How are symptoms of ovarian cancer managed?

A study of primary care physicians

Authors


  • We thank Blythe Ryerson from the Centers for Disease Control and Prevention for her early contributions to the development of the study's methods and Gilmore Research in Seattle, Washington for conducting the survey.

  • See editorial on pages 4340-2, this issue.

Abstract

BACKGROUND:

A study was undertaken to identify the diagnostic approaches that primary care physicians and gynecologists undertake in women with symptoms associated with ovarian cancer.

METHODS:

A vignette-based survey was mailed to 3200 primary care physicians from the American Medical Association Physician Masterfile. The vignette described a 55-year-old woman with symptoms associated with ovarian cancer, although ovarian cancer was never mentioned. The authors evaluated patient, physician, and practice characteristics associated with a workup that could detect ovarian cancer.

RESULTS:

The survey response rate was 61.7%. After exclusions, 1532 physicians were included. Overall, 89.5% of physicians reported that they would recommend testing that can detect ovarian cancer (71.2% ultrasound; 25.4% pelvic computed tomography; 26.5% CA125). In adjusted analysis, the only patient factor associated with ovarian cancer testing was symptom type, genitourinary versus gastrointestinal (risk ratio, 1.07; 95% confidence interval, 1.03-1.11). Physician and practice characteristics associated with recommending of ovarian cancer testing included specialty (gynecologists > family physicians and internists); type of practice (group > solo); clinical teaching (yes > no); and within Census division, location of practice, with all Central (East, West, North, and South) and Atlantic (Middle and South) areas having a lower likelihood than New England.

CONCLUSIONS:

On the basis of a vignette in which a woman reported symptoms associated with ovarian cancer, the majority of primary care physicians and gynecologists would not recommend CA125, but would recommend imaging of the pelvis. Gynecologists, physicians involved with clinical teaching, and those in group practices were significantly more likely to recommend testing that could lead to an ovarian cancer diagnosis. Cancer 2011;. © 2011 American Cancer Society.

Historically, ovarian cancer has been called a “silent killer,” because symptoms were not thought to be overtly present until advanced stages, when chances of cure are poor.1 However, recent studies have confirmed that >80% of women with early stage disease and >90% of women with advanced stage disease will have frequent symptoms of new onset before their diagnosis.2-13 This finding has important implications, as women with early stage disease have cure rates of 70% to 90%, compared with 15% to 30% for women with advanced stage disease.1 Unfortunately, the symptoms of ovarian cancer are somewhat vague and include bloating, abdominal or pelvic pain, early satiety, and urinary symptoms.2-4 These can be the presenting symptoms for a variety of diseases and are often not recognized as potentially serious by patients and providers. In 1 survey of 1700 women with ovarian cancer, the majority were diagnosed with conditions such as irritable bowel disease, constipation, cystitis, stress, and depression before their cancer diagnosis.2 Prescription medication for a condition other than ovarian cancer was given to 30% of the patients surveyed before the diagnosis of ovarian cancer.2

Although there have been concerns that recall bias may affect symptom reporting by patients as compared with controls, there have been case-control studies using chart review and International Classification of Diseases, 9th edition billing codes, before the diagnosis of ovarian cancer, that have confirmed that women with ovarian cancer are significantly more likely than controls to have specific symptoms.5, 6, 8, 12 Smith et al8 evaluated Medicare claims linked to the California Surveillance, Epidemiology, and End Results (SEER) database for 1985 ovarian cancer patients with age-matched breast cancer and noncancer controls, and found that ovarian cancer patients were significantly more likely to have provider visits for abdominal or pelvic pain, abdominal swelling, and gastrointestinal symptoms in the 6 months before diagnosis than the controls. Hamilton et al6 reviewed chart notes before diagnosis in 212 ovarian cancer patients and 1060 controls. They found that abdominal distension, urinary frequency, and abdominal pain were significantly associated with ovarian cancer even at 6 months before diagnosis. These studies suggest that there is a window of time in which an earlier diagnosis of ovarian cancer can be made, with the potential for a better prognosis.

A 2002 American College of Obstetricians and Gynecologists Committee Opinion on the Role of the Generalist in the Early Detection of Ovarian Cancer14 states that the best way to detect ovarian cancer is for both the patient and the clinician to have a high index of suspicion in symptomatic women. When evaluating a woman with symptoms that could be related to ovarian cancer, it is recommended that providers perform an exam that includes the pelvis and that imaging such as a pelvic ultrasound and laboratory tests such as CA125 be used for evaluation. Although these recommendations were published in 2002, there have been no studies evaluating whether they are being incorporated into clinical practice. In 2007, the American Cancer Society issued an evidence-based consensus statement about the importance of recognizing the symptoms of ovarian cancer.15 Our goal was to identify the diagnostic approach that primary care physicians and gynecologists undertake in women with symptoms associated with ovarian cancer using a clinical vignette.

MATERIALS AND METHODS

Study Sample

Our study sample included 3200 physicians aged 64 years and younger practicing in office or hospital settings in the United States, with equal numbers from the primary specialties of family medicine, general internal medicine, and obstetrics-gynecology. Of these, 200 received a pilot version of our questionnaire. The remaining 3000 received the final version of the questionnaire.

The physicians were randomly sampled from the 72,241 family physicians, 77,007 general internists, and 28,929 obstetrician-gynecologists listed in the August 2008 American Medical Association (AMA) Physician Masterfile.

Survey Instrument and Administration

We developed a 12-page mailed survey booklet examining physicians' care for women's health, and posing numerous questions about physician demographics, practice characteristics, attitudes, beliefs, sources of information, and cancer experience. We conducted cognitive interview testing with all 3 specialties, asked physicians at professional meetings to complete the survey with written feedback, and conducted a pilot test of 200 mailed questionnaires to physicians to refine and improve clarity and face validity. The questionnaire evaluated how physicians would treat a patient with symptoms consistent with ovarian cancer. It included a vignette of a 55-year-old women who had symptoms for 6 to 8 weeks. She had seen the physician 3 weeks earlier, but the symptoms had not improved and were occurring almost daily. We varied the patient's race (African American, Caucasian), insurance (Medicaid, private), family history (ovarian cancer in mother, coronary artery disease in father), and type of symptom (predominantly gastrointestinal [GI: bloating, abdominal pain, constipation, and fatigue] vs predominantly genitourinary [GU: abdominal and pelvic pain, urinary urgency, and fatigue]). Physicians were told that on abdominal exam there was tenderness in the left adnexa. The rectal exam had no masses, and guaiac was negative. We then provided physicians with a list of studies, imaging tests, and blood tests and asked what, if any, they would recommend at this visit.

Physicians were randomly assigned to 1 of the 16 versions of the vignette based on the 4 patient variable factors. To optimize response, we conducted the survey using the Total Design Method with modification, entailing two 2-day priority mailings, a midpoint reminder postcard/thank you, a $20 bill with the first mailing, and an encouraging handwritten note from the principal investigator with the second mailing.16

Study Variables

Outcome variable

The study outcome was recommendation of a test that could lead to a diagnosis of ovarian cancer: pelvic ultrasound/transvaginal ultrasound, pelvic computed tomography (CT) scan, and/or CA125.

Independent variables

Patient characteristics.

Patients characteristics were race, insurance, family history, and symptom type.

Physician Characteristics

Physician demographics included age (the survey year [2008] − year of birth), sex from the AMA Physician Masterfile, and race and ethnicity from the questionnaire. We used the primary physician specialty recorded on the survey. When 2 of the specialties of interest (family medicine; general internal medicine; and obstetrics and gynecology) were recorded, we used the 1 that agreed with the specialty in the AMA Physician Masterfile.

Other physician characteristics included years in practice (2008 − year graduated from medical school); involvement in clinical teaching; board certification; average number of outpatients seen weekly; location in an urban, large rural, or small/isolated small rural area (based on Rural Urban Commuting Area codes linked by physician mailing ZIP code from the AMA Physician Masterfile)17; census division of physician mailing address; primary practice setting (eg, office practice, community health center); and practice type (eg, solo, group practice). We measured attitude toward risk taking and malpractice concern using modified published measures.18-20 We measured physicians' nonprofessional experience with cancer: none, experience with family/close friend/coworker only, and the physician's own cancer experience. We created a variable indicating whether a physician identified the American College of Obstetricians and Gynecologists among the top 3 sources of information about cancer screening, as this is the only organization with formal recommendations to evaluate symptoms for early detection.

Development and Weighting of Sample

We used responses from both the pilot and main surveys to ensure the largest sample size. Of the 3200 surveyed physicians, we sequentially excluded 33 duplicates, 95 undeliverable surveys, 19 retired, disabled, or deceased respondents, and 11 not practicing or on leave, resulting in 3042 sample physicians. Of these, 1878 (61.7%) responded. We then further excluded 200 physicians not providing outpatient care to women, 71 working in settings not providing outpatient/primary care (eg, emergency rooms), 10 reporting specialties other than the 3 focused on in this study, and 23 in residency or fellowship training. After exclusions, the overall study sample included 1574 respondents. We weighted the responses of the 591 family physicians, 414 general internists, and 569 obstetrician-gynecologists to their representative number in the practicing US physician population using AMA Physician Masterfile counts proportionately adjusted to 63,418 family physicians, 62,573 general internists, and 26,676 obstetrician-gynecologists based on the exclusions noted above (total weighted N = 152,667). For this study, we excluded the 4264 weighted responses that corresponded to the 42 physicians who were missing data on the outcome variable (recommendation of a test that could lead to a diagnosis of ovarian cancer). Our final study population included 1532 physicians representing 148,403 weighted physicians nationally.

We compared respondents and nonrespondents on the variables available through the AMA Masterfile, and found differences in the response rate by present employment type (P = .02). Respondents and nonrespondents were distributed across the different present employment categories as follows: group practice, 69.3% versus 63.6%; self-employed, 17.7 % versus 22.2%; government, 6.9% versus 7.0%; and other, 6.1% versus 7.2%, respectively. We found no difference in response rate by physician specialty, sex, or age.

Analysis

We first described demographic, practice, and other characteristics of the weighted physician population. We used SUDAAN 10.0 (RTI International, Research Triangle Park, NC) to compare physicians' unadjusted rates of recommending tests likely to lead to a diagnosis of ovarian cancer overall and by patient, physician, and practice characteristics, using P ≤ .01 to denote significance because of multiple comparisons. Multivariate logistic regression analysis identified the patient, physician, and practice characteristics that were independently and significantly associated with appropriate diagnostic testing at the P ≤ .05 level. We also analyzed tests ordered by physicians who did and did not order cancer screening tests at the P ≤ .01 level.

We first entered all patient characteristics into the regression model, and then used stepwise modeling to enter physician and practice characteristics. Because appropriate diagnostic testing is a common outcome, we calculated risk ratios (RRs) based on predicted marginals within SUDAAN 10.0.

RESULTS

The physician and practice characteristics of the weighted respondent sample are shown in Table 1. Family medicine accounted for 41.4%, internal medicine 41.0%, and obstetrics-gynecology 17.6% of the sample. Approximately 40% were female, and almost half (44.5%) had been in practice for >20 years. Almost a quarter of physicians (23.5%) were in solo practice, with the majority in group (72.8%) practices. Most had experience with a family member, close friend, or coworker (78.7%) with cancer. Fear of malpractice was high in over half (57.9%) of physicians, and 31.2% listed American College of Obstetricians and Gynecologists among their top 3 sources of cancer screening information.

Table 1. Characteristics of Physician Respondents (Weighted N = 148,403)
Characteristic% of Total
  1. Abbreviation: ACOG, American College of Obstetricians and Gynecologists.

  2. Weighted N based on 1532 respondent physicians: family medicine, 572; general internal medicine, 403; obstetrics-gynecology, 557.

  3. Missing data (actual respondents/weighted respondents): race, 51/4696; Hispanic ethnicity, 26/2415; board certification, 8/1017; primary setting, 22/2047; practice type, 18/1546; weekly average number of patients, 22/2272; mean percentage of patients who are female, 18/1980; involved in clinical teaching, 9/1124; nonprofessional experience with cancer, 24/2278; level of risk taking, 24/2634; fear of malpractice, 20/2282; and listed ACOG, 23/2319.

Age 
 30-39 years23.1
 40-49 years34.3
 50-64 years42.5
Race 
 Caucasian70.9
 Asian/Pacific Islander16.1
 African American5.1
 Other, including American Indian/Alaska Native, mixed race, missing race7.9
Hispanic ethnicity4.7
Primary specialty 
 Family medicine41.4
 Obstetrics-gynecology17.6
 Internal medicine41.0
Sex female40.4
Board certified91.6
Years in practice 
 0-1018.1
 11-2037.4
 21+44.5
Primary setting 
 Office, urgicenter, hospital outpatient department, freestanding or federal government clinic87.5
 Health maintenance organization or other prepaid practice2.5
 Community health center, tribal health center/Indian Health Service, or nonfederal government clinic4.5
 Other, including institutional setting, family planning clinic, and missing primary setting5.5
Practice type 
 Solo practice23.5
 Group practice72.8
 Other, including missing practice type3.7
Weekly average number of patients 
 1-6027.9
 61-9028.9
 91+43.3
Mean percentage of patients who are female64.3
Involved in clinical teaching40.7
Nonprofessional experience with cancer 
 Family (immediate or extended), close friend, coworker78.7
 Self4.8
 None16.4
Geographic location 
 Urban84.8
 Large rural9.2
 Small/remote rural6.0
Census division 
 New England5.5
 Middle Atlantic14.1
 East North Central16.8
 West North Central8.0
 South Atlantic16.0
 East South Central5.5
 West South Central9.2
 Mountain6.7
 Pacific18.1
Level of risk taking 
 Low, 6-1758.3
 Medium, 18-2434.1
 High, 25+7.6
Fear of malpractice 
 Low, 2-414.1
 Medium, 5-728.0
 High, 8+57.9
Listed ACOG among top 3 sources of cancer screening information31.2

Overall, 89.5% of physicians, when presented with a patient complaining of symptoms that can be associated with ovarian cancer, recommended tests likely to lead to a diagnosis of ovarian cancer (71.2% transvaginal ultrasonography, 25.4% pelvic CT, and 26.5% CA125). However, only about a quarter of physicians (26.5%) are specifically targeting ovarian cancer detection with CA125. In unadjusted analysis, the only patient factor significantly associated with physicians recommending ovarian cancer testing was type of symptoms. Physicians presented with a vignette of a woman with predominantly GU symptoms were significantly more likely to recommend some type of ovarian cancer testing than those presenting with GI symptoms (92.5% vs 86.5%, P ≤ .001). Race, insurance type, and level of ovarian cancer risk (family history) did not impact physicians' recommendations for testing (Table 2).

Table 2. Rates of Ovarian Cancer Diagnostic Testing by Patient Characteristics (Weighted N = 148,403)
CharacteristicTVUPelvic CTCA125Any
  • Abbreviations: TVU, transvaginal ultrasonography; CT, computed tomography.

  • Weighted Ns and estimates are based on data from 1532 physicians resulting in a weighted N of 148,403 physicians with complete data for the 3 testing variables: TVU, pelvic CT, and CA125.

  • a

    P ≤ .001.

All women71.225.426.589.5
Race    
 Caucasian73.922.927.289.0
 African American68.727.725.890.0
Family history of ovarian cancer    
 No72.525.025.389.7
 Yes68.726.128.789.2
Insurance    
 Medicaid69.924.927.688.6
 Private72.625.825.390.5
Symptom type    
 (1) Gastrointestinal68.723.828.686.5a
 (2) Urological73.826.924.392.5

Table 3 shows the unadjusted ovarian cancer testing rates for physicians with different characteristics. Obstetrician-gynecologists were significantly more likely to recommend testing as compared with internists and family physicians (96.2% vs 88.1% and 88.1%, P ≤ .001). Physicians involved with clinical teaching were more likely to recommend ovarian cancer testing than those not involved with teaching (92.9% vs 87.7%, P ≤ .01). Physicians in group practices were more likely to recommend ovarian cancer testing than those in solo practice (91.4% vs 83.9% P ≤ .01). Those who listed American College of Obstetricians and Gynecologists as 1 of their top 3 sources of cancer screening information were more likely to offer ovarian cancer testing than those who did not (92.5% vs 88.4%, P ≤ .01). We found significant variation in testing based on the Census division of the physician's mailing address, ranging from a high of 97.6% in New England to a low of 81.4% in the West North Central area.

Table 3. Rates of Ovarian Cancer Diagnostic Testing by Physician and Practice Characteristics (Weighted N = 148,403)
CharacteristicTVU, %Pelvic CT, %CA125, %Any, %
  • Abbreviations: TVU, transvaginal ultrasonography; CT, computed tomography; ACOG, American College of Obstetrics and Gynecology.

  • Weighted Ns and estimates are based on data from 1532 physicians resulting in a weighted N of 148,403 physicians with complete data for the 3 testing variables: TVU, pelvic CT, and CA125.

  • Missing data (actual respondents/weighted respondents): board certification, 8/1017; practice type, 18/1546; weekly average number of patients, 22/2272; involved in clinical teaching, 9/1124; nonprofessional experience with cancer, 24/2278; level of risk taking, 24/2634; fear of malpractice, 20/2282; listed ACOG, 23/2319.

  • Missing data for practice type is included in the Other category for this variable to minimize the number of physicians excluded.

  • a

    P ≤ .01.

  • b

    P ≤ .001.

Total71.225.426.589.5
Age    
 30-39 years76.322.321.1a92.3
 40-49 years71.224.525.787.9
 50-64 years68.527.830.089.3
Sex    
 Female78.2b18.1b25.191.7
 Male66.530.327.488.1
Specialty    
 Family medicine71.2b23.1b23.3a88.1b
 General internal medicine62.534.027.888.1
 Obstetrics-gynecology91.710.630.996.2
Board certification    
 Yes72.225.425.790.1
 No64.524.336.585.6
Years in practice    
 0-1073.821.720.590.8
 11-2072.525.225.989.1
 21+69.127.029.489.4
Practice type    
 Solo practice64.426.927.583.9a
 Group practice73.125.225.791.4
 Other, including missing practice type76.819.835.389.0
Weekly average number of patients    
 1-6075.923.126.689.6
 61-9067.329.926.191.9
 91+71.923.427.288.3
Involved in clinical teaching    
 Yes76.6b25.026.592.9a
 No68.125.626.687.7
Nonprofessional experience with cancer    
 Family (immediate or extended)72.425.126.690.3
 Self72.822.427.190.9
 None64.729.424.786.6
Geographic location    
 Urban71.026.027.089.6
 Large rural75.122.526.090.1
 Small rural/remote rural69.220.619.088.1
Census division    
 New England83.822.038.8b97.6b
 Middle Atlantic73.029.830.090.1
 East North Central67.627.526.691.1
 West North Central68.318.127.081.4
 South Atlantic68.925.617.386.4
 East South Central72.715.618.582.8
 West South Central68.330.529.487.1
 Mountain72.424.117.295.0
 Pacific73.324.832.192.8
Level of risk taking    
 Low, 6-1771.624.727.690.1
 Medium, 18-2469.727.425.389.1
 High, 25+77.023.425.490.2
Fear of malpractice    
 Low, 2-474.423.527.490.3
 Medium, 5-772.123.727.987.0
 High, 8+70.027.025.690.7
ACOG among top 3 sources of cancer screening information    
 Yes82.8b14.8b28.792.5a
 No66.430.225.788.4

Adjusted analysis produced findings similar to those in the unadjusted analysis (Table 4). Again, the only patient factor significantly associated with physician's recommending ovarian cancer testing was type of symptom. Physicians presented with a vignette of a patient complaining of GU symptoms were 1.07 (95% confidence interval [CI], 1.03-1.11) times more likely to recommend testing compared with those presented with a vignette of a patient with GI symptoms. Significant physician and practice factors were specialty, with a RR for family medicine and internal medicine of 0.91 and 0.92, respectively, for testing (95% CI, 0.88-0.95 for each) as compared with gynecologists, involvement in clinical teaching RR 1.04 (95% CI, 1.01-1.08) as compared with not involved, and group practice RR 1.06 (95% CI, 1.01-1.12) as compared with solo practice. In the adjusted analysis, the geographic Census division in which physicians practiced remained a significant factor associated with ovarian cancer testing (Table 4).

Table 4. Adjusted Risk of Ovarian Cancer Diagnostic Testing by Patient, Physician, and Practice Characteristics (Weighted N = 147,279)
CharacteristicRisk RatioLower 95% LimitUpper 95% Limit
  1. Weighted Ns and estimates are based on data from 1523 responding physicians. The weighted Ns are lower than those in Table 3 because of missing values for 1 or more variables for 9 physicians.

Patient characteristics   
 Patient race, African American vs Caucasian1.000.961.04
 Family history of ovarian cancer, yes vs no1.000.961.04
 Insurance, Medicaid vs private0.980.951.02
 Symptom type, urological vs gastrointestinal1.071.031.11
Physician and practice characteristics   
 Census division   
  Middle Atlantic vs New England0.930.870.99
  East North Central vs New England0.940.890.99
  West North Central vs New England0.820.730.92
  South Atlantic vs New England0.890.840.95
  East South Central vs New England0.850.760.95
  West South Central vs New England0.900.840.97
  Mountain vs New England0.980.921.04
  Pacific vs New England0.960.921.01
 Specialty   
  Family medicine vs obstetrics-gynecology0.910.880.95
  General internal medicine vs obstetrics-gynecology0.920.880.95
 Practice type   
  Group practice vs solo practice1.061.011.12
  Other or missing vs solo practice1.051.001.16
 Involved in clinical teaching, yes vs no1.041.011.08

An adjusted analysis was also performed to identify what factors were predictors of physicians ordering a CA125 (data not shown). Board certification was associated with lower RR 0.69 (95% CI, 0.53-0.90) as compared with those not board certified. In analyzing Census divisions, South Atlantic, East South Central, and Mountain had significantly lower RR of ordering CA125 compared with New England.

Finally, we analyzed what other tests physicians were ordering and found that there were significant differences between those physicians who did and did not recommend ovarian cancer testing. In patients presenting with GI symptoms, those physicians who did not recommend ovarian testing were significantly more likely to order gastrointestinal tests (esophagogastroduodenoscopy, colonoscopy, sigmoidoscopy, stool for occult blood, barium enema, carcinoembryonic antigen, Helicobacter pylori, CT colonography) than those who recommended ovarian cancer testing (72.8% vs 49.6%, P ≤ .001), significantly more likely to order general abdominal tests (abdominal ultrasound, abdominal CT, abdominal x-ray) than those who recommended ovarian cancer testing (70.0% vs 55.0%, P ≤ .01), and significantly less likely to do urologic evaluations (cystoscopy, urinalysis/culture) than those who recommended ovarian cancer testing (28.2% vs 47.0%, P ≤ .001). Among physicians presented with a vignette with GU symptoms, those physicians who did not recommend ovarian cancer testing were more likely to order GI testing (58.2% vs 42.4%) and general abdominal testing (72.3% vs 58.2%) as compared with physicians who did recommend ovarian cancer testing; however, these findings were not significant at the P ≤ .01 level.

DISCUSSION

The World Health Organization characterizes ovarian cancer as a disease that would likely benefit from screening, as the cure rates for women diagnosed with early stage disease are significantly higher than for women with advanced disease.21, 22 Although there is substantial research to identify screening tests, at this time there are no screening methods that have adequate sensitivity or specificity to allow cost-effective screening programs.23 In addition, the results of the Prostate Lung Colorectal Ovarian trial showed that most of the cancers detected in the women randomized to annual transvaginal ultrasound and CA125 were diagnosed in advanced stages.24 Currently the US Preventive Services Task Force recommends against ovarian cancer screening at a periodic health exam, because more women are harmed by false-positive screening.25 Because the prevalence of ovarian cancer is only 1 per 2500 women older than 50 years, screening tests must have a very high specificity of 99.6% to have a positive predictive value of 10%.

Because no adequate ovarian cancer screening test exists, there has been recent interest in promoting early identification of symptoms among patients and physicians. Studies from multiple institutions have found that specific symptoms are significantly more common in women with ovarian cancer as compared with those without.2-13 The majority of women with early stage disease will have abdominal, urinary, or gastrointestinal symptoms. Although these symptoms are also common in women without ovarian cancer, studies have shown that when women experience these symptoms frequently (>12× per month) and the symptoms are of new onset (<6 months to 12 months), these symptoms are predictive of ovarian cancer.4, 26, 27 Currently, the American College of Obstetricians and Gynecologists endorses evaluating symptoms associated with ovarian cancer.14 In addition, the American Cancer Society along with the Society of Gynecologic Oncologists and the Gynecologic Cancer Foundation produced a consensus statement to provide information to patients and practitioners about the association of specific symptoms with ovarian cancer.15

This is the first study conducted to evaluate how a large group of primary care physicians and gynecologists respond to a vignette representing a woman with symptoms that are associated with ovarian cancer. It is reassuring to see that the majority do recommend testing that would lead to a correct diagnosis if ovarian cancer were present. Most physicians recommended imaging that included the pelvis (pelvic ultrasound or CT). Only about a quarter of physicians ordered a CA125, a specific tumor marker for ovarian cancer. However, standard practice for many physicians may include ordering CA125 as a secondary test once a mass is detected by imaging to help determine the likelihood of malignancy.14 Therefore, the lack of a CA125 test as the first diagnostic step is not necessarily a sign that ovarian cancer is not in the physician's differential diagnosis.

Fortunately, patient factors such as race, insurance, or family history did not impact physicians' recommendations of tests that would lead to a diagnosis of ovarian cancer. We did see that women who reported urinary urgency with abdominal and pelvic pain were significantly more likely to have ovarian cancer testing recommended than those complaining of bloating, constipation, and abdominal and pelvic pain. Bloating is the most common symptom associated with ovarian cancer,2-13 and this finding suggests that some physicians are unaware of the association or are inclined to test for other diseases first. We did find that, especially in patients with GI complaints, GI and general abdominal testing was commonly performed as expected. In a Medicare SEER study of 3250 women with ovarian cancer, those with GI symptoms were more likely to have later stage disease and longer time to key diagnostic tests than those with gynecologic symptoms.28 Our study may help to explain this finding.

There were multiple physician and practice factors associated with recommendations for ovarian cancer testing. As expected, gynecologists had the highest rates of testing. As ovarian cancer is the second most common gynecologic malignancy in the United States, and education about gynecologic malignancies is part of the core competencies in training, these physicians are likely to have had more educational opportunities to learn about ovarian cancer. Those involved in teaching were significantly more likely to order ovarian cancer testing; this probably is also the result of enhanced educational opportunities. Other studies evaluating adherence to guideline-recommended surgery for ovarian cancer have shown that physicians involved with teaching are significantly more likely to perform recommended surgical procedures and administer appropriate chemotherapy for ovarian cancer as compared with physicians not involved with teaching.29-31 In addition, those in group practice were more likely than those in solo practice to perform ovarian cancer testing. Studies have shown that physicians in group practices are more likely to perform recommended cancer screening tests as compared with those in solo practices, although the reason for this is unknown.32

Finally, 1 of the most interesting findings in this study is the significant variation in practice patterns based on Census division. When compared with physicians in New England, all of the divisions except Pacific and Mountain had significantly lower RR for testing. Multiple studies have shown significant variations in practice patterns and adherence to guidelines by region.33-37 Geographical variation in cholesterol management,33 prevention of atherothrombosis,34 management of malignancies,35, 36 and prenatal screening,37 despite well-publicized guidelines, has been found. Some have hypothesized that fear of malpractice lawsuits may drive physicians in some areas to more closely follow guidelines,18, 19 but we did not find that fear of malpractice lawsuits predicted a higher rate of ordering tests. The explanation for large geographic variations in practice patterns is probably multifactorial, but something that should be taken into account when designing educational programs.

There are limitations to this study, as results are based on self-report to a hypothetical patient vignette without any verification of chart notes or with standardized patients. Our response rate was quite good at 61.7%, and these results are probably generalizable to the 38.3% of physicians who did not respond to the survey. Other than minor differences in present employment, respondents had demographics and primary specialty similar to those of nonrespondents with the exception of small differences in present employment. Although the vignette type of survey is an efficient means of examining physician-reported diagnostic testing when faced with a woman with symptoms associated with ovarian cancer, this method does not allow us to examine the frequency with which physicians actually order these tests or what percentage consider ovarian cancer in the differential diagnosis.

The majority of physicians, when asked about a patient with symptoms suggestive of ovarian cancer, recommended diagnostic tests that would likely lead to a correct diagnosis. Symptoms that were predominantly GI in nature were somewhat less likely to prompt physicians to recommend ovarian cancer testing. Ovarian cancer testing was more commonly performed by gynecologists, those involved with clinical teaching, and those in group practices. There was also wide geographic variation across the United States. Further research examining best methods to disseminate information about ovarian cancer symptoms and reduce variations in practice patterns will improve our ability to identify ovarian cancer at the earliest point in time.

FUNDING SOURCES

This project was funded by the Centers for Disease Control and Prevention (CDC) through the University of Washington Health Promotion Research Centers Cooperative Agreement U48DP001911, and through the Alliance for Reducing Cancer, Northwest, funded by both the CDC (grant U48DP001911, V. Taylor, principal investigator) and the National Cancer Institute. The findings and conclusions in this journal article are those of the authors and do not necessarily represent the official position of the CDC.

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

Ancillary