Family Practice Advance Access originally published online on June 17, 2005
Family Practice 2005 22(5):548-553; doi:10.1093/fampra/cmi044
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Early symptoms of ovarian cancer: a case–control study without recall bias
a The Kaiser Permanente Division of Research, b The Kaiser Permanente Sacramento Medical Center and c The Department of Obstetrics and Gynecology, University of California at Davis Medical Center, USA
Correspondence to Dr Friedman, Kaiser Permanente Division of Research, 2000 Broadway, Oakland, CA 94612, USA; Email: gdf{at}dor.kaiser.org
Received 9 February 2005; Accepted 21 April 2005.
Friedman GD, Skilling JS, Udaltsova NV and Smith LH. Early symptoms of ovarian cancer: a case–control study without recall bias. Family Practice 2005; 22: 548–553.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
Background. Ovarian cancer is usually diagnosed after it has spread and is difficult to cure. Previous attempts to identify early symptoms have either lacked a control group or have been based on interviews of cases, with possible recall bias.
Objective. The purpose of this study was to identify early symptoms of ovarian cancer by reviewing prediagnostic medical records, free of recall bias, and comparing women with and without ovarian cancer.
Methods. In an integrated health care delivery system, symptoms recorded in medical records of 102 women with ovarian cancer during the two years before diagnosis were compared with those of 102 matched control women.
Results. More cases than controls complained of several symptoms up to one year before diagnosis. Most of these symptoms were abdominal or gastrointestinal in nature and were more prevalent in the advanced stage cases. Other symptom sites included pelvic, urinary, back, and systemic. Because case–control differences were not large and prevalence is low, positive predictive values were generally quite low.
Conclusion. Previous reports of early symptoms of ovarian cancer were confirmed in a study with a control group and free of recall bias. It is not clear that these symptoms occurred while the disease was still localized. Because hundreds of women would have to be investigated to detect one case of ovarian cancer, the clinical utility of these symptoms is uncertain. Nevertheless, health care providers should keep ovarian cancer in mind, when women present with symptoms such as abdominal pain and bloating.
Keywords. Diagnosis, early detection, ovarian neoplasms, screening, symptoms.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
Most women with ovarian cancer are diagnosed when the disease is advanced and difficult to cure.1 Research is under way to determine whether transvaginal ultrasound and measuring blood levels of the tumor marker, CA 125, are useful as screening methods,2 but at present there are no proven screening methods for identifying women who are likely to have localized and curable ovarian cancer. Thus efforts have been made to identify early symptoms of ovarian cancer in the hope that women experiencing them will seek medical care promptly and care providers, in turn, will have a high index of suspicion when patients report them.
Studies to date have identified symptoms experienced by women subsequently diagnosed with ovarian cancer. Those investigations that have relied on review of medical records have not ascertained symptoms in control groups of women without ovarian cancer.3–6 Some studies have been based on interviews or self-administered questionnaires completed by patients or their families.7–12 Two of these included control groups,9,10 but even with this desirable feature, recall bias, wherein patients are more apt to remember symptoms than controls, is a distinct possibility.
The present study ascertained symptoms from medical records and included a control group of women free of ovarian cancer but comparable to the cancer patients in some other respects.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
The Kaiser Permanente Medical Care Program in northern California is an integrated health care system that provides prepaid comprehensive inpatient and outpatient care to over three million subscribers. Its Institutional Review Board approved this study. Over 200 cases of ovarian cancer are diagnosed each year in the Program as recorded in its cancer registry. We aimed to include approximately 100 recent cases in the study, about half diagnosed with disease confined to the ovaries without capsular extension, rupture or any evidence of extra-ovarian cancer [i.e. FIGO (International Federation of Gynecology and Obstetrics) stages IA and IB, TNM stages T1a and T1b] and about half with some evidence of metastases or progression beyond the ovary (i.e. FIGO stages IC through IV, TMN stages T1c through M1). Women with stage IA and IB disease were of interest because symptoms occurring in them were clearly early, occurring while the cancer was still localized. Women with stage IC or greater cancer, i.e. regional and distant spread of the cancer, were of interest because the current system did not detect their disease early enough for most to be cured. We focused on cases diagnosed in 2001.
Of the 249 women diagnosed in that year, only 38 had stage IA and IB disease, three of these had to be excluded later because of insufficient follow-back time or the simultaneous presence of another cancer, and two because of their youth (9 and 15 years old), leaving 33. Resources permitted inclusion of 69 randomly selected women with stage IC-IV disease, making the total 102. For each case, a control subject was randomly selected from all female subscribers who matched for year of birth, length of membership in the Program, and medical facility used primarily, or likely to be used primarily, based on location of residence.
A structured record review was carried out by two medical record analysts, with duplicate readings at first and resolution of discrepancies and questions by one of us (GDF). The analysts were not blinded to the diagnosis. The period covered for both cases and controls was the two years before the date of diagnosis of the case. A numbered list of symptoms was created including symptoms often reported in the outpatient setting including both visits and phone calls, with emphasis on abdominal, gastrointestinal, pelvic, urinary, back and general systemic symptoms. The list was expanded as the analysts initially recorded all symptoms that they found in the charts. Some diagnoses, such as urinary tract infection and irritable bowel syndrome, were also included. For simplicity, these will be referred to as ‘symptoms’ below and in the tables.
We wanted both to evaluate previously reported symptoms of ovarian cancer and to detect previously unsuspected symptoms. However, about half way through the abstraction process, it was clear that much effort was being spent on recording symptoms and conditions unlikely to be related to ovarian cancer, such as upper respiratory infections and injuries. The list was then reduced to 104 symptoms, which were included in the analyses reported here.
Also recorded was whether the symptom was new or old. A symptom was classified as old if it had been recorded in the two years before it was found in the two year review period. Also noted was whether the symptom was recorded before versus on or after the ‘cut-off date’ when ovarian cancer was first suspected, as noted in the records, and the work-up for it begun. The abstractors also recorded whether the symptom or diagnosis was found in a note written by the health care provider or in the Clinical Information Presentation System (CIPS), a computerized file containing diagnoses or major complaints that is supposed to be recorded by the physician or nurse practitioner with each clinic visit.
Our analysis consisted of comparing the number and percentage of cases and controls reporting each symptom. In addition to looking at the entire period, two years before diagnosis to cut-off date, analyses were conducted for two years to one year before diagnosis, one year to six months before diagnosis and six months before diagnosis to cut-off date. Patients with either old or new diagnoses were counted in each interval. Because of relatively small numbers of subjects, we heightened sensitivity by requiring only P < 0.10 rather than the usual P < 0.05 for including symptoms in the tables to be presented. We used Fisher one-sided test results generated by the SAS system13 because it seemed reasonable that ovarian cancer would produce symptoms, not prevent them, and that the absence of a symptom would not be clinically useful. Before presenting individual symptoms, we looked at whether more symptoms were more prevalent in cases than controls than the 50 percent of differences expected by chance in two equivalent groups. For example, for all 104 symptoms in all subjects, 16 were reported by equal numbers of cases and controls, so that this analysis was applied to the 88 symptoms with case–control differences.
We also computed the likelihood ratio for each symptom with a statistically significant excess in cases. This ratio is simply the proportion of cases with, divided by the proportion of controls with, the symptom. The likelihood ratio can be used to estimate the positive predictive value of a symptom, that is the probability that a disease is present in persons with the symptom (posterior probability) if the prevalence of the disease (prior probability) is known.14 The positive predictive value determines how many persons with a symptom would have to receive further diagnostic testing to detect one person with the disease.
For a symptom prevalence of 10% in control women, our study had 80% power to detect at least a 2.1-fold increase in prevalence in all 102 cases, a 3.1-fold increase in stage IA and IB cases, and a 2.3-fold increase in stage IC-IV cases.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
Of the 102 cancers 95 (93.1%) were epithelial (including 17 borderline tumors), 3 (2.9%) were germ cell, and 4 (3.9%) were stromal in origin. The FIGO (TMN) stage distribution at diagnosis was: stage IA and IB (T1a and T1b): 33 (32.4%); stage IC (T1c): 8 (7.7%); stage II (T2): 11 (10.6%); stage III (T3): 37 (35.6%); stage IV (M1): 13 (12.5%). The mean, median, and standard deviation of age at cases' diagnosis were for both cases and controls: 58, 59, and 13.4 years. The age range was 30–87 years in cases and 29–87 years in controls. The mean, median and range of intervals between the cut-off date when ovarian cancer was first suspected and tissue diagnosis were 30, 22, and 0 to 133 days, respectively. Zero days occurred for one patient when unsuspected ovarian cancer was discovered during surgery for another condition.
Both new and old diagnoses showed case–control differences and only the combined results are reported here. The written notes revealed more symptoms than CIPS.
Of the 88 symptoms with case–control differences in reporting, there were 67 or 76% (95% confidence interval 67% to 85%) in which cases exceeded controls, clearly above the 50% expected by chance. The corresponding percentages and 95% confidence intervals for portions of the two-year ascertainment period were: 78% (68%–88%) for six months before diagnosis to cut-off date, 69% (58%–80%) for one year to six months before diagnosis, and 58% (47%–69%) for two years to one year before diagnosis. For the 33 stage IA and IB cases, none of the excess reports in cases exceeded chance expectation, the highest being 59% (47%–70%) for the entire two-year period. For the 69 stage IC–IV cases, their percentages of excess reports were slightly higher than those for all cases, the highest being 87% (79%–95%) from six months before diagnosis to the cut-off date. Thus, ovarian cancer does induce patient complaints of symptoms, but primarily after it has spread to and beyond the ovarian capsule and primarily during the year before diagnosis.
For the entire ascertainment period, the predominant symptoms in the stage IC–IV cases (Table 1) were abdominal and gastrointestinal. Other symptoms included pelvic, rectal, flank, and back pain, dysuria, unintentional weight loss, fatigue, headache, shortness of breath and menopausal symptoms. More of these ovarian cancer patients than controls also reported anxiety after six months before diagnosis and paresthesias one year to six months before diagnosis.
|
Computable likelihood ratios ranged from 1.73 to 13.0 in all periods combined (Table 1), the highest being for pain in the side of the trunk or flank. The likelihood ratio for the most common symptom, abdominal pain, was 2.92.
Among the stage IA and IB cases (Table 2), obesity was prominent. Their only notably excessive abdominal or gastrointestinal symptom was abdominal pain in the interval six months before diagnosis to the cut-off date.
|
The findings for all cases combined were similar to the stage IC–IV cases. Case excesses in a few additional symptoms attained statistical significance. These included amenorrhea (4 versus 0, P = 0.06) for all periods combined, and both change in stool size (5 versus 0, P = 0.03) and urinary incontinence (10 versus 3, P = 0.04) after six months before diagnosis.
There were no statistically significant excesses in either all cases or stage IC–IV cases in the period, two years to one year before diagnosis.
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
Our data confirm previous findings that women with ovarian cancer do experience symptoms related to the disease, at least up to one year before diagnosis. These symptoms were predominantly abdominal and gastrointestinal in our patients. We also confirmed that some pelvic, back, urinary, and systemic symptoms predate the diagnosis of ovarian cancer. The symptoms that showed up in this case–control comparison are very similar to those that have been reported previously from different settings.3–12 Abdominal and gastrointestinal symptoms have generally been the most frequently mentioned. Although our sample size was limited by the available resources, this study offered the advantages of a matched control group of women and objective ascertainment of symptoms without the possibility of recall bias.
It should be kept in mind that we made multiple comparisons and that some of our findings, although nominally statistically significant, may be due to chance. We viewed the analysis of each symptom as independent since many had been reported previously and elected not to make statistical adjustments for multiple comparisons.15
Although the categories are not strictly comparable, the prevalence of symptoms was considerably lower in our subjects than found by Olson et al.9 in another study that included a control group. For example, in the period, one year to six months before interview (their Table 2), "unusual abdominal or lower back pain" was reported by 52% of their cases and 15% of their controls, as compared to 16.3% of all cases and 5.8% of all controls reporting abdominal pain and 10.6% of all cases and 4.8% of all controls reporting low back pain in our study. At least three factors could account for this discrepancy. First, an interview with questions about specific symptoms is likely to elicit more positive responses than would come to attention when coming to a clinic, where mentioning and recording of symptoms are both required. Second, although not as prevalent as Olson et al. found, these symptoms were more commonly reported by our subjects after six months before diagnosis. It may be that Olson et al.'s subjects did not remember exactly when the symptoms first occurred and tended to assign them to earlier periods. Finally, some of our subjects may have developed symptoms earlier but not reported them until after six months before diagnosis.
A factor that might increase reporting in our controls as compared to Olson et al.'s community controls is that ours were patients at a clinic who came in usually because of symptoms. On the other hand, recording of symptoms occurring in control women could have been reduced if they came to the clinic for different complaints or problems. They might not have mentioned symptoms of less concern to them, or the provider may not have recorded them.
Goff et al.10 obtained comparative data by surveying women who visited two primary care clinics to ascertain symptoms experienced during the past year. Again, the survey method elicited much higher symptom frequency in controls than we found, for example, back pain in 45% and abdominal pain in 22%. The symptoms in that study with the most striking case–control differences were increased abdominal size, bloating, urinary urgency and pelvic pain. Based on very small numbers their early-stage patients appeared to have more symptoms than ours. However their early-stage patients included all of stage I and stage II patients, while ours included only stages IA and IB.
An important difference between studies that ascertain symptoms by questionnaire and those that review medical record relates to the use to which the findings can be put. The former would seem more relevant to patient education—alerting patients of what they should notice and report to their providers. The latter would be of more help to providers in evaluating patients' complaints and in designing systems that use data from clinics to alert providers to increase their index of suspicion that ovarian cancer may be present.
Possibly because of small numbers of subjects we found little to suggest that stage IA and IB cases could be detected much earlier than happens currently, well before cancer was suspected. This is probably because localized disease in an ovary does not involve adjacent organs. However, more advanced cancer did produce significant early symptoms. The question remains as to whether any of these symptoms occurred when the disease was sufficiently localized to permit complete surgical removal.
A possible limitation of this study is the lack of blinding of the medical record abstractors to case–control status. We doubt that this biased our results as the abstractors were not told which, if any, symptoms are thought to be associated with ovarian cancer and were asked merely to record all symptoms as they encountered them in the records. Spot-checking of their work indicated that they were following these instructions.
We computed for abdominal pain, the most common symptom, and pain in the side of the trunk or flank, the symptom with the highest likelihood ratio, the number of women with each who would have to receive diagnostic evaluations to detect one woman with ovarian cancer. In a study involving annual screening of 14 469 women by transvaginal ultrasound, investigators at the University of Kentucky found 17 ovarian cancers, for a prevalence of 0.117%,16 the prevalence we chose to multiply by the likelihood ratios to estimate positive predictive value. (Technically, prevalence should be converted to odds before multiplying by the likelihood ratio, and the resulting odds converted back to prevalence.14 However with prevalence this small, the results are virtually identical.) This is probably an over-estimate of prevalence for US women in general since the screenees consisted of both postmenopausal women age 50 years or greater and women age 30 or greater with a family history of ovarian cancer and the examinations were repeated annually. An over-estimate of prevalence would tend to reduce the apparent number of women requiring investigation to find one case.
Multiplying 0.00117 by likelihood ratio 2.92 yields a positive predictive value of 0.00342, the reciprocal of which is 293 women with abdominal pain requiring investigation. A likelihood ratio of 13.0 implies that 66 women with side or flank pain require investigation. This estimate number of women could be much larger because the likelihood ratio of 13.0 is based partly on only one control with the symptom, a small, unreliable number.
Although these low positive predictive values are not encouraging, it may be that a much larger study than ours could identify a combination of two or more symptoms that are more helpful in differentiating women with ovarian cancer from those without. We believe that such a study needs to be similar to ours, that is, based on what is recorded in the medical record. Since manual medical record review is so costly and time-consuming, it is not a practical means for the timely identification of women in the clinical setting even if such combinations of symptoms could be found, possibly with the addition of laboratory data. Hope for rapid screening comes with the increasing substitution of electronic for paper medical records. This could be useful in rapidly identifying women with relevant combinations of findings and raising the index of suspicion of health care providers.
Meanwhile, we recommend that caregivers at least consider the possibility that ovarian cancer may be present when women complain of abdominal pain and pressure symptoms. Persistence of symptoms is useful in helping to rule out acute illnesses, but other conditions to be considered in differential diagnosis, such as irritable bowel syndrome may also produce persistent symptoms. As noted by Daly and Ozols,17 good patient–physician communication is essential and "early detection of ovarian cancer will continue to challenge both the artistic skill of astute clinicians as well as their accumulated scientific acumen."
|
Declaration |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
Funding: supported by Direct Community Benefit Investment Grant #CN-02JSkil-01-H from the Kaiser Foundation Research Institute.
Ethical approval: approved by the Kaiser Foundation Institutional Review Board.
Conflicts of interest: none.
|
Acknowledgments |
|---|
We thank Margarita Magallon and Jerry Fortes for medical record abstraction.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionDeclarationReferences |
|---|
1 Ozols RF, Schwartz PE, Eifel PJ. Ovarian cancer, fallopian tube carcinoma, and peritoneal carcinoma. In DeVita VT Jr, Hellman S, Rosenberg SA. Cancer: Principles and Practice of Oncology, 6th edn. Philadelphia; Lippincott Williams & Wilkins; 2001, 1597–1632.
2 Andriole GL, Reding D, Hayes RB, Prorok PC, Gohagan JK; PLCO Steering Committee. The prostate, lung, colon, and ovarian (PLCO) cancer screening trial: status and promise. Urol Oncol 2004; 22: 358–361.[Web of Science][Medline]
3 Pearse WH, Behrman SJ. Carcinoma of the ovary. Obstet Gynecol 1954; 3: 32–45.[Medline]
4 Ranney B, Ahmad MI. Early identification, differentiation and treatment of ovarian neoplasia. Int J Gynaecol Obstet 1979; 17: 209–218.[Medline]
5 Flam F, Einhorn N, Sjövall K. Symptomatology of ovarian cancer.Eur J Obstet Gynecol Reprod Biol 1988; 27: 53–57.[CrossRef][Web of Science][Medline]
6 Wikborn C, Pettersson F, Silfersward C, Moberg PJ. Symptoms and diagnostic difficulties in ovarian epithelial cancer. Int J Gynecol Obstet 1993; 42: 261–264.[Medline]
7 Smith EM, Anderson B. The effects of symptoms and delay in seeking diagnosis on stage of disease at diagnosis among women with cancer of the ovary. Cancer 1985; 56: 2727–2732.[CrossRef][Web of Science][Medline]
8 Goff BA, Mandel L, Muntz HG, Melancon CH. Ovarian carcinoma diagnosis: results of a national ovarian cancer survey. Cancer 2000; 89: 2068–2075.[CrossRef][Web of Science][Medline]
9 Olson SH, Mignone L, Nakraseive C, Caputo TA, Barakat RR, Harlap S. Symptoms of ovarian cancer. Obstet Gynecol 2001; 98: 212–217.[CrossRef][Web of Science][Medline]
10 Goff BA, Mandel LS, Melancon CH, Muntz HG. Frequency of symptoms of ovarian cancer in women presenting to primary care clinics. J Am Med Assoc 2004; 291: 2705–2712.
11 Koldjeski D, Kirkpatrick MK, Swanson M, Everett L, Brown S. Ovarian cancer: early symptom patterns. Oncol Nurs Forum 2003; 30: 927–933.[Medline]
12 Webb PM, Purdie DM, Grover S, Jordan S, Dick M-L, Green AC. Symptoms and diagnosis of borderline, early and advanced epithelial ovarian cancer. Gynecol Oncol 2004; 92: 232–239.[CrossRef][Medline]
13 SAS/STAT Users' Guide, Version 8. Cary, NC: SAS Institute, Inc; 1999.
14 Koepsell TD, Weiss NS. Epidemiologic Methods: Studying the Occurrence of Illness. New York; Oxford University Press: 2003, 448–449.
15 Rothman KJ, Greenland S. Modern Epidemiology. 2nd edn. Philadelphia: Lippincott-Raven; 1998, 225–229.
16 DePriest PD, DeSimone CP. Ultrasound screening for the early detection of ovarian cancer. J Clin Oncol 2003; 21(10 Suppl): 194–199.
17 Daly MB, Ozols RF. Symptoms of ovarian cancer—where to set the bar? J Am Med Assoc 2004; 291: 2755–2756.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||