Main

The standard treatment for early-stage breast cancer in women of all ages was surgery until the late 1970s (Kesseler and Seton, 1978). Primary endocrine therapy was first described in the early 1980s as an alternative for older women (Preece et al, 1982). Treatment involved the sole use of tamoxifen, an oestrogen-receptor antagonist, without surgery, radiotherapy or chemotherapy. Tamoxifen is effective in around 80% of women with moderately or strongly ER-positive tumours (Gaskell et al, 1992), but the duration of local disease control is shorter than with surgery, and some women require either alternate endocrine therapies or surgery at a later date (Kenny et al, 1998).

In the UK, the trend towards treating women aged 70 and over with tamoxifen alone has been based on the premise that they are less likely to be fit for surgery because of co-morbidity (Satariano and Ragland, 1994). However, both mastectomy and wide local excision have low mortality rates (Hunt et al, 1980; Wyld and Reed, 2003). Breast surgery-related morbidity, especially where axillary surgery is involved, is quite high and may impact on quality of life. Primary endocrine therapy is not a treatment option in the USA and is rarely used in Australia (Craft et al, 2000; Diab et al, 2000). In the UK, its use is widespread, with up to 42% of all women over 70 being treated in this way, regardless of whether co-morbidity is documented (Wyld et al, 2004).

To establish whether primary endocrine therapy is justifiable for women who are fit for surgery, we systematically reviewed the evidence from randomised trials comparing primary endocrine therapy to surgery, with or without adjuvant endocrine therapy, in the management of women aged 70 years or over with operable breast cancer.

Methods

Full details of the methods employed are published elsewhere (Hind et al, 2006). The Cochrane Breast Cancer Group Specialised Register was searched on 29 June 2005. Citations coded as ‘EARLY BREAST CANCER’, ‘ENDOCRINE THERAPY’, ‘PSYCHOSOCIAL’ or ‘SURGERY’ on the specialised register were retrieved with no date or language restrictions.

Only controlled trials with the following characteristics were included. Participants were women aged 70 years or over with clinically defined operable primary breast cancer: TNM classification T1-3 and T4b where there was only minor skin involvement, N0-1, mobile lymph nodes (Union Internationale Contre le Cancer, 1987). Studies had to compare either (1) surgery alone vs primary endocrine therapy; or (2) surgery plus adjuvant endocrine therapy vs primary endocrine therapy. Mastectomy could be with or without axillary clearance, and wide local excision could be with or without radiotherapy. Primary outcomes were overall survival (OS) and progression-free survival (PFS) (interval between start of treatment and need for second-line or palliative treatment, recurrence or death from any cause). Secondary outcomes were adverse effects (number of surgical complications or endocrine therapy related side effects), local disease control (interval between start of treatment and the development of local disease), distant metastasis-free interval (interval between start of treatment and the development of metastatic disease) and quality of life (however measured). Pre-specified subgroups included the type of surgery (mastectomy or wide local excision, with or without radiotherapy).

Two reviewers, LW and DH, independently assessed each potentially eligible trial for inclusion in the review with the results section masked. The same two reviewers independently reviewed each study according to its design and by how the study was conducted to assess any bias. The checklist for quality of randomised controlled trials included: concealment of the allocation sequence, generation of the allocation sequence, comparability between groups at the baseline and inclusion of all randomised participants in the analysis. For allocation concealment, trials were graded ‘A’ (adequate concealment), ‘B’ (adequacy of concealment unclear) or ‘C’ (clearly inadequate concealment).

The most complete data set feasible was assembled from the published literature. Where necessary, we sought additional information from the principal investigator of the trial concerned. Results of eligible studies were statistically synthesised if appropriate and possible. For time-to-event analyses, combined hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated using formal methods for extracting summary statistics to perform meta-analyses of the published literature (Parmar et al, 1998). Heterogeneity between trial results was tested using the χ2 test and the I2 measurement (Higgins et al, 2003). Absolute risk reductions and numbers needed to treat were calculated using the Altman and Andersen (1999) method.

Results

The search strategy retrieved 770 citations. Of these, 742 were excluded based on information in the title or abstract. The remaining 28 citations reported on the seven potentially eligible studies for the review. None of these studies was excluded. Five additional papers, all conference abstracts, relating to the same trials were identified through informal reference tracking and contact with authors. The study selection process is illustrated in Figure 1, in accordance with the QUOROM statement (Moher et al, 1999).

Figure 1
figure 1

QUOROM flow diagram showing study selection process.

Full size image

Three eligible trials addressing surgery vs primary tamoxifen therapy were identified (Table 1), all of which reported outcome data (Gazet et al, 1994; Kenny et al, 1998; Fentiman et al, 2003). None of the studies evaluated the oestrogen receptor status of the women they recruited. Four eligible trials addressing surgery plus endocrine therapy vs primary endocrine therapy were identified, of which three have reported outcome data (Willsher et al, 1997; Capasso et al, 2000; Mustacchi et al, 2003; Fennessey et al, 2004), although there is currently no data from one in a form that can be meta-analysed (Capasso et al, 2000). Not all trials identified provided adequate information on all outcomes. Only one of the studies evaluated oestrogen receptor status, recruiting only women with moderately or strongly oestrogen receptor-positive tumours (Willsher et al, 1997).

Table 1 Study characteristics
Full size table

It was not possible to assess accurately the quality of all studies (including the quality of the randomisation process) owing to lack of information in the published articles. The quality of three trials was graded as A (Fentiman et al, 2003; Mustacchi et al, 2003; Fennessey et al, 2004) with the rest being graded as B (Gazet et al, 1994; Willsher et al, 1997; Kenny et al, 1998; Capasso et al, 2000). No potentially eligible study was excluded from the review. Proposed sensitivity analysis based on trial quality was not conducted because of the small number of trials identified. There was good agreement on study selection, quality assessment, and data extraction.

Results are presented in Table 2. Analysis of OS, based on three trials (495 women: Gazet et al, 1994; Kenny et al, 1998; Fentiman et al, 2003), showed no significant difference between interventions (HR 0.98, 95% CI 0.74–1.30, P=0.9). One trial (164 women: Fentiman et al, 2003) reported adequate summary data to show a significant difference in PFS, favouring surgery (HR 0.55, 95% CI 0.39–0.77, P=0.0006). In the surgery arm, 77% of women had died or progressed compared to 84% in the tamoxifen arm: an extra 7% of participants receiving surgery benefited from the treatment. For every 14 women treated, one death or disease progression would be prevented over 120 months.

Table 2 Surgery vs primary endocrine therapy results
Full size table

Methodological issues (discussed below) prohibited either the meta-analysis of, or dissemination of results from the three trials, which reported data on local disease control or the distant metastasis-free interval. One trial (200 women: Gazet et al, 1994) reported adverse events. No patient discontinued primary tamoxifen treatment. Eight patients had a total of 10 side effects, including hot flushes, skin rash, vaginal discharge, indigestion, breast pain, and sleepiness. No trial reported quality of life data.

Results are presented in Table 3. Three trials (1076 women: Willsher et al, 1997; Mustacchi et al, 2003; Fennessey et al, 2004) reported data on OS which could be meta-analysed. There was a nonsignificant trend in favour of surgery plus endocrine therapy (HR 0.86, 95% CI 0.73–1.00, P=0.06). Only one trial (474 women: Mustacchi et al, 2003) reported adequate data on PFS to calculate a significant difference favouring surgery plus endocrine therapy (HR 0.65, 95% CI 0.53–0.81, P=0.0001). In the surgery arm, 59% of women died or progressed compared to 80% in the tamoxifen arm: an extra 21% of participants receiving surgery benefited from the treatment. For every five women treated, one death or disease progression would be prevented over 80 months.

Table 3 Surgery plus adjuvant endocrine therapy vs primary endocrine therapy results
Full size table

Analysis of two trials (929 women: Mustacchi et al, 2003; Fennessey et al, 2004) showed a significant difference in local disease control in favour of surgery plus endocrine therapy (HR 0.28, 95% CI 0.23–0.35, P< 0.00001). There was significant heterogeneity between the two studies (χ2 2.90, P<0.09, I2 65.6%), which is discussed below, although each individually showed a statistically significant difference in treatment effect favouring the surgery arm. Data from one trial (Willsher et al, 1997) were not included in this analysis as reported results were immature compared to the other two studies. Adequate data were not available to evaluate the difference in distant metastasis-free interval. One study reported that both mastectomy and wide local excision significantly improved local control compared to primary endocrine therapy (Fennessey et al, 2004).

One trial did not quantify adverse events, only reporting that one woman from the primary endocrine therapy arm had to drop out of the trial because of endocrine therapy-related adverse effects (Fennessey et al, 2004). In another, all patients in the surgery plus tamoxifen arm who had axillary clearance had paraesthesia on the ipsilateral arm and lateral thoracic wall. Tamoxifen-related toxicity was similar between the two arms and included headache, vertigo, itching, hair loss, cystitis, vaginal bleeding, acute thrombophlebitis, nausea, and indigestion (Mustacchi et al, 2003). The other two studies did not report adverse events (Willsher et al, 1997; Capasso et al, 2000). No trial investigated differences in quality of life. The only trial to evaluate differences in psychiatric morbidity (CRC) used the General Health Questionnaire 28 (Goldberg et al, 1970), and a socio-demographic questionnaire, which investigated levels of domestic support and social isolation. At 3 months after the start of treatment the surgery group had more psychosocial morbidity (P=0.03); however, there was no difference between the surgery and primary endocrine therapy groups at 2 years (Fallowfield, 1994).

Discussion

Statement of principal findings

This study has demonstrated that primary endocrine therapy is inferior to surgery with endocrine therapy for the local control of breast cancer in ER-unselected, medically fit older women. This is independent of the type of surgery, with both mastectomy and wide excision (without adjuvant radiotherapy) achieving superior local control. The meta-analysis showed no significant difference in OS between the two treatments in this group of women. One trial showed a small but significant survival advantage for surgery with adjuvant endocrine therapy, where follow-up was extended to 13 years (Fennessey et al, 2004). There are no data on formal QoL assessments between these two groups and no data on patients’ preferences.

Strengths and weaknesses of the study

The results of this review are based on a limited number of small studies of variable methodological quality. In some cases, the internal validity of the primary studies was affected by competing risks and informative censoring, which violate the assumptions underlying the Kaplan–Meier survival analysis method (these issues are discussed fully elsewhere in Hind et al, 2006). Heterogeneity between trials, in terms of interventions and outcome assessment, also made assessment of some outcomes problematic. There was considerable variation in surgical technique to the breast and the axilla (Table 1). It was unclear whether surgical margins were adequate by modern standards in any trial. In one trial, case selection for breast conservation included women with large tumours (T3 and T4) which would be inappropriate by modern standards (Gazet et al, 1994). Two of the trials which offered women wide local excision (St Georges and CRC) did not report using adjuvant radiotherapy which is standard practice today and may have reduced the clinical effect size of surgery (Gazet et al, 1994; Fennessey et al, 2004). In the analysis of local disease control in the second comparison (surgery with adjuvant tamoxifen vs primary tamoxifen) statistical heterogeneity cannot be explained by differences in population or treatment characteristics and is likely to be an artefact of the CRC trial's longer follow-up time.

Most trials recruited women regardless of oestrogen receptor status. Only 85–90% of women in this age group have ER-positive tumours (Diab et al, 2000). For the remainder, tamoxifen is not an active intervention and their treatment with tamoxifen is not therefore in line with modern clinical practice. The inclusion of such women may also have biased the results of the meta-analysis. Had such women been excluded from included trials, the primary endocrine therapy arms may have performed better against surgery arm plus endocrine therapy. However, the one trial to recruit exclusively patients with ER-positive tumours found local control to be superior with surgery and endocrine therapy (Willsher et al, 1997). Furthermore, none of the included studies controlled for patient co-morbidity and even among those fit for surgery in this age group, a significant proportion of patients will die of co-morbid diseases so reducing the relative advantages of any breast cancer therapies (Satariano and Ragland, 1994).

Strengths and weaknesses in relation to other studies

This is the first systematic review on this topic and provides an overview of all primary studies on the subject.

Meaning of the study possible explanations and implications for clinicians and policymakers

Primary endocrine therapy should only be offered to women with ER-positive tumours who are unfit for, or who refuse, surgery. In a cohort of women with reduced life expectancy, owing to significant co-morbid disease, and ER-positive tumours, primary endocrine therapy may be an appropriate treatment choice. A national UK trial will shortly be starting to evaluate selection criteria for the use of primary endocrine therapy, (Endocrine +/− Surgical Therapy for Elderly women with Mammary cancer, ESTEEM), which will aid in decision making.

Unanswered questions and future research

Since these studies were designed, endocrine therapies other than tamoxifen have become available. Aromatase inhibitors have been shown to be superior to tamoxifen in the adjuvant setting and may by attractive as primary endocrine therapy for older women who are unfit for surgery. Trials are needed to test this hypothesis. The ESTEEM trial will use aromatase inhibitors rather than tamoxifen, which may enhance the efficacy of primary endocrine therapy.