Symptoms associated with diagnosis of ovarian cancer: a systematic review

Authors

  • Clare R. Bankhead,

    1. Cancer Research UK Primary Care Education Research Group, University of Oxford, Old Road Campus, Old Road, Headington, Oxford, UK
    Search for more papers by this author
  • Sean T. Kehoe,

    1. Nuffield Academic Department of Obstetrics and Gynaecology, University of Oxford, The Women's Centre, John Radcliffe Hospital, Headington, Oxford, UK
    Search for more papers by this author
  • Joan Austoker

    1. Cancer Research UK Primary Care Education Research Group, University of Oxford, Old Road Campus, Old Road, Headington, Oxford, UK
    Search for more papers by this author

Introduction

The overall five-year survival from ovarian cancer is currently poor at around 30%, with the majority of cancers being diagnosed at an advanced stage.1 The late stage at presentation has been blamed upon the insidious nature of the disease and the vagueness of the symptoms of ovarian cancer, which may be interpreted by patients as being normal changes in the body such as the effects of childbearing, menopause and ageing.2 In addition, many of the reported symptoms are not specific to ovarian cancer and may mimic conditions such as irritable bowel syndrome.3 The difficulties in recognising the likely symptoms of ovarian cancer are reflected in the often convoluted pathways that patients follow before being correctly diagnosed.4,5

Frequent symptoms recorded at presentation include pain and abdominal swelling, dyspepsia, vomiting, altered bowel habit and urinary symptoms of frequency or retention.3,6 These non-specific symptoms (as noted in the medical records) have been shown to be predictive of decreased survival.7 As the consequences of diagnosis at a late stage are catastrophic, the symptoms associated with the presence of ovarian cancer, and their subtleties, need to be better understood in order to facilitate appropriate and timely health-seeking behaviours and referral. Some research has been conducted in this field but has often been retrospective, limited and subject to recording and recall biases. In this study, the research to date has been identified and summarised.

Methods

Searching

The electronic databases Medline, Embase and Cinahl were searched from 1984 to April 2004 using the terms below (Box 1). Papers in all languages were considered. Two hundred and twenty unique papers were retrieved. In addition, five journals were hand searched: Gynecologic Oncology; Journal of Obstetrics and Gynaecology; Obstetrics and Gynecology; European Journal of Obstetrics and Gynecology and Reproductive Biology; and the British Medical Journal. This resulted in one additional paper, which was not retrieved by the electronic search because the term symptom* did not appear within 10 words of the ovarian cancer terms. One other paper was published in June 2004. Therefore, 222 papers were considered.

Table Box 1: . Search terms used in electronic databases
#1cancer* or malignan* or tumo?r* or neoplas* or carcinoma* or adenocarcinoma*
#2#1 near3 ovar*
#3#2 near10 symptom*

Selection

After reading titles and abstracts, papers were excluded because the focus was on other areas such as treatment of ovarian cancer and palliative care (n= 37); other conditions (n= 30); screening, prevention and risk factors (n= 27); progression or prognosis (n= 11); diagnostic techniques (n= 6); not research (n= 8); and miscellaneous (n= 17). Other papers that initially appeared to be about symptoms of ovarian cancer were excluded for the following reasons: case studies (n= 23); article, not research (n= 16); no data on symptoms (n= 6); conference proceedings with no relevant data (n= 5); not symptoms of new disease (n= 5); data relating only to childhood tumours (n= 3); letter in response to a paper (n= 2); data on delay, details of symptoms published elsewhere (n= 1); and an abstract reporting the same data as another full paper (n= 1).

Data extraction

The 24 remaining papers included data on symptoms of ovarian cancer and were data extracted independently by two reviewers (CB and JA). Any differences were resolved by discussion. In two of these papers, the source of the data was unclear and these were therefore excluded.8,9 Another paper did not distinguish between symptoms experienced in benign or malignant cases.10 Therefore, the results relate to a total of 21 published articles.

Where the original authors did not conduct significance tests these have been undertaken and the results shown where appropriate.

Quantitative data synthesis

Meta-analysis was conducted to obtain a combined estimate of the proportions of women reported as asymptomatic at time of diagnosis. The standard error of the percentage P was calculated by the normal approximation:

image

Meta-analysis was performed by combining the percentages, using the inverse-variance method to pool the studies (metan command in Stata 8.2). Meta-analyses were conducted separately for studies that collected data directly from participants and studies that utilised medical notes to collect data. Meta-analyses were not conducted for groupings of symptoms as the categorisation of these were not consistent across the different studies.

Results

Twenty-one papers reported the results of systematic studies that investigated the symptoms experienced before the diagnosis of ovarian cancer. Fourteen papers collected data directly from participants. Ten of these papers used symptom checklists, which had been devised using the literature or clinical expertise,5,11–19 two used open-ended questions in a questionnaire,20,21 one was a qualitative interview study2 and another was a qualitative analysis of written communications to an ovarian cancer group newsletter.22 Seven collected data from existing medical records.7,23–28

All but one of the studies were retrospective, and the time between diagnosis and data collection for the retrospective studies varied between 2–3 weeks and 12 years. Of the 10 studies that utilised symptom checklists, 1 was prospective with data collection prior to surgery and therefore before pathological diagnosis.12 Seven were conducted within six months of diagnosis11,14–19 and the other two had median periods of 20 and 24 months after diagnosis.5,13 The two studies that used open-ended questions to collect data were conducted in newly diagnosed patients.20,21

Experience of being asymptomatic prior to diagnosis

Fifteen of the studies reported the proportion of participants who were asymptomatic prior to diagnosis. Nine of these collected data directly from participants using symptom checklists,5,11–15,17–19 two used open-ended questions to directly collect data,20,21 and four collected data ‘indirectly’ using medical records.7,23,25,26 The proportion of patients who directly reported that they were asymptomatic ranged from 5% to 10% in all studies except one in which the participants were asked whether they had any symptoms that prompted them to seek a diagnosis. This is different from asking whether they experienced symptoms—this study reported a proportion of 31.7%.17 The proportion of asymptomatic patients was around 20% when hospital medical records were used to collect data, but around 7% when general practice notes were used.7

Meta-analysis of the proportion of women reported as asymptomatic

A meta-analysis of the 11 studies that collected data directly from women and provided information on the proportion of participants without symptoms before diagnosis resulted in high heterogeneity. Excluding the study which recorded symptoms that prompted a consultation,17 and further restricting the analysis to the eight studies that collected data before or within a year of diagnosis (to increase reliability and minimise the loss of subjects), the heterogeneity was reduced to an acceptable level (χ2 value of 11.3 on seven degrees of freedom, P= 0.13). The resultant overall proportion was 7.2% (95% CI 6.2–8.2%) as shown in Fig. 1.

Figure 1.

Proportion of patients who were asymptomatic at time of diagnosis: meta-analysis of data collected from participants.

The meta-analysis of the data collected from hospital medical notes (Fig. 2) produced an overall proportion of asymptomatic patients of 22.6% (95% CI 18.1–27.0%). The overall test for heterogeneity was not statistically significant, giving a χ2 value of 1.76 on two degrees of freedom (P= 0.42).

Figure 2.

Proportion of patients who were asymptomatic at time of diagnosis: meta-analysis of data collected from hospital notes.

Frequencies of symptoms reported from quantitative studies

Most of the studies used broad descriptions of symptoms regardless of whether the data were collected directly from participants (Table 1) or indirectly from medical records (Table 2). In one paper, it was unclear how the percentages were derived and therefore they are not presented here.17 In two studies, only one symptom was investigated, pain at time of diagnosis (experienced by 62% of women)16 and urinary symptoms (experienced by 11%)24 and therefore no other information on symptomology was available. In another paper, the focus was not specifically on symptoms associated with ovarian cancer.27

Table 1.  Studies using quantitative data collection directly from participants.
Publication detailsStudy designParticipantsMain outcomes
  • Percentages presented to the greatest accuracy available.

Smith and Anderson17 (USA)Retrospective. Interviewer completed standardised questionnaires.82/107 cancers: 36 local stage, 46 distant stage. Recently diagnosed (1–3 months) cases registered on a cancer register. 1980–1982 inclusive. Aged 20–54.Symptoms prior to diagnosis (percentages not given as difficult to see where they are derived from). Most common symptoms: gastrointestinal/urinary/vaginal bleeding. Early stage disease related to fatigue/urinary problems, irregular bleeding—late associated with abdominal pain (all ns).
Flam et al.11 (Sweden)Retrospective questioning regarding symptoms.362 cancers: 172 early stage (IA–IB),190 advanced stage (IIB–IV). Recently diagnosed, questioned immediately before treatment. 1975–1976. Ages not given.Symptoms prior to diagnosis. Abdominal swelling (25.4%), GI symptoms (19.6%), abdominal pain (13.5%), vaginal bleeding (11.9%), urinary symptoms (7.2%), fatigue and/or fever (9.7%), breathlessness and/or back pain (4.9%) and others. In early disease, the proportions were asymptomatic (10.2%),** fatigue and/or fever (4.1%),** breathlessness and/or back pain (1.8%)** and others. In advanced disease, the proportions were asymptomatic (2.1%),** fatigue and/or fever (14.6%),** breathlessness and/or back pain (7.9%)** and others (*CB calculated P < 0.05; **P < 0.01).
Portenoy et al.16 (USA)Questionnaire survey of current pain and recall of pain.151/205 cancers undergoing treatment: 111 inpatients, 40 outpatients 1992–1993. Mean age 54 (SD 13), median 55, range 23–86.Pain associated with onset of disease. Pain (62.3%), mostly abdominal/lower back and genital.
Igoe13 (USA)Retrospective 21-item symptom checklist with room for narrative.Convenience sample drawn from Internet Web sites of 50 women with ovarian cancer. Median length of time after diagnosis = 20 months. Women aged 35–70.Symptoms prior to diagnosis. Average number of symptoms per patient = 8.6 (SD = 4.3, range 1–19) and 6% were asymptomatic. Checklist data: fatigue (82.0%), abdominal swelling (78.0%), indigestion (72.0%), lower abdomen pressure/heaviness (66.0%), abdominal pain (64.0%), back pain (56.0%), early satiety (54.0%), constipation (52.0%), weight gain (46.0%), frequent urination (40.0%), breathlessness on activity (38.0%), nausea (34.0%), loss of appetite (32.0%), fever (30.0%), vaginal bleeding (26.0%), breathlessness on rest (20.0%), lower limb swelling (18.0%), weight loss (16.0%), painful urination (12.0%) and vomiting (8.0%). In narrative section non-prompted symptoms included menstrual changes (22.0%), palpable mass (10.0%) and others.
Goff et al.5 (USA and Canada)Retrospective survey using a symptom checklist.1725 women subscribers to an ovarian cancer magazine who returned survey: 1225 stage III–IV. Median time since diagnosis 24 months. Median age 52 years, range 18–84 (at time of questionnaire, not diagnosis).Symptoms prior to diagnosis. Increased abdominal size (61%), abdominal bloating (57%), fatigue (47%), abdominal pain (36%), indigestion (31%), urinary frequency (27%), pelvic pain (26%), constipation (25%), urinary incontinence (24%), back pain (23%), pain on intercourse (17%), unable to eat normally (16%), mass (14%), vaginal bleeding (13%), weight loss (11%), nausea (9%), postcoital bleeding (3%), DVT (1%) and diarrhoea (1%). Asymptomatic in stage I/II (11%) versus 3% in stage III/IV, P= 0.001.
Olson et al.15 (USA)Retrospective case–control study. Interviewer-led symptom checklist.168 new cases, 251 population controls. Of the cases, 37 stage I/II, 118 stage III/IV (diagnosed 1994–1997). Median time since diagnosis 4.7 months. Aged 18+.Symptoms experienced in the 6–12 months before diagnosis. Frequencies for cases shown, plus OR and 95% CI compared with controls 7% of cases were asymptomatic during the 6–12 months before diagnosis, compared with 58% of controls. Bloating, fullness and pressure in abdomen or pelvis (70.8%), OR = 25.3 (15.6–40.9); abdominal or lower back pain (51.8%), OR = 6.2 (4.0–9.6); unusual lack of energy (42.9%), OR = 3.9 (2.5–6.1); frequent urination, urgency or burning (32.7%), OR = 3.5 (2.2–5.7); constipation (21.4%), OR = 3.5 (2.0–6.3); lack of appetite (20.2%), OR = 8.8 (4.3–18.2); diarrhoea (16.1%), OR = 3.0 (1.6–5.7); nausea (12.5%), OR = 1.5 (0.8–2.8); other (35.1%), OR = 6.6 (3.9–11.2). Percentages for early stage and late stage, respectively, plus OR and 95% CI late stage versus early stage. Asymptomatic 11% early stage and 7% late stage. Bloating, fullness and pressure in abdomen or pelvis (64.9% and 71.2%), OR = 1.3 (0.61–2.9); abdominal or lower back pain (48.6% and 51.7%), OR = 1.1 (0.54–2.4); unusual lack of energy (27.0% and 45.8%), OR = 2.3 (1.0–5.1); frequent urination, urgency or burning (35.1% and 31.4%), OR = 0.87 (0.4–1.9); constipation (29.7% and 21.2%), OR = 0.64 (0.28–1.5); lack of appetite (8.1% and 22.0%), OR = 3.2 (0.96–10.7); diarrhoea (27.0% and 13.6%), OR = 0.42 (0.18–1.0); nausea (13.5% and 11.0%), OR = 0.79 (0.26–2.4); other (21.6% and 39.0%), OR = 2.2 (0.92–5.2).
Vine et al.19 (USA)Cases from an existing case–control study. Standardised interview with symptom checklist.767 ovarian cancer cases: 616 invasive, 151 borderline. New diagnoses 1994–1998. Women were aged 20–69.Symptoms experienced prior to diagnosis. Asymptomatic (8% of invasive vs 16% borderline, P < 0.01). Pelvic discomfort (70%: 71% of invasive vs 66% borderline, P < 0.05); bowel irregularity (45%: 47% invasive, 35% borderline, P < 0.01); urinary frequency/urgency (37%: 37% invasive, 36% borderline, ns).
Vine et al.18 (USA)Case–control study. Interviewer-led symptom checklist.267 new ovarian cancer cases, 317 population controls. 200 invasive cancers: 65 stage I/II, 135 stage III/IV; 67 borderline. Newly diagnosed cases 1999–2001. Cases aged 20–74, mean 53.9.Symptoms experienced for 2 weeks during the year before diagnosis/interview. Frequencies for cases shown, plus OR and 95% CI compared with controls. Pelvic/abdominal discomfort/pain (64%), OR = 16.2 (10.3–25.3); bloating or feeling full (62%), OR = 14.6 (9.4–22.8); distended/hard abdomen (59%), OR = 29.2 (16.5–51.8); gas, nausea, indigestion (47%), OR = 4.7 (3.2–6.9); fatigue (46%), OR = 3.9 (2.7–5.6); weight change (45%), OR = 14.2 (8.2–24.5); urinary frequency/urgency (39%), OR = 3.5 (2.4–5.2); bowel irregularity (33%), OR = 3.5 (2.3–5.4); pain during intercourse (22%), OR = 5.8 (3.2–10.6); abnormal bleeding (20%), OR = 6.3 (3.3–12.1); abnormal bleeding aged <55 (28%), OR = 5.3 (2.6–10.7); chest pain/respiratory difficulty (18%), OR = 2.2 (1.4–3.7); other (45%), OR = 6.9 (4.5–10.7). Percentages for early stage invasive and late stage invasive, respectively, plus OR and 95% CI late stage versus early stage (excluding borderline). Asymptomatic 11% and 2%, OR = 0.2 (0.0–0.8). Pelvic/abdominal discomfort/pain (62% and 67%), OR = 1.3 (0.7–2.4); bloating/feeling full (58% and 64%), OR = 1.2 (0.7–2.3); distended/hard abdomen (51% and 67%), OR = 2.0 (1.1–3.7); gas, nausea, indigestion (35% and 49%), OR = 1.7 (0.9–3.2); fatigue (35% and 50%), OR = 1.8 (1.0–3.3); weight change (35% and 50%), OR = 1.9 (1.0–3.4); urinary frequency/urgency (45% and 36%), OR = 0.7 (0.4–1.3); bowel irregularity (26% and 34%), OR = 1.5 (0.8–2.8); pain during intercourse (20% and 19%), OR = 0.9 (0.4–1.9); abnormal bleeding (22% and 15%), OR = 0.6 (0.3–1.4); abnormal bleeding aged <55 (28% and 23%), OR = 0.7 (0.3–1.6); chest pain/respiratory difficulty (15% and 20%), OR = 1.4 (0.6–3.0); other (48% and 44%), OR = 0.9 (0.5–1.5). The authors calculated the number of women with combinations of three symptoms—see text.
Chan et al.20 (Hong Kong)Open-ended questionnaire study in newly diagnosed ovarian cancer patients.80/87 patients: 43 early stage (stage I/II), 37 late stage (stage III/IV). Newly diagnosed 1999–2000 from one cancer centre. Aged 18–70, median age 44.Symptoms experienced prior to diagnosis. Abdominal pain/ discomfort (32.5%), abdominal distension (31.3%), menstrual symptoms (18.8%), palpable mass (16.3%), bowel (15.0%), urinary (5.0%). For those with early stage disease the data were as follows: asymptomatic (11.6%), abdominal pain/discomfort (25.6%), abdominal distension (32.6%), menstrual symptoms (23.3%), palpable mass (20.9%), bowel (7.0%)* and urinary (2.3%). For those with late stage disease, the data were as follows: asymptomatic (8.1%), abdominal pain/discomfort (40.5%), abdominal distension (29.7%), menstrual symptoms (13.5%), palpable mass (10.8%), bowel (24.3%)* and urinary (8.1%). *Difference between early and late stage disease P < 0.05.
Koldjeski et al.14 (USA)Retrospective symptom checklist. Part of a longitudinal study of impact of ovarian cancer.19/20 women with newly diagnosed ovarian cancer: 6 early stage (I and II), 13 late stage (III and IV). Recently diagnosed cases (previous 2–3 weeks). 1998–2000. Aged 28–73 (mean = 57.5).Symptoms prior to diagnosis. Bloating (84.2%), vague abdominal pain (68.4%), indigestion (63.2%), fatigue (57.9%), painful spots in abdomen (52.6%), lumpy abdomen (52.6%), urinary (42.1%), nausea (26.3%), respiratory problems (26.3%), abdominal fullness (21.1%), constipation (21.1%), fluid in abdomen (21.1%), lower abdominal pain (15.8%), weight loss (15.8%), bleeding/spotting (10.5%), fluid in lungs (10.5%), weight gain (10.5%), postcoital bleeding (5.3%), back pain (5.3%), elevated temperature (5.3%), insomnia (5.3%), leg cramps (5.3%) and flulike virus (5.3%). Symptoms in early stage disease started with bloating, indigestion, fatigue, lumps and vague pains in the abdomen, urinary problems and painful lumps in the abdomen. Women with late stage disease also experienced these symptoms but additionally suffered from the other symptoms listed (no figures given, presumably due to small numbers).
Webb et al.21 (Australia)Part of a case–control study. Open-ended questions regarding up to four symptoms. Later categorised in eight broad types.811/821 cancer: 146 borderline; 218 early stage (I and II); 447 advanced stage (III and IV). Sample of newly diagnosed cases from early 1990s. No information given regarding age.Symptoms prior to diagnosis. Abdominal pain/pressure (44%), abdominal swelling/tightening (39%), gastrointestinal (15%), abdominal mass (12%) and gynaecological (12%). Abdominal symptoms were the first symptom noticed by the majority of women. In cases with borderline cancer, 16% did not have any symptoms.*** Abdominal pain/pressure (43.9%), abdominal swelling/tightening (42.3%),* mass (13.0%),*** gastrointestinal (11.5%),* gynaecological (15.5%),* general malaise (4.9%),** other (5.7%) and urinary (4.9%).* In early disease, 7% were asymptomatic.*** Abdominal pain/pressure (46.0%), abdominal swelling/tightening (31.2%),* mass (19.3%),*** gastrointestinal (11.9%),* gynaecological (15.4%),* general malaise (9.4%)** and urinary (11.4%).* In advanced disease, 4% were asymptomatic.*** Abdominal pain/pressure (42.6%), abdominal swelling/tightening (40.9%),* mass (8.8%),*** gastrointestinal (17.9%),* gynaecological (9.5%),* general malaise (13.0%)** and urinary (5.6%)* (comparison between stages *P < 0.05, **P < 0.001, ***P < 0.0001).
Goff et al.12 (USA)Case–control study versus benign tumour group and clinic attenders. Prospective symptom checklist for experiences in the previous year.44 malignant ovarian cancers: 11 early stage (stage I/II), 33 late stage (stage III/IV). 84 benign tumours (including borderline tumours) and 1011 clinic attenders. Women going through diagnosis in 2001–2002. Aged 15–90, median age 45.Symptoms in 12 months prior to diagnosis. Frequencies for cases shown, plus OR and 95% CI compared with controls. Clinic controls: increased abdominal size (63.6%), OR = 7.4 (3.8–14.2); abdominal mass (% not given), OR = 5.4 (2.4–12.0); bloating (68.2%), OR = 3.6 (1.8–7.0.); difficulty eating (% not given), OR = 2.5 (1.3–5.0); urinary urgency (% not given), OR = 2.5 (1.3–4.8); abdominal pain (50.0%), OR = 2.3 (1.2–4.4); pelvic pain (40.9%), OR = 2.2 (1.2–3.9). Benign tumour controls: bloating (68.2%), OR = 3.5 (1.5–8.2); urinary urgency (% not given), OR = 3.5 (1.6–8.2); constipation (50.0%), OR = 3.5 (1.5–8.1); increased abdominal size (63.6%), OR = 3.0 (1.3–6.9). Comparisons were made between early and late stage disease (small numbers). Only significant difference was 9% of early stage women experienced indigestion compared with 45% of late stage (P= 0.03).
Table 2.  Summary table of research utilising medical records.
Publication detailsStudy designParticipantsMain outcomes
  • Percentages presented to the greatest accuracy available.

Wikborn et al.28 (Sweden)Retrospective record audit.160/162 cancers diagnosed 1981–1986. Mean age 62.6 years (range 25–87 years).Symptoms recorded in hospital notes. Pain (66%), abdominal swelling (59%), general symptoms—fatigue, nausea, etc. (55%), GI (48%), urinary (37%) and gynaecological (21%).
Petignat et al.26 (Switzerland)Retrospective cancer registry data. Symptoms were recorded from all sources of notes.119 women registered as having ovarian cancer: 27 early stage disease (IA–IB), 92 advanced stage (IC–IV). Diagnosed 1989–1995. Median age 61 years, range 26–89.Symptoms from hospital notes. Asymptomatic: 19% of early stage, 20% of advanced. Of those with symptoms (n= 96), abdominal symptoms (76.0%), gastrointestinal (34.4%), general (29.2%), urinary (25.0%) and gynaecological (7.3%). In early disease (IA and IB), the frequencies were abdominal (72.7%), gastrointestinal (31.8%), general (9.1%),* urinary (36.4%) and gynaecological (13.6%). In advanced disease (IC–IV): abdominal (77.0%), gastrointestinal (35.1%), general (35.1%),* urinary (21.6%) and gynaecological (5.4%) (*CB calculated P < 0.05).
Eltabbakh et al.23 (Canada)Retrospective review of case notes.72 cancers: 50 stage I/II, 22 borderline diagnosed 1984–1999. Mean age 51.6 years, range 16–89.Symptoms reported in hospital notes at time of presentation. Asymptomatic: 22.2%, 18.0% of early cancers and 31.8% of borderline. Abdominal/pelvic pain (34.7%), bloatedness (31.9%), vaginal bleeding (19.4%), bowel change (16.7%), urinary symptoms (16.7%), abdominal pressure (11.1%), weight gain (6.9%), weight loss (5.6%), more than one symptom (43.1%). 72.2% had a palpable mass on exam and 12.5% had ascites. CA125 values were known in 46 women (64%) and were higher than 35 U/mL in 24 (52.5% of those with a CA125 result). Stage I/II cancers: abdominal/pelvic pain (38.0%), bloatedness (28.0%), vaginal bleeding (22.0%), bowel changes (20.0%), urinary symptoms (18.0%), abdominal pressure (10.0%), weight gain (6.0%), weight loss (6.0%) and more than one symptom (46.0%). Borderline cancers: abdominal/pelvic pain (27.3%), bloatedness (40.9%), vaginal bleeding (13.6%), bowel changes (9.1%), urinary symptoms, frequency or dysuria (13.6%), abdominal pressure (13.6%), weight gain (9.1%), weight loss (4.5%) and more than one symptom (36.4%). There were no significant differences in the frequencies of recorded symptoms between borderline and early stage ovarian cancer, but this could be due to small numbers.
Nelson et al.25 SwedenRetrospective data from hospital notes.152 ovarian cancers: 91 stage I/II, 59 stage III/IV. Diagnosed 1989–1991. Young women (15–40 years of age).Symptoms recorded at time of presentation. Abdominal/back pains (32.7%), pelvic symptoms (e.g. bloating/palpable mass; 23.5%), gynecological–urological symptoms such as dysuria, abnormal bleeding, discharge (12.0%), general symptoms (e.g. dizziness; 5.3%). For those with early stage disease (I and II) the data were: asymptomatic (30.8%), abdominal/back pains (28.6%), pelvic symptoms, including bloating and palpable mass (22.0%), gynaecological–urological symptoms (14.3%) and general symptoms (4.4%). In advanced disease (III and IV), trigger symptoms were: asymptomatic (20.3%), abdominal/back pains (39.0%), pelvic symptoms (25.4%), gynaecological–urological symptoms (8.5%) and general symptoms (6.8%). No significant differences between early and advanced disease, possibly due to small numbers.
Goldberg et al.24 (USA)Retrospective review of hospital notes and telephone contact. Unclear when phone contact was made.52 consecutive cancer patients, all early stage (I and II). 43 were contacted by telephone. Surgery 1996–1997. Mean age 54.Presenting symptom recorded in notes. 11% reported urinary frequency or bladder pressure as the presenting symptom. 11% confirmed that the presenting symptom was urinary problems during the telephone contact.
Takeuchi et al.27 (Japan)Retrospective analysis of medical records of ovarian tumour patients.42 ovarian cancer cases: 38 non-pregnant, 6 pregnant. Diagnosed 1991–2001. Aged 15–44 years.Symptoms recorded in medical notes prior to surgery. Pain (83%). In non-pregnant women (who are not routinely offered ultrasound), symptoms were pain (89.5%) and for pregnant women pain (16.7%). This difference is likely to be due to routine U.S. examination during pregnancy and therefore enhanced detection of asymptomatic cancers.
Kirwan et al.7 (UK)Retrospective review of general practice records of ovarian cancer patients.135 women: 38 early stage disease (I and II), 78 late stage, 19 not staged. Diagnosed 1992–1994.Primary symptoms recorded in GP notes. Abdominal pain (48.1%), bowel change (25.2%), abdominal swelling (18.5%), vaginal bleeding (11.1%), weight loss (8.1%), backache (2.2%).

Overall, the symptoms most frequently reported by at least half the respondents were abdominal pain or discomfort, abdominal bloating and abdominal swelling. It is important to note that, while both abdominal swelling and abdominal bloating have been categorised separately in determining the most frequently cited symptoms, they were reported independently in only three papers.5,12,18 In all other papers, either one or other of these symptoms was reported, and it is possible that these terms are being used interchangeably. Generally, the proportions of women reported to have experienced specific symptoms were lower in the studies that utilised medical notes, which may be a reflection of the non-systematic way in which these data were recorded.

Comparison between ovarian cancer cases and controls

Case–control studies were conducted using symptom checklists in three instances.12,15,18 The highest odds ratios (OR) were for bloating (including fullness and pressure in abdomen/pelvis), OR 25.3 (95% CI 15.6–40.9); lack of appetite, OR 8.8 (95% CI 4.3–18.2); and abdominal/ lower back pain, OR 6.2 (95% CI 4.0–9.6) in one study15; and distended or hard abdomen, OR 29.2 (95% CI 16.5–51.8); pelvic/abdominal discomfort or pain, OR 16.4 (95% CI 10.3–25.3); and bloating or feelings of fullness, OR 14.6 (95% CI 9.4–22.8) in the second study.18 The third study used two control groups: one was women that attended a primary care clinic with a medical problem (i.e. not for routine check-up or mammography); and the other was women undergoing surgery to remove a pelvic mass but turned out to have a benign tumour. In this study, the odds ratios were lower (as you would expect when comparing two symptomatic samples), but a similar pattern was observed. The largest odds ratios when comparing ovarian cancer with clinic controls were obtained for increased abdominal size, OR 7.4 (95% CI 3.8–14.2); abdominal mass, OR 5.4 (95% CI 2.4–12.0); and bloating, OR 3.6 (95% CI 1.8–7.0). Using benign tumour controls, the highest odds ratios were observed for bloating, OR 3.5 (95% CI 1.5–8.2); urinary urgency, OR 3.5 (95% CI 1.6–8.2); and constipation, OR 3.5 (95% CI 1.5–8.1).12

Symptoms in early and advanced stages

Only one paper reported similar rates in the proportion of women who were asymptomatic prior to diagnosis with early or late stage ovarian cancer.26 The other papers that reported these data tended to show a lower proportion of asymptomatic women in late stage than in early stage cancers,5,11,15,18,20,21,25 although only three of these differences reached statistical significance.5,11,18

Six papers which used symptom checklists, or open-ended questionnaires, provided details of frequencies of symptoms experienced by women diagnosed with early and late stage disease.11,12,15,18,20,21 Symptoms that were reported significantly more frequently in advanced cancer were gastrointestinal symptoms,11,12,20,21 unusual fatigue,15,18 abdominal distension18,21 and weight change.18 There was a tendency for urinary symptoms to be reported more frequently by women with early stage disease than late stage disease in some papers,11,18,21 but these differences were only significant in one paper.21 Gynaecological symptoms (not specified) and pelvic mass were more frequently reported in early stage cancers than advanced stage.21 Similar patterns were observed in both the studies that extracted data from medical records and compared symptoms in early and late stage disease.25,26

Invasive cancer and borderline tumours

A lower proportion of women with invasive cancers were asymptomatic than those with borderline cancers. This was true when all types of invasive cancer were included19,21 and when the research focussed on early stage cancers.21,23

Significant differences were observed in the proportions experiencing pelvic discomfort, 71% of invasive cancers compared with 66% of borderline cancers (P < 0.05) and in bowel irregularity, which was reported in 47% of invasive cancers and 35% of borderline cases (P < 0.01).19 In one study, malaise was reported significantly more frequently by women with invasive cancer compared with women with borderline tumours.21

Another smaller study utilising medical notes showed no statistically significant differences between borderline and invasive cancers, although some quite large differences were evident. For example, the presence of the following symptoms was reported in the medical notes more frequently in invasive cancer than in borderline cancers: abdominal or pelvic pain (38%vs 27%); vaginal bleeding (22%vs 14%); and bowel irregularities (20%vs 9%). Conversely, abdominal bloating was recorded less frequently in the notes of women with invasive cancers compared with those with borderline tumours (28%vs 41%).23

Combinations of symptoms

Two studies examined combinations of symptoms. One looked at clusters of three symptoms and produced a set of 10 different combinations, which only included the following: bloating or fullness; distended or hard abdomen; pelvic or abdominal discomfort; fatigue; gas/nausea/indigestion; and weight change. Each of these 10 combinations were reported by 26–39% of cases, compared with 2% or less of controls (P < 0.01).18 In the other study, 43% of ovarian cancer cases experienced bloating, increased abdominal size and urinary urgency compared with 10% of those with benign masses, OR 5.3 (95% CI 2.2–12.6), and 8% of the clinic controls, OR 9.4 (95% CI 5.0–17.7). The addition of concurrent constipation marginally altered the odds ratios to 6.2 (95% CI 2.0–18.8) for the benign tumour control comparison and 8.6 (95% CI 4.2–17.4) in clinic patients.12

Results from the qualitative research papers

Proportions of women reporting salient symptoms are not given here (Table 3), as the samples were highly selected and not representative of all ovarian cancer patients. However, both papers utilising qualitative research highlighted that the diagnosis of ovarian cancer is difficult, as women tend to attribute symptoms to the normal ageing process or other, less serious, conditions2 or that despite symptoms being present ‘it is not a silent killer, more that the medical profession do not listen’.22

Table 3.  Qualitative research studies investigating symptoms associated with diagnosis of ovarian cancer.
Publication detailsStudy designParticipantsMain outcomes
Fitch et al.2 (Canada)Qualitative semistructured telephone interviews of ovarian cancer patients.Convenience sample of 18 women from two cancer centres. 12 women had been diagnosed in the previous 12 months, but range was up to 12 years after diagnosis. Aged 35–73 (mean 53.2).The interviews identified that diagnosis is difficult in ovarian cancer and that many women did not recognise the symptoms as serious and often attributed them to normal body changes associated with childbirth, menopause, ageing, etc. Common symptoms reported included bloating, weight gain around the middle, GI symptoms, fatigue, sexual problems and pain. Symptoms were often vague and mimicked gastric disorders.
Ferrell et al.22 (USA)Ethnographic qualitative analysis of 21,806 letter, cards and emails sent to the editor of a newsletter for ovarian cancer patients (same newsletter as used by Goff).21,806 letters, cards and emails that contained 677 comments relating to symptoms. Of these, 230 were pre-diagnostic symptoms (no data on the number of patients). Participants were people who had written to the newsletter for ovarian cancer patients between 1993 and 2000. No data on time since diagnosis.Reports of pre-diagnostic symptoms. A total of 230 comments were made regarding pre-diagnostic symptoms: some were only recognised after a diagnosis was given others were obvious and led to the diagnosis (often delayed). Bloating/abdominal swelling was commented on 56 times, fatigue 33 times, abdominal/pelvic pain 24 times, urinary frequency 20 times, back pain 11 times, decreased appetite 10 times, vaginal bleeding 10 times, indigestion 8 times, constipation 8 times,bowel changes/difficulty 7 times, stomach/pelvic cramping 5 times, palpable mass 5 times, diarrhoea 5 times. Other symptoms were mentioned less than five times. The general feeling was that ovarian cancer is not a silent killer, more that the medical profession do not listen.

Discussion

Research focussing on symptoms experienced prior to diagnosis of ovarian cancer has increased over the last few years, but the results of the research have not led to identification of salient predictive symptoms. The reasons for this could be the retrospective nature of most of the studies and the long duration between diagnosis and data collection. The ability to accurately recall symptoms that were experienced before diagnosis decreases with time and is less reliable. Prolonged time frames from diagnosis to data collection introduce inherent patient bias due to the poor survival rates of ovarian cancer. The recently published paper by Goff et al.12 adds greatly to the existing literature as it prospectively collected the information.

Several studies have relied solely on the symptoms that are recorded in the medical records. This is likely to underrepresent the experiences of patients, as only those which are deemed salient by the clinician tend to be documented. An indication that this occurs is the significantly different proportions that were deemed to be asymptomatic in the data collected directly from women themselves (7.2%) compared with the hospital medical records (22.6%). In addition, clinic data are unlikely to record the quality of these symptoms (such as severity, duration and subtleties), as this was not the original purpose of recording these events in the notes.

Studies have attempted to systematically record symptoms that are experienced prior to diagnosis, but the data collection techniques that have been used are frequently unvalidated symptom checklists developed by the investigators. Although these checklists have been derived using the existing literature on ovarian cancer symptomology, these may miss previously unrecorded events, which may be the key to this particular diagnostic dilemma. Also, subtle differences in experiences will not be identified when using crude checklists.

The limitations of research to date may explain the lack of definitive results. An alternative approach of collecting detailed data such as semistructured interviews with women newly diagnosed with ovarian cancer may yield more exact and definable symptoms while avoiding investigator bias. This could facilitate a more rapid and accurate referral and diagnosis of patients with suspected disease.

Acknowledgments

The authors would like to thank Christelle Kervella and Anja Tusold for their translations. Funding for this research has been provided by the English Department of Health and Cancer Research UK. The researchers and this research are independent from the funding bodies. The authors have no conflicts of interest. STK undertook no role in the journal's editorial process for this manuscript.

Ancillary