Screening for Ovarian Cancer: What We Know, What We Need to Know

Screening for Ovarian Cancer: What We Know, What We Need to Know

ABSTRACT: The majority of women with ovarian cancer present with advanced-stage disease. Women with early-stage ovarian cancer have a much better chance of achieving a cure than do women with late-stage disease. This difference makes screening for ovarian cancer, with the hope of detecting it in its presymptomatic state, an attractive concept. Unfortunately, efforts to demonstrate that screening for ovarian cancer in the general population can decrease mortality have been disappointing. Current screening techniques do not have high enough sensitivity and specificity to be applied to the general population, because the low prevalence of the disease in the general population leads to very low positive predictive values for the available screening tests. However, applying current screening strategies to certain high-risk populations (women who carry mutations in the BRCA1 or BRCA2 genes, or with strong family histories of breast/ovarian cancer) is a reasonable approach and may result in acceptably high positive predictive values. This article discusses the results of screening studies using serum CA-125, sonography, other serum markers, and combinations of these tests. Screening for women of average risk is not recommended, although such women should be encouraged to participate in clinical trials whose end points are either the demonstration of the impact of screening on mortality, or the development of novel screening strategies. Screening with twice yearly transvaginal sonography and serum CA-125 testing is recommended for women at high risk for ovarian cancer, although prospective data are needed regarding the impact of such screening on stage of cancer detected, quality of life, and psychological distress, as well as the costs—both personal and societal—of screening. [ONCOLOGY 14(11):1601-1607, 2000]


Screening for certain cancers has been shown to increase
disease-specific and overall survival, thus leading to
recommendations that certain screening tests (eg, Pap smears, annual
mammography for women over 50 years of age) be offered as standard
care for the general population. When a screening test is being
considered for use, the condition being screened for should meet the
following criteria:

(1) The prevalence of the disease in the general population should be
relatively high.

(2) The disease should cause substantial morbidity and mortality.

(3) Earlier diagnosis of the condition should result in an improved

(4) There should be an effective screening test for detection of the

It is useful to apply these criteria to the issue of screening for
ovarian cancer. More than 25,000 women are diagnosed with ovarian
cancer annually in the United States and approximately 14,500 women
die from the disease each year. Both the incidence and associated
mortality of ovarian cancer appear to be increasing. In the general
population, the lifetime risk of developing ovarian cancer is
approximately 1.4%. However, among certain women, the risk is much
higher. One of the strong-

est risk factors for developing the disease is having a close family
member with ovarian cancer. The relative risk among women with a
single first-degree relative with ovarian cancer is 3.1 (95%
confidence interval: 2.6–3.7),[1] and if more than one relative
has been affected, the relative risk is higher.[2]

Although shared exposures and sociocultural characteristics such as
child-bearing choices may explain some of the increased risks within
families, a genetic tendency to develop ovarian cancer is likely to
be present in families with multiple cases of ovarian cancer. For
example, women who are known to be carriers of germ-line mutations in
the breast/ovarian cancer predisposition genes, BRCA1 or BRCA2,
have a lifetime risk of developing ovarian cancer that is estimated
to be between 16% and 65%.[3,4] It is likely that current ovarian
cancer screening techniques may be most useful in this population of
women in whom prevalence is relatively high.

Ovarian cancer clearly causes substantial morbidity and mortality,
with 5-year survival averaging approximately 45%. The disease also
meets the third screening criterion, since 5-year survival among
women with stage I or stage II ovarian cancer is 80% to 90% vs 5% to
50% for women with stage III or IV. Nevertheless, screening efforts
for the detection of ovarian cancer in the general population have,
to date, been disappointing—the result of the lack of a highly
effective screening test, and the low prevalence of the disease.

Evaluating Screening Tests

Sensitivity vs Specificity

The effectiveness of any screening test may be measured in terms of
certain test parameters. Sensitivity refers to the ability of the
test to detect the cancer. Specificity refers to the ability of the
test to rule out cancer. Using the letter designations in Table
, sensitivity is defined as the true-positive patients (a),
divided by all the patients in the population who truly have cancer,
ie, the true positives plus the false negatives (a + b). Specificity
is defined as the true negatives (d) divided by all the patients who
truly do not have cancer, ie, the true negatives plus the false
positives (c + d).

There is a trade-off between sensitivity and specificity, such that
no test possesses both 100% sensitivity and 100% specificity. Highly
sensitive tests are unlikely to miss cancers, but many patients will
undergo further testing or treatment for benign disease. Tests that
are highly specific will give fewer false-positive results but will
falsely reassure some patients that they are cancer-free.

Positive Predictive Value

Equally important for evaluating the effectiveness of screening tests
is the prevalence of the disease in the population to be screened;
the positive predictive value (PPV) of a test incorporates this
parameter. The importance of disease prevalence in determining the
PPV can be seen in the equation for its calculation (where S =
sensitivity and P = prevalence:

  PPV = (S ´ P) / [(S ´ P) + (1 – S)(1 –

Thus, the PPV value will vary among different screening populations, even
when sensitivity and specificity are the same. For example, if an
elevated serum CA-125 value has a sensitivity of approximately 80%
for ovarian cancer and a specificity of 99% in the general
population, with an ovarian cancer prevalence of 0.0003, its PPV is
2.3%. In a high-risk population with an ovarian cancer prevalence of
0.005, the PPV of an elevated CA-125 is 28.7%.

The low PPV in the general population raises concerns that screening
the general population will generate many false-positive results that
require surgical evaluation. For example, in the general population,
50 women would need to undergo laparoscopy/laparotomy to find one
ovarian cancer. It has been suggested that a PPV of 10% is
“reasonable” for an ovarian cancer screening strategy.
Adopting a strategy with a PPV of 10% would mean that 10 women would
undergo invasive surgery for each ovarian cancer found; ie, in order
for a test to have a PPV of 10% in the general population (prevalence
of 0.0003, sensitivity of 83%), its specificity must be greater than 99.7%.

Ovarian Cancer Screening Tests

Ovarian cancer screening studies have been carried out in a variety
of patient populations, utilizing one-time, serial, or single
screening tests, or combinations thereof. However, many studies have
been limited to postmenopausal women, in whom false-positive test
results turn up less frequently (and in whom ovarian cancer incidence
is higher) than in premenopausal women. Some studies have excluded
women with strong family histories of ovarian cancer. Most studies
have not had a no-screening control group and, therefore, could not
assess the impact of screening on mortality.

Serum CA-125 Levels

Measurement of serum CA-125 is an attractive screening test for
ovarian cancer because it is readily available and noninvasive.
Levels of this antigen are elevated in approximately 90% of women
with stage II, III, and IV ovarian cancer, but in only 50% of women
with stage I disease.[5,6] Moreover, the test’s lack of
specificity limits its use, as CA-125 levels are also elevated in
many common benign conditions, including liver disease, fibroids,
endometriosis, and ovulation. In a study of 1,010 postmenopausal
women, 31 were found to have a CA-125 level > 30 IU/mL, and one of
these 31 was diagnosed with stage IA ovarian cancer. The specificity
of the serum CA-125 test was 97%.[7]

In a case-control study using stored serum from 20,305 blood donors,
37 ovarian cancers were diagnosed over 15 years of follow-up. The
sensitivity of the CA-125 level to detect these cancers was only 24%,
and specificity, 96%.[8] Skates et al observed that women with
ovarian cancer are more likely than women without ovarian cancer to
have increasing CA-125 levels over time.[9] Researchers at St.
Bartholomew’s Hospital are investigating the rate of change of
the CA-125 level in serial determinations, to see if the test’s
specificity improves. Preliminary data suggest that the rate of
change of the CA-125 level has a sensitivity of 83% and specificity
of 99.7% among postmenopausal women, although these data are derived
from only six cases of ovarian cancer in the cohort.[9]

Other Tumor Markers

• Lysophosphatidic acid (LPA) is another tumor marker
that is reportedly increased in the serum of women with ovarian
cancer. The LPA level may be elevated more often than the CA-125
level in early-stage ovarian cancer. In one small study, serum LPA
was elevated in 9 of 10 women with stage I cancer, whereas CA-125
levels were increased in only 2 of 9 cases.[10] In more advanced
ovarian cancer, LPA has not been shown to be more sensitive than

• Macrophage–colony-stimulating factor (M-CSF) levels
are reported to be complementary to CA-125 levels in detecting
ovarian cancer. Among 25 patients with known ovarian cancer and
normal CA-125 levels, M-CSF was elevated in 56%. Elevated M-CSF
levels were found in 31% of 29 women with normal CA-125 levels and
positive findings for cancer at second-look surgical evaluation.[11]

• Serum OVX-1 is another tumor marker that may be
elevated among patients with ovarian cancer.[12] In one study of
patients with normal CA-125 levels who underwent second-look
laparoscopy after primary treatment for ovarian cancer, 59% had
elevated OVX-1 levels.[13] However, another study found that only 65%
of 46 patients with stage I ovarian cancer had elevated OVX-1
levels.[14] Currently, stability and assay reliability problems make
serum OVX-1 levels difficult to adapt for screening.

• Combinations of serum markers can increase the
sensitivity of serum testing for the presence of ovarian cancer,[14]
but at the expense of decreased specificity.

Sonography as a Screening Tool

• Transabdominal pelvic sonography was evaluated as a
screening tool in 5,540 women in England.[15] Each woman was to
undergo three annual sonographic evaluations. A total of five ovarian
cancers—all stage IA—were detected, three of which were
borderline tumors. The specificity of the test was 94.6%, but the PPV
was only 2.6%. More than 25% of the false-positive sonograms had no
ovarian pathology at exploration, and 74.3% had benign ovarian
abnormalities. With this approach, 51 surgical explorations were
required for each cancer found.

• Transvaginal sonography provides a clearer morphologic
assessment of the ovaries, compared to pelvic ultrasound. In a study
of 1,300 postmenopausal women,[16] 33 had abnormal sonograms, and 27
agreed to surgical exploration. Of these 27 women, 2 were found to
have stage IA ovarian cancer. A larger study of 3,220 postmenopausal
women[17] used a specific morphology index to determine whether the
sonogram was considered abnormal. In this study, 44 women had
morphologic index-defined abnormalities. At laparotomy, 3 of the 44
were found to have ovarian cancer.

These promising results, with only 15 laparotomies required for each
cancer found, led to a larger study of 8,500 women.[18] Women with
abnormal transvaginal ultrasounds underwent the procedure again
within 4 to 6 weeks. Those with persistent abnormalities had a pelvic
examination, serum CA-125 testing, color Doppler transvaginal
ultrasound, and morphologic index assessment. Persistent
abnormalities that required surgical evaluation were found in 121
women on transvaginal ultrasound. Of these 121 women, 8 had ovarian
cancer, 5 of which were epithelial ovarian cancers (4 of the 5 were
stage I or II), and 3 of which were granulosa cell tumors. In this
study, the PPV was lower, with three epithelial ovarian cancers found
among the 121 women requiring surgical exploration for abnormal
transvaginal sonograms.

Lack of specificity, particularly in premenopausal women, is a major
problem in using transvaginal sonography for screening the general
population, with specificity reported as 98.1%[16] to 98.7%,[17]
which means PPV would be low in the general population. In an effort
to increase the pretest probability that an ovarian abnormality would
be cancerous, Bourne et al limited their sonographic screening to
women with at least one first- or second-degree relative with ovarian
cancer.[19] Of 776 women screened, 43 had an abnormal sonogram. At
surgical evaluation, three stage IA ovarian cancers were detected,
with one being a borderline tumor. The PPV was 7.7%.

• Transvaginal sonography with color Doppler may improve
the specificity of sonographic screening for ovarian cancer, based on
the premise that the vascularity of malignant tissue has a lower
impedance to blood flow than benign tissue.[20] This combination of
imaging techniques was evaluated in 14,317 women,[21] and all adnexal
masses were investigated surgically, independent of the Doppler
impedance results. Among the 680 women who required surgical
evaluation, ovarian malignancies were found in 56. Of these 56
malignancies, 54 had low impedance on Doppler. Adding Doppler imaging
to transvaginal sonography decreased sensitivity to 96.4%, but
increased specificity to 99.8%.

In a study of 1,601 women with a family history of ovarian cancer who
referred themselves for screening, Bourne et al[22] found that 61 had
abnormal sonograms, based on morphologic and Doppler characteristics.
All 61 underwent surgical exploration. Ovarian cancer was found in
six; three of these six were borderline tumors, and one was stage
III. Thus, the PPV for sonography with Doppler for detecting invasive
ovarian cancer, even in this higher-risk population, remained low.

Karlan et al summarized five of the largest studies of sonography for
the detection of ovarian cancer. These trials used various forms of
sonographic imaging, including transabdominal sonography,
transvaginal sonography, transvaginal sonography with morphologic
index, and color Doppler. Taken together, the five trials enrolled a
total of 11,283 women. Surgical evaluation by laparotomy was required
in 486 women, with 22 ovarian cancers found. Of these, 13 were stage
I , and 5 of the 13 were invasive. The overall specificity was 95.8%,
and the PPV was 3.1%.[23]

An updated report by van Nagell et al was published recently.[24]
Transvaginal sonography was performed annually from 1987 to 1999 in
women over 50 years old as well as in women over 30 with a family
history of ovarian cancer. Women with abnormal sonograms had a repeat
sonogram 4 to 6 weeks later. Women with persistently abnormal scans
then had a third sonogram with morphology indexing and Doppler flow,
and a serum CA-125 test. Among 14,469 women, 180 (1.2%) had
sonographic findings requiring surgical evaluation. Ovarian cancer
was found in 17; 11 were stage I, 3 were stage II, and 3 were stage

Among patients with normal screening sonograms, four developed
ovarian or peritoneal cancer within 12 months of a normal sonogram
(false-negative screening test results). In this population of women,
annual transvaginal sonography, with work-up of abnormal sonograms as
specified above, yielded a sensitivity for ovarian cancer of 81%,
specificity of 98.9%, and PPV of 9.4%. The sensitivity for detecting
stage I epithelial ovarian cancer (excluding the borderline and
granulosa cell tumors) was 31%.

Serum Test/Imaging Combinations

Combinations of serum tests and imaging hold promise for improving
specificity. In the previously mentioned study of 1,010
postmenopausal women, the specificity increased to 99.8% when
sonography was performed in the 31 women with elevated CA-125 levels,
and surgical intervention was limited to those with abnormal findings
on both tests.[4]

In a subsequent study, 22,000 postmenopausal women had a CA-125
screening test, and those with levels > 30 IU/mL (n = 340)
underwent sonography. Sonographic abnormalities were seen in 41
women, who then underwent surgical investigation. Of these 41 women,
11 had ovarian cancer (3 had stage I, 1 had stage II, 7 had stage III
or IV). Among the remaining 21,959 women whose screening tests did
not reveal an abnormality, 8 subsequently developed ovarian cancer.
Thus, the sensitivity of the combination of CA-125 > 30 IU/mL and
an abnormal sonogram was 78.6% at 1 year and 57.9% at 2 years,
although specificity was 99.9%, and PPV was 26.8%.[25]

Einhorn studied 5,550 women over 40 years old with a serum CA-125
measurement and, in those with elevated CA-125 levels, a
sonogram.[26] Of the 175 women who had an elevated CA-125, 6
were found to have ovarian cancer. Importantly, three women with
normal CA-125 screening levels were subsequently diagnosed with
ovarian cancer. Among the women in the study 50 years of age or
older, the specificity of a CA-125 > 35 IU/mL was 98.5%; for those
aged 40 to 49 years, the specificity was 94.5%. The PPV of an
elevated CA-125 plus an abnormal sonography was 50% among women over
age 50 years.


Loading comments...
Please Wait 20 seconds or click here to close