Ovarian cancer causes more deaths than any other gynecologic malignancy, and is the fifth-leading cause of cancer death among U.S. women.1 In 2002, the annual incidence rate of ovarian cancer in the United States was 13.1 per 100,000 women, and the death rate was 9.0 per 100,000 women.2 The annual age-specific incidence rates for ovarian cancer increase with advancing age, peaking at 58.0 per 100,000 for women aged 80–84.3 Although the 5-year relative survival rate for women with localized ovarian cancer is quite high (94%), the majority of women (81%) are diagnosed at a regional or distant stage where the 5-year survival rate is 69% and 29%, respectively.1
Because most ovarian cancers are diagnosed at an advanced stage, recent research efforts have focused on improving early recognition of the disease. Currently, there is no effective screening test recommended for use in the general population.4–6 Thus, prompt work-up of symptoms may be critical in detecting the disease earlier. However, the vast majority of symptoms are common, subtle, or appear unrelated to gynecologic conditions, which may result in delays in care seeking or extended diagnostic work-up. Although recent studies suggest that nearly all patients diagnosed with ovarian cancer are symptomatic regardless of stage,7–13 many have found that gastrointestinal symptoms are often associated with later-stage disease,13–15 and symptoms such as pelvic swelling or a palpable mass,13–16 urinary symptoms,12–14 and gynecologic symptoms such as vaginal bleeding16 are associated with early-stage disease.13–16
In addition to the types of symptoms reported, previous studies have shown that women with ovarian cancer report more frequent symptoms in the year before diagnosis than do various control groups.8,17,18 A recent study of Medicare provider claims in California found that women with ovarian cancer were more likely to have symptoms that the investigators identified as being related to ovarian cancer 6–12 months before diagnosis than patients in control groups.18 The investigators concluded that pelvic imaging and CA 125 may be good options for triaging women with symptoms that cannot be otherwise explained by routine medical evaluations.
Although many of these studies have observed patterns of symptoms that may be useful in the earlier detection of ovarian cancer, the majority of them are based on patient recall or are hampered by small sample size.19 Our study examines these previously identified patterns in two large population-based data sources: the Surveillance, Epidemiology, and End Results (SEER) cancer registries database, and Medicare enrollment and claims files for 3,250 women diagnosed with ovarian cancer. The objectives of this study were to examine the type of signs, symptoms, and diagnostic procedures reported in billing claims 12 months before diagnosis for women aged 65 years and older diagnosed with ovarian cancer by stage and to assess the association between types of symptoms and symptom-related diagnoses reported and the time to key diagnostic imaging procedures and surgical evaluation.
MATERIALS AND METHODS
The National Cancer Institute, in collaboration with the Centers for Medicare and Medicaid Services, has linked data from Medicare claims to the 1991–2002 SEER databases to produce the SEER-Medicare data. The SEER cancer registries collect data from 13 geographic regions; these data include patient demographics, cancer site, stage of disease, histology, date of diagnosis, cause of death, and date of death. Medicare claims contain covered health care services including hospital, physician, outpatient, home health, and hospice bills from the time of a person’s Medicare eligibility until death. In addition, we linked data from the American Medical Association (AMA) database so as to accurately identify physician specialty. We analyzed the SEER-Medicare data linked to AMA data for the years 1995–1999. Institutional review board exemption status was obtained by all participating organizations on the premise that the SEER-Medicare data were an existing data source stripped of all personal identifiers.
There were 8,357 women identified in the linked SEER-Medicare database with a diagnosis of ovarian cancer from 1995 through 1999. We excluded 3,689 women because they met one or more of the following criteria: women aged less than 65 years (n=1,801), those diagnosed postmortem (n=161), those with less than 12 months of continuous fee-for-service enrollment in Medicare before diagnosis, and those not continuously entitled to both Part A and Part B during the same time (n=3,305). The last exclusion included 1,137 women who were diagnosed with ovarian cancer at age 65 or those diagnosed in 1995 regardless of age because these women had fewer than 12 months of previous claims. Because our outcomes of interest included SEER stage of disease and the surgical procedure date from the claims file, an additional 1,267 women were excluded who had no surgical procedure claim matched to the SEER diagnosis date, as well as 151 women with unknown stage of disease in SEER. Those with a surgical procedure claim that matched to the SEER diagnosis date were selected by identifying admissions within a month of the SEER diagnosis date with a procedure code relevant to the diagnosis of ovarian cancer on the inpatient, outpatient, or physician claim. Ultimately, the multivariable analysis included a total of 3,250 women diagnosed from 1996 through 1999. The time-to-event analysis included only those women who had at least one encounter with one or more documented symptoms or diagnoses that we identified as previously associated with ovarian cancer (n=2,653).
Stage at diagnosis was determined from the SEER summary stage variables. Summary staging is the most basic way of categorizing how far a cancer has spread from its point of origin using all the clinical and pathological documentation of the extent of disease available in the medical record. For clinicians accustomed to the International Federation of Gynecology and Obstetrics (FIGO) classification system: FIGO stages IA, IB, and I not otherwise specified can be translated into localized disease; FIGO stages IC and II can be translated to regional disease; and FIGO stages III and IV can be translated to distant disease.20
We defined our “target” symptom and diagnosis code groupings relevant to the clinical signs of ovarian cancer using the appropriate International Classification of Disease, Ninth Revision (ICD-9) codes through a comprehensive review of the literature and in consultation with clinicians, including a gynecologic oncologist. In addition to patient-reported symptom codes, we included diagnosis codes that are frequently associated with ovarian cancer such as benign neoplasm of the ovary and ascites. Based on the natural clinical differences between many of the symptom and diagnosis codes and the associations found during our multivariable analysis, three groupings were established for the time-to-event analysis: 1) gynecologic symptoms, 2) gastrointestinal signs and symptoms, and 3) other target symptoms not specifically gynecologic or gastrointestinal in nature, or target diagnoses whose related symptoms were unclear or undefined. The categorization of these signs, symptoms, and diagnoses and corresponding ICD-9 codes can be found in the Appendix.
Women with ovarian cancer often present with symptoms that are commonly associated with other diseases or conditions. As would be expected, initial assessment and treatment is often designed to address the most likely cause of the symptoms. For this reason, the time period between specified symptoms and diagnostic testing for ovarian cancer provides insight into the prediagnostic patterns of care for women with the disease and provides information about which symptoms are most likely to raise a physician’s suspicion for this malignancy. Therefore, we were also interested in the timing of the primary diagnostic procedures which led to a diagnosis of ovarian cancer. Consistent with the National Comprehensive Cancer Network practice guidelines for ovarian cancer, we examined the time to abdominal or pelvic imaging as it is recommended for women with symptoms without other obvious sources of malignancy.21 We used appropriate ICD-9 codes and the Healthcare Common Procedure Coding System to identify both the imaging and surgical procedures of interest.
The multivariable analyses are based on claims occurring in the year before the admission date on the surgical evaluation claim. Because our outcome variable for the multivariable analysis had three possible levels of stage, we used age- and race-adjusted ordinal logistic regression to model the association among each of 13 symptoms, 11 previous diagnoses, and nine diagnostic imaging procedures on later stages of disease. This model gives cumulative odds ratios for which earlier stage is the referent group for each comparison (ie, localized disease compared with regional and distant, and localized and regional compared with distant). The statistical software package SAS 9.1 (SAS Institute, Inc, Cary, NC) was used for the logistic regression analysis.
We used a survival analytic approach with Stata 8 (StataCorp LP, College Station, TX) for the time-to-event analyses and report Kaplan-Meier failure curves and adjusted hazard ratios to examine the association between type of sign or symptom and both 1) days to pelvic or abdominal imaging procedures, and 2) surgical evaluation. The typical terms for time-to-event analyses, “survival analysis,” “hazard ratios,” and “failure,” imply that the event or events of interest are adverse, such as death. The methods, however, are equally valid for desirable or neutral events of interest.22 The index event in each model was defined as the first claim within 12 months before surgical evaluation for which one of the symptoms or diagnoses of interest was recorded. The outcomes for each model were defined as the first claim after the index event for which a pelvic or abdominal imaging procedure was recorded, and the surgical evaluation date on the medical claims file. Both models were adjusted for age, race, and the specialty of the physician seen at the index visit. The physician specialty was thought to be a potential confounder because some specialists may have easier access to imaging equipment and thus may be more likely to order or request such an evaluation.
Of the 3,250 women included in our analysis, 18.5% were diagnosed with localized disease, 6.0% had regional disease, and 75.5% were diagnosed with distant disease. The demographic characteristics of women surgically evaluated for and diagnosed with ovarian cancer by stage are shown in Table 1. The median age was 75 years at diagnosis and did not vary by stage. Only slight variations by race were observed, with a higher proportion of white women being diagnosed at later stages of disease when compared with other racial groups. Although some variations in the distribution of cases by SEER geographic location were observed (data not shown), the variations were no longer observed when we included this information in our age- and race-adjusted multivariable analysis, and we thus excluded it from our final models.
Among the women included in our analysis, 81.6% had at least one symptom or diagnosis presumed to be related to ovarian cancer on a claim within the 12 months before diagnosis. A slightly higher proportion of women with localized disease (83.5%) had a claim for at least one symptom than did women with regional (81.6%) and distant (81.2%) disease.
The frequency of women presenting with the selected symptoms or diagnoses by stage and the adjusted odds ratios relative to later stage disease for these symptoms or diagnoses are shown in Table 2. Gastrointestinal symptoms such as nausea and vomiting, abdominal distention, and constipation, diarrhea, or other digestive disorders were all associated with later-stage ovarian cancer. In addition, women diagnosed with an intestinal obstruction or ascites 12 months before diagnosis were more likely to have been diagnosed with later stage ovarian cancer. In contrast, the presence of specific gynecologic symptoms such as abnormal bleeding and genital organ pain was associated with early disease.
The frequency of women who received diagnostic pelvic or abdominal imaging procedures by stage and the adjusted odds ratios (aOR) relative to later-stage disease for these procedures are shown in Table 3. Imaging of the pelvic area including vaginal (aOR 0.52, 95% confidence interval [CI] 0.42–0.63) and pelvic (aOR 0.51, 95% CI 0.43–0.60) ultrasound examinations was associated with early-stage disease, as were the combined imaging procedures of the pelvis or abdomen (aOR 0.76, 95% CI 0.63–0.92).
Given the observed association between stage at diagnosis and claims reporting gynecologic-related symptoms and imaging procedures, versus the gastrointestinal-related signs and symptoms, we also assessed the association between the type of symptoms and diagnoses reported and the time to key diagnostic imaging procedures and surgical evaluation using Cox proportional hazards models and failure curves. The estimated time from the first recorded symptom or diagnosis to the first pelvic or abdominal imaging procedure by symptom type was analyzed for 2,653 women (81.6%) with at least one index symptom or diagnosis claim that could be clinically related to ovarian cancer (Fig. 1). The results of the two Cox proportional hazards models for time from first presumed ovarian cancer-related symptom or diagnosis claim to 1) diagnostic imaging procedure, and 2) surgical evaluation are shown in Table 4. The rate at which women with previous reports of gynecologic symptoms received pelvic or abdominal imaging procedures is higher (hazard ratio [HR] 5.5, 95% CI 5.1–6.0) than the rate for women with other nongastrointestinal target symptoms. This same association is present, albeit at a lower magnitude, for women with claims of gastrointestinal symptoms (HR 3.6, 95% CI 3.2–4.0). A similar pattern for the estimated time from first recorded symptom or diagnosis claim to surgical evaluation is illustrated in Figure 2. In general, women with claims for gynecologic-related symptoms went to surgery sooner after presenting with symptoms (HR 4.1, 95% CI 3.8–4.5) than did those with gastrointestinal symptoms (HR 2.1, 95% CI 1.9–2.3) or other target symptoms or diagnoses.
Without an effective screening test for ovarian cancer, early recognition and prompt work-up of symptoms may be critical in improving prognosis by detecting the disease at an earlier stage. This analysis provides information from a large population-based data source about the type of symptoms women experience by stage of disease and provides unique data about the time it takes for women presenting with different symptoms to undergo procedures necessary to rule out ovarian cancer. Symptoms were reported among women at all stages, including those diagnosed with early-stage ovarian cancer. Women who were diagnosed at an early stage were more likely to have claims reporting gynecologic symptoms, whereas those diagnosed at later stages were more likely to have claims reporting gastrointestinal symptoms. Compared with women with gastrointestinal symptoms, those with gynecologic symptoms received abdominal or pelvic imaging sooner and underwent surgical evaluation sooner after presenting with their symptoms.
The reason a difference in symptoms is observed in women diagnosed with early-stage versus late-stage disease remains unclear. This pattern could be due to biological differences between two distinct tumor types, differences in patients’ perception of their symptoms as severe or important enough to seek medical attention, or a delay during the diagnostic work-up of commonly encountered and seemingly benign symptoms. Early-stage disease may have a biologically distinct tumor that manifests itself differently from the late-stage disease, producing different symptoms.14 For example, a tumor diagnosed at an early stage may actually be larger, producing a palpable mass and related symptoms, but may not have metastasized as quickly as a smaller more aggressive tumor that is identified at a later stage. Alternatively, patients diagnosed with late-stage disease may have previously exhibited the gynecologic symptoms associated with early-stage disease, but they either neglected to seek prompt medical attention or presented with symptoms that did not indicate the need for a gynecologic evaluation, thereby delaying diagnosis.
Our results also indicate that women presenting with gastrointestinal symptoms associated with later-stage disease did not receive pelvic or abdominal imaging as quickly as those with the gynecologic symptoms characteristic of early disease. Diagnostics for gastrointestinal symptoms may take more time or be more extensive than those for gynecologic symptoms, or alternatively, mild gastrointestinal symptoms may be treated or tolerated for longer periods of time before pursuing additional tests. A gynecologic cause for gastrointestinal symptoms may be low on the list of differential diagnoses for physicians. More research is required to examine the diagnostic patterns for women with gastrointestinal symptoms and to evaluate whether raising physician awareness of ovarian cancer would affect both the time to pelvic imaging and the long-term outcomes for patients with persistent, unexplained symptoms.
This study is limited in several ways. First, its dependence on diagnostic and procedure codes in the Medicare claims database may underestimate the frequency of symptoms, diagnoses, and services performed. Some symptoms and diagnoses may have been exhibited and either not reported by the woman or not recorded by the clinician in the billing data. Conversely, it is unknown whether particular reported symptoms in this study were actually related to the ovarian cancer. Women may have had multiple symptoms with some or all being related to other conditions. Although we found that women diagnosed with early-stage ovarian cancer were more likely to have claims reporting gynecologic conditions, and we infer that these may have led to their diagnosis, we do not know whether the women diagnosed with late-stage disease had gynecologic symptoms at some point before diagnosis when their disease was localized. Lastly, this analysis focuses on women aged 65 years and older, reducing the generalizability for women of younger ages. However, slightly less than half of all ovarian cancer cases occur among women represented in this study, making this age group a substantial proportion of the population carrying the burden of the disease.23 Despite potential underestimates and the unique limitations that exist when using an administrative claims database for retrospective epidemiologic research, the SEER-Medicare data have been used for a variety of cancer-related health services research, and the findings of this longitudinal study using a large population-based cohort can nonetheless enhance the understanding of issues related to ovarian cancer diagnosis.24
Finally, it is worth noting that many of the symptoms found to be associated with both early- and late-stage disease can also be related to benign conditions. Given the subtlety or seemingly unrelated nature of many of the symptoms described, it is unclear how to clinically manage women presenting with such symptoms. The implications of this study are not that the symptoms of early- versus late-stage disease should necessarily be recognizable by the patient or the clinician, but that both should be aware of the potential for these symptoms to be indicators for ovarian cancer and that they may warrant further diagnostic work up. Of further interest and importance would be an examination of the specialty of the diagnosing physicians and how this affects the time to diagnosis. Future analyses of these data will focus on the specialties of the referring, diagnosing, and treating physicians presented on these claims to help understand the potential impact of physician training on the recognition and work-up of the symptoms of ovarian cancer. Additionally, work is underway to compare the experience of women being diagnosed with ovarian cancer to comparison populations without the disease. Until more is known about the frequency or severity of these symptoms in women without ovarian cancer, or the potential biologic differences of ovarian tumors presenting at earlier or later stages are identified, women and clinicians must carefully consider all the possible benign and malignant diagnoses when presented with such symptoms.
1. Jemal A, Siegel R, Ward E, Murray T, Xu J, Smigal C, et al. Cancer statistics, 2006. CA Cancer J Clin 2006;56:106–30.
2. U.S. Cancer Statistics Working Group. United States cancer statistics: 2002 incidence and mortality. Atlanta (GA): U.S. Department of Health and Human Services, Centers for Disease Control and Prevention and National Cancer Institute; 2005.
3. Ries L, Harkins D, Krapcho M, Mariotto A, Miller B, Feuer E, et al, editors. SEER cancer statistics review, 1975–2003. Available at: http://seer.cancer.gov/csr/1975_2003/
. Retrieved February 16, 2007.
4. Buys SS, Partridge E, Greene MH, Prorok PC, Reding D, Riley TL, et al. Ovarian cancer screening in the Prostate, Lung, Colorectal and Ovarian (PLCO) cancer screening trial: findings from the initial screen of a randomized trial. Am J Obstet Gynecol 2005;193:1630–9.
5. U.S. Preventive Services Task Force. Screening for ovarian cancer: recommendation statement. Ann Fam Med 2004;2:260–2.
6. Nelson HD, Westhoff C, Piepert J, Berg A. U.S. Preventive Services Task Force. Screening for ovarian cancer: brief evidence update. Rockville (MD): Agency for Healthcare Research and Quality; 2004. Available at: http://www.ahrq.gov/clinic/3rduspstf/ovariancan/ovcanup.htm
. Retrieved February 16, 2007.
7. Goff BA, Mandel L, Muntz HG, Melancon CH. Ovarian carcinoma diagnosis. Cancer 2000;89:2068–75.
8. Goff BA, Mandel LS, Melancon CH, Muntz HG. Frequency of symptoms of ovarian cancer in women presenting to primary care clinics. JAMA 2004;291:2705–12.
9. Igoe BA. Symptoms attributed to ovarian cancer by women with the disease. Nurse Pract 1997;22:122, 127–8, 130.
10. Lataifeh I, Marsden DE, Robertson G, Gebski V, Hacker NF. Presenting symptoms of epithelial ovarian cancer. Aust N Z J Obstet Gynaecol 2005;45:211–4.
11. Olson SH, Mignone L, Nakraseive C, Caputo TA, Barakat RR, Harlap S. Symptoms of ovarian cancer. Obstet Gynecol 2001;98:212–7.
12. 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;56:2727–32.
13. Webb PM, Purdie DM, Grover S, Jordan S, Dick ML, Green AC. Symptoms and diagnosis of borderline, early and advances epithelial ovarian cancer. Gynecol Oncol 2004;92:232–9.
14. Attanucci CA, Ball HG, Zweizig SL, Chen AH. Differences in symptoms between patients with benign and malignant ovarian neoplasms. Am J Obstet Gynecol 2004;190:1435–7.
15. Flam F, Einhorn N, Sjovall K. Symptomatology of ovarian cancer. Eur J Obstet Gynecol Reprod Biol 1988;27:53–7.
16. Eltabbakh GH, Yadav PR, Morgan A. Clinical picture of women with early stage ovarian cancer. Gynecol Oncol 1999;75:476–9.
17. Friedman GD, Skilling JS, Udaltsova NV, Smith LH. Early symptoms of ovarian cancer: a case-control study without recall bias. Fam Pract 2005;22:548–53.
18. Smith LH, Morris CR, Yasmeen S, Parikh-Patel A, Cress RD, Romano PS. Ovarian cancer: can we make the clinical diagnosis earlier? Cancer 2005;104:1398–407.
19. Daly MB, Ozols RF. Symptoms of ovarian cancer—where to set the bar? JAMA 2004;291:2755–6.
20. Young Jr, JL Roffers SD, Ries LAG, Fritz AG, Hurlbut AA, editors. SEER summary staging manual - 2000: codes and coding instructions. NIH Pub. No. 01-4969. Bethesda (MD): National Cancer Institute; 2001.
21. National Comprehensive Cancer Network. NCCN Practice Guidelines in Oncology. Ovarian cancer v 1. Available at: http://www.nccn.org/professionals/physician_gls/default.asp
. Retrieved February 16, 2007.
22. Introduction to survival analysis. In: Kleinbaum D, Klein M. Survival analysis: a self-learning text. 2nd ed. New York (NY): Springer; 2005. p. 1–44.
23. National Program of Cancer Registries. Cancer registry public information data: 1999–2002, WONDER On-line Database. Atlanta (GA): U.S. Department of Health and Human Services, Centers for Disease Control and Prevention; 2005.
24. Warren JE. Use of the SEER-Medicare data for cancer-related health services research. Medical Care 2002;40(8 Suppl):1–117.