Evidence review

This evidence summary includes the best available evidence to support the off-label use of adjuvant bisphosphonates to prevent recurrence and improve survival in people with early breast cancer. A literature search was conducted which identified 89 references (see search strategy for full details). These references were screened using their titles and abstracts and 6 references were obtained and assessed for relevance.

One study was included in this evidence summary, the Early Breast Cancer Trialists' Collaborative Group (EBCTCG) meta-analysis (2015). A summary of this study is shown in the table below (see evidence tables for full details).

Table 3 Summary of included study



Intervention and comparison

Primary outcomes


Meta-analysis of individual participant data from 26 RCTsa

18,766 womena with early breast cancer (not defined)

Adjuvant bisphosphonate treatment (mainly zoledronic acid 50%, clodronate 27% and ibandronic acid 16%) compared with no bisphosphonate treatment (including placebo)

Recurrence of breast cancer

Distant recurrence of cancerb

Breast cancer mortality

a 11,767 women were postmenopausal, 6171 women were premenopausal and 828 women were perimenopausal

b Recurrence in the bone or elsewhere, not in the breasts or regional lymph nodes

Abbreviations: EBCTCG, Early Breast Cancer Trialists' Collaborative Group; RCT, randomised controlled trial

The remaining 5 references were excluded and are listed in excluded studies with reasons for their exclusion. They include a 2012 Cochrane review that assessed the effects of bisphosphonates and other bone agents for breast cancer (including 12 RCTs in early breast cancer, n=10,124), which has been superseded by the EBCTCG meta-analysis.

Clinical effectiveness

Of 18,766 women included in the EBCTCG meta-analysis (2015), over a median follow-up of 5.6 woman-years, breast cancer recurred in 3,453 women (18.4%), after which 2,106 women (11.2%) died. Overall, use of bisphosphonates compared with control resulted in no statistically significant reduction in the recurrence of breast cancer, but there were statistically significant reductions in distant recurrence (mainly because of a reduction in bone recurrence) and breast cancer mortality. With bisphosphonates compared with control, the absolute 10-year risk was reduced by:

  • 1.1% for breast cancer recurrence (24.9% compared with 25.9%; rate ratio [RR] 0.94, 95% confidence interval [CI] 0.87 to 1.01; p=0.08, not statistically significant)

  • 1.4% for distant recurrence (20.4% compared with 21.8%; RR 0.92, 95% CI 0.85 to 0.99; p=0.03)

  • 1.1% for bone recurrence (7.8% compared with 9.0%; RR 0.83, 95% CI 0.73 to 0.94; p=0.004), and

  • 1.7% for breast cancer mortality (16.6% compared with 18.4%; RR 0.91, 95% CI 0.83 to 0.99; p=0.04)

  • 1.6% for all-cause mortality (20.8% compared with 22.3%; RR 0.92, 95% CI 0.85 to 1.00, p=0.06, not statistically significant).

In the prespecified subgroup analyses (which may have been post-hoc for the individual trials included in the meta-analysis), no statistically significant benefits from bisphosphonate use were seen in premenopausal women (n=6,171) compared with control. In postmenopausal women (n=11,767, natural or induced menopause), there were statistically significant reductions in breast cancer recurrence, distant recurrence (recurrence in the bone or elsewhere, not in the breasts or regional lymph nodes), bone recurrence and breast cancer mortality. The absolute reductions with bisphosphonates compared with control in postmenopausal women at 10 years were:

  • 3.0% for breast cancer recurrence (22.8% compared with 25.8%; RR 0.86, 95% CI 0.78 to 0.94; p=0.002)

  • 3.4% for distant recurrence (17.9% compared with 21.2%; RR 0.82, 95% CI 0.74 to 0.91; p=0.0003)

  • 2.2% for bone recurrence (6.6% compared with 8.8%; RR 0.72, 95% CI 0.60 to 0.86; p=0.0002) and

  • 3.3% for breast cancer mortality (14.7% compared with 18.0%; RR 0.82, 95% CI 0.73 to 0.93; p=0.002)

  • 2.3% for all-cause mortality (21.1% compared with 23.5%; RR 0.86, 95% CI 0.77 to 0.96, p=0.005).

Bisphosphonate use had no statistically significant effect on distant recurrence other than bone recurrence, or on locoregional recurrence (in the same site as the original tumour or in the regional lymph nodes) or contralateral breast cancer (in the opposite breast) in all women or in the postmenopausal subgroup.

Subgroup analyses looking at bone recurrence and breast cancer mortality suggested that the efficacy of bisphosphonates is greater in women aged 55 years or older, compared with younger women. However, menopausal status and age are closely correlated so it is difficult to know which characteristic is most relevant. Reductions in bone recurrence and breast cancer mortality did not depend on tumour characteristics (including oestrogen receptor status, nodal status or tumour grade). No subgroup analyses assessed bisphosphonates according to women's estimated risk of breast cancer recurrence or mortality.

No statistically significant difference was found between non-aminobisphosphonates (clodronate) and aminobisphosphonates (zoledronic acid and ibandronic acid). There was no apparent benefit with the non-aminobisphosphonate pamidronate (taken orally), but the number of women taking this bisphosphonate was low (n=953) and so the meta-analysis may have had insufficient statistical power to detect a benefit. The number of women taking risedronate (n=398) was also low and insufficient to assess its efficacy, and no women received alendronic acid or intravenous (IV) pamidronate in the trials included in the meta-analysis.

For bone recurrence, there was no statistically significant difference between low-intensity bisphosphonate schedules (such as 6-monthly IV zoledronic acid) and high-intensity schedules (such as monthly IV zoledronic acid, daily oral ibandronic acid or daily oral clodronate). Similar reductions with bisphosphonates were seen in the presence or absence of chemotherapy, suggesting any benefits are additive to chemotherapy.

An overview of the key results for clinical effectiveness can be found in results tables.

Safety and tolerability

The most common adverse effects of bisphosphonates are gastrointestinal effects (such as nausea, dyspepsia, mild oesophagitis and abdominal pain) and bone, joint or muscle pain. Other less common adverse effects include more serious oesophageal reactions (such as oesophagitis and oesophageal ulcers, strictures and erosions), osteonecrosis of the jaw and atypical stress fractures (NICE clinical knowledge summaries on osteoporosis - prevention of fragility fractures). Adverse effects vary among bisphosphonates. See the individual summaries of product characteristics (SPCs) for more information.

The EBCTCG meta-analysis (2015) did not report adverse effects. The bisphosphonates used most commonly in the trials in the meta-analysis were oral clodronate and IV zoledronic acid. Adverse effects reported in between 1 in 10 and 1 in 100 people taking sodium clodronate for the licensed indications are asymptomatic hypocalcaemia, diarrhoea, nausea, vomiting and increased transaminases (within the normal range) (Bonefos SPC). Adverse effects reported in between 1 in 10 and 1 in 100 people taking zoledronic acid 4 mg for the licensed indications are anaemia, headache, conjunctivitis, nausea, vomiting, decreased appetite, bone pain, myalgia, arthralgia, generalised pain, renal impairment, fever, flu-like syndrome, increased blood creatinine and urea, and hypocalcaemia. Hypophosphataemia occurs in more than 1 in 10 people taking zoledronic acid (Zometa SPC).

A 2012 Cochrane review of 34 randomised controlled trials (RCTs) that assessed the effects of bisphosphonates and other bone agents for breast cancer (including 12 RCTs in early breast cancer, n=10,124) reported adverse events. It concluded that the adverse effects of bisphosphonates were generally mild and infrequent. The main adverse effect seen with IV zoledronic acid was renal toxicity, which was related to the dose and infusion time. No significant renal toxicity was seen with IV pamidronate or IV ibandronic acid. The main adverse effect seen with oral clodronate and oral ibandronic acid was mild gastrointestinal toxicity. The authors noted that there have been reports of osteonecrosis of the jaw with long-term use of bisphosphonates. In the trials in early breast cancer included in the Cochrane review, the incidence of osteonecrosis of the jaw was found to be from 0% to 0.7%, highlighting the need for good oral care before and during treatment in people taking bisphosphonates long-term.

The MHRA has issued guidance on the use and safety of bisphosphonates, which summarises important safety issues with bisphosphonates, including oesophageal reactions, osteonecrosis of the jaw, atrial fibrillation, atypical femoral fractures and adverse effects on renal function. Since this was published in 2014, the MHRA has also warned that osteonecrosis of the external auditory canal has been reported very rarely with bisphosphonates.

Evidence strengths and limitations

The EBCTCG meta-analysis (2015) was a large, meta-analysis of individual participant data from RCTs. Of the 32 trials (n=19,291) identified that recorded recurrence, data were not obtained for 6 trials; however, these included only 525 women and would have been unlikely to change the key outcomes of the meta-analysis. Data were obtained for 26 trials (n=18,766, 97% of participants). Citations for publications for the included and excluded trials were not reported.

The methods of identifying trials, seeking collaboration, data collection, collation, checking, and presentation have not been fully reported, apart from to note these were undertaken as in previous EBCTCG reports. In addition, no assessment of quality or risk of bias of the individual trials is provided, including randomisation, allocation concealment, blinding and adjudication of outcomes. Most of the trials included in the meta-analysis were open-label; outcome assessment was often not blinded (although this will have limited effect on measures of recurrence and survival); and some of the trials were industry-sponsored. The subgroup analyses in the EBCTCG meta-analysis were prespecified, but may not have been prespecified in the individual trials: details are not reported.

The paper does not report if the populations and treatments studied were homogeneous. For example, it is unclear whether differences in the standard breast cancer management of pre- and postmenopausal women may account for the differences in benefits in these populations, rather than bisphosphonates. It is also unclear what chemotherapy was used for breast cancer, whether it was similar across the trials, and whether choice of chemotherapy reflects current clinical practice. Breast cancer survival has improved significantly over the last few decades because of better treatments and better identification of risk groups. It is possible that bisphosphonates only benefited women who had suboptimal treatment in the older trials. Nevertheless, the 2 largest trials included in the EBCTCG meta-analysis, the ABCSG-12 (n=1,083) and AZURE (BIG 01/04) trials (n=3,360) enrolled participants until 2006, and sensitivity analyses excluding these trials still found benefits of bisphosphonates in postmenopausal women.

Overall, the authors of the Cancer Care Ontario and the American Society of Clinical Oncology (CCO/ASCO) guideline on the Use of adjuvant bisphosphonates and other bone-modifying agents in breast cancer considered that the EBCTCG had strict inclusion criteria and protocols and the results of the EBCTCG meta-analysis are of high-quality, with the limitation that some outcomes have not yet been completely reported. However, they highlighted that combining the data for all bisphosphonates in the meta-analysis was controversial, as was including small trials not designed to measure the outcomes of interest; for example, some trials were designed to look at bone mineral density and did not report recurrence or survival. This also means that there may have been no adjudication of these outcomes in those trials. Furthermore, it has been questioned whether a meta-analysis of subgroups should be used to inform clinical decision making because subgroup analyses are prone to error and generally considered more appropriate for hypothesis generation for future trials than changing practice (Jacobs et al. 2015).

The EBCTCG used log-rank statistics to assess the effects of bisphosphonates compared with control on various outcomes, and to estimate rate ratios and their confidence intervals. Log-rank analyses allow analyses of the effects of treatment in separate years, which may improve medical understanding but can also carry some statistical uncertainties. Estimates for a given year can only be based on trials that have follow-up and data reported in that year. The median follow-up in the meta-analysis was 5.6 years and the interquartile range was 3.7 years to 8.0 years. This means that 75% of women were followed up for 8 years or less. It is unclear how many women were followed up for 10 years but the number is likely to be low when subgroups are considered.

Although data on safety outcomes were reportedly collected, adverse effects were not reported in the EBCTCG meta-analysis; however, the adverse effect profile of bisphosphonates is well-established. The trials in the meta-analysis included women only and no information is available on using bisphosphonates to prevent recurrence of breast cancer in men. When considering menopausal status, women with natural menopause and premature menopause induced by chemotherapy or oophorectomy were included in the same subgroup. Only 1 trial (ABCSG-12, n=1,803), which studied zoledronic acid, included women with induced menopause. The majority of participants in the meta-analysis (83%) also received adjuvant chemotherapy, and bisphosphonate treatment generally started 0–12 weeks after surgery or chemotherapy. Clodronate has not been studied specifically in people receiving aromatase inhibitors (CCO/ASCO guideline).

Little or no information is available for some commonly used bisphosphonates, such as risedronate and alendronic acid. No benefit was seen with oral pamidronate but the authors of the meta-analysis note that this might be expected because the oral formulation is poorly absorbed and pamidronate has been shown to have little or no effect on markers of bone resorption, metastatic bone disease and myeloma. Only limited information from subgroup analyses is available comparing different bisphosphonates, or different dosages of bisphosphonates (low- or high-intensity schedules). Most of the participants (97%) were in trials lasting 2–5 years, and the optimal duration of treatment is unclear.

The authors of the meta-analysis note that more reliable comparisons of different bisphosphonate regimens may emerge from ongoing trials that compare them directly; for example, the SWOG 0307 trial (NCT00127205) comparing clodronate, zoledronic acid and ibandronic acid over 3 years (n=5,400), and the SUCCESS trial (NCT02181101) comparing 5 years' and 2 years' treatment with zoledronic acid (n=3,800). Results of an interim analysis from SWOG 0307 have been published as an abstract only, suggesting there were no significant differences between the groups in 5-year disease-free survival or overall survival, but there were some differences in adverse events (Gralow et al. 2015). Full publication of the trial is needed before the relevance of the results can be assessed.

A literature search was undertaken in June 2016 for the CCO/ASCO guideline on the use of adjuvant bisphosphonates and other bone-modifying agents in breast. In addition to trials of bisphosphonates with data included in the EBCTCG meta-analysis, the search found results for SWOG S0307 and a few small trials on clodronate and zoledronic acid. However, the authors noted that data on survival and recurrence were reported in abstracts only and that the results do not currently change the conclusions of the EBCTCG meta-analysis. Results from ongoing trials (including HOBOE [NCT00412022] and TEAM-IIb [ISRCTN17633610]), and longer follow-up of the trials included in the EBCTCG meta-analysis, may provide more evidence on the effects of bisphosphonates in premenopausal women, and allow more stable estimates of the 10-year outcomes in postmenopausal women.

An overview of the quality assessment for the included study can be found in the evidence tables.