U. Menon* on behalf of the UKCTOCS team Gynaecological Cancer Research Centre, Women's Cancer, UCL EGA Institute for Women's Health, Maple House 1st Floor, 149 Tottenham Court Road, London, W1T 7DN, UK *Correspondence. (e-mail: [email protected])