The study was presented in part at the Annual Meeting of the European Society of Gynaecological Oncology in Istanbul, Turkey, September 2005.
Feasibility of screening for ovarian cancer using symptoms as selection criteria*
Article first published online: 6 DEC 2006
RCOG 2007 BJOG An International Journal of Obstetrics and Gynaecology
BJOG: An International Journal of Obstetrics & Gynaecology
Volume 114, Issue 1, pages 59–64, January 2007
How to Cite
Rufford, B., Jacobs, I. and Menon, U. (2007), Feasibility of screening for ovarian cancer using symptoms as selection criteria. BJOG: An International Journal of Obstetrics & Gynaecology, 114: 59–64. doi: 10.1111/j.1471-0528.2006.01153.x
- Issue published online: 6 DEC 2006
- Article first published online: 6 DEC 2006
- Accepted 3 October 2006.
- ovarian cancer;
Objective Retrospective studies have reported that 95% of women with ovarian cancer have symptoms prior to diagnosis and that women with these symptoms are at an increased risk of ovarian cancer. Failure to recognise these symptoms may result in a delay in referral and diagnosis. We assess the feasibility of screening for ovarian cancer using symptoms as selection criteria.
Design A randomised controlled trial.
Setting General practices in East London.
Population Three hundred and ninety GPs.
Methods GPs were randomised by practice, and those in the study group were given rapid access to ultrasound and CA125 test for women, >45 years, suffering from symptoms that may be caused by ovarian cancer.
Main outcome measures Symptoms leading to referral, ultrasound and CA125 results and ovarian cancer diagnosis.
Results Seventy nine practices containing 197 GPs were randomised to the study arm. Three hundred and seventeen women were referred, of which 315 were eligible. Women reported the following symptoms: abdominal 87%, gastrointestinal 41% and constitutional 29%. Twenty-three women had abnormal findings on ultrasound: 20 were managed conservatively and 3 surgically. Histology revealed a mucinous cystadenoma, a Brenner tumour and a serous cystadenoma. Incidental findings included endometrial pathology in 13 women and bladder pathology in 2. Ninety five percent of CA125 results were <35 units/ml.
Conclusions This pilot study confirms the feasibility of screening for ovarian cancer using symptoms as selection criteria. Specificity was high and patient compliance good. Initial concerns about referral volumes and additional investigations and referrals generated were not confirmed. No ovarian cancers were detected in this pilot study, and this may be due to the size of the cohort.
Seventy percent of women with ovarian cancer are diagnosed with stage III or IV disease, with 5-year survival of less than 15%.1 Late diagnosis is thought to be at least in part due to vague nonspecific symptoms, which can often go unrecognised for a period of time. So far, efforts to improve disease outcome by detecting the disease at an early stage have focused on screening asymptomatic women. Large randomised control ovarian cancer screening trials with mortality as an endpoint are underway in the UK (UK Collaborative Trial of Ovarian Cancer Screening; http://www.ukctocs.org.uk) and the USA (Prostate Lung Colon Ovary Trial).2
Recent retrospective studies have reported that 95% of women with ovarian cancer have symptoms, although they may not be of gynaecological in nature.3–7 Furthermore, the data suggest that symptoms are reported in early-stage disease and that the pattern and duration of symptoms differs in women with ovarian cancer compared with other women.8,9 This is an issue of increasing interest to patients, their relatives, doctors and ovarian cancer support groups such as OVACOME (http://www.ovacome.org.uk) and The Ovarian Cancer National Alliance (http://www.ovariancancer.org). The nonspecific nature and frequency of symptoms in ovarian cancer makes it challenging to develop clear guidelines that can be used to triage patients. However, if symptoms are indeed a feature of early-stage ovarian cancer, this group of women may be an important population to target for ovarian cancer screening.
No trials of ovarian cancer screening for women selected by symptoms have been reported to date, and the potential of such an approach is unknown. This pilot cluster randomised controlled trial was undertaken to assess the feasibility of such an approach. GPs in the study arm were offered the option of referring women with vague symptoms for an ovarian cancer screen using the serum CA125 levels and transvaginal ultrasound. This report describes the findings for women referred for screening in the study.
Approval for the study was obtained from the North East London Health Authority Research Ethics Committee. The general practices covering the areas of Tower Hamlets, Newham and City and Hackney within the North East London Health Authority were contacted, and their permission was obtained for randomisation into the study after presentations and provision of written details about the study. GPs were informed about the symptoms women suffer prior to diagnosis of ovarian cancer and the scientific evidence relating to this. This was done prior to randomisation so both cohorts received similar information. Randomisation was performed by practice, without stratification, using computer-generated random numbers. After randomisation, GPs allocated to the study group were again informed in detail about the study.
GPs in the study group were given the opportunity to refer women to the gynaecological cancer research unit at St Bartholomew’s Hospital for targeted screening using serum CA125 levels and transvaginal ultrasound. Eligibility criteria were women more than 45 years of age who complained of the vague symptoms that women in retrospective studies reported prior to their diagnosis with ovarian cancer. These were abdominal pain or discomfort, abdominal bloating or increased abdominal size, indigestion or heartburn, pelvic pain, urinary frequency, urinary incontinence, constipation, diarrhoea, abnormal vaginal bleeding, loss of appetite or difficulty eating, weight loss, pain during sexual intercourse, fatigue/lethargy or tiredness, nausea or vomiting, noticing an abdominal mass, other bowel symptoms or back pain. Women were ineligible if the GPs had a high degree of suspicion that the diagnosis may be ovarian cancer. In particular, it was emphasised that the National Institute for Clinical Excellence criteria for urgent referral for suspected ovarian cancer should be followed. It was emphasised that women who fit these criteria and other women in whom the GP was highly suspicious of a diagnosis of ovarian cancer should be urgently referred, as per guidelines, to the local cancer unit. GPs were also advised that where the main suspicion was other pathology, they should investigate this appropriately. Women with abnormal vaginal bleeding as the only symptom were also excluded, as this is a common gynaecological symptom in the perimenopausal period, and few ovarian cancers present in this manner. Other exclusion criteria included bilateral oophorectomy, history of ovarian cancer or an active nonovarian malignancy. Women with a history of malignancy were eligible if they had no documented persistent or recurrent disease and were not on treatment for more than 12 months. Tamoxifen prophylaxis following breast cancer was not an exclusion.
The most commonly reported symptoms in the literature were listed on a facsimile form. On receipt of the facsimile, the gynaecological cancer research unit directly contacted the woman by telephone and arranged a screening appointment within 1 week of the referral.
Women were offered a serum CA125 test and a transvaginal ultrasound scan (TVS). A transabdominal ultrasound scan was arranged for those who declined TVS. Two experienced ultrasonographers performed all ultrasound scans and documented ovarian size, morphology and the presence of any pelvic abnormalities. If an ovarian cyst was detected, further documentation included the following: cyst dimensions, cyst wall thickness structure and regularity, presence and dimensions of septae, papillations and solid areas. Cyst morphology was documented pictorially, and colour Doppler studies were performed of any abnormal areas. Abnormal ultrasound scans were divided into ovarian abnormalities and incidental findings. The ovaries were described as follows:
- 1Normal: normal ovarian echogenicity with or without a single inclusion cyst or spots of calcifications, or the ovary was not visualised despite a good view of the iliac vessels. Inclusion cysts were defined as simple cysts of less than 10 mm in diameter that did not distort the outline of the ovary.
- 2Simple cysts: one cyst distorting the outline of the ovary or ≥10 mm in diameter, with no septae or papillation and thin wall with regular internal outline. Simple cysts included a cyst with incomplete septae or a cyst with a single wall irregularity of <3 mm in height.
- 3Complex morphology: all nonuniform ovarian echogenicity excluding simple cyst. Complex morphology included more than one simple cyst, cysts with multiple wall irregularities of <3 mm in height or a multiloculated cyst.
The ovarian scan was classified as ‘abnormal’ for 2 and 3 of the above, although there was no further intervention for simple cysts of less than 5 cm in diameter in the presence of a normal serum CA125 level.
Serum CA125 was assayed using the Roche Elecsys 2010 (F. Hoffman-LARoche Ltd, Grenzacherstrasse, Basel, Switzerland), ElectroChemiLuminescence ImmunoAssay method. This uses the monoclonal antibody OC 125 as the detection antibody and monoclonal antibody M11 as the capture antibody. Abnormal CA125 results were defined as values above 35 units/ml.
Statistical analysis was carried out using Fisher’s exact test. All women were followed up from time of referral until December 2005.
The 162 general practices, comprising 390 GPs, covering the areas of Tower Hamlets, Newham and City and Hackney within the North East London Health Authority were contacted for permission to be randomised into the study in 2002. Three practices comprising nine GPs declined. After randomisation, 79 practices (197 GPs) were allocated to the study group and 80 practices (184 GPs) to the control group.
Of the 79 practices in the study arm, 40 did not make any referrals. Thirty-nine of the 79 general practices in the study group (141 GPs) referred women to the study (Table 1).
|Number of referrals||Number of practices|
Among the 39 practices that did refer women, there was significant variation between the referral patterns of individual GPs, with 98 of the 141 GPs referring women (Table 2). Overall, 49.7% of the GPs in the study group practices referred women.
|Number of referrals||Number of GPs|
The study recruited women between October 2002 and March 2005. A total of 317 women were referred from the study practices, of which 315 women were eligible for the study. Two women referred were ineligible as they were younger than 45 years of age. Of these 315 women, 290 (92%) attended for their screen. Twenty-five women either could not be contacted, decided that they did not want an appointment, cancelled or did not attend their appointment.
The median age of the women was 58 years (45–86). Of the 290 women who attended, 43 (15%) had had a previous hysterectomy. Seven women (2%) previously had cancer—five breast, one stomach and one chronic lymphocytic leukaemia. Seventeen women (6%) were taking hormone replacement therapy.
The symptoms prompting inclusion in the study are summarised in Table 3. The median number of symptoms each woman experienced was three (one to nine). There was no statistically significant difference between the symptoms of those women with abnormal ultrasound scans or CA125 results and the women with normal findings, except that women with a result of raised serum CA125 level were more likely to suffer from indigestion or heartburn (P= 0.029).
|Symptoms||Abdominal (A), gastrointestinal (G), constitutional (C) or other (O)||All referrals, n= 290||Women with ovarian abnormalities on ultrasound, n= 23||Women with raised serum CA125 level, n= 13|
|Abdominal pain or discomfort||A||213 (73)||18 (78)||10 (77)|
|Abdominal bloating or increased abdominal size||A||148 (51)||11 (48)||9 (69)|
|Abdominal combined||276 (87)||21 (91)||12 (92)|
|Indigestion or heartburn||G||63 (22)||2 (9)||6 (46)|
|Constipation||G||56 (19)||4 (17)||3 (23)|
|Diarrhoea||G||25 (9)||0||2 (15)|
|Nausea or vomiting||G||19 (7)||0||2 (15)|
|Other bowel symptoms||G||7 (2)||0||0|
|Gastrointestinal combined||131 (41)||6 (26)||8 (62)|
|Weight loss||C||19 (7)||0||1 (8)|
|Loss of appetite or difficulty eating||C||25 (9)||1 (4)||2 (15)|
|Fatigue/lethargy/tiredness||C||69 (24)||2 (9)||3 (23)|
|Constitutional combined||92 (29)||3 (13)||5 (38)|
|Pelvic pain||O||96 (33)||8 (35)||5 (38)|
|Urinary frequency||O||58 (20)||7 (30)||1 (8)|
|Urinary incontinence||O||18 (6)||1 (4)||0|
|Abnormal vaginal bleeding||O||17 (6)||1 (4)||2 (15)|
|Pain during sexual intercourse||O||14 (5)||0||3 (23)|
|Mass in the abdomen||O||5 (2)||0||0|
|Back pain||O||54 (19)||4 (17)||3 (23)|
|Other symptoms||O||39 (13)||3 (13)||1 (8)|
|Combination of three symptoms*||25 (8)||2 (9)||1 (8)|
|Combination of four symptoms**||11 (4)||0||0|
7.9 percent of women had abnormal ovarian ultrasound findings. Twelve had simple ovarian cysts of less than 5 cm in diameter and normal CA125 results. These women had no further intervention. One woman had a simple ovarian cyst of less than 5 cm in diameter and a raised CA125 of 44 units/ml. A repeat screen 6 weeks later showed the cyst to have reduced in size and the CA125 to be 38 units/ml. One woman had a 6-cm diameter, simple ovarian cyst and a normal CA125. A further screen 6 weeks later showed no increase in size of the cyst. Five women had small complex cysts of less than 5 cm in diameter and normal CA125 results. All had further screens after approximately 6 weeks, and none showed significant changes in the size or morphology of the cysts or elevations in the CA125 levels. One woman had a 6-cm diameter complex cyst, with the appearances of an endometrioma and a raised CA125 level of 38. A repeat screen showed no change in the size or morphology of the cyst, and the CA125 level was stable. All 20 of these women were managed conservatively, and no ovarian cancer was found on median follow up of 22 months (range 12–33 months).
Three women were managed surgically. One woman with a 9-cm, simple, left ovarian cyst had a laparoscopic left salpingo-oophorectomy. Histology showed a benign serous cystadenoma. One woman with an apparent 3-cm complex ovarian cyst and a rising CA125 underwent a diagnostic laparoscopy. This showed no ovarian abnormality, and a biopsy of an enlarged fallopian tube proved benign. One woman with a 5-cm complex cyst and normal CA125 elected to have a total abdominal hysterectomy and bilateral salpingo-oophorectomy. Histology showed a benign mucinous cystadenoma and a Brenner cell tumour. No complications occurred during surgery.
There were additional incidental findings on ultrasound. Thirteen women had endometrial polyps or a thickened endometrium of greater than 5 mm. These women were referred back to their GPs, and nine of them were managed conservatively as they had no abnormal vaginal bleeding. Four women underwent hysteroscopy and endometrial biopsy, and the pathology from all was benign. Two women had abnormal bladder findings and underwent cystoscopy. One woman suffered from pelvic pain, constipation, decreased appetite and fatigue. She reported no urinary symptoms but was found to have rectovaginal and vesicovaginal fistulae related to severe diverticular disease and went on to have a colostomy. A second woman suffered from abdominal and back pain as well as urinary frequency and was found to have prolapsed ureteric mucosa that was resected.
The distribution of CA125 results in 287 women who had CA125 tests performed is shown in Figure 1. Three women did not have blood tests or samples proved inadequate. Thirteen of 287 women had CA125 levels raised above 35 units/ml. Three of these women had abnormal ultrasound scans and are discussed above. One further woman was found to have a uterine fibroid which explained her symptoms and was referred to a gynaecologist. She was managed conservatively and had no cancer on follow up for 32 months. The remaining nine women with a CA125 > 35 units/ml had normal ultrasound scans. Two women suffered from rheumatoid arthritis, and this was presumed to be the cause of the raised level in the presence of a normal scan. Seven women had repeat levels after approximately 6 weeks, and the levels were found to have fallen or were stable. They had no further intervention. No ovarian cancer was found on median follow up of 23 months (range 15–33 months).
There are an increasing number of reports that most women with ovarian cancer suffer from nonspecific symptoms many months prior to their diagnosis. These reports raise the possibility that significant delays in diagnosis could be avoided by an early referral for an ultrasound scan or CA125 blood test.3–15 If so, screening women aged more than 45 years with nonspecific symptoms related to the urinary and gastrointestinal systems using serum CA125 and pelvic ultrasound may be a method of achieving earlier diagnosis. This is the first report of a prospective, pilot randomised controlled trial to assess the feasibility of such an approach. The pilot data provide important information on the feasibility of this approach, referral rates, patient compliance, incidence of false-positive results and surgical intervention rates.
At the start of the study, GPs were informed of the findings of the relevant recent studies and were encouraged to refer women with the full spectrum of symptoms in whom a firm diagnosis had not been reached. It was explained that while 77% of women with ovarian cancer reported abdominal symptoms, 70% also reported gastrointestinal symptoms and 50% constitutional symptoms. In our study population of 290 women, 87% had abdominal symptoms, whereas only 41 and 29% had gastrointestinal and constitutional symptoms, respectively, suggesting that GPs continued to focus on the traditional teaching from textbooks. More research into the best techniques for educating GPs is required if these symptoms are to be associated with a diagnosis of ovarian cancer.
The symptoms reported by women with ovarian cancer are nonspecific, and 95% of women visiting primary care physicians report at least one symptom that could represent ovarian cancer in the previous year.9 At the start of the study, there was concern that the numbers of women referred might exceed the resources available. In fact, referrals were only received from 39 of 79 practices randomised into the study group. This translated into 98 of 197 (49.8%) of the GPs. Some GPs sent more than ten women, while the majority sent none. This variation persisted despite visits, emails and letters reminding GPs about the study. GPs in smaller practices were less likely to refer women to the study. While referral patterns may reflect to some extent the patient profiles of individual GPs, i.e. proportion of women aged more than 45 years, there are clearly factors that may limit targeted screening if it were to be offered only through doctors. In future trials, it is worth exploring the use of patient information leaflets or symptoms questionnaires when women attend general practices.
92 percent of women referred attended for a screen. This is higher than the 80% attendance rate at NHS gynaecology and ultrasound clinics in this deprived area of London. It suggests that compliance with targeted screening will be high and that women are motivated to attend.
The specificity in this study was 99%, and the intervention rate was acceptable. Three women of 290 (1%) had surgery for an ovarian abnormality. Two had bilateral salpingo-oophorectomy for symptomatic benign ovarian tumours, and a further woman had a diagnostic laparoscopy, with no significant pathology detected. It needs to be emphasised that these rates were achieved in the context of a trial centre which has more than a decade of experience in ovarian cancer screening and hence significant expertise in the conservative management of ovarian abnormalities in older women.
It is also important to note that in 33 (11%) women, the tests required repeating. This is critical in achieving high specificity. It is important that if the tests are carried out by GPs that both the doctors and women are aware that one in ten tests will need to be repeated. This number is likely to rise if the symptom profile is expanded to include other symptoms.
The incidence of ovarian cancer in women more than 45 years of age is 1 in 2000 per year.16 The existing data suggest that the relative risk of ovarian cancer in women complaining of abdominal bloating is 25.3 (95% CI: 15.6–40.9).10 Based on this, the expected number of women with ovarian cancer is approximately 1 in 80 in a year for a population with abdominal bloating. No woman in this cohort of 290 was diagnosed with ovarian cancer within 1 year of the test. This may be due to the small size of the cohort or a lower relative risk for ovarian cancer in women with these symptoms than previously thought. A campaign is currently underway by ovarian charities to encourage women with abdominal bloating, increasing abdominal size and urinary frequency to visit their GPs so that CA125 and pelvic scans can be arranged. It is important to emphasise that it is still not known how many women will need to undergo testing for each case of ovarian cancer detected.
A complicating factor in pelvic ultrasound screening of the ovaries is the inadvertent detection of endometrial thickening. Of the 15 women who had incidental findings, 13 were endometrial abnormalities. It is reasonable to limit further investigation to those who are symptomatic or with particularly suspicious findings (e.g. endometrium > 10 mm), as most women have benign pathology.17 Other incidental findings warrant investigation, and the remaining two women with bladder abnormalities on ultrasound had significant pathology.
There is increasing demand from women and ovarian cancer support groups that the problem of diagnostic delay in women with ovarian cancer be addressed. This study confirms that women with nonspecific symptoms are willing to undergo ovarian cancer screening. However, it remains unclear how many women would need to be referred for each case of ovarian cancer detected, what are the best methods of educating GPs and women about symptoms and what are the resource implications of such a strategy. This study provides the basis for definitive trials to explore the impact on ovarian cancer diagnosis of screening women with abdominal, gastrointestinal, urinary or constitutional symptoms with serum CA125 and ultrasound. Any trial would need to address issues related to repeat testing, unnecessary surgery, quality of life and resource implications. It remains entirely unclear and uncertain whether or not earlier intervention by targeted screening can achieve a stage shift survival improvement or mortality reduction.
Our deepest appreciation to all the women who participated in the trial and to all the GPs whose collaboration was essential to the study. Mr Barnaby Rufford was supported by grants from Eve Appeal and The Drapers’ Company.
- 16Cancer Registration Statistics. 2003. Office of National Statistics, London, UK, 2003.