Purpose: About 60% of ovarian cancers are diagnosed at late stage, when 5-year survival is less than 30% in contrast to 90% for local disease. This has prompted search for early detection biomarkers. For initial testing, specimens taken months or years before ovarian cancer diagnosis are the best source of information to evaluate early detection biomarkers. Here we evaluate the most promising ovarian cancer screening biomarkers in prospectively collected samples from the European Prospective Investigation into Cancer and Nutrition study.

Experimental Design: We measured CA125, HE4, CA72.4, and CA15.3 in 810 invasive epithelial ovarian cancer cases and 1,939 controls. We calculated the sensitivity at 95% and 98% specificity as well as area under the receiver operator curve (C-statistic) for each marker individually and in combination. In addition, we evaluated marker performance by stage at diagnosis and time between blood draw and diagnosis.

Results: We observed the best discrimination between cases and controls within 6 months of diagnosis for CA125 (C-statistic = 0.92), then HE4 (0.84), CA72.4 (0.77), and CA15.3 (0.73). Marker performance declined with longer time between blood draw and diagnosis and for earlier staged disease. However, assessment of discriminatory ability at early stage was limited by small numbers. Combinations of markers performed modestly, but significantly better than any single marker.

Conclusions: CA125 remains the single best marker for the early detection of invasive epithelial ovarian cancer, but can be slightly improved by combining with other markers. Identifying novel markers for ovarian cancer will require studies including larger numbers of early-stage cases. Clin Cancer Res; 22(18); 4664–75. ©2016 AACR.

See related commentary by Skates, p. 4542

This article is featured in Highlights of This Issue, p. 4537

Translational Relevance

Ovarian cancer is a leading cause of cancer-related death among women. About 60% of ovarian cancers are diagnosed at late stage, when 5-year survival is less than 30% in contrast to 90% for local disease. Biomarkers for early detection are urgently needed to improve survival. Using blood samples of ovarian cancer cases and cancer-free control subjects from the European EPIC study, we examined the prospective diagnostic capacity of CA125, HE4, CA72.4, and CA15.3. All markers were significantly elevated many months before clinical manifestation of ovarian cancer. The best discrimination between cases and controls was within 6 months of diagnosis for CA125 (C-statistic = 0.92) and HE4 (0.84). Marker performance declined with longer time between blood draw and diagnosis and for earlier staged disease. Combining CA125 with HE4 and further markers modestly improved discrimination. Our study confirms CA125 as the single best marker for the early detection of invasive epithelial ovarian cancer.

Ovarian cancer generally does not exhibit specific early symptoms. About 60% of ovarian cancers are diagnosed at late stage, which is associated with a 5-year survival of less than 30%, contrasted with more than 90% survival for disease found locally (1). This has prompted extensive research to find early detection biomarkers for ovarian cancer.

Many potential serum biomarkers for ovarian cancer have been indentified (2). Candidate biomarkers are often first identified from preclinical studies using immunohistochemical testing or gene expression profiles of tumor tissue. These are called phase I studies (3). Potential biomarkers are then tested by comparing blood from cases at diagnosis of ovarian cancer with blood from either women with benign disease or healthy controls. This type of study has been described as a phase II study. Markers that have been approved using phase II data include CA125, HE4, and a panel of markers, including prealbumin, apolipoprotein A-1, β2-microglobulin, and transferrin (4, 5). CA125 has been approved for disease monitoring (6), and HE4 and a panel of markers as tools for distinguishing benign from malignant pelvic masses (4, 5). Phase III data refers to studies based on blood samples from asymptomatic women taken months or years prior to a diagnosis of ovarian cancer, while phase IV data refers to markers tested in a clinical trial in which asymptomatic women are randomized to a screening arm or to usual care.

There have been three randomized trials of screening for ovarian cancer using either CA125 alone or CA125 in combination with transvaginal ultrasound (TVUS). No reduction in ovarian cancer mortality was observed in the Prostate, Lung, Colorectal, and Ovarian cancer (PLCO) screening trial, based on a combination of TVUS and CA125 measurements for 4 years and two additional years of CA125 measurements (7). However, recent results from the UKCTOCs study showed a 15% reduction in mortality for postmenopausal women followed for change in CA125, which was marginally significant (8). One additional randomized trial in Japan (9) showed a nonsignificant increase in early-stage tumors detected in the screening arm but did not follow participants for mortality (9). To date, neither CA125 nor TVUS have been approved or recommended for screening on the basis of the randomized trials.

Although selected specimens from the phase IV studies may be and have been used in the context of discovering and testing new biomarkers, they may not be ideal for this purpose for at least two reasons. First, as CA125 was the primary screening tool, this may lead to preferential selection of CA125-expressing tumors. Second, as annual screening was employed, the natural history of the disease may have been interrupted at early stages and may not provide a true measure of the lead time, that is, the time between early diagnosis with screening and when diagnosis would have occurred in the absence of screening. Therefore, samples collected in asymptomatic women before cancer diagnosis are needed to test new biomarkers. To date, only a few case–control comparisons have been made in prospectively collected specimens obtained in asymptomatic women before clinical cancer diagnosis under usual care (phase III studies; refs. 10–12). Both the Carotene and Retinol Efficacy Trial (CARET) and Women's Health Initiative (WHI) studies were designed as randomized trials to evaluate other disease outcomes but have the advantage of closely monitoring a large group of women with banked blood samples. In the CARET study, a panel of markers including CA125, HE4, mesothelin, B7-H4, DcR3, and spondin-2 were measured on serial samples from 34 women with ovarian cancer and 70 matched controls. Of these, only CA125 and HE4 showed significant differences between cases and controls and had modest discriminatory ability that waned with increasing time between blood draw and diagnosis. Similarly, in the WHI study, CA125 and HE4 were measured in 353 ovarian cancer cases and 1,261 healthy controls and these markers significantly improved a risk prediction algorithm based on epidemiologic factors (11).

Here, we present results from a phase III study using serum samples from the EPIC cohort (European Prospective Investigation into Cancer and Nutrition). We measured CA125, CA15.3, HE4, and CA72.4 in 197 cases of ovarian cancer diagnosed within the first 3 years after blood donation and 724 matched control subjects. For 613 additional ovarian cancer patients diagnosed more than 3 years after blood draw, and for 1,215 additional control subjects, we extended the measurements of CA125 and CA15.3 for examination of a possible longer-term risk diagnostic prediction capacity of these markers and to allow more accurate analyses of possible relationships of these markers with epidemiologic risk factors for ovarian cancer. The objectives of our analyses were: (i) to examine the early detection capacity of our biomarker panel for ovarian cancer diagnoses within comparatively short time intervals (variable lag time strata ≤3 years between blood donation and diagnosis); (ii) to examine the capacity of CA125 and CA15.3 to predict ovarian cancer risk over a longer term (>3 years between blood donation and diagnosis); and (iii) to examine whether early diagnostic capacity or longer-term risk prediction by the biomarkers could be improved by integrating further information about a woman's general epidemiologic risk factor profile.

The EPIC cohort—background and collection of blood samples

The European Prospective Investigation into Cancer and Nutrition is an ongoing multicenter prospective cohort study designed to investigate the relationship between diet, nutrition, and metabolic factors with cancer. Descriptions of study design, population, and baseline data collection of the cohort have been reported in detail previously (13, 14). In brief, 519,978 participants (366,521 women) were enrolled from 1992 to 2000 in 23 centers in 10 European countries: Denmark, France, Germany, Greece, Italy, Norway, the Netherlands, Spain, Sweden, and the United Kingdom. For women, the age range at recruitment was mostly between 35 and 70 years. At baseline, comprehensive data on diet, lifestyle, reproductive and menstrual factors, current and past use of exogenous hormones [oral contraceptives (OC) and postmenopausal hormone replacement therapy (HRT)] and medical history were collected through standardized questionnaires. In addition, anthropometric measures were obtained.

A total of 385,747 study participants in the EPIC cohort (226,673 women and 159,074 men) also provided a baseline blood sample. In France, the Netherlands, the United Kingdom, Germany, Spain, Italy, and Greece, blood samples were collected according to a standardized protocol (15). From each study participant, about 30 mL of blood was drawn, and serum, plasma, erythrocytes, and buffy coat were aliquoted in 28 plastic straws of 0.5 mL each, which were heat-sealed and stored under liquid nitrogen (−196°C). In Denmark, blood fractions were aliquoted into 1-mL tubes, and stored in the vapor phase in liquid nitrogen containers (−150°C). In the Swedish Center of Umeå, blood samples were divided into 10 aliquots of 1.5-mL each: 6 plasma, 2 buffy-coat, and 2 erythrocytes, which were rapidly frozen at −80°C in standard freezers.

Ascertainment of incident cancer cases

Prospective follow-up for cancer occurrences and histologic confirmation was performed through record linkage with cancer and pathology registries (all countries except France, Germany and Greece) or through active follow-up and systematic verification of self-reports by detailed examination and coding of clinical records. In all countries, vital status was determined by regular linkages with population and mortality registers at the regional or national level.

At the time the current study was initiated, prospective follow-up was complete until the end of 2005 (France) to 2008 (Germany). Within this timeframe, and among those women who had provided a baseline blood sample, a total of 810 incident invasive cases of ovarian cancer had been identified. Case subjects were defined as women who developed incident epithelial invasive ovarian (ICD code: C569), fallopian tube (C570), or peritoneal cancer (C480, C481, C482, C488) after recruitment into the EPIC study according to the International Classification of Diseases for Oncology (ICD) 0–3 and with data on tumor histology.

More detailed information on tumor characteristics [histologic subtype [serous, endometrioid, clear cell, mucinous, not otherwise specified (NOS), grade (well, moderately, or poorly/undifferentiated), and stage (local, regional, metastatic)] were obtained from pathology reports and from cancer registries. For the 810 invasive ovarian cases, complete information on tumor grade was available for 473 patients (58%) and information on tumor stage was available for 712 patients (88%). Well-differentiated tumors were classified as low grade; moderately and poorly/undifferentiated tumors were classified as high grade. We classified cases with local disease (stage I) as low stage and cases with regional (stage II) or metastatic disease (stage III/IV) as high stage.

Design of nested case–control study

For each case subject, up to four control subjects were randomly selected among appropriate risk sets consisting of all female cohort members with a blood sample, alive and free of cancer at the time of diagnosis of the index case. An incidence density sampling protocol was used, such that, in principle, control subjects could include study participants who became a case later in time and each control subject could be sampled more than once—the control subjects actually drawn, however, did not include any of the future cases of ovarian cancer detected so far in the EPIC cohort. Case and control subjects were matched on study recruitment center, age at blood donation (±6 months), time of the day of blood collection (±1 hour), fasting status (<3, 3–6, or >6 hours), follow-up time, and menopausal status at blood collection (premenopausal, perimenopausal, postmenopausal), current use of exogenous hormones (OCs, HRT) at the time of blood draw, as well as menstrual cycle phase for premenopausal women (3–5 categories, depending on available data). Cases missing data on phase of menstrual cycle were matched to control subjects whose information on menstrual cycle phase was also missing.

Informed consent and data protection

All participants had given their consent for future analyses of their blood samples and the current study was approved by the IARC Ethics Committee and the Institutional Review Board of Brigham and Women's Hospital (Boston, MA).

Laboratory assays

Measurements of the protein levels of CA125, CA15.3, HE4, and CA72.4 were completed for a total of 197 incident cases of invasive ovarian cancer and 725 matched, cancer-free control subjects. In addition, we measured CA125 and CA15.3 in 613 cases with invasive ovarian cancer who had been diagnosed more than 3 years after blood donation, and 1,214 additional matched control subjects. Laboratory values were missing for CA125 (3 cases, 12 controls), CA15.3 (6 cases, 19 controls), HE4 (2 controls), and CA72.4 (2 controls) due to lack of sufficient sample volume. All measurements were performed in the Genital Tract Biology Laboratory following preanalytic, analytic, and postanalytic SOPs established under the laboratory's accreditation by the College of American Pathologists, using a volume-effective highly sensitive multiplex platform (Meso Scale Discovery, MSD) based on electrochemoluminiscence (ECL) detection. Human CA125 (catalog number K151WC) and Human Prototype CA15.3 (catalog number N45ZA-1) were provided by MSD in singleplex assays. The linearity range for CA125 was 10,000–0.6 U/mL, and for CA15.3 was 12,500–0.19 mU/mL. The HE4 and CA72.4 measurements were done in a custom designed duplex assay. The following reagents were a gift from Fujirebio Diagnostics, Inc.: IgHE4 antigen, which we used to generate a calibration curve with a linear range starting at 3600 pmol/L; anti-HE4 capture IgG1 (2H5 mouse hybridoma, Fujirebio catalog number 414-01S); anti-HE4 detection IgG1 (mouse hybridoma 3D8, Fujirebio catalog number 415-01); TAG72 Defined Antigen, which we used to generate a calibrator curve with a linear range starting at 2,400 U/mL; anti-72.4 capture IgG1 (mouse hybridoma CC49, Fujirebio catalog number 110-005); anti-CA72.4 IgG1 (mouse hybridoma B72.3, Fujirebio catalog number 110-000). The samples were split into batches such that matched case–control sets and samples from the same study center were kept together in the same batches. The samples were tested undiluted in the CA125 singleplex and the HE4/CA72.4 duplex, and they were tested at a 50-fold dilution in the CA15.3 assay. A quality control pool was prepared from serum samples from ovarian cancer patients with within linearity range levels of each protein and split into equal aliquots. To establish interplate variability, one aliquot of this pool was tested at multiple dilutions spanning the linearity range of each assay, three dilutions run in duplicates and two dilutions run in triplicates, providing up to six quality control data points in each assay plate. In addition, blinded, randomly chosen citrated plasma, EDTA plasma, and serum Blood Bank samples were split into aliquots (128 for CA125, 130 for CA15.3, and 104 for HE4/CA72.4) and distributed within and between plates. Coefficients of variation (CV) were calculated as 100 × SD/mean. The unblinded quality control sample pool repeatedly tested on every assay plate showed the following interplate CV and min–max range (mean) of intraplate CV: (i) CA125, 8.4% interplate CV and 0.2%–13.5% (3.4%) intraplate CV; (ii) CA15.3: 15.4% interplate CV and 0.5%–6.1% (2.3%) intraplate CV; (iii) HE4: 8.99% interplate CV and 1.6%–7.6% (3.6%) intraplate CV; (iv) CA72.4: 17.3% interplate CV and 0.9%–13% (5.5%) intraplate CV. Similarly, the blinded aliquots with values within the linearity range of each assay showed the following interplate CVs and min–max (mean) intraplate CVs: 19% and 3%–20% (9%) for CA125, 22% and 3%–5% (4%) for CA15.3, 9% and 4%–10% (6%) for HE4, 16% and 1%–16% (6%) for CA72.4. As the majority of the blinded aliquots for CA72.4 fell below the lower limit of detection, blinded CA72.4 CVs were based on the remaining 13 aliquots, ranging in CA72.4 value from 1.15 to 1.87 U/mL.

Statistical analyses

First, we evaluated the distribution of each biomarker for normality and outliers. As 81% of the samples had CA72.4 values below the lower limit of detection for this assay (1.119 U/mL), we assigned these values to the midpoint between zero and the lower limit of detection for future analyses. Other markers assessed did not have any values below the lower limit of detection. Locally estimated scatterplot smoothing (LOESS) curves were used to describe mean levels of each marker among cancer cases and control subjects at different lag-times until ovarian cancer diagnosis. The discrimination between cases and control subjects was described using ROC) curves, with the AUC, also known as the C-(concordance) statistic, as an overall measure for discrimination capacity. We estimated the diagnostic sensitivities (SE95 and SE98, respectively) of each marker at cut-off points corresponding to 95% and 98% specificity, determined in our full dataset for all control subjects (N = 1,939 for CA125 and CA15.3; N = 725 for HE4 and CA72.4).

The diagnostic sensitivity, specificity, and C-statistics were calculated for risk scores based on the associations between biomarker levels and ovarian cancer risk, overall, and by strata of lag-time between blood donation and cancer diagnosis, conditional logistic regression models were used, accounting for the matched study design. Models were fitted for continuous biomarker measurements after log2-transformation, to achieve approximate normality of their distributions. Basic analyses focused on single markers. Additional multivariate models were developed to examine the discrimination capacity of multiple markers in combination, and of markers combined with an epidemiologic risk prediction algorithm, including age at menopause, duration of hormone replacement therapy, body mass index, unilateral ovariectomy, duration of oral contraceptive use, and number of full-term pregnancies that we developed previously on the basis of the full EPIC cohort data (16).

To examine how the early detection and/or risk prediction capacities of the biomarkers changed with time between blood draw and clinical cancer diagnosis, all analyses were performed within variable strata of lag-time (≤6 months, ≤12 months, 1–2 years, 2–3 years, 3–6 years, >6 years). To examine heterogeneity of diagnostic prediction capacity by tumor stage at diagnosis or by histologic tumor subtypes, likelihood-ratio tests were used comparing the model fit for logistic regression models with and without corresponding interaction terms. For all risk models, the discrimination between cases and control subjects was described using ROC curves.

For multimarker discrimination models, the statistical fit of nested models was compared with likelihood-ratio tests, and bootstrapping methods were used to correct for model overfitting and overoptimism in the estimation of discrimination capacity. In addition, measures of continuous net reclassification improvement were calculated, which represents the percent of case and control subjects correctly reclassified as a result of the added marker (17). Analyses were conducted in SAS (version 9.3, SAS Institute).

Baseline characteristics of ovarian cancer case patients by tumor characteristics are presented in Table 1. Of the 810 case patients examined in this study, 752 (93%) had the ovary classified as primary tumor site, whereas in 33 (4%) the primary site was the fallopian tube and in 25 patients (3%) it was the peritoneum. More than half of the tumors (55%) were of serous histology (n = 445), 12% endometrioid (n = 96), 7% mucinous (n = 58), 5% other (malignant epithelial neoplasms, carcinoma, malignant mixed Müllerian or malignant Brenner tumors; n = 38), 4% clear cell (n = 37), and 17% not otherwise specified (NOS; n = 136). Of the 712 case subjects with information about tumor stage at diagnosis, 115 were classified as stage I, 128 as stage II, and 469 as stage III and higher (stage III/IV). Compared with case patients diagnosed at stage II and higher, there was a relative over-representation of mucinous, clear cell, and endometrioid tumors among the stage I patients, whereas serous tumors were predominantly represented among the patients with cancer in stage II and higher (see Supplementary Table S1). Overall, the median age at cancer diagnosis was 62.7 years (range: 30.6–86.5 years), and varied according to the histologic subtypes (Supplementary Table S1).

Table 1.

Characteristics of cases and controls in the EPIC cohort [median (min–max) or n (%)]

Cases (n = 810)Controls (n = 1,939)P
Age at blood draw, years 56.4 (29.9–80.7) 56.7 (30.1–79.3)  
Age at blood draw, years    
 <50 166 (20%) 405 (21%)  
 50–55 187 (23%) 430 (22%)  
 55–60 183 (23%) 430 (22%)  
 60–65 185 (23%) 445 (23%)  
 ≥65 89 (11%) 229 (12%)  
Menopausal status    
 Pre 132 (16%) 329 (17%)  
 Peria 118 (15%) 274 (14%)  
 Post 560 (69%) 1 336 (69%)  
BMI 25.1 (17.2–45.4) 25.0 (14.9–50.6) 0.03 
Smokingb   0.12 
 Never 432 (55%) 1 103 (58%)  
 Former 185 (23%) 435 (23%)  
 Current 177 (22%) 368 (19%)  
Parousb 617 (83%) 1 585 (89%) <0.0001 
Number of childrenb,c   0.31 
 1 114 (19%) 277 (18%)  
 2 297 (49%) 733 (48%)  
 >2 191 (32%) 531 (34%)  
Hysterectomyb 69 (11%) 176 (11%) 0.35 
Case characteristics 
Age at diagnosis 62.7 (30.6–86.5) —  
Lag time (years) 6.1 (0–16.0) —  
Cancer site    
 Ovary 752 (93%) —  
 Fallopian tube 33 (4%) —  
 Peritoneum 25 (3%) —  
Histology    
 Serous 445 (55%) —  
 Mucinous 58 (7%) —  
 Endometrioid 96 (12%) —  
 Clear cell 37 (4%) —  
 NOS 136 (17%) —  
 Other 38 (5%) —  
Cancer gradeb    
 Well differentiated 45 (9%) —  
 Moderately differentiated 164 (35%) —  
 Poorly differentiated/undifferentiated 264 (56%) —  
Disease spreadb    
 Localized (stage I) 115 (16%) —  
 Regional (stage II) 128 (18%) —  
 Metastatic (stage III/IV) 469 (66%) —  
Markerd 
CA125 (U/mL) 27.6 (26.2–29.1) 20.2 (19.5–20.8) <0.0001 
HE4 (pmol/L) 29.1 (26.9–31.6) 18.9 (18.1–19.7) <0.0001 
CA72.4 (U/mL)e 5.5 (4.4–6.9) 2.4 (2.0–2.9) 0.004 
CA15.3 (mIU/mL) 624.3 (600.3–649.2) 600.6 (585.6–616.0) 0.10 
Cases (n = 810)Controls (n = 1,939)P
Age at blood draw, years 56.4 (29.9–80.7) 56.7 (30.1–79.3)  
Age at blood draw, years    
 <50 166 (20%) 405 (21%)  
 50–55 187 (23%) 430 (22%)  
 55–60 183 (23%) 430 (22%)  
 60–65 185 (23%) 445 (23%)  
 ≥65 89 (11%) 229 (12%)  
Menopausal status    
 Pre 132 (16%) 329 (17%)  
 Peria 118 (15%) 274 (14%)  
 Post 560 (69%) 1 336 (69%)  
BMI 25.1 (17.2–45.4) 25.0 (14.9–50.6) 0.03 
Smokingb   0.12 
 Never 432 (55%) 1 103 (58%)  
 Former 185 (23%) 435 (23%)  
 Current 177 (22%) 368 (19%)  
Parousb 617 (83%) 1 585 (89%) <0.0001 
Number of childrenb,c   0.31 
 1 114 (19%) 277 (18%)  
 2 297 (49%) 733 (48%)  
 >2 191 (32%) 531 (34%)  
Hysterectomyb 69 (11%) 176 (11%) 0.35 
Case characteristics 
Age at diagnosis 62.7 (30.6–86.5) —  
Lag time (years) 6.1 (0–16.0) —  
Cancer site    
 Ovary 752 (93%) —  
 Fallopian tube 33 (4%) —  
 Peritoneum 25 (3%) —  
Histology    
 Serous 445 (55%) —  
 Mucinous 58 (7%) —  
 Endometrioid 96 (12%) —  
 Clear cell 37 (4%) —  
 NOS 136 (17%) —  
 Other 38 (5%) —  
Cancer gradeb    
 Well differentiated 45 (9%) —  
 Moderately differentiated 164 (35%) —  
 Poorly differentiated/undifferentiated 264 (56%) —  
Disease spreadb    
 Localized (stage I) 115 (16%) —  
 Regional (stage II) 128 (18%) —  
 Metastatic (stage III/IV) 469 (66%) —  
Markerd 
CA125 (U/mL) 27.6 (26.2–29.1) 20.2 (19.5–20.8) <0.0001 
HE4 (pmol/L) 29.1 (26.9–31.6) 18.9 (18.1–19.7) <0.0001 
CA72.4 (U/mL)e 5.5 (4.4–6.9) 2.4 (2.0–2.9) 0.004 
CA15.3 (mIU/mL) 624.3 (600.3–649.2) 600.6 (585.6–616.0) 0.10 

aDefined as women between the ages of 42 and 55 years who have missing or incomplete questionnaire data, reported irregular menstrual cycles in the past 12 months, or had a prior hysterectomy without oophorectomy.

bData were missing on smoking for 16 cases and 33 controls, on parity for 64 cases and 151 controls, on number of children for 79 cases and 195 controls, on hysterectomy for 178 cases and 436 controls, on cancer grade for 337 cases, on the dualistic model for 385 cases, and on disease spread for 98 cases.

cAmong parous women (n = 2,202).

dPresented as geometric mean (5th–95th percentile); Data were missing on CA125 for 3 cases and 12 controls, on HE4 for 2 controls, on CA72.4 for 2 controls, and on CA15.3 for 6 cases and 19 controls.

eBased on 67 (34%) cases and 109 (15%) controls with CA72.4 above the detection limit (CA72.4 was measured in 197 cases and 725 controls).

Visual inspection of LOESS curves suggests that none of the biomarkers were increased over normal (control) values earlier than about 2 years prior to diagnosis, and more than 6 months prior to diagnosis increased levels were discernable only for case patients who were diagnosed with ovarian cancer at advanced stage (stage II or III; Fig. 1).

Figure 1.

LOESS curves over time prior to diagnosis. The blue line represents controls, the green line stage I cases, the yellow line stage II cases, and the red line stage III/IV cases. For CA125 and CA15.3, data are shown only for lag-times between blood donation and cancer diagnosis up to 6 years.

Figure 1.

LOESS curves over time prior to diagnosis. The blue line represents controls, the green line stage I cases, the yellow line stage II cases, and the red line stage III/IV cases. For CA125 and CA15.3, data are shown only for lag-times between blood donation and cancer diagnosis up to 6 years.

Close modal

For the predefined variable lag times between blood donation and date of diagnosis, the ability of the early detection markers to discriminate between case patients and control subjects is indicated by C-statistics and estimated sensitivities at specificity cut-off points of 95% (SE95) and 98% (SE98; Table 2). In addition, ROC curves are shown in Fig. 2. For blood samples taken ≤6 months prior to diagnosis, the highest C-statistic was observed for CA125 (C = 0.92), followed by HE4 (C = 0.84), CA72.4 (C = 0.77), and CA15.3 (C = 0.73). Correspondingly, within the first 6 months, values for SE95 and SE98 were fairly high for CA125 (0.81 and 0.77, respectively) and HE4 (0.67 and 0.59), and modest for CA72.4 (0.56 and 0.37) and CA15.3 (0.31 and 0.23). For all markers, the capacity to discriminate between future case patients and noncases dropped rapidly with increasing time lags between blood donation and tumor diagnosis (Table 2; Fig. 2). For example, for a time lag between 1 and 2 years, C-statistic values were 0.72 for CA125, 0.65 for HE4, 0.61 for CA72.4, and 0.52 for CA15.3. At time lags between 3 and 6 years, the two markers that were tested for longer-term prediction of ovarian cancer, CA125 and CA15.3 had C-statistics of only 0.55 and 0.53, respectively (Table 2).

Table 2.

Sensitivity at 95% and 98% specificity and C-statistics by time between blood draw and diagnosis

# of SetsSensitivity at 95% specificitySensitivity at 98% specificityC-statistic (95% CI)Pheta
CA125  Cut-off point: 56.64 U/mL Cut-off point: 77.92 U/mL   
Overall 807 0.14 (0.11–0.18) 0.10 (0.07–0.13) 0.58 (0.56–0.60)  
 ≤6 months 26 0.81 (0.61–0.92) 0.77 (0.57–0.89) 0.92 (0.86–0.98)  
 ≤12 months 61 0.59 (0.46–0.71) 0.52 (0.39–0.66) 0.82 (0.76–0.88)  
 >1–2 years 75 0.27 (0.18–0.38) 0.20 (0.12–0.31) 0.72 (0.65–0.78)  
 >2–3 years 58 0.10 (0.05–0.22) 0.03 (0.01–0.13) 0.56 (0.48–0.64)  
 >3–6 years 200 0.08 (0.05–0.13) 0.06 (0.03–0.11) 0.55 (0.50–0.60)  
 >6 years 413 0.09 (0.06–0.13) 0.04 (0.02–0.07) 0.54 (0.51–0.57)  
Stage I 115 0.16 (0.1–0.24) 0.10 (0.06–0.18) 0.58 (0.51–0.64)  
 ≤6 months 0.43 (0.14–0.77) 0.29 (0.07–0.68) 0.69 (0.47–0.91)  
 ≤12 months 16 0.38 (0.18–0.63) 0.31 (0.13–0.57) 0.65 (0.49–0.80)  
 >1–2 years 0.14 (0.02–0.58) 0.14 (0.02–0.58) 0.55 (0.31–0.80)  
 >2–3 years 0 (0–0.34) 0 (0–0.34) 0.52 (0.30–0.74)  
 >3–6 years 32 0.13 (0.05–0.29) 0.09 (0.03–0.26) 0.60 (0.48–0.72)  
 >6 years 51 0.14 (0.07–0.26) 0.06 (0.02–0.17) 0.57 (0.47–0.66)  
Stage II 128 0.14 (0.09–0.22) 0.09 (0.05–0.16) 0.61 (0.55–0.67)  
 ≤6 months     
 ≤12 months 0.75 (0.38–0.94) 0.50 (0.20–0.80) 0.91 (0.79–1.02)  
 >1–2 years 10 0.30 (0.10–0.63) 0.30 (0.10–0.63) 0.75 (0.58–0.93)  
 >2–3 years 11 0.09 (0.01–0.44) 0 (0–0.28) 0.66 (0.48–0.84)  
 >3–6 years 37 0.05 (0.01–0.19) 0.05 (0.01–0.20) 0.53 (0.41–0.64)  
 >6 years 62 0.10 (0.04–0.20) 0.05 (0.02–0.14) 0.57 (0.48–0.66)  
Stage III/IV 467 0.14 (0.10–0.18) 0.09 (0.06–0.13) 0.57 (0.54–0.60) 0.26 
 ≤6 months 15 0.93 (0.65–0.99) 0.93 (0.64–0.99) 0.96 (0.92–1.01) 0.09 
 ≤12 months 31 0.68 (0.49–0.82) 0.65 (0.46–0.80) 0.87 (0.80–0.94) 0.045 
 >1–2 years 50 0.26 (0.15–0.40) 0.16 (0.08–0.30) 0.72 (0.64–0.80) 0.77 
 >2–3 years 36 0.14 (0.06–0.30) 0.06 (0.01–0.20) 0.57 (0.46–0.67) 0.20 
 >3–6 years 106 0.07 (0.03–0.13) 0.04 (0.01–0.10) 0.52 (0.45–0.59) 0.68 
 >6 years 244 0.07 (0.05–0.12) 0.04 (0.02–0.07) 0.53 (0.48–0.57) 0.34 
HE4  Cut-off point: 39.41 pmol/L Cut-off point: 54.01 pmol/L   
Overall 197 0.24 (0.17–0.32) 0.18 (0.11–0.27) 0.67 (0.63–0.71)  
 ≤6 months 27 0.67 (0.46–0.82) 0.59 (0.39–0.77) 0.84 (0.76–0.92)  
 ≤12 months 62 0.48 (0.35–0.62) 0.39 (0.25–0.54) 0.79 (0.73–0.85)  
 >1–2 years 76 0.17 (0.10–0.28) 0.14 (0.07–0.26) 0.65 (0.58–0.72)  
 >2–3 years 59 0.07 (0.02–0.17) 0 (0–0.06) 0.56 (0.48–0.65)  
Stage I 32 0.13 (0.05–0.30) 0.09 (0.03–0.27) 0.59 (0.48–0.70)  
 ≤6 months 0.14 (0.02–0.59) 0.14 (0.02–0.59) 0.51 (0.27–0.76)  
 ≤12 months 16 0.19 (0.06–0.45) 0.19 (0.06–0.46) 0.57 (0.41–0.73)  
 >1–2 years 0 (0–0.41) 0 (0–0.41) 0.52 (0.28–0.77)  
 >2–3 years 0.11 (0.02–0.51) 0 (0–0.34) 0.71 (0.52–0.90)  
Stage II 29 0.10 (0.03–0.28) 0.07 (0.02–0.25) 0.68 (0.57–0.79)  
 ≤6 months     
 ≤12 months 0.25 (0.06–0.63) 0.13 (0.02–0.55) 0.72 (0.52–0.92)  
 >1–2 years 10 0.10 (0.01–0.47) 0.10 (0.01–0.48) 0.66 (0.46–0.85)  
 >2–3 years 11 0 (0–0.28) 0 (0–0.28) 0.67 (0.49–0.85)  
Stage III/IV 119 0.31 (0.22–0.42) 0.23 (0.14–0.34) 0.69 (0.64–0.75) 0.34 
 ≤6 months 16 0.88 (0.61–0.97) 0.75 (0.48–0.91) 0.97 (0.93–1.01) 0.007 
 ≤12 months 32 0.72 (0.53–0.85) 0.56 (0.37–0.74) 0.92 (0.87–0.97) 0.01 
 >1–2 years 50 0.22 (0.12–0.37) 0.18 (0.09–0.33) 0.67 (0.58–0.75) 0.07 
 >2–3 years 37 0.08 (0.03–0.23) 0 (0–0.09) 0.53 (0.42–0.63) 0.09 
CA72.4  Cut-off point: 2.46 U/mL Cut-off point: 5.96 U/mL   
Overall 197 0.23 (0.17–0.32) 0.13 (0.08–0.21) 0.61 (0.56–0.65)  
 ≤6 months 27 0.56 (0.36–0.74) 0.37 (0.20–0.58) 0.77 (0.68–0.87)  
 ≤12 months 62 0.45 (0.32–0.59) 0.26 (0.15–0.40) 0.72 (0.65–0.79)  
 >1–2 years 76 0.20 (0.12–0.32) 0.11 (0.05–0.21) 0.61 (0.54–0.68)  
 >2–3 years 59 0.05 (0.02–0.15) 0.02 (0–0.12) 0.53 (0.44–0.61)  
Stage I 32 0.22 (0.10–0.40) 0.06 (0.01–0.23) 0.58 (0.47–0.69)  
 ≤6 months 0.14 (0.02–0.59) 0 (0–0.41) 0.61 (0.37–0.85)  
 ≤12 months 16 0.25 (0.09–0.52) 0.13 (0.03–0.4) 0.61 (0.46–0.77)  
 >1–2 years 0.14 (0.02–0.59) 0 (0–0.41) 0.57 (0.32–0.81)  
 >2–3 years 0.22 (0.05–0.58) 0 (0–0.34) 0.52 (0.30–0.74)  
Stage II 29 0.21 (0.09–0.40) 0.14 (0.05–0.33) 0.59 (0.47–0.71)  
 ≤6 months     
 ≤12 months 0.38 (0.12–0.72) 0.25 (0.06–0.63) 0.69 (0.48–0.90)  
 >1–2 years 10 0.30 (0.10–0.63) 0.20 (0.05–0.55) 0.67 (0.48–0.86)  
 >2–3 years 11 0 (0–0.28) 0 (0–0.28) 0.56 (0.37–0.76)  
Stage III/IV 119 0.24 (0.16–0.34) 0.14 (0.08–0.24) 0.61 (0.56–0.67) 0.40 
 ≤6 months 16 0.69 (0.43–0.87) 0.50 (0.26–0.74) 0.80 (0.68–0.92) 0.29 
 ≤12 months 32 0.56 (0.38–0.73) 0.34 (0.19–0.54) 0.76 (0.67–0.85) 0.80 
 >1–2 years 50 0.18 (0.09–0.32) 0.10 (0.04–0.23) 0.61 (0.52–0.70) 0.35 
 >2–3 years 37 0.03 (0–0.17) 0.03 (0–0.18) 0.52 (0.41–0.62) 0.39 
CA15.3  Cut-off point: 1,372 mIU/mL Cut-off point: 1,610 mIU/mL   
Overall 804 0.07 (0.05–0.09) 0.04 (0.02–0.06) 0.51 (0.49–0.54)  
 ≤6 months 26 0.31 (0.16–0.51) 0.23 (0.11–0.43) 0.73 (0.62–0.84)  
 ≤12 months 61 0.16 (0.09–0.28) 0.13 (0.06–0.25) 0.58 (0.50–0.66)  
 >1–2 years 74 0.14 (0.07–0.24) 0.08 (0.04–0.17) 0.52 (0.45–0.60)  
 >2–3 years 58 0 (0–0.06) 0 (0–0.06) 0.54 (0.45–0.62)  
 >3–6 years 200 0.04 (0.02–0.08) 0.02 (0–0.05) 0.53 (0.48–0.58)  
 >6 years 411 0.06 (0.04–0.09) 0.03 (0.02–0.05) 0.51 (0.48–0.55)  
Stage I 114 0.07 (0.03–0.14) 0.04 (0.02–0.11) 0.52 (0.46–0.59)  
 ≤6 months 0.29 (0.07–0.68) 0 (0–0.41) 0.78 (0.59–0.97)  
 ≤12 months 16 0.19 (0.06–0.45) 0.06 (0.01–0.34) 0.66 (0.51–0.81)  
 >1–2 years 0.17 (0.02–0.63) 0.17 (0.02–0.63) 0.67 (0.42–0.91)  
 >2–3 years 0 (0–0.34) 0 (0–0.34) 0.56 (0.34–0.77)  
 >3–6 years 32 0.03 (0–0.19) 0 (0–0.11) 0.61 (0.49–0.73)  
 >6 years 51 0.06 (0.02–0.17) 0.06 (0.02–0.17) 0.55 (0.46–0.65)  
Stage II 128 0.06 (0.03–0.12) 0.02 (0–0.06) 0.54 (0.49–0.6)  
 ≤6 months     
 ≤12 months 0 (0–0.37) 0 (0–0.37) 0.52 (0.29–0.75)  
 >1–2 years 10 0.30 (0.10–0.63) 0.1 (0.01–0.47) 0.59 (0.39–0.79)  
 >2–3 years 11 0 (0–0.28) 0 (0–0.28) 0.50 (0.30–0.69)  
 >3–6 years 37 0 (0–0.09) 0 (0–0.09) 0.56 (0.44–0.67)  
 >6 years 62 0.08 (0.03–0.18) 0.02 (0–0.11) 0.55 (0.46–0.64)  
Stage III/IV 465 0.06 (0.04–0.09) 0.04 (0.02–0.07) 0.50 (0.46–0.53) 0.37 
 ≤6 months 15 0.27 (0.10–0.54) 0.27 (0.10–0.54) 0.71 (0.57–0.86) 0.51 
 ≤12 months 31 0.16 (0.07–0.34) 0.16 (0.07–0.34) 0.55 (0.43–0.66) 0.64 
 >1–2 years 50 0.10 (0.04–0.22) 0.08 (0.03–0.20) 0.50 (0.41–0.59) 0.16 
 >2–3 years 36 0 (0–0.1) 0 (0–0.1) 0.56 (0.45–0.67) 0.32 
 >3–6 years 106 0.04 (0.01–0.10) 0.02 (0–0.07) 0.50 (0.44–0.57) 0.16 
 >6 years 242 0.06 (0.04–0.10) 0.03 (0.01–0.06) 0.53 (0.48–0.57) 0.22 
# of SetsSensitivity at 95% specificitySensitivity at 98% specificityC-statistic (95% CI)Pheta
CA125  Cut-off point: 56.64 U/mL Cut-off point: 77.92 U/mL   
Overall 807 0.14 (0.11–0.18) 0.10 (0.07–0.13) 0.58 (0.56–0.60)  
 ≤6 months 26 0.81 (0.61–0.92) 0.77 (0.57–0.89) 0.92 (0.86–0.98)  
 ≤12 months 61 0.59 (0.46–0.71) 0.52 (0.39–0.66) 0.82 (0.76–0.88)  
 >1–2 years 75 0.27 (0.18–0.38) 0.20 (0.12–0.31) 0.72 (0.65–0.78)  
 >2–3 years 58 0.10 (0.05–0.22) 0.03 (0.01–0.13) 0.56 (0.48–0.64)  
 >3–6 years 200 0.08 (0.05–0.13) 0.06 (0.03–0.11) 0.55 (0.50–0.60)  
 >6 years 413 0.09 (0.06–0.13) 0.04 (0.02–0.07) 0.54 (0.51–0.57)  
Stage I 115 0.16 (0.1–0.24) 0.10 (0.06–0.18) 0.58 (0.51–0.64)  
 ≤6 months 0.43 (0.14–0.77) 0.29 (0.07–0.68) 0.69 (0.47–0.91)  
 ≤12 months 16 0.38 (0.18–0.63) 0.31 (0.13–0.57) 0.65 (0.49–0.80)  
 >1–2 years 0.14 (0.02–0.58) 0.14 (0.02–0.58) 0.55 (0.31–0.80)  
 >2–3 years 0 (0–0.34) 0 (0–0.34) 0.52 (0.30–0.74)  
 >3–6 years 32 0.13 (0.05–0.29) 0.09 (0.03–0.26) 0.60 (0.48–0.72)  
 >6 years 51 0.14 (0.07–0.26) 0.06 (0.02–0.17) 0.57 (0.47–0.66)  
Stage II 128 0.14 (0.09–0.22) 0.09 (0.05–0.16) 0.61 (0.55–0.67)  
 ≤6 months     
 ≤12 months 0.75 (0.38–0.94) 0.50 (0.20–0.80) 0.91 (0.79–1.02)  
 >1–2 years 10 0.30 (0.10–0.63) 0.30 (0.10–0.63) 0.75 (0.58–0.93)  
 >2–3 years 11 0.09 (0.01–0.44) 0 (0–0.28) 0.66 (0.48–0.84)  
 >3–6 years 37 0.05 (0.01–0.19) 0.05 (0.01–0.20) 0.53 (0.41–0.64)  
 >6 years 62 0.10 (0.04–0.20) 0.05 (0.02–0.14) 0.57 (0.48–0.66)  
Stage III/IV 467 0.14 (0.10–0.18) 0.09 (0.06–0.13) 0.57 (0.54–0.60) 0.26 
 ≤6 months 15 0.93 (0.65–0.99) 0.93 (0.64–0.99) 0.96 (0.92–1.01) 0.09 
 ≤12 months 31 0.68 (0.49–0.82) 0.65 (0.46–0.80) 0.87 (0.80–0.94) 0.045 
 >1–2 years 50 0.26 (0.15–0.40) 0.16 (0.08–0.30) 0.72 (0.64–0.80) 0.77 
 >2–3 years 36 0.14 (0.06–0.30) 0.06 (0.01–0.20) 0.57 (0.46–0.67) 0.20 
 >3–6 years 106 0.07 (0.03–0.13) 0.04 (0.01–0.10) 0.52 (0.45–0.59) 0.68 
 >6 years 244 0.07 (0.05–0.12) 0.04 (0.02–0.07) 0.53 (0.48–0.57) 0.34 
HE4  Cut-off point: 39.41 pmol/L Cut-off point: 54.01 pmol/L   
Overall 197 0.24 (0.17–0.32) 0.18 (0.11–0.27) 0.67 (0.63–0.71)  
 ≤6 months 27 0.67 (0.46–0.82) 0.59 (0.39–0.77) 0.84 (0.76–0.92)  
 ≤12 months 62 0.48 (0.35–0.62) 0.39 (0.25–0.54) 0.79 (0.73–0.85)  
 >1–2 years 76 0.17 (0.10–0.28) 0.14 (0.07–0.26) 0.65 (0.58–0.72)  
 >2–3 years 59 0.07 (0.02–0.17) 0 (0–0.06) 0.56 (0.48–0.65)  
Stage I 32 0.13 (0.05–0.30) 0.09 (0.03–0.27) 0.59 (0.48–0.70)  
 ≤6 months 0.14 (0.02–0.59) 0.14 (0.02–0.59) 0.51 (0.27–0.76)  
 ≤12 months 16 0.19 (0.06–0.45) 0.19 (0.06–0.46) 0.57 (0.41–0.73)  
 >1–2 years 0 (0–0.41) 0 (0–0.41) 0.52 (0.28–0.77)  
 >2–3 years 0.11 (0.02–0.51) 0 (0–0.34) 0.71 (0.52–0.90)  
Stage II 29 0.10 (0.03–0.28) 0.07 (0.02–0.25) 0.68 (0.57–0.79)  
 ≤6 months     
 ≤12 months 0.25 (0.06–0.63) 0.13 (0.02–0.55) 0.72 (0.52–0.92)  
 >1–2 years 10 0.10 (0.01–0.47) 0.10 (0.01–0.48) 0.66 (0.46–0.85)  
 >2–3 years 11 0 (0–0.28) 0 (0–0.28) 0.67 (0.49–0.85)  
Stage III/IV 119 0.31 (0.22–0.42) 0.23 (0.14–0.34) 0.69 (0.64–0.75) 0.34 
 ≤6 months 16 0.88 (0.61–0.97) 0.75 (0.48–0.91) 0.97 (0.93–1.01) 0.007 
 ≤12 months 32 0.72 (0.53–0.85) 0.56 (0.37–0.74) 0.92 (0.87–0.97) 0.01 
 >1–2 years 50 0.22 (0.12–0.37) 0.18 (0.09–0.33) 0.67 (0.58–0.75) 0.07 
 >2–3 years 37 0.08 (0.03–0.23) 0 (0–0.09) 0.53 (0.42–0.63) 0.09 
CA72.4  Cut-off point: 2.46 U/mL Cut-off point: 5.96 U/mL   
Overall 197 0.23 (0.17–0.32) 0.13 (0.08–0.21) 0.61 (0.56–0.65)  
 ≤6 months 27 0.56 (0.36–0.74) 0.37 (0.20–0.58) 0.77 (0.68–0.87)  
 ≤12 months 62 0.45 (0.32–0.59) 0.26 (0.15–0.40) 0.72 (0.65–0.79)  
 >1–2 years 76 0.20 (0.12–0.32) 0.11 (0.05–0.21) 0.61 (0.54–0.68)  
 >2–3 years 59 0.05 (0.02–0.15) 0.02 (0–0.12) 0.53 (0.44–0.61)  
Stage I 32 0.22 (0.10–0.40) 0.06 (0.01–0.23) 0.58 (0.47–0.69)  
 ≤6 months 0.14 (0.02–0.59) 0 (0–0.41) 0.61 (0.37–0.85)  
 ≤12 months 16 0.25 (0.09–0.52) 0.13 (0.03–0.4) 0.61 (0.46–0.77)  
 >1–2 years 0.14 (0.02–0.59) 0 (0–0.41) 0.57 (0.32–0.81)  
 >2–3 years 0.22 (0.05–0.58) 0 (0–0.34) 0.52 (0.30–0.74)  
Stage II 29 0.21 (0.09–0.40) 0.14 (0.05–0.33) 0.59 (0.47–0.71)  
 ≤6 months     
 ≤12 months 0.38 (0.12–0.72) 0.25 (0.06–0.63) 0.69 (0.48–0.90)  
 >1–2 years 10 0.30 (0.10–0.63) 0.20 (0.05–0.55) 0.67 (0.48–0.86)  
 >2–3 years 11 0 (0–0.28) 0 (0–0.28) 0.56 (0.37–0.76)  
Stage III/IV 119 0.24 (0.16–0.34) 0.14 (0.08–0.24) 0.61 (0.56–0.67) 0.40 
 ≤6 months 16 0.69 (0.43–0.87) 0.50 (0.26–0.74) 0.80 (0.68–0.92) 0.29 
 ≤12 months 32 0.56 (0.38–0.73) 0.34 (0.19–0.54) 0.76 (0.67–0.85) 0.80 
 >1–2 years 50 0.18 (0.09–0.32) 0.10 (0.04–0.23) 0.61 (0.52–0.70) 0.35 
 >2–3 years 37 0.03 (0–0.17) 0.03 (0–0.18) 0.52 (0.41–0.62) 0.39 
CA15.3  Cut-off point: 1,372 mIU/mL Cut-off point: 1,610 mIU/mL   
Overall 804 0.07 (0.05–0.09) 0.04 (0.02–0.06) 0.51 (0.49–0.54)  
 ≤6 months 26 0.31 (0.16–0.51) 0.23 (0.11–0.43) 0.73 (0.62–0.84)  
 ≤12 months 61 0.16 (0.09–0.28) 0.13 (0.06–0.25) 0.58 (0.50–0.66)  
 >1–2 years 74 0.14 (0.07–0.24) 0.08 (0.04–0.17) 0.52 (0.45–0.60)  
 >2–3 years 58 0 (0–0.06) 0 (0–0.06) 0.54 (0.45–0.62)  
 >3–6 years 200 0.04 (0.02–0.08) 0.02 (0–0.05) 0.53 (0.48–0.58)  
 >6 years 411 0.06 (0.04–0.09) 0.03 (0.02–0.05) 0.51 (0.48–0.55)  
Stage I 114 0.07 (0.03–0.14) 0.04 (0.02–0.11) 0.52 (0.46–0.59)  
 ≤6 months 0.29 (0.07–0.68) 0 (0–0.41) 0.78 (0.59–0.97)  
 ≤12 months 16 0.19 (0.06–0.45) 0.06 (0.01–0.34) 0.66 (0.51–0.81)  
 >1–2 years 0.17 (0.02–0.63) 0.17 (0.02–0.63) 0.67 (0.42–0.91)  
 >2–3 years 0 (0–0.34) 0 (0–0.34) 0.56 (0.34–0.77)  
 >3–6 years 32 0.03 (0–0.19) 0 (0–0.11) 0.61 (0.49–0.73)  
 >6 years 51 0.06 (0.02–0.17) 0.06 (0.02–0.17) 0.55 (0.46–0.65)  
Stage II 128 0.06 (0.03–0.12) 0.02 (0–0.06) 0.54 (0.49–0.6)  
 ≤6 months     
 ≤12 months 0 (0–0.37) 0 (0–0.37) 0.52 (0.29–0.75)  
 >1–2 years 10 0.30 (0.10–0.63) 0.1 (0.01–0.47) 0.59 (0.39–0.79)  
 >2–3 years 11 0 (0–0.28) 0 (0–0.28) 0.50 (0.30–0.69)  
 >3–6 years 37 0 (0–0.09) 0 (0–0.09) 0.56 (0.44–0.67)  
 >6 years 62 0.08 (0.03–0.18) 0.02 (0–0.11) 0.55 (0.46–0.64)  
Stage III/IV 465 0.06 (0.04–0.09) 0.04 (0.02–0.07) 0.50 (0.46–0.53) 0.37 
 ≤6 months 15 0.27 (0.10–0.54) 0.27 (0.10–0.54) 0.71 (0.57–0.86) 0.51 
 ≤12 months 31 0.16 (0.07–0.34) 0.16 (0.07–0.34) 0.55 (0.43–0.66) 0.64 
 >1–2 years 50 0.10 (0.04–0.22) 0.08 (0.03–0.20) 0.50 (0.41–0.59) 0.16 
 >2–3 years 36 0 (0–0.1) 0 (0–0.1) 0.56 (0.45–0.67) 0.32 
 >3–6 years 106 0.04 (0.01–0.10) 0.02 (0–0.07) 0.50 (0.44–0.57) 0.16 
 >6 years 242 0.06 (0.04–0.10) 0.03 (0.01–0.06) 0.53 (0.48–0.57) 0.22 

aHeterogeneity of discrimination capacity by tumor stage was examined with likelihood-ratio tests comparing the model fit for logistic regression models with and without corresponding interaction terms.

Figure 2.

ROC curves and C-statistics for diagnosis ≤6 months, ≤12 months, 1–2 years, and 2–3 years after blood collection.

Figure 2.

ROC curves and C-statistics for diagnosis ≤6 months, ≤12 months, 1–2 years, and 2–3 years after blood collection.

Close modal

Within the first 12 months after blood donation, for all markers except CA15.3 the ability to predict future cancer diagnosis was clearly stronger for advanced tumors (stage II and III/IV) and relatively weak for stage I tumors (Supplementary Fig. S1), and this heterogeneity was statistically significant for CA125 and HE4 (Phet < 0.05; Table 2). Regarding tumor histology, CA125, HE4, and CA72.4 showed fairly strong discrimination of serous ovarian cancer patients from their matched controls, especially within short lag-times after blood donation (Supplementary Table S2); for the other histologic subtypes, the numbers of patients were too small to obtain reliable estimates.

Among the control subjects, no meaningful correlations between markers were observed (r = −0.15–0.18). Among the case subjects only, and especially among those with lag-times since blood donation below 1 or 2 years, moderately strong correlations were observed between CA125, HE4, and CA72.4 (e.g., within 1-year's lag-time: r = 0.23–0.74), whereas CA15.3 showed somewhat weaker associations with the other markers (within 1-year's lag-time: r = 0.11–0.24; Supplementary Table S3).

In a stepwise forward selection strategy, focusing on variable lag-time strata within the first 3 years after blood donation, the overall model fit for a logistic risk model improved statistically significantly with successive additions of CA125, HE4, CA72.4, and CA15.3 as prediagnostic predictors of future ovarian cancer diagnosis, although the statistical significance for CA15.3 was lowest and largely restricted to lag-times less than 6 months (Table 3). However, the overall improvements in the overall discrimination, assessed by C-statistic (Table 3) or NRI (Supplementary Table S4), were small compared with a model based on any of the markers CA125, HE4, or CA72.4 alone.

Table 3.

Stepwise forward selection of early detection markers in combined prediagnostic models for variable lag-times since blood donationa

Model 1: CA125Model 2: CA125 + HE4Model 3: CA125 + HE4 + CA72.4Model 4: CA125 + HE4 + CA72.4 + CA15.3
# SetsC-Statistic (95% CI)PbC-Statistic (95% CI)PcC-Statistic (95% CI)PdC-Statistic (95% CI)Pe
All women 197 0.70 (0.66–0.74) <0.001 0.71 (0.67–0.75) <0.001 0.71 (0.67–0.75) 0.002 0.71 (0.67–0.75) 0.05 
 Corrected Cf  —  0.71  0.71  0.71  
≤6 months 27 0.92 (0.86–0.98) <0.001 0.89 (0.83–0.96) 0.58 0.90 (0.83–0.96) 0.36 0.92 (0.86–0.97) 0.06 
 Corrected Cf  —  0.89  0.90  0.92  
≤1 year 62 0.82 (0.76–0.88) <0.001 0.82 (0.77–0.88) 0.005 0.84 (0.78–0.89) 0.002 0.84 (0.78–0.89) 0.55 
 Corrected Cf  —  0.81  0.83  0.83  
1–2 years 76 0.72 (0.65–0.78) <0.001 0.73 (0.66–0.79) 0.004 0.73 (0.67–0.80) 0.03 0.73 (0.67–0.80) 0.08 
 Corrected Cf  —  0.72  0.72  0.72  
2–3 years 59 0.56 (0.48–0.64) 0.42 0.58 (0.50–0.67) 0.18 0.58 (0.50–0.66) 0.57 0.58 (0.50–0.66) 0.35 
 Corrected Cf  —  0.53  0.53  0.53  
Model 1: CA125Model 2: CA125 + HE4Model 3: CA125 + HE4 + CA72.4Model 4: CA125 + HE4 + CA72.4 + CA15.3
# SetsC-Statistic (95% CI)PbC-Statistic (95% CI)PcC-Statistic (95% CI)PdC-Statistic (95% CI)Pe
All women 197 0.70 (0.66–0.74) <0.001 0.71 (0.67–0.75) <0.001 0.71 (0.67–0.75) 0.002 0.71 (0.67–0.75) 0.05 
 Corrected Cf  —  0.71  0.71  0.71  
≤6 months 27 0.92 (0.86–0.98) <0.001 0.89 (0.83–0.96) 0.58 0.90 (0.83–0.96) 0.36 0.92 (0.86–0.97) 0.06 
 Corrected Cf  —  0.89  0.90  0.92  
≤1 year 62 0.82 (0.76–0.88) <0.001 0.82 (0.77–0.88) 0.005 0.84 (0.78–0.89) 0.002 0.84 (0.78–0.89) 0.55 
 Corrected Cf  —  0.81  0.83  0.83  
1–2 years 76 0.72 (0.65–0.78) <0.001 0.73 (0.66–0.79) 0.004 0.73 (0.67–0.80) 0.03 0.73 (0.67–0.80) 0.08 
 Corrected Cf  —  0.72  0.72  0.72  
2–3 years 59 0.56 (0.48–0.64) 0.42 0.58 (0.50–0.67) 0.18 0.58 (0.50–0.66) 0.57 0.58 (0.50–0.66) 0.35 
 Corrected Cf  —  0.53  0.53  0.53  

aConditional logistic regression model with log2-transformed markers as continuous variables.

bLikelihood ratio test, comparing Model 1 with empty model.

cLikelihood ratio test, comparing Model 2 with Model 1.

dLikelihood ratio test, comparing Model 3 with Model 2.

eLikelihood ratio test, comparing Model 4 with Model 3.

fC-Statistic in multivariate models corrected for over optimism obtained by bootstrap sampling.

Finally, we examined whether the overall discrimination between case patients and control subjects could be improved by combining the biomarkers with an epidemiologic stratification algorithm that was developed previously using the full EPIC cohort data (16). Because some of the key epidemiologic risk variables included in the algorithm (age, menopausal status, use of OCs or HRT), overlapped with some of the matching factors for the present nested case–control study, the risk model showed a lower discrimination (C = 0.56) in our case–control set as compared with our previous full cohort analysis (corrected for over-optimism, C = 0.64). For lag-times below 2 years, combining the risk model with the biomarkers did not improve overall discrimination as compared with each of the biomarkers alone (results not shown). In contrast, for lag times greater than 3 years, the longer-term prediction of future ovarian cancer diagnosis was moderately but significantly improved when CA125 was added to the model (C = 0.57 vs. C = 0.55), whereas adding CA15.3 showed no improvement.

In our evaluation of four potential ovarian cancer screening biomarkers measured in prospectively collected samples from women with ovarian cancer and matched controls in the EPIC cohort, we observed the best sensitivity and specificity for CA125, followed by HE4, CA72.4, and finally CA15.3. The ability of these biomarkers to distinguish cases from controls declined with increasing time between blood draw and diagnosis, as well as with earlier stage at diagnosis. These observations suggest that, generally, these markers are best at identifying advanced disease close to diagnosis, but their ability to detect early disease that is amenable to interventions that can improve survival may be limited. Addition of a previously established risk prediction model did not improve the performance of markers in women who went on to develop clinically manifest ovarian cancer less than three years in advance of diagnosis. In contrast, adding CA125 (but not CA15.3) to the risk prediction model did slightly improve the longer-term prediction of ovarian cancer occurrence over a time interval of about 3 to 6 years after blood donation.

Our results are consistent with those from large randomized trials and prospective assessments of these markers in other populations. All four of these biomarkers were included in an ancillary study of 49 biomarkers previously evaluated in phase II studies, and all four were also among the best 35 that were subsequently examined within the prospective PLCO cohort (18). Results from the PLCO study were similar to what we observed in EPIC, with the best performance in cases diagnosed 6 months or less after blood draw (depending on the time between blood draw and diagnosis), C-statistics ranged from 0.83 to 0.96 for CA125, 0.78 to 0.88 for HE4, 0.80 for CA72.4, and 0.72 for CA15.3. As in our current study, the discriminatory ability of these markers in the PLCO cohort declined rapidly for samples collected more than 6 months prior to diagnosis. An important difference between PLCO and our study, however, is that PLCO participants had been annually screened for ovarian cancer by CA125 plus ultrasonography, thereby reducing the occurrence of further ovarian cancer diagnoses over time periods longer than 12 months. In another prospective study, Anderson and colleagues measured CA125 and HE4 among other markers in a total of 34 incident cases of ovarian cancer who had provided serial blood samples up to 18 years prior to diagnosis and noted a similar decline in marker performance over time with an AUC at <2 years of 0.74 and 0.71 and from 2 to 4 years of 0.68 and 0.67 for CA125 and HE4, respectively (10).

The ultimate goal of screening is to identify cancer at a stage at which medical intervention has the highest chances of providing a cure or prolonging survival. None of the four markers that we tested showed a clear capacity for predicting disease that was diagnosed in stage I, even for follow-up times of less than 6 months. This finding, however, may have been confounded by tumor histology, as among patients diagnosed with stage I disease, there was an over-representation of mucinous and clear cell tumors, which are generally slowly growing tumor subtypes and diagnosed at earlier stages. In contrast, within less than 12 months of prospective follow-up especially, CA125 and HE4 showed substantial discriminatory capacity for tumors that had been subsequently diagnosed at a more advanced stage, and the discrimination of these markers was also significantly stronger for the more aggressive type II tumors as compared with type I tumors. A major limitation of prospective studies such as EPIC and PLCO is that they do not inform about a patient's tumor stage at the time of blood donation; hence, no direct information is available whether those women whose tumor might have been diagnosed 6–12 months before clinical diagnosis through CA125 or HE4 screening would have actually shown a sufficient shift towards an earlier tumor stage to allow speculations about significant improvements in survival.

Despite having the best performance among various candidate markers considered for ovarian cancer screening in various studies (10, 18, 19), annual CA125 measurement (combined with transvaginal ultrasound) in the PLCO randomized trial showed no mortality benefit (7). This lack of benefit is likely related to insufficient sensitivity of a single CA125 measurement for detecting the more aggressive forms of ovarian cancer in an early stage of disease, as suggested by the findings in both PLCO and our study. However, several studies have shown that the use of serial measurements over time can improve the diagnostic performance for ovarian cancer detection (20–23). In the world's largest ongoing, randomized screening trial for ovarian cancer, the United Kingdom Collaborative Trial on Ovarian Cancer Screening (UKCTOCS), consideration of change in CA125 over time using the ROCA algorithm improved marker performance from C-statistic = 0.87 for a single CA125 to C = 0.92 and doubled the number of screen-detected invasive epithelial ovarian cancers compared with CA125 screening with a fixed cutoff (22). Furthermore, recent mortality results from the UKCTOCs revealed a 15% reduction in mortality for women screened using the ROCA algorithm among incident cases (P = 0.02; ref. 8).

Results from various phase II and other clinical studies have suggested that combinations of multiple biomarkers may be better at distinguishing malignant from benign tumors than CA125 alone. For example, improved discrimination has been documented for the combination of CA125 and HE4, the two strongest discriminating biomarkers in our analyses, as compared either marker alone (12, 24–31). While our analyses confirm that biomarker combinations improve prediction of future ovarian cancer diagnosis, the absolute gain in classification appeared to be small in our data, and a similar observation was made in the PLCO cohort (19). Thus, the addition of biomarkers can improve the discriminatory ability of CA125 but current biomarkers may not improve performance to the degree required for population screening.

Analyses in the PLCO study as well as ours, show substantial discrepancy between the often promising findings from phase II discovery studies based on clinical case–control comparisons and their lack of replication in prospective evaluations based on prediagnostic blood samples. This observation has triggered recommendations that greater care should be taken in selecting the appropriate sample set for screening biomarker discovery. In particular, it was recommended that prospective cohort studies should be used for new biomarker discovery rather than simply validation of known candidate biomarkers (32–34). One advantage of such an approach would be that it ensures rigorous internal validity for the evaluation of systematic differences between case and control subjects. Another possible advantage of the prospective design is that by focusing on blood samples collected months prior to cancer diagnosis, one would avoid a bias toward markers exclusively associated with advanced disease (32). While attractive from a methodologic perspective, however, the use of prospective cohorts for biomarker discovery may have several limitations in practice. In our study, among 366,521 women mostly aged 35–70 years at blood donation, there was an annual incidence of about 35 ovarian cancer cases. Thus, assuming an early detection time window of 6–18 months prior to diagnosis (excluding the first 6 months of follow-up to reduce the presence of advanced disease), studies for marker discovery would be based on a very limited, yet etiologically diverse sample set. This basic observation illustrates that even very large prospective cohorts may not have a sufficient number of cases for biomarker discovery studies focusing on early-stage disease. Moreover, as already noted, the tumor grade and stage at the time of blood sampling would remain unknown. Therefore, uncertainty will remain as to whether those patients whose tumor would have had elevated biomarkers 6–18 months prior to diagnosis (and hence potentially detected) would actually benefit from detection at that timepoint. In light of these limitations, we believe that, as a complement to prospective cohort studies, bio-banking initiatives in large clinical networks will remain needed for the collection of samples especially from well-characterized early-stage patients to allow large-scale comparisons with samples from cancer-free individuals.

In summary, CA125 and HE4 continue to hold potential for ovarian cancer screening but lack sensitivity and specificity needed to detect early-stage disease. New biorepositories of early-stage disease and matched controls are needed to identify novel markers that focus on the disease timepoint where intervention can make the biggest improvement in mortality and morbidity.

No potential conflicts of interest were disclosed.

Conception and design: K.L. Terry, A. Hüsing, T. Johnson, K. Overvad, A. Tjønneland, G. Severi, H. Boeing, H.B. Bueno-de-Mesquita, E. Weiderpass, M.-J. Sanchez, E. Ardanaz, C. Navarro, D.W. Cramer, R. Kaaks

Development of methodology: K.L. Terry, T. Johnson, P.H. Peeters, E. Weiderpass, E. Ardanaz

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.S. Yamamoto, K. Overvad, A. Tjønneland, M.-C. Boutron-Ruault, G. Severi, H. Boeing, V. Benetou, P. Lagiou, A. Trichopoulou, V. Krogh, S. Panico, H.B. Bueno-de-Mesquita, P.H. Peeters, E. Weiderpass, M.-J. Sanchez, E. Ardanaz, N. Etxezarreta, C. Navarro, A. Idahl, E. Lundin, K. Jirström, J. Manjer, N.J. Wareham, K.-T. Khaw, R.C. Travis, M.J. Gunter, R. Kaaks

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.L. Terry, H. Schock, R.T. Fortner, A. Hüsing, H.S. Yamamoto, A.F. Vitonis, V. Benetou, P.H. Peeters, I.T. Gram, E. Weiderpass, E. Ardanaz, R. Kaaks

Writing, review, and/or revision of the manuscript: K.L. Terry, H. Schock, R.T. Fortner, A. Hüsing, H.S. Yamamoto, K. Overvad, A. Tjønneland, M.-C. Boutron-Ruault, S. Mesrine, G. Severi, L. Dossus, S. Rinaldi, H. Boeing, V. Benetou, P. Lagiou, A. Trichopoulou, V. Krogh, E. Kuhn, S. Panico, H.B. Bueno-de-Mesquita, N.C. Onland-Moret, P.H. Peeters, I.T. Gram, E. Weiderpass, E.J. Duell, M.-J. Sanchez, E. Ardanaz, N. Etxezarreta, C. Navarro, A. Idahl, E. Lundin, J. Manjer, N.J. Wareham, K.-T. Khaw, K.S. Byrne, R.C. Travis, M.J. Gunter, M.A. Merritt, E. Riboli, D.W. Cramer, R. Kaaks

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Schock, H.S. Yamamoto, A.F. Vitonis, T. Johnson, K. Overvad, H.B. Bueno-de-Mesquita, P.H. Peeters, E. Weiderpass, E. Ardanaz, E. Lundin, J. Manjer, K.-T. Khaw, R. Kaaks

Study supervision: E. Weiderpass, E. Ardanaz, R. Kaaks

Other (Development and performance of laboratory assays): R.N. Fichorova

Other (Principal Investigator on the grant, co-senior author): D.W. Cramer

The authors thank Fujirebio Diagnostics for supplying the reagents for the HE4 and CA72.4 assays.

This work was supported by the NIH (R01CA158119). The coordination of the European Prospective Investigation into Cancer and Nutrition is financially supported by the European Commission (DG-SANCO) and the International Agency for Research on Cancer. The national cohorts are supported by Danish Cancer Society (Denmark); Ligue Contre le Cancer, Institut Gustave Roussy, Mutuelle Générale de l'Education Nationale, Institut National de la Santé et de la Recherche Médicale (INSERM; France); Deutsche Krebshilfe, Deutsches Krebsforschungszentrum and Federal Ministry of Education and Research (Germany); Hellenic Health Foundation (Greece) and the Stavros Niarchos Foundation; Italian Association for Research on Cancer (AIRC) and National Research Council (Italy); Dutch Ministry of Public Health, Welfare, and Sports (VWS), Netherlands Cancer Registry (NKR), LK Research Funds, Dutch Prevention Funds, Dutch ZON (Zorg Onderzoek Nederland), World Cancer Research Fund (WCRF), Statistics Netherlands (the Netherlands); ERC-2009-AdG 232997 and Nordforsk, Nordic Centre of Excellence programme on Food, Nutrition, and Health (Norway); Health Research Fund (FIS), Regional Governments of Andalucía, Asturias, Basque Country, Murcia (no. 6236) and Navarra, ISCIII RETIC (RD06/0020; Spain); Swedish Cancer Society, Swedish Scientific Council, and Regional Government of Skåne and Västerbotten (Sweden); Cancer Research UK (grant no. 14136, to K.-T. Khaw; grant no. C570/A16491, to R.C. Travis), Medical Research Council (United Kingdom; grant no. G1000143).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
American Cancer Society
.
Cancer Facts & Figures.
Atlanta, GA
:
American Cancer Society
; 
2015
.
2.
Terry
KL
,
Sluss
PM
,
Skates
SJ
,
Mok
SC
,
Ye
B
,
Vitonis
AF
, et al
Blood and urine markers for ovarian cancer: a comprehensive review
.
Dis Markers
2004
;
20
:
53
70
.
3.
Pepe
MS
,
Etzioni
R
,
Feng
Z
,
Potter
JD
,
Thompson
ML
,
Thornquist
M
, et al
Phases of biomarker development for early detection of cancer
.
J Natl Cancer Inst
2001
;
93
:
1054
61
.
4.
U.S. Food and Drug Administration
. 
510(K) Substantial equivalence determination decision summary
.
Available from
: http://www.accessdata.fda.gov/cdrh_docs/reviews/K081754.pdf.
6.
U.S. Food and Drug Administration
. 
510 (k) Substantial equivalence determination decision summary
.
Available from
: http://www.accessdata.fda.gov/cdrh_docs/reviews/K042731.pdf.
7.
Buys
SS
,
Partridge
E
,
Black
A
,
Johnson
CC
,
Lamerato
L
,
Isaacs
C
, et al
Effect of screening on ovarian cancer mortality: the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Randomized Controlled Trial
.
JAMA
2011
;
305
:
2295
303
.
8.
Jacobs
IJ
,
Menon
U
,
Ryan
A
,
Gentry-Maharaj
A
,
Burnell
M
,
Kalsi
JK
, et al
Ovarian cancer screening and mortality
in the UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS): a randomised controlled trial
.
Lancet
2016
;
387
:
945
56
.
9.
Kobayashi
H
,
Yamada
Y
,
Sado
T
,
Sakata
M
,
Yoshida
S
,
Kawaguchi
R
, et al
A randomized study of screening for ovarian cancer: a multicenter study in Japan
.
Int J Gynecol Cancer
2008
;
18
:
414
20
.
10.
Anderson
GL
,
McIntosh
M
,
Wu
L
,
Barnett
M
,
Goodman
G
,
Thorpe
JD
, et al
Assessing lead time of selected ovarian cancer biomarkers: a nested case-control study
.
J Natl Cancer Inst
2010
;
102
:
26
38
.
11.
Urban
N
,
Hawley
S
,
Janes
H
,
Karlan
BY
,
Berg
CD
,
Drescher
CW
, et al
Identifying post-menopausal women at elevated risk for epithelial ovarian cancer
.
Gynecol Oncol
2015
;
139
:
253
60
.
12.
Urban
N
,
Thorpe
JD
,
Bergan
LA
,
Forrest
RM
,
Kampani
AV
,
Scholler
N
, et al
Potential role of HE4 in multimodal screening for epithelial ovarian cancer
.
J Natl Cancer Inst
2011
;
103
:
1630
4
.
13.
Bingham
S
,
Riboli
E
. 
Diet and cancer–the European Prospective Investigation into Cancer and Nutrition
.
Nat Rev Cancer
2004
;
4
:
206
15
.
14.
Riboli
E
,
Hunt
KJ
,
Slimani
N
,
Ferrari
P
,
Norat
T
,
Fahey
M
, et al
European Prospective Investigation into Cancer and Nutrition (EPIC): study populations and data collection
.
Public Health Nutr
2002
;
5
:
1113
24
.
15.
Hainaut
P
,
Vozar
B
,
Rinaldi
S
,
Riboli
E
,
Caboux
E
. 
The European Prospective Investigation into Cancer and Nutrition biobank
.
Methods Mol Biol
2011
;
675
:
179
91
.
16.
Li
K
,
Husing
A
,
Fortner
RT
,
Tjønneland
A
,
Hansen
L
,
Dossus
L
, et al
An epidemiologic risk prediction model for ovarian cancer in Europe: the EPIC study
.
Br J Cancer
2015
;
112
:
1257
65
.
17.
Pencina
MJ
,
D'Agostino
RB
 Sr
,
Steyerberg
EW
. 
Extensions of net reclassification improvement calculations to measure usefulness of new biomarkers
.
Stat Med
2011
;
30
:
11
21
.
18.
Cramer
DW
,
Bast
RC
 Jr
,
Berg
CD
,
Diamandis
EP
,
Godwin
AK
,
Hartge
P
, et al
Ovarian cancer biomarker performance in prostate, lung, colorectal, and ovarian cancer screening trial specimens
.
Cancer Prev Res (Phila)
2011
;
4
:
365
74
.
19.
Zhu
CS
,
Pinsky
PF
,
Cramer
DW
,
Ransohoff
DF
,
Hartge
P
,
Pfeiffer
RM
, et al
A framework for evaluating biomarkers for early detection: validation of biomarker panels for ovarian cancer
.
Cancer Prev Res (Phila)
2011
;
4
:
375
83
.
20.
Drescher
CW
,
Shah
C
,
Thorpe
J
,
O'Briant
K
,
Anderson
GL
,
Berg
CD
, et al
Longitudinal screening algorithm that incorporates change over time in CA125 levels identifies ovarian cancer earlier than a single-threshold rule
.
J Clin Oncol
2013
;
31
:
387
92
.
21.
Lu
KH
,
Skates
S
,
Hernandez
MA
,
Bedi
D
,
Bevers
T
,
Leeds
L
, et al
A 2-stage ovarian cancer screening strategy using the Risk of Ovarian Cancer Algorithm (ROCA) identifies early-stage incident cancers and demonstrates high positive predictive value
.
Cancer
2013
;
119
:
3454
61
.
22.
Menon
U
,
Ryan
A
,
Kalsi
J
,
Gentry-Maharaj
A
,
Dawnay
A
,
Habib
M
, et al
Risk algorithm using serial biomarker measurements doubles the number of screen-detected cancers compared with a single-threshold rule in the United Kingdom Collaborative Trial of Ovarian Cancer Screening
.
J Clin Oncol
2015
;
33
:
2062
71
.
23.
Skates
SJ
,
Menon
U
,
MacDonald
N
,
Rosenthal
AN
,
Oram
DH
,
Knapp
RC
, et al
Calculation of the risk of ovarian cancer from serial CA-125 values for preclinical detection in postmenopausal women
.
J Clin Oncol
2003
;
21
:
206s
10s
.
24.
Azzam
AZ
,
Hashad
DI
,
Kamel
NA
. 
Evaluation of HE4 as an extrabiomarker to CA125 to improve detection of ovarian carcinoma: is it time for a step forward?
Arch Gynecol Obstet
2013
;
288
:
167
72
.
25.
Ghasemi
N
,
Ghobadzadeh
S
,
Zahraei
M
,
Mohammadpour
H
,
Bahrami
S
,
Ganje
MB
, et al
HE4 combined with CA125: favorable screening tool for ovarian cancer
.
Med Oncol
2014
;
31
:
808
.
26.
Karlsen
MA
,
Hogdall
EV
,
Christensen
IJ
,
Borgfeldt
C
,
Kalapotharakos
G
,
Zdrazilova-Dubska
L
, et al
A novel diagnostic index combining HE4, CA125 and age may improve triage of women with suspected ovarian cancer - An international multicenter study in women with an ovarian mass
.
Gynecol Oncol
2015
;
138
:
640
6
.
27.
Karlsen
MA
,
Sandhu
N
,
Hogdall
C
,
Christensen
IJ
,
Nedergaard
L
,
Lundvall
L
, et al
Evaluation of HE4, CA125, risk of ovarian malignancy algorithm (ROMA) and risk of malignancy index (RMI) as diagnostic tools of epithelial ovarian cancer in patients with a pelvic mass
.
Gynecol Oncol
2012
;
127
:
379
83
.
28.
Kristjansdottir
B
,
Levan
K
,
Partheen
K
,
Sundfeldt
K
. 
Diagnostic performance of the biomarkers HE4 and CA125 in type I and type II epithelial ovarian cancer
.
Gynecol Oncol
2013
;
131
:
52
8
.
29.
Moore
RG
,
Brown
AK
,
Miller
MC
,
Skates
S
,
Allard
WJ
,
Verch
T
, et al
The use of multiple novel tumor biomarkers for the detection of ovarian carcinoma in patients with a pelvic mass
.
Gynecol Oncol
2008
;
108
:
402
8
.
30.
Van Gorp
T
,
Cadron
I
,
Despierre
E
,
Daemen
A
,
Leunen
K
,
Amant
F
, et al
HE4 and CA125 as a diagnostic test in ovarian cancer: prospective validation of the Risk of Ovarian Malignancy Algorithm
.
Br J Cancer
2011
;
104
:
863
70
.
31.
Yurkovetsky
Z
,
Skates
S
,
Lomakin
A
,
Nolen
B
,
Pulsipher
T
,
Modugno
F
, et al
Development of a multimarker assay for early detection of ovarian cancer
.
J Clin Oncol
2010
;
28
:
2159
66
.
32.
Jacobs
I
,
Menon
U
. 
The sine qua non of discovering novel biomarkers for early detection of ovarian cancer: carefully selected preclinical samples
.
Cancer Prev Res
2011
;
4
:
299
302
.
33.
Pepe
MS
,
Li
CI
,
Feng
Z
. 
Improving the quality of biomarker discovery research: the right samples and enough of them
.
Cancer Epidemiol Biomarkers Prev
2015
;
24
:
944
50
.
34.
Schully
SD
,
Carrick
DM
,
Mechanic
LE
,
Srivastava
S
,
Anderson
GL
,
Baron
JA
, et al
Leveraging biospecimen resources for discovery or validation of markers for early cancer detection
.
J Natl Cancer Inst
2015
;
107
:
djv012
.