The Association between Early-Onset Diagnosis and Clinical Outcomes in Triple-Negative Breast Cancer: A Systematic Review and Meta-Analysis

Simple Summary Epidemiologic studies have observed higher risks of breast cancer recurrence and death in women diagnosed <40 years of age compared to ≥40 years. There is ongoing clinical debate as to whether this is due to the overrepresentation of advanced and aggressive disease features or unique disease biology. Younger women are more likely to be diagnosed with more aggressive subtypes, including triple-negative breast cancer. Herein, we conduct a systematic review and meta-analysis of studies assessing the association between age <40 years at diagnosis and clinical outcomes in triple-negative breast cancer. We present pooled risk estimates for recurrence-free survival, breast cancer-specific and overall survival, and pathological complete response. Our findings highlight the prognostic significance of age in triple-negative breast cancer and may point to a need for tailored local and systemic treatment strategies in women of younger ages. Abstract Early-onset diagnosis, defined by age <40 years, has historically been associated with inferior outcomes in breast cancer. Recent evidence suggests that this association is modified by molecular subtype. We performed a systematic review and meta-analysis of the literature to synthesize evidence on the association between early-onset diagnosis and clinical outcomes in triple-negative breast cancer (TNBC). Studies comparing the risk of clinical outcomes in non-metastatic TNBC between early-onset patients and later-onset patients (≥40 years) were queried in Medline and EMBASE from inception to February 2023. Separate meta-analyses were performed for breast cancer specific survival (BCSS), overall survival (OS), and disease-free survival (DFS), locoregional recurrence-free survival (LRRFS), distant recurrence-free survival (DRFS), and pathological complete response (pCR). In total, 7581 unique records were identified, and 36 studies satisfied inclusion criteria. The pooled risk of any recurrence was significantly greater in early-onset patients compared to later-onset patients. Better BCSS and OS were observed in early-onset patients relative to later-onset patients aged >60 years. The pooled odds of achieving pCR were significantly higher in early-onset patients. Future studies should evaluate the role of locoregional management of TNBC and the implementation of novel therapies such as PARP inhibitors in real-world settings, and whether they improve outcomes.

observed to have inferior outcomes compared to patients of older ages, while no age-related differences have been observed in the HER2 subtype since the introduction of targeted therapies [11,19,20]. Many studies have assessed the independent effects of a young age at diagnosis on outcomes in TNBC, but results in the current body of literature remain inconsistent and have yet to be systematically reviewed. Ascertaining the prognostic significance of young age in TNBC may have important implications for future research and clinical practice. It may provide insights as to whether TNBC in women <40 years is biologically unique, and whether tailored treatments are needed to improve outcomes in young women. We conducted a systematic review and meta-analysis to synthesize qualitative evidence of the association between early-onset diagnosis and clinical outcomes in non-metastatic TNBC.

Materials and Methods
The study was registered in PROSPERO (CRD42022308683). Reporting of this systematic review and meta-analysis adhered to the guidelines of Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) and Meta-analysis of Observational Studies in Epidemiology (MOOSE) [21,22]. Abstract screening and full-text review were completed independently in duplicate by three investigators (K.C., D.K., and M.K.). All conflicts were resolved by R.B.B. Quality assessment and data extraction were also completed independently in duplicate by three investigators (K.C., A.L., R.B.B.). All conflicts were solved by the 3 investigators via consensus.

Search Strategy
The Medline and Excerpta Medica Database (EMBASE) was queried from inception to 9 February 2023, the date of the last conducted search. We restricted our search to English language studies and no restrictions were made on publication date. The search strategy was developed by R.B.B., in consultation with librarians from the University of Calgary and Alberta Health Services. Search terms related to four key concepts: breast cancer, triple-negative subtype, clinical outcomes, and early-onset diagnosis. The search strategy is fully detailed in Table S1 of the Supplemental Material.

Eligibility Criteria
Eligibility criteria were established in our PICOs (Population, Intervention, Comparator, and Outcomes) statement and prior to conducting the search. To be included, studies had to have compared clinical outcomes between early-onset and later-onset age groups in an adult (≥18 years) population of non-metastatic TNBC patients. An age of diagnosis of 40 years was used as a cut-off to define early-onset and later-onset age groups. Clinical outcomes included overall and breast cancer-specific survival; local, locoregional, and distant recurrences/events/metastases; and pathological complete response (pCR) following neoadjuvant chemotherapy. Prospective and retrospective cohort studies, case-control studies, cross-sectional studies, and pooled analyses of trial data were eligible for inclusion. In full-text screening, studies were excluded if they (1) were a published abstract, (2) did not include a non-metastatic TNBC population/subgroup, (3) defined early-onset patients by age >40 years, (4) did not compare clinical outcomes between early-onset and later-onset age groups in the TNBC group, and (5) did not report an estimate to quantify the association of early-onset age and outcomes or did not provide values necessary to derive an effect estimate.

Data Extraction
A data extraction form was created prior to conducting the search. Extracted information included author, year of publication, geographic location, data sources, study design, sample size, definitions of early-onset and late-onset age groups, outcomes examined, clinical stage, lymph node status, and receipt of surgery, radiotherapy, and neo/adjuvant chemotherapy. The referent later-onset age group, association estimates, and 95% confi- 15, 1923 4 of 19 dence intervals (CIs) for early-onset age groups were also extracted. Extracted association estimates included hazard ratios (HR) and odds ratios (OR). Only adjusted estimates were extracted; however, unadjusted estimates were extracted if no adjusted estimates were available. All the extracted information was stored in an Excel file (Microsoft, Redmond, WA) and was checked for accuracy by R.B.B.

Quality Assessment
The quality of included studies was evaluated using the Newcastle Ottawa Scale (NOS) [23]. The NOS assesses three domains of internal validity: (1) selection of study groups (4 questions, 1 point each); (2) comparability of study groups (1 question, 2 points max); and (3) assessment of exposure and outcome (3 questions, 1 point each). A cumulative score of 0 indicates a study of low quality and a score of 9 indicates a study of high quality. For the assessment of comparability of study groups, a score of 0 was assigned for unadjusted estimates, a score of 1 was assigned for estimates adjusted for clinicopathological variables, and a score of 2 was assigned to estimates adjusted for clinicopathological and treatment variables.

Statistical Analysis
Random-effects models with DerSimonian and Laird weighting were employed to pool association estimates for all meta-analyses. Separate meta-analyses were conducted for each clinical outcome that was examined in 3 or more studies. If studies made comparisons with multiple early or later-onset age groups, then these estimates were pooled prior to meta-analysis. We identified 6 primary outcomes to obtain pooled estimates for (1) Diseasefree survival (DFS)-time from diagnosis to first breast cancer related event/recurrence or death; (2) Locoregional recurrence-free survival (LRRFS)-time from diagnosis to local or locoregional recurrence; (3) Distant recurrence-free survival (DRFS)-time from diagnosis to distant recurrence or metastasis; (4) Breast cancer-specific survival (BCSS)-time from diagnosis to death due to breast cancer; (5) Overall survival (OS)-time from diagnosis to death due to any cause; (6) Pathological complete response (pCR)-the absence of invasive cancer in the breast or axilla.
Pooled hazard ratio (HR) estimates with 95% confidence intervals (95% CI) were obtained for DFS, LRRFS, DRFS, BCSS, and OS while pooled odds ratio (OR) estimates and 95% CIs were obtained for pCR. Individual study and pooled estimates were visualized using forest plots, and heterogeneity was quantified with the Q-test and I 2 statistics. In anticipation of heterogeneity, we performed subgroup analyses among studies that defined later-onset age as >60 years if at least three estimates were available. It was hypothesized that later-onset age groups defined by >60 years would include a large portion of patients with comorbid conditions who received less intensive treatment or declined compared to later-onset age groups defined by ≥40 years. Therefore, to capture potential heterogenous effects, subgroup analysis was performed wherein one subgroup later-onset age was defined by age groups ≥40 years (<40 years vs. ≥40 years), and in the other subgroup lateronset age was defined by age groups >60 years (<40 years vs. >60 years). Meta-regressions, by definition of later-onset age group (>60 years vs. ≥40 years) and adjustment for confounding (scored 2 in the comparability domain of the NOS vs. 0 or 1), were performed to statistically test whether these variables contributed to between-study heterogeneity. A funnel plot was used to visually assess the potential of publication bias and was statistically tested using Egger's regression. The "trim and fill" method was applied where significant publication was present. A p value of <0.05 was used to define significance for statistical tests. All analyses were performed using Stata version 16.

Figure 6.
Forest plot for meta-analysis of pathological complete response (pCR). Odds ratios less than 1 show lower odds of achieving pCR in early-onset age group; Odds ratios greater than 1 show higher odds of achieving pCR in early-onset age group. References [24,26,37,[43][44][45]. Abbreviations: CI = confidence interval; OR = odds ratio. Forest plot for meta-analysis of pathological complete response (pCR). Odds ratios less than 1 show lower odds of achieving pCR in early-onset age group; Odds ratios greater than 1 show higher odds of achieving pCR in early-onset age group. References [24,26,37,[43][44][45]. Abbreviations: CI = confidence interval; OR = odds ratio.

Discussion
Early-onset diagnosis has historically been considered a poor prognostic factor in breast cancer. However, this association has been questioned with improved understanding of the biological heterogeneity of breast tumours, recognition of the predictive role of tumour subtype, and awareness that young women are more likely to develop more aggressive phenotypes. Our study expands on this body of evidence by systematically reviewing the literature on the association between the age of diagnosis and outcomes within the TNBC subtype. Overall, 36 studies were included in our meta-analyses, and we demonstrated that patients diagnosed with TNBC under 40 years of age are at greater risk of locoregional and distant recurrences compared to those diagnosed older than 40 years. We also show significantly better BCSS and OS among early-onset patients when compared to later-onset age groups defined by >60 years, but worse BCSS and OS when compared to later-onset age groups defined by ≥40 years. Finally, we found that early-onset patients are more likely to achieve pCR following neoadjuvant chemotherapy.
Given the aggressive features and poor prognosis of TNBC, there is concern that more aggressive treatment approaches should be considered. The absence of molecular markers for targeted therapy stresses the importance of locoregional management. Wang et al. performed a meta-analysis to evaluate the effectiveness of breast-conserving surgery (BCS) versus mastectomy in controlling locoregional recurrences for TNBC. Rates of locoregional recurrence and distant metastasis were 25% and 32% lower in the BCT group compared to the mastectomy group, respectively [59]. The benefit of BCS in TNBC may be attributable to the receipt of postoperative radiation. A meta-analysis by O'Rorke et al. demonstrated a significantly lower risk of locoregional recurrence in patients receiving adjuvant radiotherapy irrespective of surgery type versus mastectomy alone [60]. Improvements in locoregional recurrence were present in T1-2N0 and T3-4N2-3 subgroups, and survival benefits were observed in the <40 years subgroup [60]. Despite these findings, the role of postmastectomy radiation is not specific to subtype and often indicated for larger (T3) and node positive tumour, among other prognostic considerations. In the context of our results, the higher risk of locoregional and distant recurrence in early-onset patients may be driven by the fact that younger women are more likely to choose mastectomy [61,62] and may not be indicated for postmastectomy radiation for smaller, node negative disease. Cancello et al. describe T1N0 disease as a spectrum from patients at very low risk for whom there is little evidence supporting the use of adjuvant therapy, to those with higher risk disease where an approach including chemotherapy plus targeted therapy appears clearly justified [63]. In their study, both triple negativity and age <35 years were associated with worse LRRFS in T1N0 disease [63]. Studies included in our review did not report on the receipt of postmastectomy radiation, so it was not possible to ascertain whether this explained differences in recurrence rates between early and later-onset patients. Future studies should explore the role of postmastectomy radiation in young women with TNBC to further clarify optimal locoregional treatment approaches in this patient population.
High heterogeneity (>80%) was observed in pooled effect estimates for BCSS and OS, and was in part explained by varying age cut-offs to define later-onset diagnosis. In this review, we used an age cut-off of 40 years to define early-onset diagnosis, which is consistent with previous guidelines and reflects that these women have specific issues including those related to fertility, genetics and psychosocial concerns that often deserve a different approach compared to older premenopausal and postmenopausal women [11]. Most included studies defined early-onset age as <35 or <40 years. Conversely, the later-onset age groups varied more in definition, as some studies used an upper limit for the age range of the later-onset group and others did not. For example, studies in which the later-onset group was defined by an age range of 40-49 likely included higher proportions of premenopausal patients compared to later-onset groups defined as ≥40 years. Likewise, studies which used a high lower limit for defining the later-onset group, such as 60 or 75 years, included higher proportions who may have received less intensive (neo)adjuvant treatment or declined [64]. This likely explains why BCSS and OS were better among early-onset patients compared to later-onset patients >60 years in our subgroup analysis. Further, early-onset patients are more likely to undergo more aggressive management of recurrences. A multicentre French cohort including over 14,403 metastatic breast cancer patients, 28% of which were de novo diagnoses, reported a higher uptake of first-line chemotherapy in patients <40 years compared to >60 years (96.3% vs 90.2%) [65]. A study of patients with regional and distant recurrences in China observed similar findings, with an 88.7% uptake of first line chemotherapy in patients <40 years compared to 76% in those >65 years [66]. Early-onset patients are less likely to have pre-existing comorbid conditions or take multiple medications. Thus, they are generally healthier and can receive multiple rounds of combination chemotherapy and tolerate treatment-related toxicities that could result in non-compliance or declining treatment in others. Minor increases in the risk of breast cancer specific and all-cause mortality were observed in early-onset patients in subgroup analysis when compared to later-onset patients defined by ≥40 years. This may be due to the increased risk of distant recurrence/metastasis, which are considered not curable. TNBCs and young age have a higher propensity for metastases in the central nervous system, soft tissue organs, and multiple metastatic sites compared to bone alone, which is associated with better survival after metastasis [7,65,[67][68][69]. If these occur more commonly in early-onset patients, poorer BCSS and OS would be expected. Previous studies have attributed this phenomenon to the "seed and soil" hypothesis, which states that tumor cells favour different microenvironments of distant organs, which provide ideal conditions for their invasion and proliferation [70,71]. The underlying molecular differences between early and later-onset TNBC patients should be explored in future studies to better understand the mechanisms of distant metastasis and whether these contribute to survival differences between these age groups.
Young women are underrepresented in clinical trials for novel therapies and outcome prediction tools to guide tailored treatment decisions [72]. Few actionable molecular targets exist for triple-negative disease, so systemic chemotherapy is generally recommended in addition to locoregional management. However, the optimal chemotherapy regime in young women in the early-stage setting is currently unknown [11]. Multicentre pooled analyses of early-stage patients treated with neoadjuvant chemotherapy demonstrate that young women with TNBC benefit most from preoperative treatment, given the prognostic value of achieving pCR [73,74]. Agreeable findings were observed in our meta-analysis of pCR. Recent data on the efficacy of adding platinum-based agents to neoadjuvant chemotherapy in early-stage TNBC has emerged from the BrighTNess trial, which showed improved EFS with the addition of carboplatin, with greater improvements in those <50 years of age versus ≥50 years [75]. Further, the KEYNOTE-522 trial showed that the addition of pembrolizumab, programmed death-1 (PD-1) targeted immunotherapy, to neoadjuvant chemotherapy improved EFS similarly in pre-and postmenopausal early-stage TNBC patients [76]. Of relevance to young women, who are more likely to harbour germline BRCA1/2 mutations, are emerging data on the use of poly(ADP-ribose) polymerases (PARPs) inhibitors in early-stage breast cancer. The OLYMPIA trial randomized BRCA1/2 mutation carriers to receive one year of adjuvant olaparib or placebo following completion of chemotherapy, surgery, and radiation [77]. The median age of the trial population was 43 years, and 82% of participants had triple-negative disease. Olaparib significantly improved invasive DFS (3- [77]. These trials demonstrate the discovery and implementation of novel therapies in TNBC with promising results in younger women; however, real-world studies will be needed in the future to assess whether they are generalizable to all early-onset TNBC patients. There is increasing attention in research and practice being paid to younger women with breast cancer, including a series of international consensus guidelines that encourage participation in clinical trials in the recurrent and advanced setting. Recent studies have observed higher participation for women <50 years in cancer clinical trials relative to 15-20 years ago [78,79]. Greater access to clinical trials and novel therapies may attenuate outcome differences in the future. It is possible that tumors arising in young women have different biologic characteristics than do those that arise in older women, even within tumor subtypes. TNBCs are a highly diverse group of cancer, with differential responses to treatment [80]. Lehmann et al. were among the first to identify distinct subtypes of TNBC and their implications on selection for neo/adjuvant treatment [81]. Their comprehensive analyses demonstrated four subtypes of TNBC with unique gene expression profiles-basal-like 1 and 2 (BL1 and BL2), mesenchymal (M), and luminal androgen receptor (LAR) type. Non-basal TNBC tumors were diagnosed in older women relative to basal TNBC (58.5 vs. 52.6, p < 0.0001) [81]. The LAR subtype was diagnosed in women of older ages compared to all other subtypes (59.5 vs. 52.7, p < 0.0001) [81]. Concordant findings were observed by Prat et al., who showed that TNBCs can be grouped into claudin-low (high expression of mesenchymal processes), basal-like, and luminal/HER2 enriched subgroups [82]. In their data, the mean age at diagnosis of basal-like versus non-basal-like TNBC was found to be significantly different (50.7 vs. 57.1 years; p < 0.0001) [82]. Gulbache et al. compared the distribution of TNBC subtypes by age groups of <50, 50-64, and ≥65 years. A significantly higher proportion of basal-like TNBCs were observed in the <50 year age group, while a higher proportion of luminal and HER2 enriched subtypes were observed in the 50-64 and ≥65 year age groups [83]. They also used qRT-PCR and evaluated a set of 10 proliferationrelated genes: ANLN, CDC20, CENPF, CEP55, KIF2C, RRM2, UBE2C, MKI67, CCNB1, and MYBL2. These proliferation genes correlate with Ki67 expression, and the last four genes are also included in the set of genes of commonly used and commercially available Oncotype DX testing. All 10 proliferation genes had significantly lower expression among older women with TNBC [83]. These results are consistent with studies using IHC staining, which report that older women with TNBC had a higher frequency of cytokeratin 5/6 expression, lower expression of EGFR, a lower rate of Ki67 labeling index and cytokeratin 7/8 positivity, and a higher rate of Bcl2 and cytokeratin 18 positivity [83][84][85]. A higher proportion of the young TNBC patients harbor a BRCA1/2 mutation; however, its role is debatable and a larger prospective trial showed that survival after two years is more favorable among BRCA mutation carriers with TNBC compared with wild-type TNBC, but not at five years [48,86].
A major strength of this review is that all included studies were published in 2010 or later and likely reflect the administration of modern-day treatment regimens. Another strength is that most included studies scored moderate to high in our quality assessment. The majority of studies included representative study samples, mutually adjusted effect estimates, and the accurate collection of exposure and outcome data. Despite their high quality, the observational nature of these studies makes them prone to unmeasured confounding and biases. For example, unmeasured confounding is likely present in studies that did not account for treatment and dose of treatments. Another limitation includes the varying definitions of older-onset age groups among included studies, increasing the heterogeneity of pooled estimates, affecting the validity of their clinical interpretations.

Conclusions
Our systematic review provides evidence that patients diagnosed <40 years have worse LRRFS, DDFS, BCSS and OS compared to patients diagnosed ≥40 years in TNBC. However, when compared to later-onset patients >60 years, BCSS and OS are significantly better in early-onset patients. We also show greater odds of achieving a pCR in early-onset patients. While we speculate that differences in recurrence risk may be due to age-related patient preferences for mastectomy and physician bias in radiotherapy recommendation, further prospective data are needed to tailor locoregional treatments by age and molecular subtype. The tolerability of early-onset patients for more aggressive management, particularly in recurrent settings, may explain better survival outcomes compared to patients >60 years. Survival in TNBC remains poor irrespective of age, and the discovery of molecular markers has led the development of novel targeted therapies which must be further evaluated in real-world practice to determine if they are improving outcomes. Age-related differences in the biology of TNBC are being discovered, but their clinical relevance in terms of pathology and targets for treatment remains to be seen.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/cancers15071923/s1, Table S1: Search strategy concepts and terms used in Medline and EMBASE, Table S2a-e: Meta-regression by later-onset age definition and confounding adjustment (based on the Q5 score of the Newcastle-Ottawa Scale) for meta-analyses of (a) disease-free survival, (b) locoregional recurrence-free survival, (c) distant recurrence-free survival, (d) breast cancer-specific survival, and (e) pathological complete response, Figure S1a-e: Funnel plot to assess publication bias for meta-analysis of (a) disease-free survival, (b) locoregional recurrence-free survival, (c) distant recurrence-free survival, (d) breast cancer-specific survival, and (e) pathological complete response.