• ovarian cancer symptoms;
  • ovarian cancer detection;
  • Surveillance;
  • Epidemiology and End Results;
  • Medicare


  1. Top of page
  2. Abstract


Patients with ovarian cancer often report having symptoms for months before diagnosis, but such findings are subject to recall bias. The aim of this study was to provide an objective evaluation of symptoms that precede a diagnosis of ovarian cancer.


Medicare provider claims linked to records in the California Surveillance, Epidemiology, and End Results data base were utilized to extract diagnosis and procedure codes for 1985 women age 68 years or older who resided in California with ovarian cancer, 6024 elderly women with localized breast cancer, and 10,941 age-matched, Medicare-enrolled women without cancer. Prevalence of rates of symptom-related diagnoses and procedure codes in Medicare claims records were obtained during 3-month periods up to 36 months before diagnosis of ovarian cancer.


From 1 month to 3 months before patients were diagnosed with ovarian cancer, the frequency and adjusted odds ratios (ORs) with 95% confidence intervals (95%CIs) for 4 “target symptom” code groups were: abdominal pain (frequency, 30.6%; OR, 6.0; 95%CI, 5.1–6.9), abdominal swelling (frequency, 16.5%; OR, 30.9; 95%CI, 21.4–44.8), gastrointestinal symptoms (frequency, 8.4%; OR, 2.3; 95%CI, 1.8–3.0), and pelvic pain (frequency, 5.4%; OR, 4.3; 95%CI, 2.8–6.7). The adjusted odds for abdominal swelling codes was elevated 10–12 months before diagnosis (OR, 2.4; 95%CI, 1.2–4.6) for abdominal pain codes 7–9 months before diagnosis (OR, 1.3; 95%CI, 1.1–1.7). Abdominal imaging (frequency, 7.0%; OR, 1.3; 95%CI, 1.0–1.7) and pelvic imaging/CA125 (frequency, 3.7%; OR, 2.4; 95%CI, 1.7–3.4) showed an elevated frequency and adjusted odds 4–6 months before diagnosis. Patients with claims codes for “target symptoms” 4–36 months before diagnosis were more likely to have abdominal imaging (61.1%) or gastrointestinal procedures (30.8%) than pelvic imaging/CA125 (25.3%).


Patients with ovarian cancer were more likely than patients with breast cancer and women in a cancer-free control group to have target symptom codes (particularly abdominal swelling and pain) > 6 months before diagnosis. The evaluation of women with unexplained “target symptoms” should include pelvic imaging and/or CA125. Cancer 2005. © 2005 American Cancer Society.

For most of the 20th century, ovarian cancer was depicted as an insidious, silent disease that offered few clues to early diagnosis.1 A view prevailed that ovarian malignancies often grew painlessly,2 frequently making it impossible to recognize the primary growth.3 This mind set was challenged by Ranney and Ahmad in 1979,4 who reviewed office records and determined that only 1 in 44 patients with ovarian cancer was asymptomatic prior to diagnosis and that an average of 13 months passed between the appearance of symptoms and surgical treatment. Later studies based on either retrospective review of medical records5 or patient questionnaires6, 7 suggested that most patients experienced symptoms, often of a nongynecologic nature, for months prior to diagnosis. Patients with ovarian cancer often believe that their symptoms began well before their diagnosis, even with localized disease.8–11 However, recall bias may affect questionnaire studies, resulting in the overestimation of symptom severity.12 Goff and colleagues13 addressed this concern by administering symptom questionnaires prospectively to women who were about to undergo surgery for ovarian masses and compared the results with women who visited primary care clinics. Despite these studies, questions remain regarding the timing of appearance of early symptoms of ovarian cancer. The duration of symptoms before the correct diagnosis is established may be crucial to the prognosis. Most women with localized ovarian cancer are cured, whereas most women with advanced ovarian cancer die of their disease. Women with advanced disease who have lesser amounts of residual disease after debulking surgery (“optimal cytoreduction”) have better survival than women with larger amounts of residual disease.14 Thus earlier clinical diagnosis could shift patients into a more favorable prognostic group.

The objective of this study was to evaluate the pattern of symptoms and the associated diagnostic tests documented in women with ovarian cancer over 36 months prior to the date of diagnosis. Linked case reports from the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Program and from Medicare Part B claims for 1985 California women who developed ovarian cancer were compared with the records from 2 control groups: women with early-stage breast cancer and a randomly selected group of age-matched women who were cancer-free. We address the following questions: 1) How long before ovarian cancer diagnosis did claims codes for symptoms in the ovarian cancer group exceed those in the control groups? 2) In patients who had symptom codes in their claims records, what diagnostic tests were used for evaluation? 3) Is there evidence, based on claims records, that an earlier clinical diagnosis could be made in many patients with ovarian cancer?


  1. Top of page
  2. Abstract

Study Design

This was a population-based, retrospective case–control study in which we used the SEER-Medicare linked data base15 to examine the occurrence of symptoms and medical services during the 36 months preceding the diagnosis of ovarian cancer.

Source of Data

The data for this study were obtained from the merged files of the SEER Program and the Centers for Medicare and Medicaid Services (CMS) Medicare Master Enrollment file. A detailed description is available elsewhere ( California's entire population is part of the SEER Program, which is a population-based system of cancer registries that collects data on incident cancers among individuals residing in SEER reporting areas at the time of their diagnosis. Data reported to SEER include the date of cancer diagnosis, tumor stage and grade, demographic characteristics, whether the patient has a prior history of cancer, treatment, and follow-up information. The Medicare Master Enrollment File contains information on Part A and Part B entitlement, health maintenance organization enrollment status by month, and demographic characteristics for each Medicare beneficiary. A unique identifier in the SEER-Medicare file allowed for patients to be linked with National Claims History records. These noninstitutional claims are largely from physicians, who bill for services provided in the office, hospital, or other sites. Each billed procedure is identified by a Common Procedure Terminology (CPT) code and must be accompanied by at least 1 diagnosis code (International Classification of Diseases, 9th revision, Clinical Modification [ICD-9-CM]) that describes the reason for the service or visit, all of which were included in the current analysis.

Study Population

All women who were included in the case group were California residents who were age 68 years or older, were enrolled in fee-for-service Medicare through 2002, and were diagnosed with ovarian cancer (American Joint Committee on Cancer [AJCC] Stage IC or higher) from 1994 through 1999. Two control groups were utilized. The first included women who were diagnosed with early-stage breast cancer (AJCC in situ or Stage I tumors) during the same period. The second control group included three randomly selected, Medicare-enrolled, cancer-free patients who were age-matched to each ovarian cancer case by year of birth. Both patients with ovarian cancer and patients with breast cancer were identified through the SEER-Medicare Patient Entitlement and Diagnosis Summary File, whereas noncancer population controls were identified through the Summarized Denominator file15 (drawn from a random 5% sample of Medicare beneficiaries residing in the SEER areas). Because individuals become eligible for Medicare coverage at age 65 years, selecting women with a minimum age of 68 years at diagnosis ensured that all women in the study were eligible for Medicare coverage during the 36 months prior to their diagnosis of breast or ovarian cancer (i.e., the index date). For women in the noncancer control group, this index date was defined as the date of diagnosis of the corresponding woman in the ovarian cancer case group to whom the woman was matched. We excluded patients who were not entitled to both Medicare Part A and B or who had not been enrolled continuously in the fee-for-service program during the 36 months prior to the index date. Case and control patients who were under managed care plans were excluded, because claims from these programs are submitted unevenly to CMS. In addition to the above criteria, patients with ovarian and breast cancer were excluded if this tumor was not their first primary tumor or if the sole reporting source for that tumor was an autopsy or death certificate.

Definition of Study Variables

Based on symptoms that were reported commonly by patients with ovarian cancer in previously published surveys,4–11, 13 we chose groups of diagnostic codes from the ICD-9-CM index that were likely to reflect the presence of relevant symptoms. These diagnostic codes (see Table 1) were grouped into categories as follows: abdominal pain, gastrointestinal symptoms, pelvic pain, abdominal swelling, malaise/fatigue, and urinary symptoms. The ICD-9-CM code for ovarian cancer (183.0) also was included in the analysis to identify when a diagnosis of ovarian cancer was recorded first on a Medicare claim. Medical tests and procedures (CPT codes) were classified into 5 categories: 1) pelvic imaging/CA125, including pelvic ultrasound, computed tomography, magnetic resonance imaging, or serum CA125; 2) abdominal imaging, including plain X-rays, contrast studies, ultrasound, computed tomography, and magnetic resonance imaging; 3) gastrointestinal procedures, including endoscopy of the stomach, small intestine, and colon with or without biopsy; 4) hematology/chemistry, including complete blood count, metabolic panel, electrolyte levels, and hemostasis tests; and 5) urine tests, including urinalysis. Office visit (evaluation and management) codes are defined in Table 1. Ovarian cancers were AJCC Stage IC or higher (with Stage IC tumors defined as those with capsular rupture, tumor on the ovarian surface, or malignant cells in ascites or peritoneal washings). Early-stage breast cancers were defined as AJCC Stage 0 or I (in situ tumors or invasive tumors measuring ≤ 2 cm in greatest dimension without lymph node involvement). Women were classified racially as white or nonwhite. Education was imputed based on the percent of women age 25 years or older with a high school diploma in the patient's zip code of residence (> 25% vs. ≤ 25%). The urban character of the patient's residence area, which was categorized as metropolitan/urban or less urban/rural, was based on Health Service Area (2002 Area Resource File) definitions.

Table 1. Diagnostic and Procedure Codes
Abdominal pain789.0–789.09
 Right upper quadrant789.01
 Left upper quadrant789.02
 Right lower quadrant789.03
 Left lower quadrant789.04
 Other specified sites789.09
Gastrointestinal symptoms536.0–787.91
 Flatulence, eructation, and gas pain787.3
 Change in bowel habits787.9
 Constipation (atonic, simple, spastic)564.0
 Irritable bowel536.9
Pelvic pain625.0, 625.5, 625.9
 Pelvic congestion syndrome625.5
 Unspecified symptoms associated with female genital organs625.9
Abdominal swelling789.3–789.39
 Abdominal or pelvic swelling mass or lump789.3
 Abdominal or pelvic swelling mass or lump, unspecified site789.30
 Abdominal or pelvic swelling mass or lump, right upper quadrant789.31
 Abdominal or pelvic swelling mass or lump, left upper quadrant789.32
 Abdominal or pelvic swelling mass or lump, right lower quadrant789.33
 Abdominal or pelvic swelling mass or lump, left lower quadrant789.34
 Abdominal or pelvic swelling mass, periumbilic789.35
 Abdominal or pelvic swelling mass, epigastric789.36
 Abdominal or pelvic swelling mass, other specified site789.39
Malaise/fatigue780.7, 783.2
 Malaise and fatigue780.7
 Abnormal loss of weight and underweight783.2
Urinary symptoms788.3, 788.1, 788.4
 Urinary incontinence788.3
 Frequency of urination and polyuria788.4
Pelvic imaging/CA125 (including pelvic ultrasound, computed tomography, magnetic resonance imaging, or serum CA125)76830, 76856, 76857, 72192–72198, 86304
Abdominal imaging (including plain X-rays with or without contrast studies, ultrasound, computed tomography with and without contrast, and magnetic resonance imaging)74000, 74010, 74020, 74022, 74150, 74160, 74170, 74181, 74240, 74241, 74245, 74270, 76700, 76705, 76770, 76775
Gastrointestinal procedures (including endoscopy of stomach, small intestine, and colon with or without biopsy)43200, 43202, 44360, 44388, 44389, 45300, 45305, 45308, 45309, 45315, 45330, 45331, 45378, 45380, 45383, 45384, 45385
Hematology/chemistry (including complete blood count, metabolic panel, electrolyte levels, and hemostasis tests)80049–80051, 80054, 80058, 80061, 80072, 82374, 82435, 82565, 82947, 84132, 84295, 84443, 84520, 85007–85009, 85013, 85014, 85018, 85021, 85022–85025, 85027, 85031, 85041, 85044, 85045, 85048, 85585, 85590, 85595, 85651
Urine tests (including urinalysis)81000–81003, 81005, 81007, 81015, 81020, 81099
Office visit (evaluation and management)99201–99205, 99211–99215, 99242–99245, 99261–99263, 99271–99273, 99275, 99279

Statistical Analysis

Frequencies, median values, and ranges were obtained for numeric variables such as age at diagnosis, the number of office visits or services provided, and the number of symptom episodes per patient and per period. Logistic regression was used to model the likelihood of symptoms (or medical procedures) among the patients with ovarian cancer compared with the early-stage breast cancer control group and the age-matched, cancer-free control group (in which case-conditional logistic regression models were used). The likelihood of experiencing symptoms or of receiving medical procedures was measured by odds ratios (ORs) with 95% confidence intervals (95%CIs). To calculate these ORs, we defined patients with at least one Medicare claim during each time segment as the group at risk (patients without Medicare claims were excluded from the analysis of that time segment). To estimate the likelihood of symptoms or procedures at different points in time, logistic regression models were run separately for each 3-month time segment, from 1 month to 36 months prior to the definitive diagnosis (or index date for cancer-free controls). Beyond 18 months, we aggregated time segments into 6-month periods in which no significant differences in ORs could be detected between the corresponding 3-month periods. Univariate logistic models were used initially to obtain the unadjusted OR by time segment for each symptom group or medical procedure described above as well as for race, age at diagnosis (68–77 yrs vs. 78 yrs and older), education (at the zip code level), and urban residence. Using the univariate results as a guide, we constructed multivariate models to obtain the OR of having claims with symptom codes (or receiving medical procedures) within each prediagnosis time segment, adjusting for all other study variables. The SAS System (release 9.1; SAS Institute Inc., Cary, NC) was used for all data analyses, with a significance threshold of 0.05. This study was approved by the Human Subjects Committee at the University of California, Davis.


  1. Top of page
  2. Abstract

Table 2 shows demographic and clinical characteristics of the women with ovarian cancer (cases), the breast cancer control group, and the noncancer control group. Of 1985 women with ovarian cancer, 73.2% were categorized with Stage III or IV disease, and 12.3% had no assigned stage. Histologically, 89.2% of ovarian cancers were epithelial carcinomas (including 4.4% categorized as borderline tumors), 3.8% were of sex cord/stromal origin, and 7% were “malignant neoplasm, not otherwise specified.” Demographically, women with ovarian cancer had a median age similar to both sets of controls. Among the women with ovarian cancer, the percent white race and the percent residing in more educated or metropolitan/urban areas were slightly lower than in the breast cancer control group and slightly higher than in the noncancer control group.

Table 2. Demographic Characteristics of Cases and Controls
Ovarian cancerBreast cancerNoncancer
  • a

    Stage IA and IB ovarian cancers were not included in the study.

  • b

    In situ.

No. of patients198510,9416024 
Stage at diagnosis  2333b21.3
Age in yrs      
Zip code of residence      
 More educated135068.0782071.5382763.5
 Less educated63532.0312128.5212835.3
Area of residence      
 Less urban/rural844.23903.63185.3

Diagnostic codes that reflected symptoms commonly reported by patients with ovarian cancer were grouped into the following “target symptom” categories: gastrointestinal symptoms, abdominal pain, pelvic pain, and abdominal swelling. Occurrence frequencies (per unique patient) and adjusted ORs were calculated for each of the four major “target symptom” categories using both breast cancer and noncancer control groups, within multiple prediagnosis time segments defined by the SEER date of ovarian cancer diagnosis (Table 3). Data on fatigue/malaise and urinary symptoms were analyzed, but are not presented in Table 3 because of minimal differences between cases and controls. Fatigue/malaise codes were associated with a maximum OR of 1.4 for ovarian cancer in the 1–3 months before diagnosis, and urinary symptom codes had a maximum OR of 1.9 in the 7–9 months before diagnosis.

Table 3. Frequency, Adjusted Odds Ratios, and 95% Confidence Intervals for Claims Codes for Symptoms among Medicare-Enrolled Patients with Ovarian Cancer During the 36-Month Period Prior to Diagnosis (Index date) Compared with Breast Cancer Controls and an Age-Matched Control Population Without Cancera
Mos before diagnosisOvarian cancer (n = 1985) (%)Breast cancer controls (n = 10,941)Noncancer controls (n = 6024)
  • OR: odds ratio; 95%CI: 95% confidence interval.

  • a

    ORs were adjusted to other ovarian cancer symptom groups (including International Classification of Diseases, 9th revision code 183.0), race, urbanicity, and education level at the zip code of residence.

  • b

    Bolded values have 95% confidence intervals above 1.0.

Gastrointestinal symptoms (mos)       
Abdominal pain (mos)       
Pelvic pain (mos)       
Abdominal swelling (mos)       

During the period 1–3 months preceding ovarian cancer diagnosis, claims records from 863 patients (47.5%) showed codes for at least 1 target symptom. The frequency of codes for the abdominal pain category was highest during the 1–3 months prediagnosis (30.6%), followed by abdominal swelling (16.5%), gastrointestinal symptoms (8.4%), and pelvic pain (5.4%). The adjusted ORs for each of these 4 categories were elevated significantly using the breast cancer control group and the noncancer control group, with the highest adjusted ORs for abdominal swelling (OR, 30.9 and 39.2, respectively), followed by abdominal pain (OR, 6.0 and 6.2, respectively), pelvic pain (OR, 4.3 and 4.2, respectively), and gastrointestinal symptoms (OR, 2.3 and 2.0, respectively). The adjusted OR for abdominal swelling codes also was elevated significantly using both control groups in the periods 4–6 months prediagnosis and 10–12 months prediagnosis (but not significantly in the period 7–9 months prediagnosis). The adjusted OR for abdominal pain codes was elevated significantly using the breast cancer control group in the periods 4–6 months and 7–9 months prediagnosis and, using the noncancer control group, in the period 4–6 months prediagnosis. The adjusted OR for gastrointestinal symptom codes was elevated significantly using both control groups in the period 7–9 months prediagnosis. Pelvic pain codes showed elevated adjusted ORs using both control groups in the periods 4–6 months and 10–12 months prediagnosis. In the entire 4–36 months prediagnosis period, claims records from 785 patients (40.1%) had ≥ 1 target symptom code. During the period 10–12 months prediagnosis, 7.4% of patients with ovarian cancer had claims codes for ≥ 1 target symptom.

The ICD-9-CM code for ovarian cancer (183.0) was included in the analysis to identify when the diagnosis first was recorded on a Medicare claim. This code appeared before the date of SEER diagnosis in the claims of 163 patients (8.2%). Of 199 claims from these patients that contained ICD-9-CM code 183.0, 159 claims (79.9%) were entered in the period 1–3 months prediagnosis, and 40 claims (20.1%) were entered in the period 4–36 months prior to the SEER diagnosis. Consequently, < 2% of the patients with ovarian cancer had a claims diagnosis of ICD-9-CM code 183.0 > 3 months before the SEER diagnosis date. Excluding patients with ovarian cancer who had a claims diagnosis of ICD-9-CM code 183.0 prior to the SEER date of diagnosis did not alter our results (not shown).

Because ICD-9-CM codes on Medicare claims may not reflect the actual frequency and nature of symptoms preceding a diagnosis of ovarian cancer, we investigated the occurrence of clinical interventions (e.g., tests, office visits) as an objective indication that symptoms may have triggered diagnostic investigation. Four categories of medical intervention were tested: pelvic imaging/CA125, abdominal imaging, gastrointestinal procedures, and office evaluation/management. Data on urinalysis and hematology/blood chemistry were analyzed but are not presented here because of minimal differences between cases and controls. Occurrence frequencies (per unique patient) and adjusted ORs for each of the remaining four intervention categories using the noncancer control group within multiple prediagnosis time segments, as defined by the SEER date of ovarian cancer diagnosis, are shown in Table 4 (breast cancer control comparison resulted in similar data; not shown). Among the test groups, abdominal imaging had the greatest frequency within the period 1–3 months prediagnosis (34.7%), followed by pelvic imaging/CA125 (26.6%), and gastrointestinal procedures (8%). Adjusted ORs based on the noncancer control group were elevated significantly for all 4 intervention categories in the period 1–3 months prediagnosis, with the highest OR for pelvic imaging/CA125 (10.3). Pelvic imaging/CA125 also showed a significantly elevated adjusted OR in the period 4–6 months prediagnosis, although the frequency of testing diminished greatly (3.7%). Abdominal imaging showed a significantly elevated adjusted OR in the periods 4–6 months and 7–9 months prediagnosis. The adjusted OR for office evaluation/management was elevated significantly in the periods 1–3 months and 4–6 month prediagnosis and was elevated slightly in earlier periods.

Table 4. Adjusted Odds Ratio, 95% Confidence Intervals, and Proportion of 1985 Patients with Ovarian Cancer who Received Medical Procedures During the 36-Month Period Prior to their Diagnosis (Index date) Compared with 6024 Age-Matched, Noncancer Population Controls
Mos before diagnosisPelvic imaging/CA125Abdominal imagingGastrointestinal proceduresOffice E&M
Ov Ca (%)Control (%)OR (95%CI)aOv Ca (%)Control (%)OR (95%CI)Ov Ca (%)Control (%)OR (95%CI)Ov Ca (%)Control (%)OR (95%CI)
  • Office E&M: office visit (evaluation and management): Ov Ca: ovarian cancer; OR: odds ratio; 95%CI: 95% confidence interval

  • a

    ORs were adjusted to other procedures received, age (breast cancer controls only), race, urbanicity, and education level at zip code of residence.

  • b

    Bolded values have 95% confidence intervals above 1.0.

1–3 mos26.61.110.35 (7.81–13.71)b34.73.75.47 (4.52–6.61) (0.96–1.80)41.325.31.30 (1.08–1.58)
4–6 mos3.71.22.41 (1.69–3.45) (1.03–1.68) (0.83–1.67)27.525.41.35 (1.14–1.60)
7–9 mos1.70.91.56 (0.98–2.48) (1.04–1.79) (0.77–1.59)26.526.21.17 (0.99–1.39)
10–12 mos1.61.41.14 (0.73–1.79) (0.65–1.15) (0.73–1.61)25.625.51.22 (1.04–1.45)
13–15 mos1.51.31.22 (0.77–1.93) (0.75–1.35) (0.65–1.43)25.724.21.09 (0.92–1.29)
16–18 mos1.10.91.25 (0.74–2.11) (0.72–1.24) (0.73–1.49)24.524.61.25 (1.05–1.49)
19–24 mos1.81.71.09 (0.74–1.62) (0.80–1.23) (0.76–1.34)40.738.81.09 (0.98–1.22)
25–30 mos1.91.81.07 (0.72–1.58) (0.79–1.22) (0.72–1.27)39.439.21.02 (0.92–1.13)
31–36 mos1.81.61.15 (0.77–1.73) (0.72–1.15) (0.59–1.07) (0.89–1.08)

The data shown in Table 4 reflect the frequency and ORs for medical interventions, regardless of indication. To investigate the possibility that medical interventions associated with specific indications (e.g., target symptoms) may provide a more reliable view of prediagnosis evaluation in patients with ovarian cancer, we analyzed abdominal imaging in more detail. Similar to the results shown in Table 4, ORs for abdominal imaging associated with any of the “target symptom” codes (using the noncancer control group) were elevated significantly through the period 7–9 months prediagnosis (data not shown). In the period 1–3 months before they were diagnosed with ovarian cancer, 70.2% of patients with target symptom codes had abdominal imaging, whereas 54.3% of such patients had pelvic imaging/CA125. In the period 4–36 months before they were diagnosed with ovarian cancer, 61.1% of patients with codes for target symptoms had abdominal imaging, whereas 30.8% had gastrointestinal procedures, and only 25.3% had pelvic imaging/CA125.


  1. Top of page
  2. Abstract

The principal findings of this population-based, retrospective case–control study using the SEER-Medicare-linked data base include the following: 1) Patients with ovarian cancer diverged from the control groups, having a greater frequency of codes for target symptoms many months prior to cancer diagnosis, especially for abdominal pain and swelling. 2) Abdominal imaging and pelvic imaging/CA125 were used frequently within the 3 months prior to diagnosis of ovarian cancer, but relatively few patients had such studies 4–36 months prior to diagnosis.

Although the natural history of ovarian oncogenesis has been documented poorly, clinical studies of women who had normal-appearing ovaries at hysterectomy and who subsequently developed ovarian cancer have shown that advanced disease can evolve within 13–24 months.16, 17 Retrospective patient questionnaire studies suggest that symptoms associated with ovarian cancer, even in patients with localized disease, often start many months prior to diagnosis, perhaps paralleling the time course of disease progression.6, 8–11 Our current findings comparing the frequency of codes for target symptoms among patients with ovarian cancer relative to controls within the 1–3 months prediagnosis generally are consistent with prior questionnaire studies that used noncancer control groups. For example, Olson et al.,10 based on interviews with 168 patients who had ovarian cancer within 9 months after surgery, found an OR for “unusual abdominal or lower back pain” of 6.2; and Goff et al.,13 based on preoperative interviews with 44 patients, found an OR of 2.3 for abdominal pain. In the 1–3-month period, we observed an OR of 5.9–6.2 for abdominal pain codes. Olson et al.10 also found an OR of 25.3 for “unusual bloating, fullness, and pressure in the abdomen and pelvis”; whereas Goff et al.13 found an OR of 3.6 for “bloating” and an OR of 7.4 for “increased abdominal size.” We observed an OR of 31–39 associated with claims codes for “abdominal or pelvic swelling or mass” in the 1–3 months prediagnosis.

Because we ascertained symptoms from Medicare claims codes, it is difficult to compare the duration of symptoms with that reported in prior patient questionnaire studies. Olson et al.10 found that patients with advanced ovarian cancer reported mean durations of 5.6 months of “unusual abdominal or lower back pain,” and 4.5 months of “unusual bloating, fullness, and pressure in the abdomen or pelvis” before diagnosis. Goff et al.13 reported a median duration of 4 months of abdominal pain (range, 1–11 mos) and 3 months of bloating (range, 1–6 mos). We found that codes for both abdominal pain and abdominal imaging were significantly more frequent among patients with ovarian cancer than among the control groups in the 7–9 month prediagnosis period. We also observed a significantly increased odds of abdominal swelling codes in the 10–12 month prediagnosis period. These observations provide objective evidence that patients with ovarian cancer, as a group, are distinguishable symptomatically from controls at least 6 months prior to diagnosis.

Similarly, it is difficult to compare our findings regarding the proportion of patients with ovarian cancer who manifested symptoms prior to diagnosis with the findings from previous studies. Goff et al.8 observed that 57% of patients with ovarian cancer reported “abdominal bloating”; Olson et al.10 observed that 71% of patients reported “unusual bloating.” In the current study, only 16.5% of patients with ovarian cancer had abdominal swelling codes in the 1–3 months prediagnosis. However, the measurement of symptom code frequencies cannot be equated with the measurement of actual symptom frequencies. ICD-9-CM coding policy encourages use of well defined diagnostic codes; codes for “ill defined conditions and symptoms” should be used only when a definitive diagnosis cannot be determined. Indeed, retrospective comparisons of Medicare claims records with physician office records have confirmed that ICD-9-CM symptom codes greatly underestimate the frequency of many symptoms.18 Even though our analysis presumably underestimated the frequency of symptoms reported to physicians, such bias occurred for both cases and controls. Therefore, the current results may provide reasonable estimates of when such symptoms become severe enough to require medical attention, as reflected in adjusted ORs using both control groups.

Delayed diagnosis of ovarian cancer has been attributed in part to physicians and health care providers.4, 6, 8, 11 Delayed diagnosis may occur if diagnostic testing is initiated but fails to include tests that are capable of rendering the correct diagnosis. When target symptoms do attract enough attention to trigger a testing algorithm, consideration should be given to ovarian cancer in the differential diagnosis. Although both pelvic imaging/CA125 and abdominal imaging were used frequently in the 3 months preceding diagnosis, abdominal imaging and gastrointestinal procedures were the most common modalities used to investigate target symptoms 4–36 months prior to diagnosis. Abdominal imaging may be sufficient to make a diagnosis of advanced ovarian cancer; however, when the disease is confined to the pelvis, abdominal imaging may not lead to a correct diagnosis. Although not recommended for the “screening” of asymptomatic patients, pelvic imaging and CA125 are reasonable options for women who have target symptoms that are not explained by routine medical evaluation. Only 25.3% of patients with ovarian cancer in the current study whose claims records showed target symptom codes 4–36 months before diagnosis had pelvic imaging or CA125. Pelvic imaging and CA125 are not completely sensitive tests,19 but performing such tests may have resulted in an earlier diagnosis for some patients who were symptomatic 4–36 months before diagnosis.

The current study is limited due to its retrospective design and dependence on diagnostic codes in the Medicare Part B claims data base. Even though the SEER-Medicare data base has been useful in a variety of studies on cancer treatment and surveillance,20 administrative and health claims data bases were designed for other purposes and must be used with caution.21, 22 We believe that our use of two independent control groups and the close agreement in results using the two groups, enhanced the validity of our findings. Our focus on women age 68 years and older who were enrolled into Medicare for a minimum of 3 years before a SEER diagnosis of ovarian cancer may reduce the generalizability of the results for women of other ages. However, the peak incidence of ovarian cancer is in the 7th decade, and > 30% of patients with ovarian cancer are age 65 years or older at the time of diagnosis.23 In addition, postmenopausal control women have a lower background rate of symptoms from ovulatory, menstrual, and other benign reproductive processes (e.g., endometriosis) than younger women.13 Because of their very small numbers, patients with Stage IA and IB ovarian cancer were excluded from the current analysis, because a much larger study would be required to assess for the presence of symptoms associated with localized disease. Similarly, a comparison of symptoms that developed among patients who eventually were diagnosed with early-stage disease (Stage I and II) versus late-stage disease (Stage III and IV) was not attempted here but is feasible using the approach outlined in this study and could add further to our understanding of the evolution of symptoms associated with ovarian cancer.

Given the findings of the current study, can the diagnosis of ovarian cancer be made earlier? We observed that the claims records in some patients with ovarian cancer had target symptom codes up to 10–12 months prior to diagnosis and that the diagnostic evaluation of such patients usually included abdominal imaging. The inclusion of pelvic imaging and/or CA125 in the diagnostic algorithm of such symptomatic patients may lead to an earlier diagnosis in some patients, although the sensitivity of these tests for early-stage disease is not high. Within 3 months prior to diagnosis, we found that almost 50% of the claims records of women with ovarian cancer were associated with the appearance of target symptom codes, and most patients with symptom codes received both abdominal imaging and pelvic imaging/CA125. However, improvements in the timeliness of clinical diagnosis in this short time frame have uncertain benefit. Our findings suggest that ovarian cancer could be diagnosed earlier in some patients whose diagnosis currently is delayed by at least 4 months, because physicians order abdominal imaging or perform gastrointestinal procedures before they order a test that is more likely to diagnose ovarian cancer, such as pelvic imaging and/or CA125. Future research is needed to confirm these findings and to evaluate prospectively the clinical impact of more timely recognition and work-up of early symptoms of ovarian cancer.


  1. Top of page
  2. Abstract
  • 1
    Novak ER, Jones GS, Jones HW Jr. Novak's textbook of gynecology, 7th ed. Baltimore: Williams and Wilkins, 1965.
  • 2
    TeLinde RW. Operative gynecology, 1st ed. Philadelphia: JB Lippincott, 1946.
  • 3
    Curtis AH. A textbook of gynecology. Philadelphia: WB Saunders, 1946.
  • 4
    Ranney B, Ahmad MI. Early identification, differentiation and treatment of ovarian neoplasia. Int J Gynecol Obstet. 1979; 17: 209218.
  • 5
    Wikborn C, Pettersson F, Silfverswärd C, Moberg PJ. Symptoms and diagnostic difficulties in ovarian epithelial cancer. Int J Gynecol Obstet. 1993; 42: 261264.
  • 6
    Flam F, Einhorn N, Sjövall K. Symptomatology of ovarian cancer. Eur J Obstet Gynecol Reprod Biol. 1988; 27: 5357.
  • 7
    Smith EM, Anderson B. The effects of symptoms and delay in seeking diagnosis on stage of disease at diagnosis among women with cancers of the ovary. Cancer. 1985; 56: 27272732.
  • 8
    Goff BA, Mandell L, Mintz HG, Melancon CH. Ovarian carcinoma diagnosis: results of a national survey. Cancer. 2000; 89: 20682075.
  • 9
    Vine MF, Ness RB, Calingaert B, Schildkraut JM, Berchuck A. Types and duration of symptoms prior to diagnosis of invasive or borderline ovarian tumor. Gynecol Oncol. 2001; 83: 466471.
  • 10
    Olson SH, Mignone L, Nakraseive C, Caputo TA, Barakat RR, Harlap S. Symptoms of ovarian cancer. Obstet Gynecol. 2001; 98: 212217.
  • 11
    Webb PM, Purdie DM, Grover S, Jordan S, Dick M-L, Green AC. Symptoms and diagnosis of borderline, early and advance epithelial ovarian cancer. Gynecol Oncol. 2004; 92: 232239.
  • 12
    Koren G, Maltepe C, Navioz Y, Wolpin J. Recall bias of the symptoms of nausea and vomiting of pregnancy. Am J Obstet Gynecol. 2004; 190: 485488.
  • 13
    Goff BA, Mandel LS, Melancon CH, Muntz HG. Frequency of symptoms of ovarian cancer in women presenting to primary care clinics. JAMA. 2004; 291: 27052712.
  • 14
    Hoskins WJ, McGuire WP, Bardy MF, et al. The effect of diameter of largest residual disease on survival after primary cyto-reactive surgery in patients with suboptimal residual epithelial ovarian carcinoma. Am J Obstet Gynecol. 1994; 170: 974979.
  • 15
    Warren JL, Klabunde CN, Schrag D, Bach PB, Riley GF. Overview of the SEER-Medicare data: content, research applications, and generalizability to the United States elderly population. Med Care. 2002; 40( Suppl IV): 3–18.
  • 16
    Sightler SE, Boike GM, Estape RE, Averette HE. Ovarian cancer in women with prior hysterectomy: a 14 year experience at the University of Miami. Obstet Gynecol. 1991; 78: 681684.
  • 17
    McGowan L. Ovarian cancer after hysterectomy. Obstet Gynecol. 1987; 69: 386389.
  • 18
    Fowles JB, Lawthers AG, Weiner JP, et al. Agreement between physician's office records and Medicare Part B claims data. Health Care Finance Rev. 1995; 16: 189199.
  • 19
    Horiuchi A, Itoh K, Shimizu M, et al. Toward understanding the natural history of ovarian carcinoma development: a clinicopathologic approach. Gynecologic Oncol. 2003; 88: 309317.
  • 20
    Warren J, editor. Use of the SEER-Medicare data for cancer-related health services research. Med Care. 2002; 40( Suppl IV): 1117.
  • 21
    Solz H, Gilbert K. Health Claims Data as a strategy and tool in disease management. J Ambul Care Manage. 2001; 24: 6985.
  • 22
    Quam L, Ellis LBM, Venus P, Clouse J, Taylor CJ, Leatherman S. Using claims data for epidemiologic research. Med Care. 1993; 31: 498507.
  • 23
    DiSaia PJ, Creasman WT. Clinical gynecologic oncology, 5th ed. Mosby, 1997.