For the first time after years of frustrating failures, a rigorous study has found that a blood test for ovarian cancer can save lives, scientists running the largest-ever such trial reported on Thursday in the Lancet.
The screening test — for a blood protein linked to ovarian cancer — was far from perfect, failing to detect a large fraction of ovarian cancers and flagging as possible cancers some physiological changes that were benign, as scores of women found out after they had their ovaries surgically removed. Even more women whose ovarian cancer was detected early died anyway. But the finding that screening can avert one in five deaths from ovarian cancer, nevertheless, offered a long-sought ray of hope for a disease that kills some 60 percent of its victims within five years of diagnosis.
“It’s not time to go off to the races yet and make this screening routine, but these results do lift our eyes to a horizon where it’s possible to effectively screen for ovarian cancer,” said Robert Smith, director of cancer screening at the American Cancer Society, who was not involved in the study.
article continues after advertisement
Risk of false alarms
The 14-year UK Collaborative Trial of Ovarian Cancer Screening, led by researchers at University College London, was the last real hope for women, physicians, and scientists searching for a way to identify ovarian cancers early, when treatment (surgery, chemotherapy, radiation, hormone therapy) is more likely to succeed. In general, only 30 percent of women with ovarian cancer are diagnosed early, usually from symptoms such as abdominal bloating, pain, or lumps.
A 2007 study failed to find that testing for the blood protein called CA-125 detected ovarian cancer significantly earlier than no screening. That study has not yet reported whether CA-125 screening saved lives. A 2011 study called the Prostate, Lung, Colorectal, and Ovarian trial did not find that screening detected cancers earlier, let alone help more women survive the disease.
As a result of these and other negative findings, both private and government-sponsored medical groups strongly recommend against screening healthy women for ovarian cancer, warning that it doesn’t help and can hurt, mostly by raising false alarms. (Monitoring CA-125 levels in women already diagnosed with ovarian cancer, however, can track how well treatment is working.)
The UK trial was the only one left standing, and investigators were determined to give screening the best shot they could. In addition to enrolling more women — 202,638, ages 50 to 74 — than any other, they used a sophisticated algorithm to interpret the CA-125 results. Rather than using a blunt cutoff, interpreting any CA-125 level above that as a red flag (which the 2011 study did), it also factored in a woman’s age, risk of the disease, and other clues. “This algorithm improved the test’s sensitivity and especially its specificity,” said the cancer society’s Smith, making it less likely to miss real cancers or incorrectly flag absent ones.
The women were randomly assigned to one of three groups: annual transvaginal ultrasound, which tries to detect suspicious masses; an annual blood test for CA-125 plus ultrasound; or no screening.
At first blush, it seemed that screening failed again: women receiving CA-125 plus ultrasound had a 15 percent lower risk of dying from ovarian cancer than unscreened women from 2001 through 2014 — 29 percent versus 34 percent. But the benefit was so small it could have been due to chance. That is, it was not statistically significant.
But digging deeper turned up something more encouraging. When the scientists focused only on the last seven years, they found that deaths from ovarian cancer were 23 percent lower in the CA-125 group than the unscreened one. That result was statistically significant, or less likely to be a fluke. And when the researchers excluded women who (unbeknownst to anyone) had ovarian cancer from the start of the study, screening also showed a mortality benefit of about 20 percent, and 28 percent in years seven through 14.
Playing statistical games is considered a no-no in clinical trials, but in this case it was justified, Smith and other outside experts said. That’s because the British scientists said at the start that they were going to analyze their data this way (rather than fishing around later for a statistical approach that would show a benefit). Also, it’s only logical that women who already have cancer are unlikely to be helped by early detection; they’re typically well past “early.”
“If you can save one in five women who would otherwise die of ovarian cancer that’s great,” Smith said. “And with continued follow-up the results should be even better.”
The future of screening
The American Cancer Society estimates that 14,180 women in the United States will die of ovarian cancer in 2015.
To avert one death from ovarian cancer, 641 women had to be screened annually for 14 years, found the study, which was led by Dr. Ian Jacobs of University College London and Steven Skates of Massachusetts General Hospital. In preventive medicine, that is considered a reasonable “number needed to screen.”
CA-125 was, nevertheless, far from a “home run,” said Dr. Karen Lu, chair of gynecologic oncology at MD Anderson Cancer Center, who was not involved in the UK study. It detected only 59 percent of ovarian cancers (the rest were found when the women had symptoms). And for every 10,000 CA-125-plus-ultrasound screenings, 14 women underwent ovary-removing surgery only to be told they never had cancer.
Still, Lu said, “there has never before been a study that shows screening results in more early-stage ovarian cancers being detected and in fewer deaths. I see this as incredibly compelling and as bringing us closer to screening women in the US.”
That won’t happen tomorrow, she and others said. Standardization and quality control for CA-125 have to improve: as it is, Lu said, readings from two different labs can differ by 100 percent. Doctors must be educated not to use CA-125 alone but, like the British study, pair it with a smart yardstick to reduce the chances of crying wolf. And insurers and government programs such as Medicare will “have to be persuaded by the data,” Smith said.