Ovarian cancer screening with annual transvaginal sonography

Findings of 25,000 women screened

Abstract

BACKGROUND

Ovarian cancer has the highest mortality rate of all gynecologic malignancies, and most women present with advanced-stage disease. The current investigation was performed to determine the efficacy of annual transvaginal sonography (TVS) as a screening method for ovarian cancer.

METHODS

Annual TVS screening was performed on 25,327 women from 1987 to 2005. Asymptomatic women aged ≥50 years and women aged ≥25 years who had a family history of ovarian cancer were eligible for participation in this trial.

RESULTS

Among 364 patients (1.4%) with a persisting ovarian tumor on TVS who underwent exploratory laparoscopy or laparotomy with tumor removal, 35 primary invasive ovarian cancers, 9 serous ovarian tumors of low malignant potential, and 7 cancers metastatic to the ovary were detected. Stage distribution was as follows: 28 patients had stage I disease, 8 patients had stage II disease, and 8 patients had stage III disease. Four patients died of disease, 2 patients died of other causes, and 38 patients were alive and well from 0.5 years to 15.8 years after diagnosis (mean, 5.3 years). Nine women developed ovarian cancer within 12 months of a negative screen (false-negative results), and 3 of these patients died of disease. TVS screening had a sensitivity of 85.0%, specificity of 98.7%, positive predictive value of 14.01%, and negative predictive value of 99.9%. After 107,276 screening years, there have been 7 ovarian cancer deaths in the annually screened population and 3 ovarian cancer deaths among women who were noncompliant. Excluding patients with nonepithelial or borderline ovarian malignancies, the survival of patients with ovarian cancer in the annually screened population was 89.9% ± 10.1% at 2 years and 77.2% ± 22.8% at 5 years.

CONCLUSIONS

TVS screening, when it was performed annually, was associated with a decrease in disease stage at detection and with case-specific ovarian cancer mortality, but it was not effective in detecting ovarian cancers in women who had normal ovarian volume. Cancer 2007; © 2007 American Cancer Society.

Despite advances in radical surgery, postoperative care, and chemotherapy, ovarian cancer remains the fourth leading cause of cancer mortality among women in the United States.1 This year, over 16,000 deaths from ovarian cancer will be reported in the United States alone. Early-stage ovarian cancer produces few specific symptoms, and pelvic examination is notably inaccurate in detecting subtle changes in ovarian size and morphology, particularly in postmenopausal women.2 Consequently, most patients continue to present with advanced-stage disease, for which the prognosis is poor. Early-stage ovarian cancer, however, is highly curable when it is treated by conventional therapy.3, 4 Therefore, screening asymptomatic, high-risk women has been proposed as a means to detect ovarian cancer at an earlier and more curable stage.5–13

In 1987, the University of Kentucky Ovarian Cancer Screening Project was initiated to assess the efficacy of annual transvaginal sonography (TVS) as a screening method for ovarian cancer. Since then, free screening has been provided to >25,000 women. The screening algorithm used has remained the same, except that, now, women with persisting, unilocular, cystic ovarian tumors that measure <5 cm in greatest dimension are followed at 6-month intervals with ultrasound instead of undergoing laparoscopic tumor removal. The current report summarizes the data from the trial.

MATERIALS AND METHODS

Patients

Patients who enrolled in the University of Kentucky Ovarian Cancer Screening Project from January 1987 to May 2005 were evaluated. Eligibility criteria included 1) all asymptomatic women aged ≥50 years and 2) asymptomatic women aged ≥25 years with a documented family history of ovarian cancer in at least 1 primary or secondary relative. Genetic testing was not performed routinely as part of this trial. All study participants completed a questionnaire that included past medical history, surgical history, menopausal status, hormonal use, and family history of cancer. Menopause was defined as the absence of menses for 12 months. Any woman with a known ovarian tumor or a personal history of ovarian cancer was excluded from this investigation.

After informed consent was obtained, each patient underwent screening according to the algorithm illustrated in Figure 1. TVS was performed initially with Aloka 620 and Aloka 680 ultrasound units (Aloka, Tokyo, Japan)with a 5-mHz vaginal probe, as described previously.12 Since 2000, TVS has been performed using General Electric (Milwaukee, Wis) Logiq 400 units with a 5-mHz vaginal probe. Doppler flow and 3-dimensional sonography were performed in selected patients using a General Electric Voluson 730 ProV unit with a 5- to 9-mHz vaginal probe. All sonographic images were reviewed by at least 1 of the authors. At each screen, both ovaries were measured in 3 dimensions. Ovarian volume was calculated by using the prolate ellipsoid formula (length × width × height × 0.523). All screening information was entered into a Medlog database on a local network. Criteria for abnormality included an ovarian volume >20 cm³ for premenopausal women and >10 cm³ for postmenopausal women. These values were used because they were >2 standard deviations (SD) above the mean volume of normal ovaries in premenopausal and postmenopausal women.14 In addition, any cystic ovarian tumor with a solid or papillary projection into its lumen was considered abnormal. Women who had a normal screen were scheduled to return in 12 months for a repeat screen. Women who had an abnormal screen underwent a repeat sonogram in 4 to 6 weeks. Women who had an abnormal second screen had a serum CA-125 determination, tumor morphology indexing, Doppler flow sonography, and (more recently) serum proteomic analysis. Morphology indexing was performed according to the classification of Ueland and colleagues.15 Each tumor was given a score from 1 to 10 according to increasing morphologic complexity and volume (Fig. 2). Women who had unilocular cystic ovarian tumors that measured ≤5 cm in greatest dimension and a normal serum CA-125 were followed with repeat sonography at 6-month intervals. Women who had persisting, complex ovarian tumors or women who had a cystic ovarian tumor and an elevated serum CA-125 level were advised to undergo surgical removal of the tumor. During the past 10 years of this trial, every effort was made to perform laparoscopic tumor removal as the initial step in the operative evaluation of patients with persisting ovarian tumors detected at screening.

Figure 1. Study algorithm.

Figure 2. Morphology index.

At the time of laparoscopy, ovarian tumors were placed in an endoscopic bag intra-abdominally and removed through a midline subumbilical incision. Patients with ovarian cancer on frozen section or patients with obvious metastatic disease at laparoscopy underwent immediate exploratory laparotomy and staging. Tumors were classified histologically according to the World Health Organization (WHO) system and were staged according to the International Federation of Gynecology and Obstetrics (FIGO) system. After surgery, patients were treated according to the cell type, histologic differentiation, and stage of each tumor, usually with a 6-month course of combination chemotherapy. Patients with ovarian cancer were followed at monthly intervals during treatment, every 3 months for 2 years, and every 6 months thereafter. Follow-up data on all patients were coordinated with the American Cancer Society and the Kentucky State Department of Vital Statistics. Three hundred eighty patients with ovarian cancer who were entered into the University of Kentucky Tumor Registry from 1987 to 2005 and who did not receive screening served as a historic control group for this investigation.

Statistical Methods

Proportions were compared by using chi-square statistics or Fisher exact tests. Means were compared by using 2-sample t statistics. Statistical significance was determined at the .05 level. Patient survival was estimated using the Kaplan-Meier method.

RESULTS

There were 25,327 women who enrolled in the study from January 1987 to May 2005, and a family history of ovarian cancer was documented in 5868 of those women (23.2%). The clinical characteristics of the patients screened are illustrated in Table 1. The mean age of these patients was 56 years (SD, 10.5 years; range, 25–92 years). The mean height was 162.56 cm (SD, 6.6 cm; range, 127–195.58 cm), and the mean weight was 72.21 kg (SD, 15.9 kg; range, 36–166.9 kg). The mean parity was 2.4 (SD, 1.4; range, 0–17), and 3705 women (14.6%) were nulliparous. Women enrolled in the study underwent a total of 120,569 scans (mean, 4.8 scans per participant). One or both ovaries were visualized in 101,299 scans (84%). Sonographically undetectable ovaries were considered normal for the purposes of this investigation.

Table 1. Clinical Characteristics of Patients Screened (n = 25,327)

Variable	Mean	Range	No. of patients (%)
Age, y	56	25–92
Parity	2.4	0–17
Weight, kg	71.21	36–166.9
Height, cm	162.56	127–195.58
Family history
Ovarian cancer			5868 (23.2)
Breast cancer			9396 (37.1)
Colon cancer			6066 (24)
History of hormone replacement			8021 (31.7)
Hormone replacement on last visit			5397 (21.3)
Nulliparous			3705 (14.6)

Three hundred sixty-four women (1.4%) with a persisting ovarian tumor on TVS underwent surgery. Histologic diagnoses of the 364 ovarian tumors removed are presented in Table 2. The most common diagnosis was ovarian serous cystadenoma followed by primary ovarian cancer and endometrioma. Forty-four women had primary ovarian cancer, including 39 epithelial neoplasms and 5 ovarian stromal neoplasms. Patients who had a positive family history of ovarian cancer did not have a statistically higher frequency of ovarian carcinoma or benign ovarian tumors than patients without this history. All ovarian malignancies except 1 had solid components or papillary projections from the tumor wall. One ovarian carcinoma was predominantly cystic, but it had an abnormally thickened wall. Morphology index (MI) scores of the tumors removed varied from 0 to 10. The mean MI value for benign tumors was 3.5 (SD, 2.3; range, 0–10) compared with a mean MI value of 6 (SD, 1.9; range, 3–10) for ovarian malignancies (P < .001). One hundred fifty-three tumors were unilocular without wall abnormalities, and none were malignant.

Table 2. Histologic Findings in Patients With Persistently Abnormal Screens (n = 364)

Finding	No. of patients
Primary ovarian cancer	44
Nonovarian cancer	7
Appendix	3
Primary peritoneum	2
Breast	1
Endometrium	1
Serous cystadenoma	153
Endometrioma	30
Mucinous cystadenoma	19
Cystic teratoma	18
Fibroma/thecoma/Brenner tumor	25
Leiomyoma	4
Hydrosalpinx/paratubal cyst	25
Other	39

Clinical and pathologic findings in the 44 patients who had primary ovarian cancer detected by annual screening are presented in Table 3. Stage distribution of these patients was as follows: 28 patients had stage I disease, 7 patients had stage II disease, and 8 patients had stage III disease. Nine patients had serous ovarian tumors of borderline malignancy. Five patients with stage I ovarian cancer (18%) had a palpable abnormality on clinical examination. Nine patients with stage I invasive ovarian cancer had preoperative CA-125 determinations, and antigen levels were elevated (>35 U/mL) in 1 patient (11%). Twenty-one of 44 primary ovarian cancers (48%) were detected on the first screen, and the remaining primary ovarian cancers were detected on subsequent screens. Four of 44 patients (9%) who had primary ovarian cancers detected by annual screening have died of disease. Three of those patients had stage III ovarian cancer at the time of their first or second screen and died of disease from 1.8 years to 7.1 years after diagnosis. One patient (No. 32) had a poorly differentiated stage IIA ovarian serous cystadenocarcinoma detected on the 12th screen and died of disease 1.3 years after she underwent surgical debulking and received carboplatin/taxol chemotherapy. Thirty-eight patients remain alive and well with no evidence of disease from 0.5 years to 15.8 years (mean, 5.3 years) after diagnosis, and 2patients without evidence of disease died of other causes 3.1 years and 8.0 years after diagnosis.

Table 3. Clinicopathologic Findings in Patients With Primary Ovarian Cancer Detected by Screening (n = 44)

Patient	Age at Dx, y	Family history	No. of screens	Physical examination	TVS finding	TVS volume, cm^3	TVS MI	CA-125, U/mL	Histology	Tumor stage	Treatment	Status	Follow-up, y
Dx indicates diagnosis; TVS, transvaginal sonography; MI, morphology index; S, surgery; C, chemotherapy; NED, no evidence of disease; DOC, dead from other causes; DOD, dead of disease.
1	72	Yes	3	Normal	Complex mass	198	5	7	Granulosa cell tumor grade 2	IA	S and C	NED	15.8
2	50	No	1	Normal	Complex mass	19	8	14	Adenocarcinoma grade 3	IA	S and C	NED	15.2
3	42	Yes	2	Palpable	Cystic mass	27	3	14	Serous cystadenocarcinoma grade 3	IA	S and C	NED	12.3
4	36	No	1	Normal	Complex mass	32	5	—	Endometrioid carcinoma grade 1	IA	S	NED	12.3
5	55	Yes	1	Normal	Complex mass	15	6	—	Granulosa cell tumor grade 2	IA	S and C	NED	11.5
6	58	No	5	Palpable	Complex mass	94	6	—	Serous ovarian tumor of borderline malignancy	IA	S	NED	9.9
7	66	No	1	Normal	Complex mass	40	6	37	Endometrioid carcinoma grade 2	IA	S and C	NED	9.9
8	64	Yes	1	Normal	Complex mass	25	3	—	Endometrioid carcinoma grade 2	IA	S and C	NED	9.5
9	63	No	1	Normal	Complex mass	223	8	12	Serous ovarian tumor of borderline malignancy	IA	S	NED	8.3
10	65	No	6	Normal	Complex mass	366	9	25	Endometrioid adenocarcinoma grade 2	IA	S and C	NED	5.7
11	63	No	1	Normal	Complex mass	164	5	—	Adenocarcinoma grade 1	IA	S	NED, DOC	3.1
12	57	No	3	Normal	Complex mass	3	3	10	Granulosa cell tumor grade 3	IA	S	NED	6.1
13	69	No	10	Normal	Complex mass	35	5	—	Adenocarcinoma grade 3	IA	S and C	NED	4.4
14	60	No	1	Normal	Complex mass	51	6	14	Adenocarcinoma grade 3	IA	S	NED	1.6
15	58	Yes	7	Normal	Complex mass	20	5	13	Granulosa cell tumor grade 2	IA	S	NED	1.4
16	59	No	1	Normal	Solid mass	118	7	5	Ovarian tumor of borderline malignancy	IA	S	NED	0.6
17	71	No	8	Palpable	Solid mass	163	7	13	Endometrioid carcinoma grade 2	IA	S and C	NED	0.5
18	60	No	1	Normal	Complex mass	91	5	23	Endometrioid adenocarcinoma grade 2	IB	S and C	NED	4.7
19	53	No	3	Normal	Complex mass	103	7	4	Clear cell adenocarcinoma grade 3	IB	S and C	NED	2.7
20	63	No	1	Normal	Complex mass	33	5	16	Serous ovarian tumor of borderline malignancy	IB	S and C	NED	1.5
21	71	No	11	Normal	Complex mass	20	6	47	Serous ovarian tumor of borderline malignancy	IB	S	NED	0.2
22	57	No	3	Normal	Complex mass	26	9	7	Granulosa cell tumor grade 1	IC	S	NED	11.0
23	48	Yes	7	Palpable	Complex mass	122	5	5	Serous tumor of borderline malignancy	IC	S	NED	7.0
24	54	No	3	Palpable	Complex mass	141	7	—	Endometrioid adenocarcinoma grade 3	IC	S and C	NED	3.3
25	72	No	7	Normal	Solid mass	32	5	27	Serous ovarian tumor of borderline malignancy	IC	S and C	NED	1.1
26	64	No	1	Normal	Complex mass	34	4	21	Serous ovarian tumor of borderline malignancy	IC	S	NED	0.6
27	63	No	1	Normal	Complex mass	27	5	26	Papillary serous adenocarcinoma grade 1	IC	S and C	NED	0.6
28	53	No	2	Normal	Complex mass	10	4	6	Serous ovarian tumor of borderline malignancy	IC	S	NED	0.3
29	79	No	12	Normal	Complex mass	19	5	259	Endometrioid adenocarcinoma grade 3	IIA	S and C	NED	5.4
30	67	No	1	Normal	Complex mass	16	3	—	Adenocarcinoma grade 3	IIA	S and C	NED	4.6
31	70	No	2	Palpable	Complex mass	90	5	11	Papillary serous carcinoma grade 2	IIA	S and C	NED	0.1
32	66	No	12	Palpable	Complex mass	129	7	7	Serous cystadenocarcinoma grade 3	IIA	S and C	DOD	1.3
33	56	No	2	Normal	Solid mass	20	5	—	Adenocarcinoma grade 3	IIB	S and C	NED	1.9
34	61	Yes	5	Normal	Complex mass	28	5	16	Adenocarcinoma grade 3	IIC	S and C	NED, DOC	8.0
	8
35	63	No	2	Normal	Complex mass	239	10	279	Adenocarcinoma grade 3	IIC	S and C	NED	11.2
36	62	No	1	Palpable	Solid mass	326	8	33	Adenocarcinoma grade 3	IIC	S and C	NED	1.1
37	69	No	2	Palpable	Complex mass	52	2	49	Carcinosarcoma grade 3	IIIA	S and C	DOD	1.8
38	52	Yes	3	Palpable	Complex mass	35 ml	4	35	Serous ovarian tumor of borderline malignancy	IIIA	S	NED	3.0
39	66	No	1	Normal	Complex mass	154	9	16	Adenocarcinoma grade 3	IIIB	S and C	DOD	7.1
40	77	No	1	Palpable	Complex mass	87	7	16	Papillary serous carcinoma grade 2	IIIB	S and C	NED	1.0
41	52	No	1	Palpable	Solid mass	313	10	59	Adenocarcinoma grade 3	IIIC	S and C	DOD	2.4
42	72	No	4	Palpable	Complex mass	384	9	134	Mucinous cystadenocarcinoma grade 3	IIIC	S and C	NED	7.5
43	45	Yes	8	Palpable	Complex mass	335	9	1550	Adenocarcinoma grade 3	IIIC	S and C	NED	4.5
44	63	Yes	15	Normal	Complex mass	34	5	214	Endometrioid carcinoma grade 3	IIIC	S and C	NED	3.2

Nine patients developed ovarian cancer within 12 months of a negative screen (false-negative results). Three of those patients had normal sized ovaries at the time of surgical exploration 3 months, 12 months, and 12 months after a normal screen but had extraovarian metastases resulting in a diagnosis of stage IIA ovarian cancer (1 patient) and stage IIIC ovarian cancer (2 patients). Six patients had increased ovarian volumes at the time of surgery from 2 months to 11 months after a normal screen and had stage IIC ovarian cancer (1 patient), stage IIIB cancer (1 patient), and stage IIIC cancer (4 patients). Seven of those patients had poorly differentiated ovarian carcinomas, and 2 had moderately differentiated tumors. Serum CA-125 determinations were obtained in all 9 of those patients at the time of surgery and were elevated (>35 U/mL) in 8 patients. All patients were treated with a combination of surgery and chemotherapy. Five patients were alive without evidence of disease from 0.5 years to 5.5 years after diagnosis, and 3 patients died of disease from 0.7 years to 7.0 years after diagnosis. One patient who had no evidence of disease died of a myocardial infarction 9.8 years after diagnosis.

Seven women had metastatic cancer to the ovary detected by screening (Table 4). It was believed that all of those women had a primary ovarian neoplasm prior to surgical exploration. Three of these patients had primary appendiceal cancer with metastases to the ovary and underwent hemicolectomy and ovarian tumor debulking. They remain alive without evidence of disease from 1 year to 6.3 years after treatment. Two patients had primary peritoneal cancer detected by the presence of paraovarian fluid. Both of those patients had peritoneal and ovarian surface involvement by a poorly differentiated epithelial adenocarcinoma and had no evidence of ovarian stromal invasion. One patient remains alive with disease 1.7 years after diagnosis, and the other patient has no evidence of disease 3.7 years after surgery and chemotherapy. Finally, there was 1 patient with ovarian metastasis from breast carcinoma who died of disease 7.9 years after ovarian tumor removal, and 1 patient with endometrial cancer and spread to the ovary is alive and well 4.4 years after surgery. To date, there have been no cases of operative mortality in patients who underwent surgery as part of this trial.

Table 4. Clinicopathologic Findings in Patients With Metastatic Cancer to the Ovary Detected by Transvaginal Sonography (n = 7)

Case	Age at Dx, y	Family history	No. of screens	Physical examination	TVS finding, (volume, cm^3)	Site of primary cancer	Treatment	Status	Follow-up, y
Dx indicates diagnosis; TVS, transvaginal sonography; S, surgery; C, chemotherapy; DOD, dead of disease; NED, no evidence of disease; AWD, alive with disease.
45	60	No	8	Normal	Ovarian tumor (18)	Breast	S and C	DOD	7.9
46	56	No	4	Normal	Ovarian tumor (70)	Appendix	S and C	NED	6.3
47	70	No	10	Normal	Paraovarian fluid (99)	Peritoneum	S and C	NED	3.7
48	67	No	9	Normal	Paraovarian fluid (>100)	Peritoneum	S and C	AWD	1.7
49	48	No	6	Pelvic mass	Ovarian tumor (76)	Appendix	S and C	NED	1.9
50	71	No	11	Pelvic mass	Ovarian tumor (100)	Appendix	S and C	NED	1
51	78	No	3	Pelvic mass	Ovarian tumor (76)	Endometrium	S and C	NED	4.4

Four hundred fifty-six patients (1.8%) with normal screens underwent hysterectomy and bilateral salpingo-oophorectomy for a variety of indications that were unrelated to screening. Ovarian volume and morphology were normal in all of these patients at the time of surgery, and there were no cases of ovarian neoplasia. There were 3746 postmenopausal women with 6513 unilocular cystic ovarian tumors who were followed at 3- to 6-month intervals by TVS without surgery. Duration of follow-up ranged from 4 months to 16.5 years (mean, 4.6 years). Spontaneous resolution occurred in 5381 of these cysts (83%), usually within 6 months, and no patient in this group has developed ovarian cancer.

The statistical definitions used in this investigation are presented in Table 5. There were 51 patients with positive screens who had either primary ovarian cancer or cancer metastatic to the ovary (true-positive results). There were 313 patients with positive screens who had benign ovarian tumors (false-positive results). Likewise, there were 23,706 patients with negative screens, and 9 women developed ovarian cancer within 12 months of a negative scan (false-negative results). Using these data, the sensitivity of TVS screening was 85%, and the specificity was 98.7%. The positive predictive value (PPV) of an abnormal screen was 0.1401, and the negative predictive value (NPV) of a normal screen was 0.999. Screening statistics related to date of entry into the trial, family history, and menopausal status are presented in Table 6. The screening algorithm was changed on January 1, 2001, such that patients with unilocular cystic ovarian tumors measuring ≤5 cm were followed sonographically at 6-month intervals rather than undergoing surgery. The PPV of screening increased from 8.8% (1987–2000) to 27.1% (2001–2006) as a result of this change. Likewise, the PPV of screening increased from 8.8% in premenopausal patients to 15.4% in postmenopausal patients. In patients receiving TVS screening, there have been 107,276 screening years and 10 deaths from ovarian cancer. Seven of the patients who died of ovarian cancer (4 with true-positive results and 3 with false-negative results) were compliant with the screening algorithm and were screened at yearly intervals. Two of the four women in the true-positive group who died of disease had stage III ovarian cancer at the time of their initial screen. There were 3 noncompliant patients who died of ovarian cancer. These patients returned to their local physicians with symptoms of abdominal swelling after having no screening for 21 months, 51 months, and 61 months. All 3 of those patients had stage III disease at the time of their diagnosis and died of disease from 1 year to 4.6 years after diagnosis.

Table 5. Statistical Definitions in Ovarian Cancer Screening*

Term	Screen	Findings	No. of study patients
* Sensitivity (TP/TP + FN) = 85.0%; specificity (TN/TN + FP) = 98.7%; positive predictive value (TP/TP + FP) = 14.0%; negative predictive value (TN/TN + FN) = 99.9%.
True positive (TP)	Positive	Histology confirms ovarian cancer	51
False positive (FP)	Positive	Benign ovarian histology	313
True negative (TN)	Negative	No evidence of disease 12 mo after negative screen	24,954
False negative (FN)	Negative	Ovarian cancer diagnosed within 12 mo of negative screen	9

Table 6. Statistical Evaluation of Screening According to Year of Trial Entry, Family History of Ovarian Cancer, and Menopausal Status

Variable	Total no. of cases	1987 to 2001	2001 to present	No family history	Family history	Premenopausal	Postmenopausal
TP indicates true positive; FP, false positive; FN, false negative; TN, true negative; Sen, sensitivity; Spec, specificity; PPV, positive predictive value; NPV. negative predictive value. * The screening algorithm changed, so that unilocular ovarian cysts that measured ≤5 cm in greatest dimension were followed sonographically and were not removed surgically.
Total	25,327	16,242	9085	19,270	6057	7078	18,249
TP	51	23	28	42	9	7	44
FP	313	238	75	227	86	72	241
FN	9	4	5	6	3	1	8
TN	24,954	15,977	8977	18,995	5959	6998	17,956
Sen	0.85	0.851	0.849	0.875	0.75	0.875	0.846
Spec	0.987	0.985	0.992	0.988	0.986	0.989	0.986
PPV	0.140	0.088	0.271	0.156	0.095	0.088	0.154
NPV	0.9996	0.9997	0.9994	0.9997	0.9995	0.9998	0.9996
Screening, y*	107,276	90,540	16,736	78,397	28,879	37,683	82,811

The survival rate of patients with ovarian cancer in the annual screening group was 92.1% ± 7.9% at 2 years and 82.4% ± 17.6% at 5 years. When we excluded the 5 patients who had granulosa cell tumors and the 9 patients who had serous ovarian tumors of borderline malignancy, the survival rate was 89.9% ± 10.1% at 2 years and 77.2% ± 22.8% at 5 years. The survival rate of all patients with ovarian cancer in the study group, including those who had cancer detected by screening and those who were diagnosed >12 months after a negative screen by clinical examination, was 88.2% ± 11.8% at 2 years and 76.6% ± 23.4% at 5 years. This compares favorably with a 2-year survival rate of 70.7% ± 2.6% and a 5-year survival rate of 48.7% ± 3.5% for patients with ovarian cancer in the University of Kentucky Tumor Registry who did not receive screening (P < .001).

DISCUSSION

Analysis of data from nearly 20 years of screening allows us to reach certain conclusions about the benefits and limitations of TVS as a screening method for ovarian cancer. First, TVS is safe, time efficient, and well accepted by patients. In over 100,000 screens, <10 women have complained that TVS is uncomfortable. A complete screening examination usually takes from 5 minutes to 10 minutes, and no medical or surgical complications from TVS have been observed. When TVS is used in high-volume settings, its cost is well within the range of other accepted screening tests, and interobserver variation is minimal.16–18 Although there is a definite learning curve for technologists performing TVS, 1 or both ovaries were visualized in 84% of women, and ovarian tumor dimensions predicted by TVS correlated closely with recorded measurements at the time of surgery.

For a screening test to be effective, it should be sensitive and specific, and it should have both a high PPV and a high NPV.19 Furthermore, periodic screening should decrease the disease stage at detection and increase disease-specific survival in the screened population.20 The major limitations of TVS as a screening method for ovarian cancer are its moderate sensitivity and relatively low PPV. Although TVS screening accurately predicted the presence of cancer involving the ovary in 51 asymptomatic women, there were 9 patients who developed ovarian cancer within 12 months of a normal scan (false-negative results). Six of those patients had minimally enlarged ovaries at the time of surgery from 3 months to 12 months (mean, 8 months) after a normal scan but had extraovarian spread. Five of these 6 ovarian cancers were poorly differentiated. TVS did not detect an abnormality in these women, because their ovarian volume was increased minimally. Whether the addition of a serum CA-125 determination to TVS as a primary test in the screening algorithm would have resulted in the earlier detection in these women cannot be answered from these data because CA-125 measurement was obtained as a secondary test only in those patients who had a persisting abnormality on TVS. However, serum CA-125 levels were elevated in 8 of those 9 patients (88%) at the time of surgery, and it is logical to assume that marker elevation would have preceded clinical detection in some of these women. Other screening trials that were designed to answer this question are in progress.7, 13 In the current investigation, serum CA-125 levels were increased (>35 U/mL) at the time of detection in 13 of 15 patients (87%) who had stage III epithelial ovarian cancer but in only 3 of 15 patients (20%) who had stage I or II disease.

A second area of concern related to TVS screening for ovarian cancer is its relatively low PPV. Whereas TVS is accurate in identifying ovarian tumors, it is less effective in differentiating benign lesions from malignant lesions. In the current investigation, 14% of persisting complex ovarian tumors were malignant. Tumor morphology21–24 and serum biomarker patterns25–28 both have been useful in predicting the risk of malignancy in sonographically confirmed ovarian tumors. In the current investigation, all but 1 of the screen-detected ovarian malignancies were either solid tumors or cystic tumors with solid areas or papillary projections from the cyst wall. There were 153 unilocular cystic ovarian tumors that measured <5 cm in greatest dimension (MI, ≤4), and all were benign. In 2001, the screening algorithm at the University of Kentucky was changed, such that unilocular cystic ovarian tumors that measured ≤5 cm in greatest dimension were followed sonographically rather than removed surgically. Since that time, the PPV of TVS screening has increased to 27.1%, which is well within the range of other accepted screening tests, such as mammography and cervical cytology.29–31 None of these patients has developed ovarian cancer, thereby confirming the safety of following small unilocular ovarian cysts sonographically and avoiding surgery in these women. It is our current policy to follow all patients with unilocular cystic ovarian tumors that measure ≤5 cm in greatest dimension at 6-month intervals with TVS rather than proceeding with operative intervention.

The use of a single biomarker value to distinguish benign ovarian tumors from early-stage ovarian cancer has been problematic. In the current investigation, for example, serum CA-125 measurements were obtained at the time of surgery in 9 patients with stage I invasive epithelial ovarian cancer detected by screening and was elevated (>35 U/mL) in only 1 patient. Recent data suggest that serial marker determinations may be more effective than a single threshold value in identifying ovarian malignancies. Generally, serum CA-125 values rise over time in patients with ovarian cancer, whereas they remain stable or decrease in patients with benign ovarian tumors. Using a risk calculation based on progressively rising serum CA-125 levels, Skates and colleagues32 were able to increase the sensitivity of ovarian cancer detection from 62% to 86%. Therefore, serum CA-125 determinations at 2- to 4-week intervals can be helpful in establishing the risk of malignancy in a specific ovarian tumor.

Despite the aforementioned limitations, annual TVS screening decreases disease stage at detection and increases case-specific ovarian cancer survival. In the current trial, 82% of women who had ovarian cancer detected by screening had stage I or stage II disease versus 34% of women in the unscreened historic control group (P < .0001). In 107,276 screening years, 53 primary ovarian cancers were detected (44 true-positive results and 9 false-negative results). Seven deaths occurred in women who were compliant with the screening algorithm. Two of those patients had advanced-stage disease at the time of their initial screen and died of disease after treatment with surgery and chemotherapy. The survival rate for all patients with invasive epithelial ovarian cancer in the screening group was 89.9% at 2 years and 77.2% at 5 years, which was significantly higher (P < .001) than the survival rate of patients with ovarian cancer in the control group who did not have screening (2-year survival rate, 70.9%; 5-year survival rate, 48.7%).

Ovarian cancer poses specific challenges for screening. Although ovarian cancer is curable when it is detected early, it is not common, and rates of progression, even among epithelial cancers, are variable. In this trial, ovarian cancer screening was performed at yearly intervals. However, it may be necessary to increase the frequency of screening in certain high-risk women. It should be possible, as more complete molecular genetic data concerning the etiology of ovarian cancer become available, to construct more accurate risk profiles for each patient and to match screening frequency to risk of malignancy.33 Finally, investigation must continue to identify proteins or other biomarkers that are specific to ovarian cancer that can be used for early diagnosis.34, 35 This is particularly important for the detection of cancers that involve only the ovarian surface epithelium and do not produce ovarian enlargement, thus evading sonographic surveillance.

With more effective chemotherapy now available for patients with ovarian cancer, even a modest advance in the time to diagnosis should be associated with a significant survival advantage.36 In the absence of early detection, most women with ovarian cancer will continue to present with advanced-stage disease, for which the cost of treatment is high and the cure rate is extremely low. The protective effect of annual sonographic screening on ovarian cancer mortality observed in the current trial should only increase as more specific biomarkers are added to TVS in screening algorithms.