Medications to reduce breast cancer risk: a network meta-analysis of randomized controlled trials

Abstract

Background

Given the rising incidence of breast cancer, especially in premenopausal women, there is an urgent need to identify additional risk-reducing medications to accelerate prevention, as only a few are currently approved. We, therefore, performed network meta-analysis (NMA) to identify and compare the efficacy of medications for primary breast cancer prevention.

Methods

We performed a literature search completed on November 16, 2023, in Embase, Ovid-Medline, Scopus, and Cochrane Library for randomized controlled trials (RCTs) evaluating risk-reducing medications in women without a history of invasive breast cancer. Two reviewers independently screened and extracted data based on predefined criteria, following the Preferred Reporting Items for Systematic reviews and Meta-Analyses guidelines, and assessed the risk of bias using the Revised Cochrane Risk of Bias tool. The primary outcome was overall breast cancer incidence, with secondary outcomes including invasive breast cancer and ductal carcinoma in situ. NMA was performed using a random-effects model, measuring efficacy with risk ratios (RR) and number needed to treat (NNT). Medications were ranked using the Surface Under the Cumulative RAnking curve (SUCRA). We performed subgroup analyses by menopause status, primary versus secondary/other outcomes, follow-up, and intervention duration.

Results

Out of 8,598 studies screened, 43 RCTs (n = 337,240 women) met inclusion criteria. Six medications reduced overall breast cancer risk compared to placebo: sulfonylurea (RR = 0.18, 95% CI = 0.04–0.91, NNT = 44.1, SUCRA = 0.90), thiazolidinediones (RR = 0.25, 95% CI = 0.08–0.78, NNT = 48.3, SUCRA = 0.80), third-generation selective estrogen receptor modulators (SERMs) (RR = 0.46, 95% CI = 0.33–0.66, NNT = 67.3, SUCRA = 0.62), aromatase inhibitors (AIs) (RR = 0.50, 95% CI = 0.39–0.66, NNT = 73.0, SUCRA = 0.55), raloxifene (RR = 0.63, 95% CI = 0.47–0.84, NNT = 96.9, SUCRA = 0.37), and tamoxifen (RR = 0.76, 95% CI = 0.65–0.88, NNT = 149.7, SUCRA = 0.23). AIs (RR = 0.48, 95% CI = 0.33–0.71), tamoxifen (RR = 0.63, 95% CI = 0.51–0.78), and raloxifene (RR = 0.63, 95% CI = 0.47–0.86), were effective for invasive breast cancer. Third-generation SERMs (RR = 0.46, 95% CI = 0.32–0.67), AIs (RR = 0.51, 95% CI = 0.40–0.64), raloxifene (RR = 0.61, 95% CI = 0.46–0.82), and tamoxifen (RR = 0.76, 95% CI = 0.66–0.86) were effective in studies with breast cancer as a primary outcome, while thiazolidinediones (RR = 0.25, 95% CI = 0.07–0.84) were effective in studies with breast cancer as a secondary/other outcome.

Conclusions

This NMA confirms the efficacy of tamoxifen, raloxifene, and AIs, and identifies thiazolidinediones and third-generation SERMs as promising agents for breast cancer prevention, though not currently included in guidelines. These findings extend prior evidence and highlight the need for trials in premenopausal and racially diverse populations to address existing gaps.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13058-025-02059-w.

Introduction

Breast cancer remains a significant health concern, with rising incidence, particularly in premenopausal women [1, 2]. Despite advances in treatment, prevention strategies are important to reduce the disease burden [3]. One promising approach is the use of risk-reducing medications, particularly selective estrogen receptor modulators (SERMs) and aromatase inhibitors (AIs) [4, 5]. Tamoxifen, the first SERM approved for breast cancer prevention, was shown in large randomized controlled trials (RCTs) to reduce breast cancer incidence by > 50% in high-risk women [6]. This was followed by the approval of raloxifene, another SERM, which showed similar efficacy with fewer side effects [7]. More recently, AIs such as exemestane and anastrozole have demonstrated effectiveness in postmenopausal women [8].

Risk-reducing medication approvals are largely based on findings from RCTs, which are challenging to perform in the prevention setting for several reasons, including the need for long follow-ups and large sample sizes [9]. As a result, research on newer medications is limited, and long-term data are still being gathered. The U.S. Preventive Services Task Force recommends offering risk-reducing medications to women at increased risk for breast cancer and low risk for adverse effects [10]. However, their use in clinical practice remains low because of concerns about side effects, difficulty in identifying suitable candidates, lack of clear benefit in overall survival in most RCTs, and primary care physicians’ unfamiliarity with the medications within the context of primary prevention [11].

As breast cancer prevention research continues to develop, new risk-reducing medications are being investigated [12, 13]. However, the difficulty with performing large-scale RCTs for many of these medications has limited their approval for breast cancer prevention purposes. The most comprehensive evidence on the efficacy of these medications comes from a 2016 study [14]. The study provided critical insights but did not stratify the analysis by menopause status, follow-up, and intervention duration, and did not examine ductal carcinoma in situ (DCIS) as an outcome. A 2019 systematic review and meta-analysis focused on approved medications, including tamoxifen, raloxifene, and AIs, and found them to be effective in reducing breast cancer incidence [15]. This review suggested that AIs might be more effective than SERMs, although it highlighted the limited long-term toxicity data for AIs [15]. However, the 2019 review did not examine newer medications.

Given these limitations, there is a clear need for updated and more comprehensive analyses to synthesize the available evidence and provide a broader understanding of the efficacy of different risk-reducing medications. We therefore performed a network meta-analysis (NMA) to pool together estimates from RCTs and identify medications associated with a reduction in breast cancer risk. By examining a wide range of risk-reducing medications, including those not yet approved for breast cancer prevention, our study aims to provide a comprehensive comparative analysis of risk-reducing medications. Findings from this analysis can guide future research in breast cancer prevention, inform clinical decision-making, and contribute to the development of updated guidelines on breast cancer prevention.

Methods

Literature search and study eligibility

We searched the published literature using strategies designed by a medical librarian for the concepts of breast cancer prevention/prophylaxis and different risk-reducing medications, with related synonyms. The search strategies were developed using a combination of controlled vocabulary terms and keywords and were executed in Embase.com, Ovid-Medline All, Scopus, and the Cochrane Library. We limited the results to English-language publications. Animal studies, conference abstracts, conference papers, and book chapters were excluded from search results. All database searches were completed on November 16, 2023. Complete search strategies are provided in Supplementary Methods.

We removed duplicates using EndNote software. We then imported the remaining citations into Covidence, a tool for managing systematic reviews, for further screening and data extraction. We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) workflow (Fig. 1). Two reviewers independently screened all titles and abstracts for eligibility. Full-text articles of potentially eligible studies were then reviewed independently by two reviewers. Discrepancies at both stages were resolved by consensus or a third reviewer.

Fig. 1.

Preferred reporting items for systematic reviews and meta-analyses (PRISMA) flow diagram

We followed the PICO (Population, Intervention, Comparison, Outcome) framework to define our inclusion and exclusion criteria. Our inclusion criteria were: 1) Population: women without a history of invasive breast cancer; 2) Intervention: pharmacological agents aimed at preventing the development of invasive breast cancer; 3) Comparison: no intervention, placebo, or other pharmacological agent; 4) Outcome: incidence of invasive breast cancer; 5) Study design: randomized controlled trials (RCTs). Our exclusion criteria were: 1) Population: women with a history of invasive breast cancer; 2) Intervention: lifestyle interventions (e.g., exercise, diet), unspecified multivitamins, herbal drugs without identified active ingredients and dosages; 3) Study design: observational studies, commentaries, editorials, reviews, abstracts, meta-analyses, systematic reviews; 4) Language: non-English publications; 5) Publication type: animal studies, conference abstracts, conference papers, book chapters. Two reviewers independently extracted the following data from each study using a standardized form: study characteristics (entry and exclusion criteria, trial name, title, study duration, country and study centers, follow-up time, number of participants in each arm, loss to follow-up), participant demographics (age, body mass index (BMI), menopause status, and race/ethnicity), intervention details (name of intervention and comparison, dosage, mode of administration, and duration of intervention), and outcomes (overall breast cancer, invasive breast cancer, and DCIS). Discrepancies were resolved by consensus or a third reviewer if needed.

Risk of bias assessment

We used the Revised Cochrane risk-of-bias tool for randomized trials (RoB 2) to assess the risk of bias for each RCT across five domains, including randomization, assignment to intervention, missing outcome data, outcome measurement, and selection of the reported result. For each domain, risks were classified as low risk, high risk, or some concerns. RCTs were randomly assigned to two reviewers. We identified and analyzed the earliest published paper and the corresponding research protocol (when available) for each RCT to perform the risk of bias assessment. The RoB 2 decision tree was followed to make the decision for each domain. Finally, an overall risk of bias judgement was made according to the RoB 2 guideline (Table S1). The summary plots were generated using robvis tool [16].

Statistical analysis

The medications were evaluated regarding the primary outcome of overall breast cancer incidence and the secondary outcomes of invasive breast cancer and DCIS. Network plots were used to present all reviewed studies. The medications included tamoxifen, raloxifene, third-generation SERMs (arzoxifene, bazedoxifene, lasofoxifene), AIs (anastrozole, exemestane), statins (pravastatin, simvastatin, lovastatin), thiazolidinediones (rosiglitazone, pioglitazone), sulfonylurea, metformin, metformin plus sulfonylurea, fenretinide, tamoxifen plus fenretinide, calcium, calcium plus vitamin D, multivitamin, vitamin E, betacarotene, goserelin plus ibandronate, aspirin, conjugated equine estrogen (CEE), CEE plus medroxyprogesterone acetate (MPA), and tibolone. NMA methods were used to summarize the efficacy of the multiple medications. For each outcome, only medications that were compared to other medications more than once were included in the NMA. The NMA used random-effects model and combined both direct and indirect comparisons to compare the efficacy between two medications (including placebo) [17]. A common between-study heterogeneity parameter (${\tau }^2$) was assumed across all comparisons in the random-effects model. The compared efficacy was measured using risk ratios (RR) with 95% confidence intervals (CI) and predictive intervals (PrI), presented in forest plots. The PrI included both the fixed-effect standard error and the random-effect standard error ($\tau$) when calculating the width. As a result, it provides information on the magnitude of heterogeneity and is the interval within which the estimate of a future study is expected to be. Additionally, the absolute effect of treatment was evaluated by the number needed to treat (NNT), i.e. number of patients needed to be treated to prevent one breast cancer case and was calculated as the inverse of risk difference against placebo. The ranking of the medications with significant efficacy compared to placebo was evaluated by the rankogram and Surface Under the Cumulative RAnking curve (SUCRA) [18]. The SUCRA value lies between 0 and 1, with greater value representing a higher rank for efficacy. Subgroup analyses were performed by menopause status, outcome type, follow-up, and intervention duration. All analyses were conducted in the R software (version 4.3.2) [19], and the netmeta package (version 2.9–0) [20].

Results

Study inclusion

Our comprehensive literature search, completed on November 16, 2023, identified 13,642 citations. After removing duplicates and including 26 additional references from citation searching, we screened 8,598 studies. Following title and abstract screening, we assessed 177 full-text articles and included 43 studies that met eligibility criteria (Fig. 1).

Study characteristics and risk of bias assessment

Among the 43 included RCTs, most compared risk-reducing medications versus placebo, except one comparing raloxifene and tamoxifen [41], one comparing different doses of raloxifene [25], one comparing metformin plus sulfonylurea with rosiglitazone [22], and one comparing metformin plus sulfonylurea with sulfonylurea alone [23] (Table 1). The mean age of women in these RCTs ranged from 34.4 to > 75 years (Table 2.). The mean BMI values ranged from 22.5 kg/m² to > 32 kg/m². Race/ethnicity data were inconsistently reported; among the RCTs reporting race (n = 14), most participants were non-Hispanic White. Among the studies reporting menopause status (n = 21), 11 included only postmenopausal women, 3 included only premenopausal women, and 7 included mixed menopause status.

Table 1.

Study characteristics of the eligible randomized controlled trials

Study ID	Trial name	Intervention, N	Control, N	Intervention dose	Median follow-up (months)
Breast cancer as a primary outcome
Lee 1999 [24]	WHS betacarotene	Betacarotene, 19939	Placebo, 19937	50 mg every other day	49.2
Cauley 2001 [25]	MORE	Raloxifene, 5129	Raloxifene, 2576	60 or 120 mg/d	47.4
Hercberg 2004 [26]	SU.VI.MAX	Multivitamin, 3844	Placebo, 3869	120 mg/d of ascorbic acid, 30 mg/d of vitamin E, 6 mg/d of betacarotene, 100 µg/d of selenium, 20 mg/d of zinc	90.5
Martino 2004 [27]	CORE	Raloxifene, 2725	Placebo, 1286	60 mg/d	NA
Anderson 2004 [28]	WHI CEE-alone	CEE, 5310	Placebo, 5429	0.625 mg/d	141.6
Lee 2005 [29]	WHS vitamin E	Vitamin E, 19937	Placebo, 19939	600 IU every other day	121.2
Fisher 2005 [30]	NSABP P-1	Tamoxifen, 6597	Placebo, 6610	20 mg/d	74.0
Veronesi 2006 [31]	NA	Fenretinide, 872	NA, 867	200 mg/d	175.2
Barrett-Connor 2006 [32]	RUTH	Raloxifene, 5044	Placebo, 5057	60 mg/d	66.7
Lappe 2007 [33]	NA	Calcium, 445a	Placebo, 288	1400–1500 mg/da	48.0
Veronesi 2007 [34]	Italian	Tamoxifen, 2700	Placebo, 2708	20 mg/d	134.5
Powles 2007 [35]	Royal Marsden	Tamoxifen, 1250	Placebo, 1244	20 mg/d	158.0
Cuzick 2007 [36]	IBIS-I	Tamoxifen, 3579	Placebo, 3575	20 mg/d	192.0
Cummings 2008 [37]	LIFT	Tibolone, 2249	Placebo, 2257	1.25 mg/d	34.0
Chlebowski 2008 [38]	WHI calciumplus vitamin D	Calcium plus vitamin D, 18176	Placebo, 18106	Calcium 1000 mg/d plus vitamin D 400 IU/d	84.0
DeCensi 2009 [39]	NA	Tamoxifen, 58b	Placebo, 58	5 mg/db	66.0
LaCroix 2010 [40]	PEARL	Lasofoxifene, 2729c	Placebo, 2740	0.25 mg/dc	60.0
Vogel 2010 [41]	NSABP STAR P-2	Raloxifene, 9754	Tamoxifen, 9736	Raloxifene 60 mg/d and tamoxifen 20 mg/d	81.0
Goss 2011 [42]	MAP.3	Exemestane, 2285	Placebo, 2275	25 mg/d	35.0
Avenell 2012 [43]	RECORD (Vitamin D)	Vitamin D, 1306d	Placebo, 1332	800 IU/dd	74.4
Powles 2012 [44]	GENERATIONS	Arzoxifene, 4676	Placebo, 4678	20 mg/d	NA
Allred 2012 [45]	NSABP B-24	Tamoxifen, 902	Placebo, 902	20 mg/d	174.0
Cook 2013 [46]	WHS aspirin	Aspirin, 19934	Placebo, 19942	100 mg every other day	120.0
DeCensi 2013 [47]	HOT	Tamoxifen, 938	Placebo, 946	5 mg/d	74.4
DeCensi 2019 [48]	TAM-01	Tamoxifen, 253	Placebo, 247	5 mg/d	61.2
Cuzick 2020 [49]	IBIS-II	Anastrozole, 1920	Placebo, 1944	1 mg/d	131.0
Chlebowski 2020 [50]	WHI CEE plus MPA	CEE plus MPA, 8506	Placebo, 8102	CEE 0.625 mg/d plus MPA 2.5 mg/d	226.8
Breast cancer as a secondary/other outcome
Downs 1998 [51]	AFCAPS/TexCAPS	Lovastatin, 499	Placebo, 498	20–40 mg/d	62.4
Trial 2002 [52]	ALLHAT-LLT	Pravastatin, 2511	NA, 2540	40 mg/d	57.6*
Miscellany 2002 [53]	HPS	Simvastatin, 2542	Placebo, 2540	40 mg/d	60.0
Trivedi 2003 [54]	NA	Vitamin D, 326	Placebo, 323	100,000 IU every 4 months for 5 years	NA
Strandberg 2004 [55]	4S	Simvastatin, 407	Placebo, 420	20 mg/d	124.8
Von Minckwitz 2011 [56]	GISS	Goserelin plus ibandronate, 15	NA, 15	Goserelin 3.6 mg every 4 weeks and ibandronate 2 mg every 3 months for 2 years	NA
Howell 2018 [57]	RAZOR	Goserelin plus raloxifene, 38	NA, 37	Goserelin 3.6 mg every 4 weeks and raloxifene 60 mg/d	108.0
Breast cancer as an adverse effect
Sacks 1996 [58]	CARE	Pravastatin, 290	Placebo, 286	40 mg/d	60.0
Miscellany 1998 [59]	NA	Pravastatin, 756	Placebo, 760	40 mg/d	73.2
Gregorio 1999 [23]	NA	Sulfonylurea, 45	Metformin plus sulfonylurea, 47	Glibenclamide up to 15 mg/d or gliclazide up to 240 mg/d	NA
Shepherd 2002 [60]	PROSPER	Pravastatin, 1495	Placebo, 1505	40 mg/d	NA
Kahn 2006 [61]	ADOPT	Rosiglitazone, 645e	NA	4 mg/de	48.0
Home 2009 [22]	RECORD (Rosiglitazone)	Rosiglitazone, 2220f	Metformin plus sulfonylurea, 2227	4–8 mg/d rosiglitazonef	66.0
Archer 2009 [62]	NA	Bazedoxifene, 1886g	Placebo, 1885	Bazedoxifene 20 mg/dg	NA
Erdmann 2014 [63]	PROactive	Pioglitazone, 870	Placebo, 905	15–45 mg/d	69.6
Leng 2021 [64]	NA	Metformin, 60	Placebo, 60	NA	NA

Abbreviations: 4S Scandinavian Simvastatin Survival Study, AFCAPS/TexCAPS Air Force/Texas Coronary Atherosclerosis Prevention Study, ADOPT A Diabetes Outcome Progression Trial, ALLHAT-LLT Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial, CARE Cholesterol and Recurrent Events, CEE Conjugated equine estrogen, CORE Continuing Outcomes Relevant to Evista, GENERATIONS Global Evaluation of a Raloxifene Analog in the Prevention of Osteoporosis and Breast Cancer, GISS Trial of Screening Plus Goserelin, Ibandronate, Versus Screening Alone, HPS Heart Protection Study, HOT Hormone Replacement Therapy Opposed by Low Dose Tamoxifen Study, IBIS International Breast Cancer Intervention Study, IU/d International units per day, LIFT Long-Term Intervention on Fractures with Tibolone, MAP.3 Mammary Prevention.3 trial, mg/d Milligrams per day, MORE Multiple Outcomes of Raloxifene Evaluation, MPA Medroxyprogesterone acetate, NA Not available, NSABP National Surgical Adjuvant Breast and Bowel Project, PEARL Postmenopausal Evaluation and Risk-Reduction with Lasofoxifene, PROactive Prospective Pioglitazone Clinical Trial in Macrovascular Events, PROSPER Pravastatin in Elderly Individuals at Risk of Vascular Disease, RAZOR Raloxifene and Zoladex Research Study, RECORD (Vitamin D) Randomized Evaluation of Calcium or Vitamin D in Reducing Fractures and Cancer Risk, RECORD (Rosiglitazone) Rosiglitazone Evaluated for Cardiovascular Outcomes and Regulation of Glycemia in Diabetes, RUTH Raloxifene Use for The Heart, STAR Study of Tamoxifen and Raloxifene, SU.VI.MAX Supplémentation en Vitamines et Minéraux Antioxydants, μg/d Micrograms per day, WHS Women's Health Study, WHI Women's Health Initiative, RECORD Rosiglitazone evaluated for cardiac outcomes and regulation of glycemia in diabetes , mg milligram, µg microgram, GENERATIONS Global Evaluation of a raloxifene analog in the prevention of osteoporosis and breast cancer, PROactive Prospective pioglitazone clinical trial in macrovascular events, 4S Scandinavian Simvastatin Survival Study, PROSPER Pravastatin in elderly individuals at risk of vascular disease, ADOPT Diabetes Outcome Progression Trial, NA Not available, ALLHAT-LLT Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial, CARE Cholesterol and Recurrent Events, AFCAPS/TexCAPS Air Force/Texas Coronary Atherosclerosis Prevention Study, WHS Women’s Health Study, IU International Unit, SU.VI.MAX Supplémentation en Vitamines et Minéraux Antioxidants, LIFT Long-Term Intervention on Fractures with Tibolone, PEARL Postmenopausal Evaluation and Risk-Reduction with Lasofoxifene, GISS Trial of Screening Plus Goserelin, Ibandronate, versus Screening Alone, STAR Study of Tamoxifen and Raloxifene, CORE Continuing Outcomes Relevant to Evista, RAZOR Raloxifene and Zoladex Research Study, MAP.3 Mammary Prevention.3 trial, NSABP National Surgical Adjuvant Breast and Bowel Project, HOT Hormone Replacement Therapy Opposed by Low Dose Tamoxifen study, IBIS International Breast Cancer Intervention Study, WHI Women’s Health Initiative, CEE plus MPA Conjugated equine estrogen plus Medroxyprogesterone acetate, MORE Multiple outcomes of raloxifene evaluation, RUTH Raloxifene Use for The Heart, HPS Heart Protection Study

^*Mean follow-up months

^aThis trial included three arms: Calcium 1400–1500 mg/d, N = 445; Calcium 1400–1500 mg/d plus vitamin D 1100 IU/d, N = 446; Placebo, N = 288. The calcium and placebo arms are presented in the table.

^bThis trial included four arms: Tamoxifen 5 mg/d, N = 58; Fenretinide 200 mg/d, N = 59; Tamoxifen 5 mg/d plus fenretinide 200 mg/d, N = 60; Placebo, N = 58. The tamoxifen and placebo arms are presented in the table.

^cThis trial included three arms: Lasofoxifene 0.25 mg/d, N = 2729; Lasofoxifene 0.5 mg/d, N = 2745; Placebo, N = 2740. The 0.25 mg/d and placebo arms are presented in the table.

^dThis trial included four arms: Vitamin D 800 IU/d, N = 1306; Calcium 1000 mg/d, N = 1343; Calcium 1000 mg/d plus vitamin D 800 IU/d, N = 1311/; Placebo, N = 1332. The vitamin D and placebo arms are presented in the table.

^eThis trial included three arms: Rosiglitazone (initial dose: 4 mg/d, titrated to max 8 mg/d), N = 645; Metformin (initial dose: 500 mg/d, max 2000 mg/d), N = 590; Glyburide (initial dose: 2.5 mg/d, max 15 mg/d), N = 605. Only the rosiglitazone arm is presented in the table.

^fThe starting doses of metformin and sulfonylurea varied by local practice and were adjusted after 8 weeks. Maximum permitted doses were: metformin 2550 mg/d, glibenclamide 15 mg/d, gliclazide 240 mg/d, and glimepiride 4 mg/d. 

^gThis trial included four arms: Bazedoxifene 20 mg/d, N = 1886; Bazedoxifene 40 mg/d, N = 1872; Bazedoxifene 60 mg/d, N = 1849; Placebo, N = 1885. The 20 mg/d and placebo arms are presented in the table.

Table 2.

Socio-demographic characteristics of women included in the eligible randomized controlled trials

		Intervention/ Control
Study ID		Mean age (years)		Mean BMI (kg/m²)	Menopause status	Race, Black (%)	Race, White (%)	Race, Hispanic (%)	Race, Asian (%)	Race, Other (%)
Breast cancer as a primary outcome
Lee 1999		54.6/ 54.6		26.0/ 26.0	NA	NA	NA	NA	NA	NA
Cauley 2001		66.4/ 66.6		25.2/ 25.2	Postmenopause	NA	95.7/ 95.7	NA	NA	NA
Hercberg 2004		46.6/ 46.6		22.8/ 22.9	Mixed	NA	NA	NA	NA	NA
Martino 2004		NA		NA	NA	NA	NA	NA	NA	NA
Anderson 2004		63.6/ 63.6		30.1/ 30.1	Postmenopause	14.7/ 15.4	75.5/75.1	6.1/6.1	1.6/ 1.4	NA
Lee 2005		54.6/ 54.6		26.0/ 26.0	Mixed	NA	NA	NA	NA	NA
Fisher 2005		53.9/ 54.0		NA	NA	1.7/ NA	96.5/ 96.3	NA	NA	NA
Veronesi 2006		51.0/ 51.0		NA	Mixed	NA	NA	NA	NA	NA
Barrett-Connor 2006		67.5/ 67.5		28.8/ 28.7	Postmenopause	NA	84.0/ 84.0	NA	NA	NA
Lappe 2007		66.7a		29.0a	NA	NA	NA	NA	NA	NA
Veronesi 2007		51.0a		NA	NA	NA	NA	NA	NA	NA
Powles 2007		47.0/ 47.0b		NA	Mixed	NA	NA	NA	NA	NA
Cuzick 2007		50.7/ 50.8		27/ 26.9	Mixed	NA	NA	NA	NA	NA
Cummings 2008		68.3/ 68.2		25.7/ 25.7	NA	NA	NA	NA	NA	NA
Chlebowski 2008		62.5/ 62.6		28.1/ 28.0	Postmenopause	9.3/ 9.0	82.8/ 83.4	4.3/ 4.0	2.0/ 2.0	1.2/ 1.2
DeCensi 2009		46.2/ 46.2/ 46.9/ 46.5c		24.1/ 23.1/ 23.7/ 23.1c	Premenopause	NA	NA	NA	NA	NA
LaCroix 2010		67.5/ 67.3/ 67.5d		25.2/ 25.4/ 25.4d	NA	0.9/ 1.0/ 0.9d	74.0/ 73.9/ 74.3d	4.8/ 5.0/ 4.9d	18.6/ 18.2/ 18.3d	1.6/1.8/ 1.6d
Vogel 2010		58.5a		20.0a	Postmenopause	2.5/ 2.4	93.4/ 93.5	2.0/ 2.0	NA	2.1/ 2.1
Goss 2011		62.5/ 62.4b		27.9/ 28.1b	Postmenopause	NA	93.6/ 93.3	NA	NA	NA
Avenell 2012		78.0/ 77.0/ 77.0/ 77.0e		NA	NA	NA	99.4/ 99.1/ 99.4/ 99.1e	NA	NA	NA
Powles 2012		67.5/ 67.4		26.7/ 26.7	Postmenopause	NA	NA	NA	NA	NA
Allred 2012		56.5/ 56.5		NA	NA	NA/ 6.0	85.0/ 87.0	NA	NA	6.0/ 5.0
Cook 2013		54.6/ 54.6		26.0/ 26.0	Mixed	NA	NA	NA	NA	NA
DeCensi 2013		53.5/ 53.1		NA	Postmenopause	NA	NA	NA	NA	NA
DeCensi 2019		54.0/ 54.0		25.7/ 25.3	Mixed	NA	NA	NA	NA	NA
Cuzick 2020		59.5/ 59.4		NA	Postmenopause	NA	NA	NA	NA	NA
Chlebowski 2020		63.6/ 63.2/ 63.6/ 63.3f		30.1/ 28.5	NA	14.7/ 6.4/ 15.4/ 7.1f	75.5/ 84.0/ 75.1/ 84.0f	6.0/ 5.5/ 6.1/ 5.1f	1.6/ 2.3/ 1.4/ 2.1f	NA
Breast cancer as a secondary/other outcome
Downs 1998		62.0/ 63.0		26.4/ 26.4	Postmenopause	NA	NA	NA	NA	NA
Trial 2002		66.4/ 66.3g		29.8/ 29.9g	NA	NA	NA	NA	NA	NA
Miscellany 2002		NA		NA	NA	NA	NA	NA	NA	NA
Trivedi 2003		73.7/ 73.6		24.4/ 24.3	NA	NA	NA	NA	NA	NA
Strandberg 2004		60.5/ 60.5		NA	NA	NA	NA	NA	NA	NA
Von Minckwitz 2011		39.6/ 37.6		22.5/ 24.8	Premenopause	NA	NA	NA	NA	NA
Howell 2018		NA		NA	NA	NA	NA	NA	NA	NA
Breast cancer as an adverse event
Sacks 1996		59.0/ 59.0		28.0/ 28.0	NA	NA	NA	NA	NA	NA
Miscellany 1998		NA		NA	NA	NA	NA	NA	NA	NA
Gregorio 1999		75.7/ 75.4g		NA	NA	NA	NA	NA	NA	NA
Shepherd 2002		75.4/ 75.3		26.8/ 26.8	NA	NA	NA	NA	NA	NA
Kahn 2006		56.3/ 57.9/56.4h		32.2/ 32.1/ 32.2h	NA	4.2/ 3.7/ 4.2h	87.2/ 89.1/ 89.0h	5.2/3.8/ 4.2h	2.7/ 2.4/ 2.2h	0.7/ 1.0/ 0.3h
Home 2009		57.0/ 57.2/ 59.8/ 59.7g, i		32.8/ 32.7/ 30.3/ 30.1g, i	NA	NA	NA	NA	NA	NA
Archer 2009		65.0/ 64.3/ 65.3/ 64.8j		NA	Postmenopause	6.9/ 10.3/ 6.6/ 4.6j	86.9, 80.6/ 84.1/ 86.1j	0.6/ 2.4/ 2.0/ 1.7j	NA	5.6, 6.7/ 7.6/ 7.5j
Erdmann 2014		61.9/ 61.6g		30.7/ 31.0g	NA	NA	NA	NA	NA	NA
Leng 2021		34.4/ 34.4		26.1/ 26.6	Premenopause	NA	NA	NA	100.0/ 100.0	NA

Abbreviations: BMI Body mass index, NA Not available

^aIntervention and control groups combined

^bMedian value reported

^cTamoxifen plus placebo/ Fenretinide plus placebo/ Tamoxifen plus fenretinide/ Placebo plus placebo

^dLasofoxifene 0.25 mg/ Lasofoxifene 0.5 mg/ Placebo

^eVitamin D/ Calcium/ Vitamin D plus calcium/ Placebo

^fConjugated Equine Estrogen (CEE)/ Placebo (CEE-alone trial)/ CEE plus Medroxyprogesterone Acetate (MPA)/ Placebo (CEE plus MPA trial)

^gIncludes both men and women

^hRosiglitazone/ Metformin/ Glyburide

ⁱBackground metformin plus rosiglitazone/ Background metformin plus sulfonylurea/ Background sulfonylurea plus rosiglitazone/ Background sulfonylurea plus metformin

^jBazedoxifene 20 mg/ Bazedoxifene 40 mg/ Bazedoxifene 60 mg/ Placebo

The majority of RCTs presented a low risk of bias. Five RCTs had a high overall risk of bias, mainly due to issues in domain four (outcome measurement) [23, 45, 52, 57, 65] (Fig. S1, Fig. S2). Three RCTs were judged as having some concerns [22, 56, 63]. This suggests that the overall quality of evidence in our analysis is robust, with low risk of systematic errors affecting the interpretation of results.

Network meta-analysis

Overall breast cancer efficacy analysis

The network plot (Fig. 2) represents the medications (including placebo) that were directly compared in RCTs for overall breast cancer outcome, with node size representing the number of participants treated by each medication and edge width representing the number of trials per comparison. The NMA for overall breast cancer incidence included 35 studies evaluating 13 medications. Six medications reduced overall breast cancer risk compared to placebo: sulfonylurea (RR = 0.18, 95% CI = 0.04–0.91), thiazolidinediones (RR = 0.25, 95% CI = 0.08–0.78), third-generation SERMs (RR = 0.46, 95% CI = 0.33–0.66), AIs (RR = 0.50, 95% CI = 0.39–0.66), raloxifene (RR = 0.63, 95% CI = 0.47–0.84), and tamoxifen (RR = 0.76, 95% CI = 0.65–0.88) (Fig. 3, Table S2). We observed no significant inconsistency between direct and indirect comparisons in the NMA (Q = 5.06, d.f. = 7, p = 0.65), supporting the validity of combining different sources of evidence in our analysis. We found low between-study heterogeneity (τ² = 0.0066, I² = 8.9%), which increases confidence in the pooled effect estimates and rankings of medications from the NMA. The NNT values were 44.1 for sulfonylurea, 48.3 for thiazolidinediones, 67.3 for third-generation SERMs, 73.0 for AIs, 96.9 for raloxifene, and 149.7 for tamoxifen. Rankogram analysis provided SUCRA values (Fig. S3). Each rankogram shows the cumulative probabilities of a medication ranking from most effective to least effective based on the NMA results, further supporting the efficacy rankings: sulfonylurea (0.90), thiazolidinediones (0.80), third-generation SERMs (0.62), AIs (0.55), raloxifene (0.37), and tamoxifen (0.23). Comparative analyses indicated that third-generation SERMs (RR = 0.61, 95% CI = 0.42–0.89) and AIs (RR = 0.67, 95% CI = 0.49–0.90) were more effective than tamoxifen in reducing overall breast cancer incidence (Fig. 3, Table S2). The contribution matrix for the six effective medications is presented in Table S3. The matrix quantifies the percentage contribution of each direct comparison to the NMA estimate for every treatment comparison. Comparisons involving tamoxifen, for instance, were mainly based on direct placebo-controlled evidence. In contrast, comparisons of third-generation SERMs and AIs with tamoxifen were informed by indirect evidence through placebo-controlled arms. For example, the comparison of third-generation SERMs versus tamoxifen draws heavily on indirect evidence—mainly through third-generation SERMs vs. placebo (46.3%) and tamoxifen vs. placebo (48.8%), with no direct head-to-head trial informing this estimate. Similarly, the AIs vs. tamoxifen comparison relies on equal contributions (50% each) from AIs vs. placebo and tamoxifen vs. placebo, again highlighting the absence of direct comparisons. In contrast, tamoxifen’s efficacy is based entirely on direct comparisons with placebo (100%), providing strong evidence for its effect size. However, third-generation SERMs and AIs also have direct placebo-controlled data (92.5% and 100%, respectively), and their favorable rankings reflect direct and consistent evidence of greater efficacy. Comparisons between raloxifene and tamoxifen are supported by both direct comparisons (49.2%) and indirect comparisons (47.6%), which help validate that effect estimate.

Fig. 2.

Network of risk-reducing medications for overall breast cancer incidence. All medications were compared with placebo at least once, and some were compared with each other, as indicated by the numbers along the edges. Node size represents the number of participants and edge width represents the marked number of trials. Third-generation SERMs include arzoxifene, bazedoxifene, and lasofoxifene; ARINs include anastrozole and exemestane; SU includes glyburide and gliclazide, TZDs include rosiglitazone and pioglitazone, and statins include pravastatin, simvastatin, and lovastatin. Abbreviations: ARINs, Aromatase inhibitors; CEE, Conjugated equine estrogen; MPA, Medroxyprogesterone acetate; SU, Sulfonylurea; TZDs, Thiazolidinediones; 3rd-gen SERMs, Third-generation selective estrogen receptor modulators

Fig. 3.

Comparative efficacy of risk-reducing medications for overall breast cancer incidence. Treatment vs. placebo: Thirteen risk-reducing medications that have been compared to other medications more than once are compared with placebo by network meta-analysis; Treatment vs. Treatment: Six risk-reducing medications that have significant efficacy (SU, TZDs, 3rd-gen SERMs, ARINs, raloxifene, and tamoxifen) are compared with each other by network meta-analysis. Effects are measured by risk ratios (RRs, ratio of breast cancer incidence in the two arms). Predictive intervals (PrIs) represent the expected range of effects in a future study, considering both the current effect estimate and between-study heterogeneity. Abbreviations: ARINs, Aromatase inhibitors; CI, Confidence interval; PrI, Predictive interval; RR, Risk ratio; SU, Sulfonylurea; TZDs, Thiazolidinediones; 3rd-gen SERMs, Third-generation selective estrogen receptor modulators

Fig S4 shows an extended network plot including all 27 medications and placebo. The medications presented here are the same as those in the overall analysis. However, we have considered the individual medications separately rather than grouping those from the same family and different doses together. Fig. S5 presents the NMA of 18 medications without grouping medications of the same family and different doses. This analysis shows that lasofoxifene is effective (RR = 0.21, 95% CI = 0.08–0.56) compared to placebo for overall breast cancer prevention and shows greater efficacy compared to raloxifene (RR = 0.33, 95% CI = 0.12–0.95) and tamoxifen (RR = 0.27, 95% CI = 0.10–0.73).

Invasive breast cancer and DCIS efficacy analysis

For invasive breast cancer, 16 studies evaluating 10 medications are presented in the network plot (Fig. 4). The NMA included 9 studies and evaluated that AIs (RR = 0.48, 95% CI = 0.33–0.71), tamoxifen (RR = 0.63, 95% CI = 0.51–0.78), and raloxifene (RR = 0.63, 95% CI = 0.47–0.86), were effective compared to placebo (Fig. 4). However, inconsistency was observed in this NMA (Q = 7.48, d.f. = 1, p = 0.0063), with moderate between-study heterogeneity (τ² = 0.0323, I² = 58.8%). This inconsistency suggests that the direct and indirect evidence for these comparisons may not be in complete agreement. The NNT values were 71.3 for AIs, 99.8 for tamoxifen, and 100.8 for raloxifene. The SUCRA values were 0.91 for AIs, 0.54 for tamoxifen, and 0.54 for raloxifene (Fig. S6).

Fig. 4.

Network and comparative efficacy of risk-reducing medications for invasive breast cancer incidence. Top: Network of risk-reducing medications evaluated for invasive breast cancer incidence. Bottom: Three risk-reducing medications (ARINs, tamoxifen, and raloxifene) are compared with placebo by network meta-analysis. All three medications are effective compared with placebo. Abbreviations: ARINs, Aromatase inhibitors; CEE, Conjugated equine estrogen; CI, Confidence interval; PrI, Predictive interval; RR, Risk ratio; 3rd-gen SERMs, Third-generation selective estrogen receptor modulators

For DCIS, 12 studies evaluating 8 medications are presented in the network plot (Fig. S7). The NMA included 7 studies and found that none of the medications showed significant efficacy compared to placebo (Fig. S7). The NMA inconsistency was not significant (Q = 2.29, d.f. = 1, p = 0.1298), although we observed moderate heterogeneity across studies (τ² = 0.1678, I² = 47.9%).

Subgroup analyses

We performed subgroup analyses stratified by whether breast cancer was a primary or secondary/other outcome, menopause status, follow-up time, and duration of intervention. The subgroup analyses used overall breast cancer incidence as the outcome.

In studies where breast cancer was the primary outcome, third-generation SERMs (RR = 0.46, 95% CI = 0.32–0.67), AIs (RR = 0.51, 95% CI = 0.40–0.64), raloxifene (RR = 0.61, 95% CI = 0.46–0.82), and tamoxifen (RR = 0.76, 95% CI = 0.66–0.86) were effective compared to placebo. However, in studies where breast cancer was considered a secondary/other outcome, only thiazolidinediones (RR = 0.25, 95% CI = 0.07–0.84) were effective compared to placebo. Sulfonylurea were evaluated in two studies with breast cancer as a secondary/other outcome and were not effective (Fig. 5).

Fig. 5.

Comparative efficacy of risk-reducing medications in studies with breast cancer as a primary vs. secondary/other outcome. Sixteen studies evaluated breast cancer as a primary outcome, with four effective medications identified. Fourteen studies evaluated breast cancer as a secondary/other outcome with four medications evaluated, and only TZDs were effective. Sulfonylurea was only evaluated in two studies with breast cancer as a secondary/other outcome. Abbreviations: ARINs, Aromatase inhibitors; BC, Breast cancer; CI, Confidence interval; PrI, Predictive interval; RR, Risk ratio; TZDs, Thiazolidinediones; 3rd-gen SERMs, Third-generation selective estrogen receptor modulators

In studies that included only postmenopausal women, third-generation SERMs (RR = 0.44, 95% CI = 0.29–0.67), AIs (RR = 0.51, 95% CI = 0.40–0.64), raloxifene (RR = 0.70, 95% CI = 0.50–0.97), and tamoxifen (RR = 0.74, 95% CI = 0.60–0.90) were effective compared to placebo. In contrast, only tamoxifen was assessed in studies involving premenopausal women (Fig. S8).

In studies with ≥ 5 years of follow-up (only 5 studies had shorter follow-up) and breast cancer as a primary outcome, AIs (RR = 0.52, 95% CI = 0.36–0.75), third-generation SERMs (RR = 0.52, 95% CI = 0.28–0.97), and tamoxifen (RR = 0.76, 95% CI = 0.63–0.91) were effective compared to placebo (Fig. S9).

In studies with an intervention duration ≥ 5 years, third-generation SERMs (RR = 0.42, 95% CI = 0.25–0.68), AIs (RR = 0.51, 95% CI = 0.40–0.64), and tamoxifen (RR = 0.75, 95% CI = 0.66–0.85) were effective compared to placebo (Fig. S10).

Discussion

We performed a comprehensive NMA of 43 RCTs assessing 13 medications for primary breast cancer prevention. Findings support the effectiveness of currently approved agents, including tamoxifen, raloxifene, and AIs, while also revealing promising results for third-generation SERMs and thiazolidinediones. These findings provide a detailed hierarchy of the most effective preventive interventions and offer valuable insights for clinical decision-making and future research in breast cancer prevention. In addition to our main findings, we confirmed that most evidence for risk-reducing medications to date is in postmenopausal women, with only tamoxifen being evaluated in premenopausal women, highlighting an urgent need for more prevention trials in younger women.

Results from this NMA align with and expand upon previous systematic reviews and meta-analyses in this field. The 2019 Cochrane review by Mocellin et al., focused on approved medications, including tamoxifen, raloxifene, and AIs [15]. The NMA showed that tamoxifen and AIs reduce the risk of breast cancer in women at above-average risk and suggested that AIs could be more effective than tamoxifen (RR = 0.67, 95% CI = 0.46–0.98). Our study confirms these findings while providing additional information by ranking the efficacy of these agents and including a broader range of medications, secondary outcomes (invasive breast cancer and DCIS), and comprehensive subgroup analyses. These updates allow for a better understanding of the populations most likely to benefit from each medication. The previous review did not include third-generation SERMs or the antidiabetic agents highlighted in our analysis. However, our study did not compare toxicity outcomes, which were assessed in the 2019 Cochrane review.

In addition to this, the 2016 NMA by Mocellin et al. identified AIs, arzoxifene, lasofoxifene, raloxifene, tamoxifen, and tibolone to be effective for breast cancer prevention, with AIs to be more effective than SERMs. Their reported RRs for lasofoxifene (RR = 0.208), arzoxifene (RR = 0.415), AIs (RR = 0.468), raloxifene (RR = 0.572), and tamoxifen (RR = 0.708) align closely with our findings [14]. While their NMA included six medications, our analysis expanded theirs by including 13 medications, identifying six as effective for breast cancer prevention. In addition, we included an expanded and updated dataset (337,240 women vs. 271,161), performed a more comprehensive analysis that evaluated invasive breast cancer and DCIS outcomes, and incorporated detailed subgroup analyses. This allowed us to demonstrate, for example, that third-generation SERMs and AIs are effective when breast cancer is a primary outcome, while thiazolidinediones show efficacy when breast cancer is a secondary/other outcome. The 2016 study ranked agents by both efficacy and acceptability, finding that third-generation SERMs (arzoxifene, lasofoxifene, raloxifene) offered the best benefit-risk ratio, followed by AIs and tamoxifen. However, our study focused on efficacy metrics.

Our findings on the efficacy of third-generation SERMs, including lasofoxifene, arzoxifene, and bazedoxifene, expand on findings from previous studies and demonstrate their potential. Mocellin et al. also found arzoxifene and lasofoxifene to be promising [14]. While bazedoxifene has not been evaluated in RCTs with breast cancer as a primary outcome and should be evaluated, newer SERMs in general may offer both improved efficacy and better side effect profiles compared to first- and second-generation SERMs. These results suggest the potential for personalized use of third-generation SERMs in appropriate candidates, mainly based on their safety benefits and patient preferences.

The identification of thiazolidinediones as potentially effective medications for breast cancer prevention would have been difficult to ascertain from pairwise meta-analyses alone and is an important finding showing the value of our NMA. These emerged in our NMA for overall breast cancer incidence, yet breast cancer was a secondary outcome in these diabetes trials. Analyzing studies where breast cancer was a secondary/other outcome, sulfonylurea was not found to be effective, but thiazolidinediones were effective. Thiazolidinediones, primarily used as antidiabetic medications, act as peroxisome proliferator-activated receptor gamma (PPAR-γ) agonists and have shown potential anticancer effects mainly through apoptosis and cell cycle regulatory proteins [66]. While previous studies have suggested mixed associations between thiazolidinediones and breast cancer risk in diabetic patients [67, 68], our findings support the potential for clinical trials investigating their use for breast cancer prevention in high-risk women. These medications, already used for diabetes management, have well-established safety profiles, potentially facilitating their investigation in breast cancer prevention. These results should be interpreted cautiously due to the limited number of studies and different settings; however, they open new avenues for research and could potentially lead to the development of novel preventive strategies. Dedicated prevention trials are essential before clinical applications.

Our examination of secondary outcomes revealed that, in contrast to invasive breast cancer, none of the medications are effective for DCIS prevention. While tamoxifen has shown promise [65, 69], our results do not confirm its benefit. This finding indicates a need for further research into preventive options specifically for DCIS. The moderate heterogeneity we observed in this analysis suggests variability in study results, emphasizing the necessity for more focused research on DCIS prevention.

Given the heterogeneity in breast cancer prevention trials (including differences in menopause status, trial setting, follow-up, and intervention duration), we addressed this by extensive subgroup analyses. Subgroup analysis stratified by menopause status revealed third-generation SERMs, AIs, raloxifene, and tamoxifen significantly reduce risk in postmenopausal women, consistent with current guidelines [70]. However, our analysis also highlights a significant gap: the limited research on risk-reducing medications for premenopausal women. Among the evaluated medications, only tamoxifen has been assessed in premenopausal women, which underscores the urgent need for more studies focusing on breast cancer prevention strategies for younger, premenopausal women at high risk. The efficacy of thiazolidinediones in studies where breast cancer was a secondary outcome suggests a potential for repurposing antidiabetic drugs for breast cancer prevention, although further trials are needed. Studies with follow-up ≥ 5 years confirmed the efficacy of AIs, third-generation SERMs, and tamoxifen, which highlights the durability of risk reduction with these medications over longer periods of follow-up. In studies where the intervention duration was ≥ 5 years, third-generation SERMs, AIs, and tamoxifen were effective compared to placebo; but there was limited evidence to suggest an effect when used for < 5 years. Overall, our findings highlight the importance of considering trial design, patient population, and intervention duration when selecting risk-reducing medications for breast cancer prevention. They also emphasize the need for more research in premenopausal women and for studies that evaluate newer agents in diverse populations and across longer follow-ups.

Our study has several strengths. First, we provide a comprehensive and updated analysis of risk-reducing medications, presenting data from 43 RCTs and evaluating 13 different medications. This wide-range comparison improves our understanding of the relative efficacies of different risk-reducing medications and represents, to our knowledge, the first such comprehensive comparison. We applied SUCRA for medication ranking and calculated NNT values that provide clinically relevant context for decision-making. Our detailed subgroup analyses based on menopause status, RCT setting, follow-up, and intervention duration provide more tailored insights into the efficacy of medications. In addition to overall breast cancer incidence, we evaluated the efficacy of medications in reducing invasive breast cancer and DCIS risk. However, our study has limitations. Despite our extensive search strategy, it’s possible that some relevant RCTs were missed and not included in our analysis. While NMA integrates both direct and indirect comparisons to estimate relative effects among medications, this approach may lead to heterogeneity, and as a result, the findings should be interpreted with caution. Finally, while CEE and tibolone have been reported to have breast cancer prevention benefits, they could not be included in NMA because they lacked sufficient network connectivity, though they have shown promise in other contexts.

Future studies should aim to address the limitations of previously conducted RCTs, such as the relatively short follow-ups, which may not be sufficient to fully capture the long-term effects of the evaluated medications, especially newer agents. Additionally, most women in the included studies were non-Hispanic White, potentially limiting the generalizability to other racial/ethnic groups. Future research should prioritize more diverse recruitment. Additionally, clinical trials investigating the potential of thiazolidinediones for breast cancer prevention in high-risk women could be considered. The potential for combination strategies, given the different mechanisms of action between thiazolidinediones and current approved medications, should also be evaluated. Also, the limited availability of survival data and the lack of survival benefit in existing RCTs is a limitation in assessing the full impact of risk-reducing medications.

In conclusion, this NMA confirms the efficacy of established medications and identifies promising new options for breast cancer prevention. Notably, third-generation SERMs and thiazolidinediones, which are not currently included in clinical guidelines, offer significant potential. Also, there is a critical need for more studies in premenopausal women and racially diverse populations. These insights should inform clinical practice and guide future research.

Authors’ contributions

ATT and GP conceptualized the study. AH conducted the comprehensive literature search and assisted with data organization. ML, SSS, and GP performed the screening of studies and data extraction. CL conducted the statistical analysis and prepared the figures. GP and ATT interpreted the findings. ML and SSS assessed the risk of bias and prepared the tables. GP prepared the first draft of the manuscript. All authors critically revised the manuscript and approved the final version for submission.

Funding

This study was supported by NIH/NCI R37CA235602. The funders had no role in the design of the study, data collection, analysis, decision to publish, or preparation of the manuscript.

Data availability

The data analyzed in this study and the code will be made available upon request.

Declarations

Ethics approval and consent to participate

Not applicable.

Competing interests

The authors declare no competing interests.