Introduction
Antipsychotic medications are widely prescribed for people living with mental disorders such as schizophrenia, bipolar disorder, major depressive disorder and dementia, with an increasing trend of off-label use also observed worldwide in recent decades (Hálfdánarson et al., Reference Hálfdánarson, Zoëga, Aagaard, Bernardo, Brandt, Fusté, Furu, Garuoliené, Hoffmann, Huybrechts, Kalverdijk, Kawakami, Kieler, Kinoshita, Litchfield, López, Machado-Alba, Machado-Duque, Mahesri, Nishtala, Pearson, Reutfors, Saastamoinen, Sato, Schuiling-Veninga, Shyu, Skurtveit, Verdoux, Wang, Yahni and Bachmann2017; Ng et al., Reference Ng, Man, Gao, Chan, Lee, Hayes and Wong2021). Despite a more tolerable safety profile of second-generation antipsychotic medications (Herrmann et al., Reference Herrmann, Mamdani and Lanctôt2004), metabolic and endocrinologic abnormalities associated with antipsychotic use have been observed (De Hert et al., Reference De Hert, Detraux, van Winkel, Yu and Correll2012). These abnormalities may represent pathomechanisms underlying the known association of antipsychotic use with a range of relatively rare adverse events such as stroke and myocardial infarction (Douglas and Smeeth, Reference Douglas and Smeeth2008; Lai et al., Reference Lai, Guthrie, Mercer, Smith, Yip, Chung, Lee, Chung, Chau, Wong, Yeoh and Wong2020).
Some studies have also reported an elevated cancer incidence related to the use of antipsychotics (Dalton et al., Reference Dalton, Johansen, Poulsen, Nørgaard, Sørensen, McLaughlin, Mortensen and Friis2006; Nielsen et al., Reference Nielsen, Lolk, Rodrigo-Domingo, Valentin and Andersen2017). It has been shown women living with schizophrenia and bipolar disorder have a higher risk of developing breast cancer compared with the general population (Chou et al., Reference Chou, Wang, Lin and Kao2017; Anmella et al., Reference Anmella, Fico, Lotfaliany, Hidalgo-Mazzei, Soto-Angona, Giménez-Palomo, Amoretti, Murru, Radua, Solanes, Pacchiarotti, Verdolini, Cowdery, Dodd, Williams, Mohebbi, Carvalho, Kessing, Vieta and Berk2021) and antipsychotic use may potentially explain at least part of this increased risk. This is supported by a widely adopted working hypothesis of the hyperprolactinaemia-inducing property of certain antipsychotics such as pimozide, risperidone and clomipramine (De Hert et al., Reference De Hert, Peuskens, Sabbe, Mitchell, Stubbs, Neven, Wildiers and Detraux2016b; Johnston et al., Reference Johnston, Bu, Hein, Garcia, Camacho, Xue, Qin, Nagi, Hilsenbeck, Kapali, Podsypanina, Nangia and Li2018). Other possible mechanisms may include poorer lifestyles regarding self-care and health consciousness among antipsychotic users (Bly et al., Reference Bly, Taylor, Dalack, Pop-Busui, Burghardt, Evans, McInnis, Grove, Brook, Zöllner and Ellingrod2014), as well as the commonly reported antipsychotic-mediated weight gain (Balt et al., Reference Balt, Galloway, Baggott, Schwartz and Mendelson2011). With complex mechanisms and likely multiple interacting risk factors, existing evidence remains inconclusive, and no definitive conclusion could be drawn regarding this association. Furthermore, although safety monitoring is an integral component of randomised controlled trials, the study design's inherent weaknesses such as insufficient sample size for rare outcomes, discrepancies in adverse event reporting and inadequate follow-up period to capture cancer incidence (Hughes et al., Reference Hughes, Cohen and Jaggi2014; Phillips et al., Reference Phillips, Hazell, Sauzet and Cornelius2019) pose as a challenge to investigate this association. Longitudinal observational data are therefore considered much more suitable for this enquiry.
A synthesis of the existing published data is important to inform clinical practices with regards to the prescription of antipsychotic medications in consideration of the potentially elevated risk of breast cancer. This synthesis will inform the risk–benefit assessment of antipsychotic use in facilitation of an optimal prescription decision and treatment outcome. In this study, we aim to systematically review and conduct a meta-analysis on the existing evidence to determine the association of antipsychotic use with breast cancer.
Methods
Search strategy and eligibility
We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist in conducting this review (Page et al., Reference Page, McKenzie, Bossuyt, Boutron, Hoffmann, Mulrow, Shamseer, Tetzlaff, Akl, Brennan, Chou, Glanville, Grimshaw, Hróbjartsson, Lalu, Li, Loder, Mayo-Wilson, McDonald, McGuinness, Stewart, Thomas, Tricco, Welch, Whiting and Moher2021). As this meta-analysis was based on published data, ethics approval was not required. In accordance with a protocol registered with PROSPERO (Ref: CRD42022307913), we performed preliminary scoping searches to identify databases with substantial pharmacoepidemiologic evidence on the topic. Based on the results of our preliminary searches, we conducted a systematic search of articles published in English in peer-reviewed scholarly journals in respective electronic databases, namely PubMed, Embase and Web of Science™ from inception. The last search was conducted on 27 January 2022. The search strategy was developed based on two subjects: antipsychotics and breast cancer. Search terms and combinations of Medical Subject Headings (MeSH), keywords and text words were derived from previously published systematic reviews (Moja et al., Reference Moja, Tagliabue, Balduzzi, Parmelli, Pistotti, Guarneri and D'Amico2012; Indave et al., Reference Indave, Minozzi, Pani and Amato2016; Krause et al., Reference Krause, Zhu, Huhn, Schneider-Thoma, Bighelli, Nikolakopoulou and Leucht2018) on the two subjects and were selected for each database to optimise sensitivity and specificity of the search. Hand searches through the reference lists of included articles were conducted to avoid the omission of relevant research. For details of specific search keywords and strategies, refer to online Supplementary eTable 1.
All published cohort and case-control observational studies that investigated and quantified the association of antipsychotic use (v. non-use) with breast cancer in individuals aged 16 or above were considered for inclusion in the review. Studies were excluded if they were not published in English, had a study design that was neither cohort nor case-control, included participants who developed breast cancer prior to antipsychotic exposure or did not compare antipsychotic use to non-use, such as comparing between different classes of antipsychotics.
Extraction
Study eligibility was independently determined by JCNL and DWYN. Cohen's kappa was computed to indicate interrater reliability. Data extraction was completed simultaneously using a standardised data extraction form. Data regarding the context, population, intervention, outcome and measures of association of each study were extracted and recorded in the form. Discrepancies were reconciled through discussion and consultation with a senior author (FTTL).
Quality assessment of included studies
The methodological quality of each included study was assessed using the Newcastle-Ottawa Scale (NOS). Like the data extraction procedure, the quality assessment was conducted independently by JCNL and DWYN. Study quality was indicated by numbers of stars, with nine representing the highest possible methodological rigour. See online Supplementary eTable 3 for details of the quality assessment procedures. Cohen's kappa was not calculated for the quality assessment decisions, as nine studies were included and there were only a few discrepancies, which were resolved through in-depth discussions.
Pooled estimates
Upon satisfactory assessment result with regards to multivariable adjustment according to the NOS, meta-analyses of the estimates of the association, i.e. odds ratios (ORs) and hazard ratios (HRs), were conducted. Stratified by study design, i.e. cohort and case-control studies, the estimates of the association of antipsychotic use and breast cancer were pooled using a random effects model. The exposure was binarily operationalised as any antipsychotic use compared with non-use. In cases where this operationalisation was not possible, the longest-term exposure category, or the category representing the farthest extent of antipsychotic use, were used in comparison with non-use in the pooled estimates. The inverse variance weighting method was used to determine the relative importance between studies while the I 2 statistic was used to examine the heterogeneity of the estimates across studies. Upon a sufficient number of included studies, the Egger's regression test was conducted to detect any publication bias in the pooled estimates. The pooled estimates and test for heterogeneity were implemented using Cochrane Collaboration Review Manager (Version 5.4.1).
Results
As shown in Fig. 1, upon initial search, we retrieved a total of 2549 articles from electronic databases, of which 441 were removed as duplicates. The title and abstract screening process further excluded 2036 articles published in non-English languages, using a study design other than cohort or case-control, not adopting breast cancer as the outcome or not using antipsychotic use as the exposure. After carefully examining the eligibility of the remaining 72 articles by full-text, nine studies (N = 2 031 380) were included for a qualitative synthesis and quality assessment. Cohen's kappa for title and abstract screening [0.496, 95% confidence interval (CI) 0.404–0.588] and full-text selection (0.742, 95% CI 0.567–0.917) suggest moderate and substantial agreement respectively. Two studies were excluded from the meta-analysis (Mortensen, Reference Mortensen1987; Dalton et al., Reference Dalton, Johansen, Poulsen, Nørgaard, Sørensen, McLaughlin, Mortensen and Friis2006), as the effect measures summarising the association were incomparable to that of the other studies and the use of incompatible statistical methods. Seven included studies (N = 1 557 013) provided adequate data for a pooled estimate of the hypothesised association. Study characteristics and results, as well as quality assessment scores are tabulated in Tables 1 and 2.
Fig. 1. Flow chart of article selection.
Table 1. Characteristics and results of the critical appraisal of included studies (N = 9)
C, cohort; CC, case-control; GPRD, General Practice Research Database; LHID2000, Longitudinal Health Insurance Database 2000; RCIPD, Registry for Catastrophic Illness Patient Database; CPR, Central Population Register; WHI, Women's Health Initiative cohort; NJ Medicaid, New Jersey Medicaid; PAAD, New Jersey Pharmaceutical Assistance to the Aged and Disabled; NJ Medicare, New Jersey Medicare; NJ Cancer Registry, New Jersey Cancer Registry; AP, antipsychotics; BC, breast cancer; ICD, International Classification of Diseases.
Table 2. Results of included studies (N = 9)
FGA, first-generation antipsychotics; SGA, second-generation antipsychotics; ATC N05A, Anatomical Therapeutic Chemical Classification System (Antipsychotics); AP, antipsychotics; BC, breast cancer; HT, hormone replacement therapy; NSAID, nonsteroidal anti-inflammatory drugs; COPD, chronic obstructive pulmonary disease; WHI, Women's Health Initiative cohort; ECT, electroconvulsive therapy.
Study characteristics
The included studies have been conducted in five countries/jurisdictions: three studies in the United States (Wang et al., Reference Wang, Walker, Tsuang, Orav, Glynn, Levin and Avorn2002; George et al., Reference George, Sturgeon, Hankinson, Shadyab, Wallace and Reeves2020; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022), three studies in Denmark (Mortensen, Reference Mortensen1987; Dalton et al., Reference Dalton, Johansen, Poulsen, Nørgaard, Sørensen, McLaughlin, Mortensen and Friis2006; Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018) and one study each in Finland (Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021), Taiwan (Chou et al., Reference Chou, Wang, Lin and Kao2017) and the United Kingdom (Hippisley-Cox et al., Reference Hippisley-Cox, Vinogradova, Coupland and Parker2007). Of the nine studies, five were cohort studies (Wang et al., Reference Wang, Walker, Tsuang, Orav, Glynn, Levin and Avorn2002; Dalton et al., Reference Dalton, Johansen, Poulsen, Nørgaard, Sørensen, McLaughlin, Mortensen and Friis2006; Chou et al., Reference Chou, Wang, Lin and Kao2017; George et al., Reference George, Sturgeon, Hankinson, Shadyab, Wallace and Reeves2020; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022) and four were case-control studies (Mortensen, Reference Mortensen1987; Hippisley-Cox et al., Reference Hippisley-Cox, Vinogradova, Coupland and Parker2007; Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018; Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021). The study sample sizes range from 120 (Mortensen, Reference Mortensen1987) to over 0.6 million (Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018) individuals. All studies received a moderate to high score in the quality assessment ranging from seven to nine stars based on the criteria of NOS. Six studies (Wang et al., Reference Wang, Walker, Tsuang, Orav, Glynn, Levin and Avorn2002; Hippisley-Cox et al., Reference Hippisley-Cox, Vinogradova, Coupland and Parker2007; Chou et al., Reference Chou, Wang, Lin and Kao2017; Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018; Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022) reported a significant association between antipsychotic use (various operationalisations) and breast cancer development.
Outcome – breast cancer
All nine studies defined the outcome of interest as the first-time diagnosis of breast cancer, with five studies specifying the adopted diagnosis explicitly based on International Classification of Diseases (ICD) (Wang et al., Reference Wang, Walker, Tsuang, Orav, Glynn, Levin and Avorn2002; Chou et al., Reference Chou, Wang, Lin and Kao2017; Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018; Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022), one of which also identified first claims of breast cancer surgeries without an ICD code diagnosis as cases (Wang et al., Reference Wang, Walker, Tsuang, Orav, Glynn, Levin and Avorn2002). Either surgery, chemotherapy or hospitalisation for breast cancer in addition to the diagnosis with ICD code was adopted for one study (Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022). Three studies (Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018; Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022) used a histological or pathologic verification for the breast cancer diagnosis. Post-mortem diagnosis of breast cancer in cases who died was also used to define cases in a case-control study (Hippisley-Cox et al., Reference Hippisley-Cox, Vinogradova, Coupland and Parker2007).
Three studies (Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018; Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022) that included additional verification like histology received at least eight out of nine stars in the quality assessment. All three studies (Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018; Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022) reported a significant association. All five studies (Wang et al., Reference Wang, Walker, Tsuang, Orav, Glynn, Levin and Avorn2002; Chou et al., Reference Chou, Wang, Lin and Kao2017; Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018; Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022) that specified the diagnosis based on ICD codes supported the association. From the remaining studies that received quality assessment scores ranging from seven to nine stars (Mortensen, Reference Mortensen1987; Dalton et al., Reference Dalton, Johansen, Poulsen, Nørgaard, Sørensen, McLaughlin, Mortensen and Friis2006; Hippisley-Cox et al., Reference Hippisley-Cox, Vinogradova, Coupland and Parker2007; Chou et al., Reference Chou, Wang, Lin and Kao2017; George et al., Reference George, Sturgeon, Hankinson, Shadyab, Wallace and Reeves2020), both association and non-association were observed.
Confounder adjustment
Confounder adjustment applied in nine studies can be summarised into three main categories, namely clinical history; lifestyle and socioeconomic factors. All nine studies adjusted for covariates related to age and clinical history. In particular, the use of drugs known or suspected to modify breast cancer risk such as lithium, oral contraceptives or hormone replacement therapy were adjusted in seven out of nine studies (Dalton et al., Reference Dalton, Johansen, Poulsen, Nørgaard, Sørensen, McLaughlin, Mortensen and Friis2006; Hippisley-Cox et al., Reference Hippisley-Cox, Vinogradova, Coupland and Parker2007; Chou et al., Reference Chou, Wang, Lin and Kao2017; Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018; George et al., Reference George, Sturgeon, Hankinson, Shadyab, Wallace and Reeves2020; Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022).
Adjusted lifestyle factors include obesity, smoking, body mass index (BMI) and substance misuse. Five of the nine studies had made such adjustments (Mortensen, Reference Mortensen1987; Wang et al., Reference Wang, Walker, Tsuang, Orav, Glynn, Levin and Avorn2002; Hippisley-Cox et al., Reference Hippisley-Cox, Vinogradova, Coupland and Parker2007; Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022), of which three (Wang et al., Reference Wang, Walker, Tsuang, Orav, Glynn, Levin and Avorn2002; Hippisley-Cox et al., Reference Hippisley-Cox, Vinogradova, Coupland and Parker2007; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022) had adjusted for obesity – suggested to be associated with an increased risk of breast cancer (Iyengar et al., Reference Iyengar, Arthur, Manson, Chlebowski, Kroenke, Peterson, Cheng, Feliciano, Lane, Luo, Nassir, Pan, Wassertheil-Smoller, Kamensky, Rohan and Dannenberg2019), whilst substance misuse or smoking have been adjusted in four studies (Mortensen, Reference Mortensen1987; Hippisley-Cox et al., Reference Hippisley-Cox, Vinogradova, Coupland and Parker2007; Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022). Of the five studies with adjustment for lifestyle factors, four studies reported a significant association between antipsychotic use and breast cancer risk.
Socioeconomic factors were mostly represented by occupation, income, education status or a summarised Townsend score. Six studies (Mortensen, Reference Mortensen1987; Wang et al., Reference Wang, Walker, Tsuang, Orav, Glynn, Levin and Avorn2002; Hippisley-Cox et al., Reference Hippisley-Cox, Vinogradova, Coupland and Parker2007; Chou et al., Reference Chou, Wang, Lin and Kao2017; Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022) adjusted for socioeconomic status, of which five (Wang et al., Reference Wang, Walker, Tsuang, Orav, Glynn, Levin and Avorn2002; Hippisley-Cox et al., Reference Hippisley-Cox, Vinogradova, Coupland and Parker2007; Chou et al., Reference Chou, Wang, Lin and Kao2017; Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022) reported a significant association between antipsychotic use and breast cancer risk.
Exposure – antipsychotic use
Antipsychotic use was defined with electronic records in eight out of the nine studies (Mortensen, Reference Mortensen1987; Wang et al., Reference Wang, Walker, Tsuang, Orav, Glynn, Levin and Avorn2002; Dalton et al., Reference Dalton, Johansen, Poulsen, Nørgaard, Sørensen, McLaughlin, Mortensen and Friis2006; Hippisley-Cox et al., Reference Hippisley-Cox, Vinogradova, Coupland and Parker2007; Chou et al., Reference Chou, Wang, Lin and Kao2017; Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018; Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022), the remaining study (George et al., Reference George, Sturgeon, Hankinson, Shadyab, Wallace and Reeves2020) used self-reported antipsychotic use to determine the exposure group. All studies took any antipsychotic use into account. Exposure durations were specified in three studies, Wang et al. included participants with at least 3 months' exposure to antipsychotics prior to the index date from which the follow-up started (Wang et al., Reference Wang, Walker, Tsuang, Orav, Glynn, Levin and Avorn2002); Dalton et al. only included participants who had received at least two prescriptions (Dalton et al., Reference Dalton, Johansen, Poulsen, Nørgaard, Sørensen, McLaughlin, Mortensen and Friis2006); and Taipale et al. considered participants with prior antipsychotic exposure until 1 year before breast cancer diagnosis, with a case control design (Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021).
The following variables were used to represent the extent of exposure for further stratification of the exposed group: cumulative doses (Wang et al., Reference Wang, Walker, Tsuang, Orav, Glynn, Levin and Avorn2002; Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018), average yearly dosage (Mortensen, Reference Mortensen1987; Chou et al., Reference Chou, Wang, Lin and Kao2017), prescription count (Dalton et al., Reference Dalton, Johansen, Poulsen, Nørgaard, Sørensen, McLaughlin, Mortensen and Friis2006), duration (Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021) and prolactin-elevating propensity (Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022). Two remaining studies (Hippisley-Cox et al., Reference Hippisley-Cox, Vinogradova, Coupland and Parker2007; George et al., Reference George, Sturgeon, Hankinson, Shadyab, Wallace and Reeves2020) included participants with any use of antipsychotics without further stratifying by the extent of exposure in their exposed groups. Of the two studies that did not stratify participants by the extent of exposure, one reported a significant association (OR 1.55, 95% CI 1.08–2.23) (Hippisley-Cox et al., Reference Hippisley-Cox, Vinogradova, Coupland and Parker2007). Five out of the seven studies (Wang et al., Reference Wang, Walker, Tsuang, Orav, Glynn, Levin and Avorn2002; Chou et al., Reference Chou, Wang, Lin and Kao2017; Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018; Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022) that stratified participants by the extent of exposure reported significant associations of antipsychotic use with breast cancer.
Despite a null association with the exposure defined as any antipsychotic use, long-term use (defined as having a cumulative dose of over 10 000 mg of olanzapine equivalents) was found to have a small association with breast cancer development in Pottegård et al. (Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018). An increased risk with prolonged exposure was also suggested in two other studies (Wang et al., Reference Wang, Walker, Tsuang, Orav, Glynn, Levin and Avorn2002; Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021). Taipale et al. reported ORs 1.18 (95% CI 0.86–1.62) for 1–4 years of antipsychotic use and 1.74 (95% CI 1.38–2.21) for at least 5 years of antipsychotic use (Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021), and Wang et al. showed an increased risk with at least 6 years of antipsychotic exposure (HR 2.37, 95% CI 1.25–4.47), whereas breast cancer risk amongst antipsychotic users of less than 6 years were reported to be non-significant. In contrast, the dose–response relationship was not observed in the atypical antipsychotic subgroup of Chou et al., where an apparent association was observed with lower exposure instead of increased exposure. They reported HRs 2.49 (95% CI 1.69–3.66) and 1.05 (95% CI 0.58–1.87) for mean antipsychotic exposure of less than 28 and greater than 245 g/year, respectively.
Some studies have also investigated the prolactin-elevating properties of antipsychotics and its association with breast cancer development (Chou et al., Reference Chou, Wang, Lin and Kao2017; Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018; Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022). Exposure to antipsychotics with prolactin-elevating properties were included in Pottegård et al. (Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018), to which long-term exposure showed an increased risk of breast cancer. Rahman et al. grouped exposure according to prolactin-elevating propensity into three categories of low, medium and high propensity. They reported that users of antipsychotics with medium and high prolactin-elevating properties were significantly associated with breast cancer development (Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022). Taipale et al. (Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021) compared prolonged periods of prolactin-increasing antipsychotic use to those exposed for less than a year. The results showed an increased risk amongst those exposed for at least 5 years (OR 1.56, 95% CI 1.27–1.92), corresponding to the results seen in Pottegård et al. Prolactin-elevating antipsychotics reported in Chou et al. were defined as risperidone, paliperidone or amisulpride, the study compared schizophrenia patients exposed to said antipsychotics to a non-schizophrenia cohort as the non-exposed comparator, the results indicate a significant association in the use of the three prolactin-elevating antipsychotics with breast cancer development (HR 1.96, 95% CI 1.36–2.82) (Chou et al., Reference Chou, Wang, Lin and Kao2017).
Quality assessment scores
All nine studies received a satisfactory quality assessment score of seven to nine stars (Wang et al., Reference Wang, Walker, Tsuang, Orav, Glynn, Levin and Avorn2002; Dalton et al., Reference Dalton, Johansen, Poulsen, Nørgaard, Sørensen, McLaughlin, Mortensen and Friis2006; Hippisley-Cox et al., Reference Hippisley-Cox, Vinogradova, Coupland and Parker2007; Chou et al., Reference Chou, Wang, Lin and Kao2017; Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018; George et al., Reference George, Sturgeon, Hankinson, Shadyab, Wallace and Reeves2020; Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022). One case-control study (Mortensen, Reference Mortensen1987) received a lower score of two out of four stars in regards to the selection of cases and controls and the limited representativeness of the cases due to its small sample size. All studies had adjusted for both age and other covariates associated with the risk of breast cancer such as comorbidity or concurrent medication.
Pooled estimates of the association
Using a random effects model, we pooled the HRs and ORs of breast cancer between antipsychotic users and non-users from four cohort studies (Wang et al., Reference Wang, Walker, Tsuang, Orav, Glynn, Levin and Avorn2002; Chou et al., Reference Chou, Wang, Lin and Kao2017; George et al., Reference George, Sturgeon, Hankinson, Shadyab, Wallace and Reeves2020; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022) and three case-control studies (Hippisley-Cox et al., Reference Hippisley-Cox, Vinogradova, Coupland and Parker2007; Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018; Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021) respectively, with the I 2 estimated at 75 and 93%. Figures 2 and 3 show the forest plots for the pooled estimate as well as the estimated ratios reported by individual studies. Results suggest a moderate association of antipsychotic use (v. non-use) with breast cancer with a >30% increased risk observed, although the pooled OR did not reach statistical significance (HR 1.39, 95% CI 1.11–1.73; OR 1.37, 95% CI 0.90–2.09). As only three and four studies were included in the pooled estimates of the OR and HR, we did not conduct the Egger's regression test for publication bias.
Fig. 2. Forest plot showing HRs generated from retrieved individual cohort studies (n = 4) using Cox proportional hazard models and the pooled HR. For George et al. (Reference George, Sturgeon, Hankinson, Shadyab, Wallace and Reeves2020), the HR for atypical antipsychotic use and invasive breast cancer was used.
Fig. 3. Forest plot showing ORs generated from retrieved individual case-control studies (n = 3) using logistic regression and the pooled OR.
As one of the cohort studies (George et al., Reference George, Sturgeon, Hankinson, Shadyab, Wallace and Reeves2020) stratified the analysis by typical/atypical antipsychotics and invasive/in-situ breast cancer, we included the HR for atypical antipsychotics and invasive breast cancer in that study for the pooled estimate and replicated the analysis with all three other HRs separately as a sensitivity analysis to test for the robustness of the results. No substantial difference was observed as shown in online Supplementary eTable 2.
Discussion
Results of this review support the association between the use of antipsychotic medications and an increased risk of breast cancer. Six out of nine included studies of a good quality reported a significant association. Evidence shows a further extent of exposure to antipsychotics, such as a longer duration of use, is associated with a higher risk of breast cancer, particularly for antipsychotics with prolactin-elevating properties. Outcome definition, exposure operationalisation and quality assessment score did not have a noticeable effect on the difference in results between the studies. From the meta-analysis, we estimated a moderate positive association of antipsychotic use and breast cancer with a >30% elevated risk.
Antipsychotics are dopamine receptor antagonists and prohibit the binding of dopamine to dopamine D2 receptors (D2R), this action increases prolactin secretion (Besnard et al., Reference Besnard, Auclair, Callery, Gabriel-Bordenave and Roberge2013). Typical antipsychotics were reported to have higher occurrences of elevated serum prolactin levels (hyperprolactinaemia) in comparison with atypical antipsychotic users (Madhusoodanan et al., Reference Madhusoodanan, Parida and Jimenez2010; Manu, Reference Manu2012; Vuk Pisk et al., Reference Vuk Pisk, Matić, Gereš, Ivezić, Ruljančić and Filipčić2019; Dehelean et al., Reference Dehelean, Romosan, Papava, Bredicean, Dumitrascu, Ursoniu and Romosan2020). Compared with typical antipsychotics, the majority of atypical antipsychotics present fewer prolactin related side effects, hypothesised to be due to a shorter binding duration between the drug and D2R (Bargiota et al., Reference Bargiota, Bonotis, Messinis and Angelopoulos2013). Atypical antipsychotics have a higher risk of inducing metabolic syndrome, including central obesity and hyperlipidaemia, than typical antipsychotics (De Hert et al., Reference De Hert, Detraux, van Winkel, Yu and Correll2012; Wei Xin Chong et al., Reference Wei Xin Chong, Hsien-Jie Tan, Chong, Ng and Wijesinghe2016), both of which have been investigated to have a potentially increased risk of breast cancer (Iyengar et al., Reference Iyengar, Arthur, Manson, Chlebowski, Kroenke, Peterson, Cheng, Feliciano, Lane, Luo, Nassir, Pan, Wassertheil-Smoller, Kamensky, Rohan and Dannenberg2019; Chowdhury et al., Reference Chowdhury, Islam, Prova, Khatun, Sharmin, Islam, Hassan, Khan and Rahman2021). Moreover, studies on schizophrenia patients showed that the risk of developing cardiovascular disease as well as type-2 diabetes mellitus of individuals was higher in atypical antipsychotic drugs (Drici and Priori, Reference Drici and Priori2007; De Hert et al., Reference De Hert, Detraux, van Winkel, Yu and Correll2012), with recent literature suggesting an association between diabetes and breast cancer risk (Liao et al., Reference Liao, Li, Wei, Wang, Zhang, Li, Wang and Sun2011). Hence, the association between antipsychotic use and breast cancer may possibly be explained by more than one physiological mechanism. With a majority of the included studies in this review having made reasonable adjustments for potential confounders such as clinical history, lifestyle factors and socioeconomic background, with several studies reporting increased breast cancer risk in prolactin-elevating antipsychotics (Pottegård et al., Reference Pottegård, Lash, Cronin-Fenton, Ahern and Damkier2018; Taipale et al., Reference Taipale, Solmi, Lähteenvuo, Tanskanen, Correll and Tiihonen2021; Rahman et al., Reference Rahman, Sahrmann, Olsen, Nickel, Miller, Ma and Grucza2022), the observed association may likely be attributed to these biological mechanisms as described.
With an increasingly prevalent use of antipsychotic medications worldwide, the risk of adverse events associated with it should be investigated in more breadth and depth to inform clinical practice. This study on the potentially elevated risk of breast cancer adds to the current knowledge of adverse events associated with antipsychotic use, such as stroke and myocardial infarction were investigated previously (Douglas and Smeeth, Reference Douglas and Smeeth2008; Sørensen et al., Reference Sørensen, Jensen and Nielsen2013; Lai et al., Reference Lai, Guthrie, Mercer, Smith, Yip, Chung, Lee, Chung, Chau, Wong, Yeoh and Wong2020), and use of prolactin-inducing antipsychotics was also reported to be associated with hip fractures (De Hert et al., Reference De Hert, Detraux and Stubbs2016a). Given the potentially multifold underlying physiological mechanisms underlying the side effects, a comprehensive holistic assessment of the clinical profile of the patients should be made along with the safety profile of specific antipsychotics to optimise the treatment outcome (Huhn et al., Reference Huhn, Nikolakopoulou, Schneider-Thoma, Krause, Samara, Peter, Arndt, Bäckers, Rothe and Cipriani2019). Interestingly, the elevated breast cancer risk observed in this study may not be applicable to other cancer types. In fact, a lower risk of lung and other cancers have been found associated with the use of antipsychotics and there are ongoing efforts in drug repurposing to experiment the cancer prevention properties of antipsychotic medications (Li et al., Reference Li, Tang, Si and Xue2022). The exact mechanism of this inverse relationship is largely unclear.
The increased use of routine electronic health records in pharmacovigilance studies have contributed to the existing literature significantly, as shown in the included studies in this review. While providing a typically large sample size with realistic real-world clinical data, there are intrinsic limitations to these records. Specifically, the lack of lifestyle and other important factors might introduce bias to the estimated association. Primary data collection may provide much more detailed information but with a much-limited sample size. Therefore, both types of research are much warranted, and the evidence needs to be considered in the context of a variety of study designs with various strengths and weaknesses for a balanced overall assessment. With the benefits of record-linkage techniques with prescription registries, antipsychotic prescription practices such as antipsychotic polypharmacy in comparison with monotherapy can be addressed in future studies. One review suggested that aripiprazole use in combination with another antipsychotic was associated with better lipid profile outcomes than the use of other antipsychotic polypharmacy or monotherapy, although the quality of evidence was lacking (Ijaz et al., Reference Ijaz, Bolea, Davies, Savović, Richards, Sullivan and Moran2018). Further investigation in this area could possibly provide a more substantiated association.
Limitations
In spite of the important clinical implications, there are several limitations. First, the reviewed evidence is all generated from observational research without randomisation. There is likely unmeasured confounding effects and causal inferences need to be made with great caution. Specifically, the comparators selected for some included studies may not be entirely suitable and could be subject to potential selection bias. One example of mitigating this bias is demonstrated in Rahman et al. through the use of anticonvulsants and lithium as comparator drugs, which are also prescribed to patients with psychiatric disorders such as anxiety, depression and bipolar disorder, but with no known risk of hyperprolactinaemia (Ajmal et al., Reference Ajmal, Joffe and Nachtigall2014). Second, the rare incidence of male breast cancer cases, even in very large electronic health record databases, poses as a challenge to derive a meaningful statistical analysis. Despite having included studies with male breast cancer cases in this review, the association of antipsychotic use with breast cancer amongst the male population would be difficult to conclude.
There are also limitations specific to this review as well. First, although the meta-analysis generated consistent results across study designs, i.e. cohort and case-control, the association could not be appropriately pooled across designs to increase the precision of the estimate. Second, the number of studies is too small to provide a more precise estimate of the hypothesised association and the presence of publication bias could not be tested as a result. Third, significant heterogeneity was observed between studies even within the same design, probably due to different populations, research practice and availability of data, further studies with more accrued data should investigate factors that contribute to this heterogeneity. Recent studies reported higher basal epigenetic changes in African American women (Joshi et al., Reference Joshi, Garlapati and Aneja2022), a population found to have the highest rates of BRCA genetic mutations (Fackenthal and Olopade, Reference Fackenthal and Olopade2007), which could increase the risk of breast cancer development. Varying degrees of risk in certain breast cancer subtypes between women of Hispanic, Asian, Black and White descent were also reported (Kurian et al., Reference Kurian, Fish, Shema and Clarke2010). The variation in breast cancer risk between ethnicities is suggestive of biological heterogeneities; further exploration may be warranted for clarification on the potential differences with regards to the observed association. Fourth, we only examined studies written in English language. Further reviews including other languages may be warranted.
Conclusion
In conclusion, we found a moderate association between the use of antipsychotics and breast cancer with a more evident association observed with prolactin-elevating medications and greater extent of antipsychotic exposure. This risk, together with other known associated adverse events, should be weighed against the anticipated treatment outcomes for a balanced clinical management decision.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S2045796022000476.
Data
All data used in the systematic review and meta-analyses can be found in the included studies.
Acknowledgements
FTTL and ICKW are partially supported by the Laboratory of Data Discovery for Health (D24H) funded by the by AIR@InnoHK administered by Innovation and Technology Commission. We thank Miss Kaitlyn Zhang for her assistance in the earlier stages of the study.
Financial support
No specific funding for this research was received by the authors.
Conflict of interest
FTTL has been supported by the RGC Postdoctoral Fellowship under the Hong Kong Research Grants Council and has received research grants from the Food and Health Bureau of the Government of the Hong Kong Special Administrative Region, outside the submitted work. EWYC reports grants from Research Grants Council (RGC, Hong Kong), Research Fund Secretariat of the Food and Health Bureau, National Natural Science Fund of China, Wellcome Trust, Bayer, Bristol-Myers Squibb, Pfizer, Janssen, Amgen, Takeda and Narcotics Division of the Security Bureau of the Hong Kong Special Administrative Region, outside the submitted work. ICKW receives research funding outside the submitted work from Amgen, Bristol-Myers Squibb, Pfizer, Janssen, Bayer, GSK, Novartis, the Hong Kong Research Grants Council, the Food and Health Bureau of the Government of the Hong Kong Special Administrative Region, National Institute for Health Research in England, European Commission and the National Health and Medical Research Council in Australia; has received speaker fees from Janssen and Medice in the previous 3 years; and is an independent non-executive director of Jacobson Medical in Hong Kong. The remaining authors have nothing to disclose.
Ethical standards
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
Open access
