Algorithm Optimizes Value of CA125 II Screening for Ovarian Ca

June 1, 1995

WASHINGTON--Applying a newly developed mathematical model to an updated version of a familiar tumor marker may increase the chance of finding ovarian cancers at an earlier stage, Robert C. Knapp, MD, said at the American Cancer Society Conference on Gynecological Cancers.

WASHINGTON--Applying a newly developed mathematical model to anupdated version of a familiar tumor marker may increase the chanceof finding ovarian cancers at an earlier stage, Robert C. Knapp,MD, said at the American Cancer Society Conference on GynecologicalCancers.

"If we could detect ovarian cancer early in asymptomaticwomen," said Dr. Knapp, professor emeritus of gynecology,Harvard Medical School, "it would make a significant differencein reducing the horrible mortality from this disease."

Currently, three quarters of ovarian cancers are diagnosed atlate stages, when the 5-year survival rate is only 20%. Womenwhose cancers are found at stage I, however, have a 90% chanceof surviving at least 5 years.

Evaluate Changes Over Time

Dr. Knapp said that by evaluating changes in several CA125 radioimmunoassaystaken over a period of time, using the updated CA125 II test (Centocor,Malvern, Penn), the new algorithm has improved the assay's positivepredictive value (PPV) for the risk of ovarian cancer from approximately2% to 16%.

The new CA125 II radioimmunoassay utilizes M11, a high-affinityantibody, in combination with the OC125 antibody. Its standardcurve is stronger and sharper than the original CA125, enhancingresolution and reducing fluctuations, Dr. Knapp said.

The algorithm looks at the initial level of the CA125 marker,as well as the exponential slope of the line connecting the markerreadings. It also takes into account any assay variability thatmight produce false readings in the short term, said biostatisticianSteven Skates, PhD, who developed the algorithm and was lead investigatorof the study reported at the meeting.

A single assay does not have the same predictive power as severalassays looked at together. With use of the algorithm, "weend up with a very specific screening test, with fewer false positives,that helps identify a high-risk group for further intensive evaluation,"said Dr. Skates, assistant professor of medicine and biostatistics,Harvard Medical School.

Dr. Knapp, who helped develop the original CA125 assay in 1983,pointed out that it was never intended as a screening tool. "Wewere looking then only at patients who had already been diagnosedwith ovarian cancer," he said.

He noted that to be an effective screening tool for ovarian cancer,a minimal requirement is a PPV of at least 10%. A single measurementof CA125 has a sensitivity of 58%, a specificity of 98%, and aPPV of just 2%.

"We asked ourselves: How can we improve the positive predictivevalue of this tumor marker?" Dr. Knapp said. "Our minimalgoal was at least 75% sensitivity, at least 99.6% specificity,and a PPV of at least 10%."

The Harvard researchers, along with coinvestigators at the Universityof Texas M.D. Anderson Cancer Center and the Karolinska Institute,Stockholm, developed the algorithm by collecting blood samples,from 1986 to 1990, in a group of 4,315 healthy postmenopausalSwedish women older than age 50 who were not known to have ovariancancer at the time the blood was collected.

The women were divided into two groups: Half were used to derivethe algorithm, while the other half served to validate it in astudy that retrospectively measured levels of CA125 in their bloodsamples.

The algorithm was used to analyze the assay results, which werethen correlated with the women's incidence of ovarian cancer through1993.

The results showed a specificity for CA125 II of 99.7%, an apparentsensitivity of 83%, and a PPV of 16%--meeting Dr. Knapp's minimumcriteria for an effective screening marker. The 16% PPV meansthat 16 of every 100 women with a positive test according to thealgorithm will subsequently be diagnosed with ovarian cancer.

Large UK Trial Planned

Ian Jacobs, MD, of St. Bartholomew's Hospital, London, England,is running a trial of CA125 II with the new algorithm and a newtumor marker, OVX1. Women with positive results are sent for ultrasoundtests, he said. If those are also positive, the patient is scheduledfor surgery.

If the preliminary results using the new algorithm, which he calls"encouraging," hold up, Dr. Jacobs plans to conducta randomized clinical trial in Britain lasting 7 years and involving120,000 healthy postmenopausal women.

"We think that this model should increase sensitivity withoutsacrificing specificity. We'll try to see if it will shift thestage distribution of cases," Dr. Jacobs commented.

More Studies Needed

In their conference talks, all three researchers cautioned thatmuch work remains to be done before the system can be judged validfor screening purposes. Many years passed before even the Papsmear was shown to improve survival, Dr. Knapp noted.

"We are concerned because in the past, screening was usedwithout showing that it reduced mortality," Dr. Jacobs said."And a test should not be used unless it demonstrates itsvalue in saving lives."

Dr. Knapp agreed: "We are not recommending at this time thatthe CA125 II algorithm be used for screening women for ovariancancer. The judgment of a physician, a clinical examination, anda history are still basic to managing the patient."