Malignancy

Gynecologic

   Ovary
   Uterus

Nongynecologic

   Pancreas
   Stomach
   Colon
   Rectum

Nonmalignancy

Gynecologic

   Benign Ovarian Neoplasm
   Endometriosis
   Pregnancy
   Leiomyomata
   Salpingitis

Nongynecologic

   Cirrrhosis
   Pancreatitis
   Renal failure

Patient A has a Ca-125 test result of 65 u/ml; a normal value is less than 35 u/ml. She almost certainly has ovarian cancer.

Patient B also has a Ca-125 test result of 65 u/ml. She almost certainly does not have cancer.

Patient C has a Ca-125 test result of 12 u/ml. The likelihood of finding ovarian cancer at surgery is greater than 50%.

Patient D has had her specimen dumped down the drain. She is told that her test result is normal. This is correct 99.97% of the time.

How can the same test and test results have almost the opposite interpretations? Is this a problem with the sensitivity or specificity of the test? What is the predictive value of the test? What do these terms mean, anyway? Consider the following table displaying the test results with the presence or absence of disease:

Sensitivity is the ratio of all those truly positive to all those who have the disease:

A
---------
A+C

Specificity is the ratio of all those truly negative to all those without the disease:

D
----------
B+D

Since we have only one patient and one test result, these numbers are not much help. What we want to know is the likelihood that an elevated Ca-125 indicates the presence of disease and the likelihood that a normal result indicates absence of disease. We need the positive predictive and negative predictive values.

The positive predictive value is the probability that a positive test indicates presence of disease:

A
--------
A+B

The negative predictive value is the probability that a normal value indicates absence of disease:

 D
--------
C+D

According to Bayes’ Theorem, the positive and negative predictive values are proportional to the prevalence of disease in the population being tested. The greater the prevalence the higher the positive predictive value. The lower the prevalence the higher the negative predictive value.   

It matters not the test or the disease. When the prevalence of disease is low most of the positive results are falsely positive rather than truly positive, whereas the negative result is also almost always true. When the prevalence is high most people have the disease so a positive result is almost always true.

Does ovarian cancer and the Ca-125 test follow Bayes’ Theorem? Yes. It can be demonstrated by calculating the positive and negative predictive values for three groups of patients: those with a high prevalence of ovarian cancer, an intermediate group and a group with a low prevalence.

The group with the highest prevalence of ovarian cancer contains those already diagnosed with ovarian cancer. They have completed chemotherapy with a complete clinical response and are to undergo a second look re-staging surgery. Three studies have been combined describing 484 patients and their second look results.(1,2,3) Of the 484 patient, 315 (65%) were found to have persistent cancer.

Sensitivity = 54%   Specificity = 96%

(+) Predictive Value = 96% (-) Predictive Value = 53%

An elevated Ca-125 in this group of women reliably indicates persistent cancer. A normal test is wrong one half of the time because it does not correlate well with small volume disease.

An intermediate risk group can also be obtained by combining three studies (4,5,6). This group consists of those with pelvic masses who are to be operated. Of 495 patients, 218 were found to have cancer (44%).

Sensitivity = 72% Specificity = 78%

(+) Predictive Value = 72% (-) Predictive Value = 78%

The low prevalence population consists of self-referred volunteers who had Ca-125 determination, which, if persistently elevated, were explored (7). There were 1,082 patients screened. Thirty-six had persistently elevated levels. One cancer was detected.

Sensitivity = 100% Specificity = 97%

(+) Predictive Value = 2.8% (-) Predictive Value = 100%

In this population with an incidence of ovarian cancer of only 0.1%, a positive test is correct only 2.8% of the time. During the duration of the study no one with a negative Ca-125 turned up with cancer so it is assumed that none was present.

To go back to the beginning of this article, Patient A was previously diagnosed with stage III ovarian cancer and had an initial Ca-125 of 1,250 u/ml. She had a complete clinical response to chemotherapy and had a normal Ca-125. Within 12 months the Ca-125 became elevated to 65 u/ml. It is almost a certainty that she has recurrent/persistent cancer. In this population of women 85% will eventually die of ovarian cancer.

Patient B is 40 years old. She read about ovarian cancer being the silent killer and that a blood test was available to detect it. She is relatively asymptomatic and has a normal examination. Her elevated Ca-125 is almost certainly due to something other than ovarian cancer. In this population of women the incidence of ovarian cancer is about 0.03%, whereas the conditions that can cause an elevated Ca-125 are cumulatively about 2%.

Patient C had stage III ovarian cancer with an elevated Ca-125. She achieved a complete clinical response to chemotherapy, including a normal Ca-125. Depending on the amount of residual cancer left after her initial surgery, the likelihood of finding persistent cancer at second look is 30-80%.

Patient D is part of a group of 3,000 women who are asymptomatic and have normal examinations. They are participating in a screening program for ovarian cancer. All the specimens are dumped down the drain and reported as normal. On average, only one will have a "false negative" test. Consider the following:

There are about 25,000 new cases of epithelial ovarian cancer diagnosed each year in the United States. Two-thirds will be stage III and IV at diagnosis. Most of these 25,000 women will have symptoms or physical findings suggesting a pelvic problem and need a diagnostic procedure. Even if you assume that about 15,000 of these women have no symptoms or findings and are eligible for screening and that there are about 45,000,000 women in this age group, then the maximum incidence is only about 1 in 3,000 per year.

If 3,000 normal asymptomatic women are screened with a Ca-125, on average only one will have a cancer. But, probably 100 will have an elevated Ca-125, since it is elevated in a variety of conditions. Any condition that causes ascites will have an elevated Ca-125. Any condition that causes irritation to the peritoneum, fallopian tubes, ovaries or uterus can cause an elevated Ca-125. It is commonly elevated with endometriosis. I have seen values exceeding 2,000 u/ml in cirrhosis with ascites, pelvic infections and in one case of fallopian tube torsion. One percent of men and women has an elevated level for no apparent reason.

The result of a Ca-125 test is interpretable only by considering the context in which it was ordered. When you order a Ca-125 test you will have to estimate your patient’s risk for having ovarian cancer. If your patient can be put in a group in which the likelihood of cancer is high then a positive test is probably correct and a negative test wrong. If your patient can be placed in a low risk group then the positive test is probably wrong and the negative test meaningless. Furthermore, there is no way to evaluate a positive test. You can repeat the test and pick the best 2 out of 3; 3 out of 5; 4 out of 7, etc. Otherwise, she will be heading for surgery.

References:

Gyn Oncol 38:373, 1990

Gyn Oncol 38:181, 1990

Amer J Ob Gyn 160:667, 1989

Amer J Ob Gyn 159:873, 1988

Amer J Ob Gyn 159:341, 1988

Ob Gyn 72:159, 1988

Gyn Oncol 36:299, 1990

William M. Rich, M.D.
Gynecologic Oncology Associates