|Home | About | Journals | Submit | Contact Us | Français|
To test whether annual screening with transvaginal ultrasound and CA-125 reduces ovarian cancer mortality.
Data from the first four annual screens, denoted T0–T3, are reported. A CA-125 value at or above 35 U/ml or an abnormality on transvaginal ultrasound was considered a positive screen. Diagnostic follow-up of positive screens was performed at the discretion of participants’ physicians. Diagnostic procedures and cancers were tracked and verified through medical records.
Among 34,261 screening arm women without prior oophorectomy, compliance with screening ranged from 83.1% (T0) to 77.6% (T3). Screen positivity rates declined slightly with transvaginal ultrasound, from 4.6 at T0 to 2.9–3.4 at T1–T3; CA-125 positivity rates (range 1.4–1.8%) showed no time trend. Eighty-nine invasive ovarian or peritoneal cancers were diagnosed; 60 were screen detected. The positive predictive value (PPV) and cancer yield per 10,000 women screened on the combination of tests were similar across screening rounds (range 1.0–1.3% for PPV and 4.7–6.2 for yield); however, the biopsy (surgery) rate among screen positives decreased from 34% at T0 to 15–20% at T1–T3. The overall ratio of surgeries to screen detected cancers was 19.5:1. Seventy-two percent of screen detected cases were late stage (III/IV).
Through four screening rounds, the ratio of surgeries to screen detected cancers was high, and most cases were late stage. However, the effect of screening on mortality is as of yet unknown.
ClinicalTrials.gov, www.clinicaltrials.gov, NCT00002540
Carcinoma of the ovary continues to be the leading cause of death from gynecologic malignancies in the United States. In 2007, it is expected that 22,400 women in the United States will develop this disease and over 15,200 will die, resulting in a very high case fatality rate . Women diagnosed with early stage (I/II) ovarian cancer have substantially better survival rates compared to those diagnosed in later stages . The hypothesis behind screening for ovarian cancer is that earlier detection of the disease will result in more women being diagnosed at less advanced stages, and that this “stage shift” will translate into lower overall mortality rates for ovarian cancer.
Both transvaginal ultrasound (TVU) [3–5] and the serum biomarker CA-125 [6–8] have been utilized to screen for early ovarian cancer. Studies to date however have not demonstrated a clear impact of screening with TVU or CA-125 on mortality from ovarian cancer. The Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial has as one of its objectives to estimate, in healthy women age 55 through 74 at entry, whether screening with both CA-125 and TVU can reduce mortality from ovarian cancer. Results from the initial screening round for ovarian cancer for the 39,115 women randomized to the intervention (screening) arm of the PLCO trial were reported earlier . The salient findings from that report were that a large number of oophorectomies were performed relative to the number of invasive cancers detected (about a 30:1 ratio), and that the majority of the cancers detected (80%) were stage III and above. The present report extends that initial analysis to the first four rounds of screening in the trial, and covers test compliance, screen positivity results, diagnostic follow-up, and numbers and characteristics of cancers detected. In addition, the report also examines interval cancers, i.e., those cancers not diagnosed through screening. Of specific interest, the analysis of these data from the subsequent screening rounds will enable a determination of whether the above-cited trends, namely, the high ratio of surgeries to detected cancers and the unfavorable stage distribution of the screen detected cancers, persist past the initial round of screening.
The design of the PLCO Trial has been described in detail elsewhere . Briefly, the objective of the ovarian cancer component was to estimate, in healthy subjects aged 55–74 at entry, whether screening with CA-125 and TVU reduces mortality from ovarian cancer. Prostate, lung, and colorectal cancers are also being studied.
In this trial, half of the subjects were randomized to receive specific, scheduled PLCO cancer screening exams (intervention arm) and the other half to receive only their usual care (control arm). The ovarian cancer screening regimen in PLCO consisted of annual CA-125 tests and trans-vaginal ultrasounds (TVU) for four years (T0–T3), plus two additional rounds of screening with CA-125 only (T4, T5). Enrollment was initiated in 1993 and completed in 2001. Participants are being followed for at least 13 years from enrollment. Ten screening centers are participating: the University of Colorado Health Sciences Center; Lombardi Cancer Research Center of Georgetown University; Pacific Health Research Institute, Honolulu; Henry Ford Health System; University of Minnesota School of Public Health/Virginia L. Piper Cancer Institute; Washington University School of Medicine; University of Pittsburgh, Pittsburgh Cancer Institute and Magee-Women’s Hospital; Huntsman Cancer Institute at the University of Utah; Marshfield Clinic Research Foundation; and the University of Alabama at Birmingham. Each institution obtained local Institutional Review Board approval to carry out the trial.
The target population for the ovarian component of the PLCO trial included women from 55 to 74 years of age who had not been diagnosed previously with lung, colorectal, or ovarian cancer. Criteria for exclusion included current treatment for cancer other than non-melanoma skin cancer, and enrollment in another cancer screening or prevention trial. Initially, women who had undergone oophorectomy were ineligible, but in 1996 this restriction was lifted because a low accrual of women threatened to jeopardize screening end points for lung and colorectal cancer. Women who reported having had a bilateral oophorectomy on the baseline questionnaire were considered ineligible for ovarian tests and are therefore not included in these analyses. Women who reported having had a prior oophorectomy at any given screening year were considered ineligible for all subsequent tests.
TVU was performed by qualified sonographers using a 5–7.5 MHz transvaginal probe. The examiner imaged both ovaries in the transverse and longitudinal planes. At least five minutes were spent looking for each ovary to ensure an adequate search; however, if the iliac vessels were visualized without ovaries being seen the examiner concluded the search for the ovaries . Ovaries were measured along the major and minor axes in both transverse and longitudinal planes, and the prolate ellipsoid formula (width × height × thickness × 0.523) was used to calculate the volume of each ovary and/or cyst. The following TVU test results were classified as positive: ovarian volume >10 cm3; cyst volume >10 cm3; any solid area or papillary projection extending into the cavity of a cystic ovarian tumor of any size; or any mixed (solid/cystic) component within a cystic ovarian tumor.
CA-125 was measured on serum obtained and frozen within 2 hours of blood draw at each of the 10 screening centers. All sample testing was performed centrally at the University of California, Los Angeles Immunogenetics Laboratory. Samples were shipped on dry ice and stored at −70°C. CA-125 results ≥35 U/mL were classified as positive.
Results of the screening tests were sent to both the participants and their personal physicians within 3 weeks of specimen submission. Evaluation and follow-up of positive screening tests were at the discretion of participants’ physicians; the PLCO trial did not specify a diagnostic algorithm. Medical records of diagnostic procedures performed following positive screens were obtained by screening center study personnel and recorded on standardized reporting forms. To track cancers not diagnosed through screening, study participants were sent annual study update forms which asked about any cancer diagnoses. Pathology reports from relevant neoplasms were abstracted by trained certified tumor registrars at the respective screening centers. This report covers all confirmed ovarian and peritoneal cancers diagnosed in women in the intervention arm in the four year period following randomization.
For purposes of this report, screen detected cancers were defined as those diagnosed as a result of investigations initiated following a positive screening test, with no lapse in the diagnostic evaluation exceeding 9 months. Non-screen detected cancers in women who received prior PLCO screens were denoted as interval cancers. “Never screened” cancers were defined as cancers diagnosed in women who did not receive any PLCO screening tests for ovarian cancer.
The chi-squared test was used to assess the statistical significance of differences in proportions. A two-sided p-value of less than 0.05 was considered statistically significant.
Out of 39,115 women randomized to receive screening, 4,854 had undergone prior oophorectomy, leaving 34,261 who were considered eligible for screens at randomization; these women are the subjects of the current analysis. Demographics and medical history for these participants are shown in Table 1. Almost 65% were 55–64 years of age and most, 88.6%, were non-Hispanic white. The population was highly educated, with about 30% being college graduates. Just over a quarter (27.3%) had prior hysterectomy and a bit over half (53.6%) had a history of oral contraceptive use.
Of 34,261 women eligible for screening, 30,630 (89%) received at least one screen during the 4 rounds. The proportion of eligible women receiving both tests decreased slightly over the study years, from 83.1% at baseline (T0) to 77.6% at T3 (Table 2). Compliance rates were essentially the same for TVU and CA-125, although slightly lower for the former, and the overwhelming majority of women receiving one test received both tests. Of note, about one third (32%) of the over 5,000 women who were not screened at T0 received at least one screen during the next three rounds.
Screening results are shown in Table 3. The percentage of women who tested positive on at least one test was slightly lower in years T1–T3 (4.9, 4.6, and 4.5%) than at T0 (5.8%). In each screening year, a significantly greater proportion of TVU’s were positive (4.6%, 3.4%, 2.9% and 2.9% at T0–T3, respectively) than CA-125 tests (1.4%, 1.6%, 1.8%, and 1.7%) (p < 0.0001). Approximately half of the positive TVUs in the study years subsequent to T0 were newly positive tests. For CA-125, the proportion of positives that were newly positive decreased from 60% at T1 to 34% at T3. Across all screening rounds, among the 30,630 women who received at least one screening test, 11.1 % had at least one positive result. For the individual tests, 8.1% had at least one positive TVU and 3.4% had at least one positive CA-125 test. Screen positivity rates showed relatively little variability across the ten screening centers; the coefficient of variation of the positivity rates across centers was around 35% for TVU and 20% for CA-125.
The number of women with a positive screen, the number biopsied as a result of that screen, and the number of neoplasms and invasive cancers identified are presented in Table 4; results stratified by screening modality are presented in Table 5. Of note, the large majority of biopsies were oophorectomies performed either with laparotomy or laparoscopy. Among all women who screened positive by either TVU or CA-125, the proportion undergoing biopsy decreased significantly over the study years, going from 33.8% at T0 to 19.7% at T1, 14.9% at T2 and 13.8% at T3 (P < 0.0001). In contrast, of women who did undergo biopsy, the percentage diagnosed with invasive (ovarian or peritoneal) cancer increased significantly over the study years, going from 3.2% at T0 to 9.5% at T3 (p=0.0004). Putting these two trends together to examine the positive predictive value (PPV) of the screening tests for invasive cancer (i.e., the percentage of positive screens that resulted in a diagnosis of invasive ovarian or peritoneal cancer), it was seen that PPV was relatively constant, and quite low, over the screening years – 1.1% (95% CI 0.6–1.6), 1.0% (95% CI 0.4–1.5), 1.1% (95% CI 0.5–1.7) and 1.3% (95% CI 0.6–2.0), respectively, at T0–T3. The yield of invasive cancers per 10,000 screens was also relatively stable over time - 6.2 (95% CI 3.4–9.2), 4.7 (95% CI 2.2–7.3), 5.2 (95% CI 2.5–8.0), and 5.9 (95% CI 2.9–8.9), respectively, at T0–T3.. Over all screening rounds, of the 3,388 women who had at least one positive screening result on either test, 1,170 (34.5%) received a biopsy at some point as diagnostic follow-up; of these, 5.1% (n=60) had invasive cancer diagnosed on biopsy.
The two individual tests, CA-125 and TVU, differed significantly in terms of both the proportion of positives who underwent biopsy and the proportion of women biopsied who had invasive cancer. The biopsy rate was considerably higher following a positive TVU than a positive CA-125, although for both tests this rate significantly decreased from T0 to T1–T3; for TVU the rate dropped from 40.6% at T0 to 17–24% at T1–T3 (p < 0.0001) whereas for CA-125 the rate dropped from 15.6% at T0 to 9–12% at T1–T3 (p=0.01). Among those biopsied, the percent diagnosed with invasive cancer was much greater for CA-125 (21.0% at T0 and 18.8–28.2% at T1–T3) than for TVU (2.3% at T0 and 4.1–4.4% at T1–T3) (p < 0.0001). PPVs were 2–3 fold higher for CA-125 (range 2.1–3.2) than for TVU (range 0.7–1.1) (p < 0.0001); for neither test was there a clear trend over time. The yields of the two tests were generally similar, ranging from 2.0 to 4.2 (per 10,000) for TVU and from 3.2 to 4.8 for CA-125.
Of 17 low malignant potential (LMP) tumors, 14 (82%) were screen detected. Of the screen detected LMPs, 12 had been positive on TVU alone, one on CA-125 alone, and one had been positive on both tests. Nine of the 14 were detected at T0.
The stage of cancers according to screening result and study year of diagnosis is shown in Table 6. The 17 low malignant potential (LMP) tumors are not included in the table; all but two were stage I, with one being stage II and one stage III. Eighty-nine women in the screening arm were diagnosed with (invasive) ovarian cancer during the T0–T3 time period. Sixty (67%) of these cancers were screen detected. Nineteen screen-detected cancers (32%) had abnormalities in both TVU and CA-125 in the year of diagnosis; of these, 15 (79%) were stage III or greater. Forty-one cancers were discordantly positive for either CA-125 only (n=27) or TVU only (n=14); 89% of the CA-125 (only) detected cancers were Stage III or greater as compared to only 29% of the TVU (only) detected cancers (p < 0.0001). Of note, 15 of 18 (83%) screen detected cancers at T0 were stage III or greater; this proportion decreased modestly, and not statistically significantly, to 67% (28 of 42) at T1–T3. Of 19 interval cancers, 16 (84)% were Stage III or greater (one was of unknown stage). Of these 16, the median CA-125 value at the last screen was 13 U/mL; 14 were primary ovarian and 2 were peritoneal cancers.
The majority of cancers (61%) were serous cystadenocarcinomas; the proportion with this histology was similar among both the screen detected (58%) and the non-screen detected (66%) cancers. Twelve cancers (13.5%) were peritoneal; 8 of these were screen detected, all by CA-125.
At the baseline screening round, a total of 566 surgeries were performed as diagnostic follow-up to positive screens, resulting in a diagnosis of 18 invasive cancers – the ratio of surgeries to invasive cancers was thus 31 to 1. Of the 18 cancers, 83% were Stage III or IV. Over three subsequent annual rounds of screening, there were 604 additional surgeries following positive screens and 42 more screen detected invasive cancers, giving a ratio of surgeries to cancers of 14 to 1; 67% of these cancers were stage III or above. Thus, during subsequent screening rounds, although the number of surgeries required to detect a cancer was halved from the baseline round, this ratio remained somewhat elevated, and the stage distribution of the detected cancers was only minimally improved. Over all four rounds of screening, the surgery to detected cancer ratio was 19.5 to 1, and 72% of screen detected cancers were late stage.
The above statistics reflect the complementary nature of the two screening tests being evaluated. The high rate of surgeries derives primarily from positive TVUs. At baseline, 94% of surgeries (biopsies) performed following positive screens were done in women with positive TVUs; this decreased marginally to around 80% for study years T1–T3. At baseline, the surgery to detected cancer ratio following a positive TVU was 44 to 1, as compared to 23 to 1 during subsequent screening rounds. In contrast, following a positive CA-125 test, the ratio of surgeries to cancers was about 4.5 to 1 at both the baseline and subsequent screening rounds. However, while the rate of “unnecessary” surgeries was much increased with TVU as compared to CA-125, TVU was also the test that detected the earlier stage cancers.
In the original report based on the baseline screening round results only, there were only 18 (screen detected) cancers, which limited the ability to perform sub-analyses. Now, with 60 screen detected cases, the pattern is clearer. The cases detected through TVU only (i.e., with normal CA-125), tended to be early stage; 71% of an admittedly small sample of 14 cases were stage I or II. Cases with elevated CA-125 however, regardless of whether they had an abnormal TVU or not, were primarily late stage; 89% of 27 cases with normal TVU and 79% of 19 cases with abnormal TVU. Of the 17 screen detected stage I/II cases, 10 (59%) were detected by TVU alone.
An interesting finding here was that the biopsy rate following a positive screen decreased from T0 to T1–T3. For TVU the decrease was from 40.6% at baseline to 17–24% at T1–T3, while for CA-125 the drop-off was from 15.6% at T0 to 9–12% at T1–T3. For CA-125, this decrease is explained by examining first positive screens over subsequent rounds. Among women with a first positive CA-125 screen at T1–T3, the biopsy rate was 14.8%, essentially equivalent to the T0 biopsy rate of 15.6%; thus first positive CA-125 tests at later rounds were followed up similarly to baseline round positives. In contrast, among repeat CA-125 positives, the average biopsy rate over T1–T3 was only 7%. For TVU however, even among the first positives at later rounds, the biopsy rate remained substantially lower (on average 23%) than the rate at baseline of 40.6%. The biopsy rate for repeat positive TVUs averaged 16%. We are currently investigating whether the specific findings on abnormal TVUs (e.g., cyst size, ovary size) may help explain the discrepancy in biopsy rates from baseline to subsequent rounds.
The large number of surgeries (primarily oophorectomies) prompted by a positive screen and not resulting in a cancer diagnosis here, 1,086 (or 3.5% of all women screened) should be seen in context of the background rate of oophorectomies in women of this age group. An essentially equivalent number (1,080) of oophorectomies (without a cancer finding) were performed on this cohort during this time period for reasons other than follow-up of a positive screen. Thus, while oophorectomies for other purposes were not un-common, this screening program utilizing CA-125 and TVU effectively doubled the oophorectomy rate in this cohort during the period of screening.
Several recent studies of screening with CA-125 and TVU had produced results generally similar to those observed here. In a trial of ovarian cancer screening in the U.K., Menon et al. utilized a sequential regimen of CA-125 followed (in some instances) by TVU . On the basis of CA-125 and age, women were assigned a risk score and those with high enough scores were then offered repeat CA-125 and/or TVU. Out of 6,532 women (median age 59.7) receiving an initial screen, 16 underwent surgery based on screening findings and 5 were diagnosed with ovarian malignancy. These numbers translate into a cancer yield of 7.7 per 10,000 women screened and a biopsy to detected cancer ratio of 3.2 to 1. Since all women in the U.K. study had to have elevated CA-125 to proceed to diagnostic follow-up, these figures are perhaps most appropriately compared to the corresponding PLCO results from women with a positive CA-125, which were a yield (at baseline) of 5.2 per 10,000 and a biopsy to detected case ratio of about 4.5 to 1. Thus the two studies demonstrated quite similar findings.
In a Japanese trial of CA-125 and TVU screening for ovarian cancer, among 41,688 women (median age 58) randomized to an intervention group, the yield of ovarian cancer was 3.1 per 10,000 women screened at the initial round and 3.8–7.4 per 10,000 women screened at later rounds (2 through 5) . This range for the yield in later rounds was quite comparable to that seen in PLCO (4.7 – 5.9), although the initial round yield was lower than the PLCO yield of 6.2. The overall ratio of surgeries to detected cancers was 33 to 1, a bit higher than the 19.5 to 1 ratio for PLCO. A total of 67% of the Japanese screen-detected cases (n=27) were stage I/II, compared to 29% in PLCO.
In conclusion, over three subsequent post-baseline annual screening rounds, the ratio of surgeries to detected cancers decreased somewhat from that observed in the baseline round, but still remained rather high at 14 to 1. As at baseline, the majority of screen detected cancers were stage III and above. TVU continued to account for most of the unnecessary surgeries in the subsequent rounds, but also for most of the early stage cancers. These data are consistent with the current guidelines of the U.S. Preventive Services Task Force that state that ovarian screening with CA-125 and TVU is not recommended. A determination on whether screening with these two modalities will reduce ovarian cancer mortality must await the final results of the PLCO Trial.
Supported by National Cancer Institute intramural and extramural funds and by individual contracts from the National Cancer Institute to each of the 10 screening centers and to the coordinating center.
Financial Disclosure: Dr. Kessel has received research funding from Vivus (Mountain View, CA), Wyeth (Madison, NJ), and Proctor & Gamble (Cincinatti, OH). He has been a consultant to Novartis (Basel, Switzerland), Merck, and Eli-Lilly & Co. (Indianapolis, IN). He has also served on the speakers bureau for Bayer (Leverkusen, Germany) and Merck. The other authors did not report any potential conflicts of interest.
For the authors represented by the Prostate, Lung, Colorectal and Ovarian Cancer Project Team who participated in this study, see the Appendix online at http://links.lww.com/xxx.