Factors influencing uptake and timing of risk reducing salpingo-oophorectomy in women at risk of familial ovarian cancer: a competing risk time to event analysis

Authors


Dr R Manchanda, Gynaecological Cancer Research Centre, EGA Institute for Women’s Health, First floor, Maple House, 149 Tottenham Court Road, London W1T 7NF, UK. Email r.manchanda@ucl.ac.uk

Abstract

Please cite this paper as: Manchanda R, Burnell M, Abdelraheim A, Johnson M, Sharma A, Benjamin E, Brunell C, Saridogan E, Gessler S, Oram D, Side L, Rosenthal A, Jacobs I, Menon U. Factors influencing uptake and timing of risk reducing salpingo-oophorectomy in women at risk of familial ovarian cancer: a competing risk time to event analysis. BJOG 2012;119:527–536.

Objective  To evaluate factors affecting uptake of risk-reducing salpingo-oophorectomy (RRSO) over time in women at high-risk of familial ovarian cancer.

Design  Prospective observational cohort.

Setting  Tertiary high-risk familial gynaecological cancer clinic.

Population/sample  New clinic attendees between March 2004 and November 2009, fulfilling the high-risk criteria for the UK Familial Ovarian Cancer Screening Study.

Methods  Risk management options discussed included RRSO and ovarian surveillance. Outcome data were analysed from a bespoke database. The competing risk method was used to model the cumulative incidence function (CIF) of RRSO over time, and the sub-hazard ratio (SHR) was used to assess the strength of the association of variables of interest with RRSO. Gray’s test was used to evaluate the difference in CIF between two groups and multivariable competing risk regression analysis was used to model the cumulative probabilities of covariates on the CIF.

Results  Of 1133 eligible women, 265 (21.4%) opted for RRSO and 868 (69.9%) chose screening. Women undergoing RRSO were older (49 years, interquartile range 12.2 years) than those preferring screening (43.4 years, interquartile range 11.9 years) (P < 0.0005). The CIF for RRSO at 5 years was 0.55 (95% CI 0.45–0.64) for BRCA1/2 carriers and 0.22 (95% CI 0.19–0.26) for women of unknown mutation status (P < 0.0001); 0.42 (95% CI 0.36–0.47) for postmenopausal women (P < 0.0001); 0.29 (95% CI 0.25–0.33) for parity ≥1 (P = 0.009) and 0.47 (95% CI 0.39–0.55) for a personal history of breast cancer (P < 0.0001). Variables of significance from the regression analysis were: a BRCA1/2 mutation (SHR 2.31, 95% CI 1.7–3.14), postmenopausal status (SHR 2.16, 95% CI 1.62–2.87)) and a personal history of breast cancer (SHR 1.5, 95% CI 1.09–2.06).

Conclusions  Decision-making is a complex process and women opt for surgery many years after initial risk assessment. BRCA carriers, postmenopausal women and women who had breast cancer are significantly more likely to opt for preventative surgery.

Introduction

Mutations in the BRCA1/2 genes contribute to most of the known ovarian cancer risk in women at increased risk for familial ovarian cancer, with a number of moderate to low penetrance variants accounting for the residual familial risk. Women carrying a BRCA1 or BRCA2 mutation have up to a 49–65% risk of developing breast cancer and an 18–40% risk of developing ovarian cancer up to 70 years of age.1,2 Higher penetrance estimates have been reported in series of high-risk families with multiple incidences of cancer ascertained through genetic clinics.3–6

Risk-reducing salpingo-oophorectomy (RRSO) has been shown to be the most effective option for preventing tubal/ovarian cancer, with a hazard ratio of 0.21 (95% CI 0.12–0.39)7 having been reported in a meta-analysis of known BRCA carriers. Oophorectomy has also been found to halve the risk of subsequent breast cancer in premenopausal women who have not undergone prophylactic mastectomy.7 Screening for ovarian cancer in this population is still of unproven benefit and is currently recommended only within the context of a research study. The advantage of reduction in ovarian cancer risk with RRSO must be weighed against the as yet unproven benefit of screening in this population and the anxiety associated with false-positive surveillance results as well as the potential surgical risks8–10 and the residual risk of primary peritoneal cancer.11 Despite the lack of evidence of benefit, many women opt for screening and RRSO uptake rates have been found to vary considerably within centres as well as between countries.12,13

In addition, in premenopausal women RRSO leads to the onset of premature menopause and the loss of subsequent fertility. Premature menopause has been associated with a higher risk of cardiovascular disease,14–16 potential cognitive impairment and Parkinsonism,17–19 osteoporosis, vasomotor symptoms and detrimental impact on quality of life.20,21 A potential impact on mortality22 has also been described. Risks seem to be higher for women who undergo the procedure under the age of 45 years and who do not take hormone replacement therapy.21,22 Hence, the timing of surgery is of significant importance and the choices that high-risk women make may change over time. However, only three of the previous reports evaluating uptake of preventative surgery in BRCA carriers report a time-to-event analysis.23–25 A study of 306 Dutch BRCA1/BRCA2 carriers found a 75% RRSO uptake rate over a 10-year period.24 A study from Chicago found a 70% uptake over a 7-year period in 88 BRCA1/BRCA2 carriers.25 A Manchester-based study of 212 BRCA1/2 carriers reported higher uptake in BRCA1 (52%) than in BRCA2 carriers (28%) over a 7-year period.23 The median time to surgery in these studies varied from 12.5 to 34 months.

Here we undertake a time-to-event analysis to report on the factors affecting uptake of RRSO in high-risk women attending a tertiary multidisciplinary gynaecological familial cancer clinic. The uniqueness of our cohort includes the presence of a large number of women from high-risk families for whom genetic testing is unavailable in the UK because of the absence of a live affected relative. Moreover, for the first time in such an analysis we use a competing risk method, which reduces the potential bias related to censoring that is associated with the Kaplan–Meier26 and standard Cox27 models in earlier reported time-to-event analyses.

Methods

The familial gynaecological cancer clinic at University College London Hospital (UCLH) is a tertiary-level clinic for managing women at ‘high-risk’ for familial gynaecological cancer. Women were identified from the clinic’s bespoke database as high-risk on the basis of the inclusion criteria (family history/mutation status) for the United Kingdom Familial Ovarian Cancer Screening Study (UKFOCSS) (see Supplementary material, Figure S1).9,28BRCA gene testing within the UK National Health Service is primarily available to cancer-affected individuals from high-risk families (≥20% carrier probability) or individuals from a family with a confirmed BRCA mutation. Hence, a number of high-risk women in the UK are unable to access gene testing and are of unknown mutation status.

Women are managed within the context of a multidisciplinary team, which includes gynaecological oncologists, a radiologist, a clinical geneticist, a clinical psychologist, a clinical nurse specialist, minimal access gynaecologists and a pathologist.9 All women attending the clinic undergo a pedigree-based clinical risk assessment and receive comprehensive advice on the advantages and disadvantages of RRSO and ovarian cancer screening as well as reproductive and lifestyle issues. The primary recommendation for high-risk women is RRSO after the age of 40 years, if her family is completed. Premenopausal women undergoing surgery are generally advised short-term hormone replacement therapy until the age of 50 years. Screening for ovarian cancer is available in the context of a national trial, UKFOCSS, for those >35 years.

Before RRSO, all high-risk women undergo a preoperative serum CA125 measurement and transvaginal ultrasound scan. Surgery involves removal of both tubes and ovaries (or all remaining adnexae in women who have undergone previous partial removal), peritoneal washings for cytological examination and endometrial sampling.

Prospectively collected demographic, clinical and pathology data were stored in a bespoke database and used for the current analysis. Where necessary, hospital case notes as well as pathology records were reviewed. The database was searched for high-risk women from breast and ovarian cancer families who had their first clinic visit between April 2004 and November 2009. Women who had amenorrhoea for 12 months (excluding those with a medical or physiological explanation such as Mirena intrauterine system, hormonal therapy or breastfeeding) were considered to be postmenopausal.

Statistical analysis

The effect of individual variables on RRSO was initially evaluated using univariate analysis. The Mann–Whitney non-parametric test was used to compare age distributions between groups after reviewing histograms. Chi-square test with Yates’ continuity correction and Fisher’s exact test were used to calculate the difference between proportions. Two-sided P-values are reported for all statistical tests.

Competing risk analysis

In a competing risks setting, the main disadvantage of standard survival analysis methods relates to censoring. Popular methods, such as the Kaplan–Meier estimator or the Cox proportional hazards model assume that censoring is non-informative and independent.26,27 Patients who withdraw or are lost to follow up during the study are classified as censored and are assumed to have the same risk of RRSO (event of interest) as others who are alive and have not undergone RRSO at the end of the study. However, women who undergo a competing risk event such as death from unrelated causes during the study are also treated as censored within a Kaplan–Meier analysis. However, as they are deceased they are no longer at risk of RRSO. Some women who are undergoing screening will have to undergo surgery because of a screen-detected abnormality. This would not be true ‘prophylactic surgery’. Hence, there may be a number of reasons (competing risks) as a result of which women cannot subsequently opt for RRSO. Using the traditional Kaplan–Meier product-limit method in such situations gives a false or over-inflated picture of the cumulative incidence of the event of interest (RRSO). Hence, in the presence of competing risks, instead of the traditional Kaplan–Meier method26 we have used a competing risk/actuarial cumulative incidence analysis that considers cause-specific hazard functions (i.e. for each competing risk separately).29

In this analysis, we have used a competing risk method to model the cumulative incidence function (CIF) of RRSO over time. The CIF gives the cumulative probability of occurrence of a particular event type in the presence of other (=competing) events and is a function of both the survival function and cause-specific hazard function at time t. Withdrawals because of death; bilateral salpingo-oophorectomy resulting from a screen-detected abnormality; and negative genetic test for a known predisposing mutation in the family were treated as competing risks. Individuals were censored at the point of all other reasons for withdrawal or at last follow up (study end).

The impact of individual variables on the CIF was calculated for RRSO and competing risk events. In addition to CIF plots for different factor levels, the significance was assessed univariately using Gray’s test for subhazard distributions. This is similar to the familiar log-rank test, except that in the latter test a woman with a competing event would exit the ‘at-risk’ set whereas in Gray’s test the woman remains ‘at-risk’ forever.30,31

It is expected that many of the identified covariates will be correlated, and provide similar information. We chose to identify those factors that have a uniquely strong relationship with time to RRSO. A competing risks version of the Cox proportional model allows a regression of multiple variables on time to RRSO. In this model, the exponentiated coefficients are known as the subhazard ratios (SHR), and were used to assess the strength of association for a variable with the primary event’s subhazard distribution, which is directly related to the CIF. As with the standard Cox model, the assumption of proportional subhazards means the effects of the SHRs work multiplicatively on the baseline subhazard. Selection of variables was via a forward stepwise regression with inclusion set at P = 0.05 and exclusion at P = 0.1. This analysis was undertaken using stata 11.0. (StataCorp LP, College Station, TX USA) and the ‘cmprsk’ package written for R [CRAN, License GPL (>= 2), B Gray, Harvard, USA]. Two-sided P-values are reported for all statistical tests.

Results

Between April 2004 and November 2009, 1241 high-risk women from breast and ovarian cancer families attended clinic (initial visit) and underwent risk assessment and counselling. Of these, 108 (8.7%) were <35 years old and deferred decision making. Of the remaining 1133 women by November 2009, 265 (21.4%) underwent RRSO and 868 (69.9%) opted for screening within UKFOCSS. Of the women being screened, 105 (12.1%) withdrew during the study period. Of these, 43 (4.95%) underwent surgery for a screen-detected abnormality, 27 (3.1%) tested negative for a known familial BRCA mutation, nine (1%) moved residence, ten (1.1%) changed their mind and 16 (1.8%) gave no reason for withdrawal. Detailed characteristics of the cohort are described in Table 1.

Table 1.   Baseline characteristics of the cohort
CharacteristicRRSO (n = 265)Screening (n = 868)P
  1. FDR, first-degree relative; FH, family history; HBC, high-risk breast cancer only family; HBOC, high-risk breast and ovarian cancer family; HOC, high-risk ovarian cancer only family; Self breast cancer, personal history of breast cancer; UMS, unknown mutation status.

  2. Using a Bonferroni correction for multiple testing the P-values should be compared with a critical value of α = 0.003.

  3. *Mann–Whitney U test.

  4. **Chi-square test.

  5. ***Fisher’s exact test.

Median age (IQR)49 (12.2)43.4 (11.9)< 0.0005*
BRCA1/2 carriers111 (41.9%)179 (20.6%)< 0.0005**
BRCA163 (23.8%)97 (11.2%)< 0.0005**
BRCA248 (16.7%)85 (7.7%)< 0.0005**
BRCA1 + 203  1***
UMS154 (58.1%)689 (79.4%)< 0.0005**
Postmenopausal138 (52.1%)     251 (28.9%)< 0.0005**
Parity ≥1189/225 (84%)624/818 (76.3%)0.013**
Jewish ancestry54/254 (18%)203/847 (19.5%)0.371**
Family history
HBC76/259 (29.3%)195/847 (23%)0.038**
HBOC150/259 (57.9%)517/847 (61%)0.369**
HOC31/259 (12%)133/846 (15.7%)0.162***
FDR breast cancer138/258 (53.5%)443/846 (52.4%)0.752**
FDR ovarian cancer125/258 (48.4%)455/845 (53.8%)0.129**
FH of ovarian cancer <50 years64/257 (24.9%)271/846 (32%)0.029**
FH of breast cancer <45 years148/257 (57.6%)466/845 (55.1%)0.491**
Self breast cancer97/258 (37.6%)157/845 (18.6%)< 0.0005**

Of the 1133 women, 157 were BRCA1 carriers, 130 were BRCA2 carriers and three carried both a BRCA1 and a BRCA2 mutation. A total of 843 women had unknown mutation status, of whom 43% were from breast cancer only families, 83% from breast and ovarian cancer families and 95% from ovarian cancer only families. Women undergoing RRSO were older (median age 49 years, interquartile range [IQR] 12.2 years) than those opting for screening (median age 43.4 years, IQR 11.9 years) (P < 0.0005). The median time to RRSO was 36.53 months (IQR 17.65–52.64 months). Initial univariate analysis showed that women who carried a BRCA1/2 mutation were postmenopausal, had a personal history of breast cancer and were from breast cancer only families were more likely to opt for RRSO over screening (Table 1).

On competing risk analysis the overall cumulative probability of undergoing prophylactic surgery in the entire cohort over 60 months was 0.29 (95% CI 0.26–0.32). The cumulative probability for undergoing RRSO at 5 years was 0.55 (95% CI 0.45–0.64) for BRCA1/2 carriers and 0.22 (95% CI 0.19–0.26) for women of unknown mutation status. Gray’s test showed this difference between BRCA carriers and women of unknown mutation status to be highly significant (P < 0.0001) for RRSO but not for competing risk events (P = 0.111) (Table 2, Figure 1). Similarly, the CIF for RRSO was significantly different between premenopausal and postmenopausal groups, women with and without a personal history of breast cancer, those with and without a history of ovarian cancer before the age of 50 years in the family, nulliparous women and those with parity ≥1, as well as between women from breast cancer only families and those from breast and ovarian/ovarian only families (Table 3). Menopausal status (P = 0.251), a personal history of breast cancer (P = 0.327), history of early onset ovarian cancer in the family (P = 0.698), parity (P = 0.396) and a family history of breast cancer (P = 0.191) did not have any significant affect on competing risk events (Figure 2A–E, respectively). We also found the CIF to be significantly different for RRSO between the age groups 30–40 years, 40–50, 50–60, 60–70 and 70–80 years (P < 0.0001, Figure 2F), but not for the competing risk events (P = 0.553).

Table 2.   Cumulative RRSO probability (CIF) by BRCA1/2 status over time
 Time (months)Gray’s test
1224364860
  1. UMS, unknown mutation status.

BRCA1/2
CIF0.2990.3810.4290.4820.545RRSO incidence BRCA1/2 versus UMS: P < 0.0001
95% CI0.243–0.3550.319–0.4430.362–0.4960.406–0.5570.449–0.641
UMS
CIF0.1140.1730.1920.2040.222Competing risk incidence BRCA1/2 versus UMS: P < 0.111
95% CI0.092–0.1360.146–0.2000.163–0.2210.174–0.2340.189–0.256
Figure 1.

 Cumulative incidence function (CIF) for RRSO and competing risk by BRCA1/2 status. BRCA yes, BRCA1/BRCA2 carriers; BRCA no, unknown mutation status; Surgery, risk-reducing salpingo-oophorectomy.

Table 3.   Cumulative RRSO probability (CIF) at 5 years
VariableCumulative RRSO probability95% CIGray’s test P-valueFigure
  1. FH, family history; h/o, history of.

Premenopausal0.230.19–0.27< 0.00012A
Postmenopausal0.420.36–0.47
Personal h/o breast cancer0.470.39–0.55< 0.00012B
No personal h/o breast cancer0.240.21–0.28
FH of early onset ovarian cancer0.230.18–0.280.0062C
No FH of early onset ovarian cancer0.320.28–0.37
Nulliparous0.200.14–0.270.0092D
Parity ≥10.290.25–0.33
FH: breast cancer only family0.350.43–0.280.0062E
FH: breast and ovarian cancer/ovarian cancer only families0.270.24–0.31
Figure 2.

 Cumulative incidence function (CIF) for RRSO and competing risk. (A) CIF for RRSO and competing risk by menopausal status; (B) CIF for RRSO and competing risk by personal history of breast cancer; (C) CIF for RRSO and competing risk by family history of ovarian cancer before 50 years of age; (D) CIF for RRSO and competing risk by parity; (E) CIF for RRSO and competing risk by high-risk breast cancer family; (F) CIF for RRSO and competing risk by age groups 30–40, 40–50, 50–60, 60–70 and 70–80 years. Surgery, risk-reducing salpingo-oophorectomy; PM, postmenopausal; parity 1+, parity ≥1; parity 0, nulliparous.

A competing risk regression assuming proportional subhazards was undertaken to identify the key covariates from Table 1 that remained significant for RRSO (Table 4). In the final model, the SHR were 2.31 (95% CI 1.7–3.14) for BRCA1/2 carriers, 2.16 (95% CI 1.62–2.87) for postmenopausal women and 1.5 (95% CI 1.09–2.06) for those with a personal history of breast cancer. The SHR of 1.43 (95%CI 0.99–2.06) for parity ≥1 neared statistical significance (P = 0.056) and remained part of the final equation (Table 4). All SHRs were greater than one, indicating that an increase (or presence) in this factor increased the subhazard and hence the CIF for RRSO.

Table 4.   Competing risk multivariable regression analysis for RRSO
CovariateSHRSEzP > z95% CI
  1. *Self breast cancer, personal history of breast cancer.

Parity ≥11.4280.2662211.910.0560.990813–2.05776
Postmenopausal2.1580.3141725.28< 0.00011.621955–2.870272
BRCA1/22.3140.3619775.36< 0.00011.70296–3.14422
Self breast cancer*1.5010.2435022.50.0121.0921–2.062774

Discussion

Our study highlights that counselling and decision making for women at high risk of familial ovarian cancer are complex processes and women continue to opt for surgery many years after their initial risk assessment. It re-emphasises the previously reported dynamic nature of decision making, which changes over time.23–25BRCA carriers, postmenopausal women and those who have had breast cancer are significantly more likely to opt for risk-reducing surgery.

The cumulative probability of undergoing prophylactic surgery in our cohort was 0.29 (95% CI 0.26–0.32). This is less than most reports in the literature, where varying RRSO uptake rates ranging from 15% to 78% have been reported, but the majority are over 48%.12 However, most of these reports include BRCA carriers in the main and are limited by not accounting for time in the analysis. The RRSO rates reported in previous time-to-event analyses vary from 45% to 75%.23–25 A significant factor accounting for our lower uptake is the larger proportion of women of unknown mutation status (CIF 22.2% at 60 months) in our cohort. This low level of uptake in untested women has been reported in one previous small series,32 and is in keeping with previous reports of a positive BRCA genetic test result being a predictor of RRSO uptake.32,33 The CIF for RRSO of 54.5% at 60 months found in BRCA1/2 carriers in our cohort is consistent with one previous time-to-event analysis23 but is lower than other reports in the literature.24,25,34 In addition to restricted access to genetic testing in the UK these differences may also be the result of heterogeneity of populations, individual preferences or other psychosocial factors. The ability to opt for a national ovarian cancer screening study at our centre may also have contributed to these results.

The strengths of our study include its large size, a mixed cohort of women with known BRCA mutations and unknown mutation status, longitudinal nature of follow up, prospectively collected data and use of the competing risk method for analysis. To the best of our knowledge ours is the largest series with high-risk women who were unable to, or chose not to, undergo genetic testing for BRCA1/2 mutations. Our study is different from previous time-to-event analyses24,25 because it helps to highlight the differences in genetic testing practices in the UK and elsewhere in the world. Our data showing that BRCA carriers are 2.3 times more likely to opt for preventative surgery suggests that uptake of prophylactic oophorectomy may vary with level of proven ovarian cancer risk. Such a finding of risk-linked uptake of preventative surgery has previously been reported for prophylactic mastectomy.23 One time-to-event analysis23 reported an increased RRSO rate in BRCA1 carriers who are known to have a higher risk compared with BRCA2 carriers.23,35 However, consistent with the Dutch study,24 and most other analyses, we did not find a significant difference in RRSO rates between BRCA1 and BRCA2 carriers (P = 0.54). Increasing access to genetic testing in the UK with resultant confirmation of risk may lead to higher RRSO uptake rates with the potential to reduce ovarian/tubal cancer incidence in high-risk women.

Our study is one of the few to explore time as a factor in the uptake of preventative surgery. Another advantage of our study is the use of competing risk methodology. Sixty (57%) of the 105 withdrawals in our study were the result of a competing risk event. These women could not have subsequently undergone RRSO. Within a routine Kaplan–Meier analysis these women would be considered at similar risk of subsequent events to other women with continued follow up. In fact most other studies do not report details of reasons for salpingo-oophorectomy. Competing risks have not been reported in three previous time-to-event analyses of preventative surgery for BRCA carriers.23–25 It is possible that this bias may have contributed to the higher rates of prophylactic salpingo-oophorectomy reported in those series.

We found that women continue to opt for RRSO many months/years after their initial decision. A significant proportion of BRCA carriers underwent surgery >12–24 months after their initial counselling appointment following the results of genetic testing (Table 2). This contrasts with most previous reports suggesting that BRCA carriers undergo surgery within a year of learning their genetic test result36–38 but is consistent with three recent time-to-event analyses24,25 indicating that BRCA carriers continue to opt for surgery many years later. Our data indicate that this finding also holds true for women with unknown mutation status, with only half of those women opting for surgery doing so within 12 months of their initial consultation (Table 2). The overall median time to RRSO in our study was greater than the Chicago study25 but similar to a Dutch study.24

Consistent with the findings of others,25,32,36,39 including two previous time-to-event analyses,24,25 we found that increasing age (Figure 2F) and having children (Figure 2D) were factors associated with RRSO uptake. The median age of women opting for RRSO in our study is slightly older than most other reports in the literature, including the Dutch and Chicago studies.24,25 The Manchester study,23 like ours, found a significant difference in RRSO uptake across age groups. However, they reported higher uptake rates with time in younger women, whereas we found an increasing RRSO uptake with increasing age (Figure 2F). Postmenopausal women in our study were 2.16 times more likely to opt for RRSO. Although menopause was not reported as an independent risk factor in previous time-to-event analyses it has been found to be of importance in other studies.34 Premenopausal women are more likely to be younger, nulliparous and have concerns regarding detrimental effects of the menopause and hence, delay surgery.23,25

Our finding of a personal history of breast cancer being associated with increased RRSO uptake was not reported in earlier time-to-event analyses23–25 but has been described by other series.13,21,38,40,41 However, in contrast with some other reports, we did not find that having a first-degree relative with ovarian cancer35,42 or a family history of early-onset breast cancer were significant predictors of RRSO. The finding that a history of early-onset ovarian cancer (<50 years) in the family was inversely associated with uptake of preventive surgery on univariate analysis is likely to be a confounding effect or chance finding because it was not maintained following multivariable competing risk regression analysis. Although the Chicago study reported a family history of ovarian cancer to be significantly associated with RRSO uptake, this was not observed in our study or the other time-to-event analyses.23,24 We did not find Jewish ethnicity to be a factor affecting uptake of risk-reducing surgery in our cohort. A lower surgical uptake has been reported in some minority populations such as African-American populations.25

Multivariable regression analysis (Table 4) indicated that the main factors affecting decision making were having a BRCA gene mutation, being postmenopausal, a personal history of breast cancer and having children. The fact that these factors did not have any significant effect on competing risk events (Figure 1, 2A–D) is reassuring because it suggests that competing risk events in the cohort occurred independently of covariates of significance. Although Gray’s test showed a significant difference in RRSO uptake across age groups, age was not part of the final model, as the effect of this variable was probably accounted for by menopausal status in the equation. It is interesting to note that the SHR for postmenopausal status was similar to that for having BRCA carrier status, indicating that the magnitude/contribution of these factors towards decision making was similar in high-risk women. Limitations of our analysis are that we lacked data on factors such as psychosocial factors, perceived risk, cancer worry and fear of surgery, which have been shown to affect uptake of preventative surgery.34,43,44 In addition, it was not possible to assess whether decision making varied depending on the individual clinician (from the familial clinic team) seen at each consultation.

The study has important implications for counselling/management and for planning/commissioning of services for women at high-risk of familial ovarian/tubal cancer, particularly in the UK and other countries with restricted access to genetic testing. It adds to the knowledge base related to factors influencing RRSO in high-risk women and the amount of time that this decision-making process can take. It also highlights the large number of high-risk families with no living cancer-affected relatives who could benefit from expanded genetic testing to clarify their cancer risks as well as access to risk management options.

Conclusion

A large number of high-risk women find bilateral salpingo-oophorectomy to be an acceptable option for reducing their risk of ovarian and tubal cancer. Decision making is a complex and dynamic process that changes over time. Women continue to opt for surgery many years after their initial counselling and risk assessment. Clinicians should pursue follow-up opportunities with their high-risk patients as many will delay decision making. A number of different factors affect uptake of risk reducing surgery in these women. Uptake of RRSO is risk dependent with lower uptake rates in high-risk women who are unaware of their mutation status. A number of women delay surgery until they have completed their families or reached the menopause. Known BRCA carriers and women who have had breast cancer are more likely to opt for preventative surgery. Recognition and appreciation of these matters can assist in planning and commissioning of services for high-risk women. Relaxation of BRCA testing criteria in the UK may lead to greater access to genetic testing, detection of more carriers and increased RRSO uptake. Risk management options need to be individualised for each woman and it is important for clinicians to be aware of these issues when counselling women at increased risk.

Disclosure of interests

IJ has consultancy arrangements with Becton Dickinson, who have an interest in tumour markers and ovarian cancer. IJ and UM have a financial interest in Abcodia, Ltd, a company formed to develop academic and commercial development of biomarkers for screening and risk prediction. IJ is a member of the board of Abcodia Ltd and Women’s Health Specialists Ltd. ANR has received honoraria from Fujirebio Diagnostics for giving lectures and attending meetings on the use of biomarkers in ovarian cancer management, but none were directly related to this work. ES received honoraria from Ethicon for provision of training to healthcare professionals; this was not related to this work. The other authors declare no conflict of interest.

Contribution to authorship

RM, AA and MJ were involved in initial data collection. RM, UM and IJ were involved in analysis, drafting and writing of the paper. MB and RM performed the statistical analysis and contributed to writing the statistical sections of the manuscript. ANR, LS, ES, AA, AS, DO, SG, MJ, CB and EB contributed to the writing of the manuscript and UM, IJ, RM, ANR, CB, LS, SG, ES, EB, AS and DO were responsible for the clinical care of the patients. The final draft was prepared by RM, UM and IJ and approved by the others.

Details of ethics approval

The project was referred to the Chair of the Research Ethics committee (National Hospital for Neurology and Neurosurgery and Institute of Neurology Joint REC, reference number 07L 173). Under the Research Governance Framework the project was deemed to be a clinical audit, and permission for data analysis and submission for publication was given.

Funding

This work has not been directly funded by any commercial organisation, charity or other sources. A large portion of this work was carried out at UCLH/UCL within the ‘women’s health theme’ of the NIHR UCLH/UCL comprehensive biomedical research centre supported by the Department of Health.

Acknowledgements

We are particularly grateful to all our patients. We acknowledge the support and help provided by The Eve Appeal and Ms S. Chamberlain (clinic secretary). We are grateful to all the surgeons who have been involved in the care of our patients (T. Mould, M. Widschwendter, A. Olaitan, N. MacDonald, A. Cutner, A. Vashisht, N. Aslam, D. Jurkovic, R. Salim, G. Pandis, R. Hadwin).

Ancillary