Introduction

In 2020, 2.3 million women were diagnosed with breast cancer (BC) globally, making it the most prevalent cancer type (excluding non-melanoma skin cancer) in the world1. Despite 5-year survival rates approaching 90% in North America for patients with BC, nearly 700,000 women died from BC in 2020 worldwide1,2. Alongside, the prevalence of obesity (defined as a body mass index (BMI) ≥ 30 kg/m2) increased excessively worldwide, rising from 7 to 16% among women between 1975 and 20163. Obesity is associated with an increased risk of developing at least 15 types of cancer, including postmenopausal BC4. Also, obesity is a prognostic disadvantage and is associated with shorter disease-free survival (DFS) and overall survival (OS) in BC5,6.

Inflammation is a hallmark of cancer7 and systemic inflammation is associated with poor BC prognosis8,9. Obesity is associated with low-grade inflammation10 and elevated levels of C-reactive protein (CRP)11, also among BC patients12. Thus, both obesity and systemic inflammation are associated with inferior BC outcomes. Yet, it is not clear if BC patients with obesity and high levels of systemic inflammation have a poorer prognosis compared with patients with obesity and lower levels of systemic inflammation.

CRP is an acute-phase protein released from the liver upon stimulation from cytokines such as interleukin 613. CRP is part of the inflammatory cascade and a marker of systemic inflammation14,15. CRP levels are increased in people with obesity compared with their normal weight counterparts, and in cancer patients compared with healthy controls or patients with benign diseases16. A systematic review by Savioli et al. concluded that high pre-operative CRP levels were associated with an increased risk of BC-specific mortality and all-cause mortality17. Likewise, in a meta-analysis from 2011, Han et al. reported an association between elevated CRP levels and lower OS and DFS18. A meta-analysis by Mikkelsen et al. found that high CRP was an indicator of poor prognosis in metastatic BC, but the prognostic value in non-metastatic early BC could not be confirmed19.

As such, CRP levels may be prognostic in BC but only three studies have investigated this relationship across BMI groups, and they reported conflicting results20,21,22. Therefore, we investigated the prognostic potential of CRP in BC patients according to BMI groups. We hypothesized that higher circulating CRP levels were associated with poorer BC prognosis and that such an association was most pronounced in patients with obesity.

Materials and methods

Data sources

All data were merged through a unique identification number for each patient, linking all data with 100% accuracy. All patients with BC treated in Denmark are registered in the Danish Breast Cancer Group (DBCG) database23. The completeness of the DBCG database exceeds 95%24. From DBCG and through systematic investigation of medical records, we retrieved baseline data concerning patient-, tumor-, and treatment characteristics. All variables from DBCG were retrieved from “The Danish Clinical Quality Program—National Clinical Registries” (RKKP), which constitutes the infrastructure of the Danish clinical quality registries and the Danish Multidisciplinary Cancer Groups (DMCG)25. Information on emigration was retrieved through RKKP from the Civil Registration System26. Information on comorbidities was retrieved through RKKP from the National Patient Registry27. Information on height, weight, and thereof BMI was extracted from both medical records and the Danish Anesthesia Database (DAD)28. BMI was obtained by merging data from the medical records and the DAD. Regarding follow-up data, we extracted information based on a prespecified codebook. We reviewed all patient electronic medical records, which included pathological reports and digital imaging, to register all recurrences, contralateral BCs, other malignancies, and deaths.

Study population

Our study cohort was women diagnosed with stage I-III BC between 2010 and 2020, who were referred for surgery at Aarhus University Hospital (AUH), for primary BC. At this department, all patients with BC were invited to donate blood for future research to the Regional Bio- and Genome Bank Denmark (RBGB)29 at the time of BC diagnosis. The blood was drawn between March 2010 and August 2020, median of seven days (interquartile range (IQR): 6–11 days) after the primary invasive BC diagnosis before breast surgery. Patients who received neoadjuvant systemic treatment were excluded from the analyses, as data on tumor characteristics are registered after neoadjuvant treatment and differ substantially from patients receiving up-front breast surgery. The final study population consisted of 2673 patients, as illustrated in Fig. 1.

Figure 1
figure 1

We identified 3424 female patients with primary invasive breast cancer who donated blood when seen for breast cancer surgery at Aarhus University Hospital between 2010 and 2020. We excluded patients who withdrew their consent. Patients with previous cancer or co-existing cancer (defined as a diagnosis of other malignancy before or within 90 days from the final primary surgery date) were excluded. Non-melanoma skin cancer was not classified as previous or co-existing cancer. As recurrence was an endpoint in the statistical analyses, patients who did not receive surgery or had distant metastasis at baseline were also excluded. Patients were excluded if a blood sample was drawn more than 6 months (182 days) before or after the date of invasive BC diagnosis. If the final primary surgery date was after the last date of follow-up (15th of November 2021) we excluded the patients. Finally, patients receiving neoadjuvant systemic therapy were excluded. After exclusion, 2673 patients were included in the statistical analyses. BC Breast cancer.

Full size image

C-reactive protein analyses

After the blood draw, serum was separated and subsequently stored at − 80 °C at the RGBG. In late 2020, serum samples were identified and released for analyses at the Department of Clinical Biochemistry, AUH. The CRP levels were analyzed with a high-sensitivity CRP (hs-CRP) test (Siemens ADVIA® Chemistry XPT system with “ADVIA® Chemistry CardioPhase™ High Sensitivity C‑Reactive Protein (hsCRP)-reagents”)30. The lowest detectable level was 0.2 mg/L.

Definitions of analytic variables

CRP levels were categorised in quartiles (CRP-Q1: \(\le\) 0.60 mg/L, CRP-Q2: 0.61–1.33 mg/L, CRP-Q3: 1.34–3.18 mg/L, CRP-Q4: \(\ge\) 3.19 mg/L). In the statistical analyses, the lowest CRP quartile (CRP-Q1) served as the reference. It should be noted that CRP below 3 mg/L is considered to be within the normal range31. BMI was grouped according to the World Health Organization (WHO) definitions32: underweight (BMI < 18.5 kg/m2), normal-weight (18.5 ≤ BMI < 25 kg/m2), overweight (25 ≤ BMI < 30 kg/m2), and obesity (BMI ≥ 30 kg/m2). The closest registered BMI measurement from the date of the blood sample was used. Registrations within 182 days before or after blood draw were included (median same date as blood draw, IQR: 0–2 days after date of blood draw).

Age and menopausal status (defined according to DBCG guidelines) refer to the time of primary invasive BC diagnosis. For tumor size and nodal status, we categorized patients in groups according to the American Joint Committee on Cancer Staging 8th edition33. The histological grade was classified using the standardization from the Nottingham Group33. The histological classification followed the WHO Classification of Breast Tumors 3rd Edition34. Tumors without registration of invasive ductal or lobular carcinoma were categorized as “Other.” Estrogen receptor (ER) status was classified as “Negative,” if tumor cells showed no ER expression, and “Positive” if 1% or more of the tumor cells expressed ER. Human Epidermal Growth Factor Receptor 2 (HER2) expression was defined as either “Positive” or “Negative” through a combination of immunohistochemistry and Fluorescence In Situ Hybridization (FISH)-ratio according to the American Society of Clinical Oncology guidelines35.

The type of breast surgery was categorized as “mastectomy” (including patients with mastectomy after lumpectomy) or “lumpectomy,” based on the final surgery for the primary BC. Adjuvant systemic treatment (endocrine therapy, HER2-targeted treatment, and chemotherapy) and radiotherapy were handled as intention-to-treat variables according to DBCG protocols.

Recurrence refers to any recurrent invasive BC in the breast, lymph nodes, or elsewhere in the body (apart from the contralateral breast) ≥ 3 months after the final primary surgery date (defined as the last date of surgery for the primary BC). The hierarchy for recurrence date was whatever came first in stating a recurrence in the pathological report, clinical charts, or imaging information. When defining the recurrence as local, regional, or distant, we followed the clinical guidelines from DMCG36. Registration of malignancy in the contralateral breast in the pathological report after final surgery was classified as contralateral BC. Other malignancies apart from non-melanoma skin cancer in the pathological report and death were registered, too.

Follow-up and statistical analysis

Disease-free survival (DFS). Follow-up for DFS began on the date of final primary surgery and continued until the first of the following: BC recurrence, contralateral BC, death, other malignancy, emigration, or end-of-follow-up (15th November 2021). We treated BC recurrence, contralateral BC, and death as events. We censored patients at other malignancies, emigration, or end-of-follow-up. However, if recurrence, contralateral BC, or death occurred within 30 days after diagnosis of other malignancy, the event was included in the analyses.

Overall survival (OS). Follow-up for all-cause mortality began at the date of final primary surgery and continued until the first of any death, emigration, or end-of-follow-up (15th November 2021). We treated any death as an event. We censored patients at emigration or end-of-follow-up.

We calculated person-years, number of events, and incidence rate per 1000 person-years (with a 95% confidence interval (95% CI)) for each CRP quartile. We used Cox regression models to estimate crude and adjusted hazard ratios (HRs) with 95% CIs for DFS and OS. Patients were followed for a maximum of 10 years in the regression models. We adjusted for patient-, tumor-, and treatment characteristics in the adjusted analysis. Only patients with complete data in all regressed variables were included (N = 2485). We included the following variables: age (continuous), menopausal status (dichotomous), comorbidities (Charlson Comorbidity Index (CCI)) (categorical), BMI (categorical), histological grade (categorical), histological classification (categorical), tumor size (categorical), nodal status (categorical), ER status (dichotomous), HER2 status (dichotomous), surgery type (dichotomous), intended adjuvant systemic treatment (dichotomous), and intended adjuvant radiotherapy (dichotomous).

To explore whether the association between CRP and outcomes differed across BMI groups, we performed DFS and OS analyses stratified for BMI groups as described above. In the stratified analyses, we created CRP quartiles within each BMI group. We adjusted for patient characteristics in the adjusted model. In the stratified analyses, patients with underweight were excluded. In the stratified analyses, we also presented Aalen-Johansen estimates on DFS (events: BC recurrence, contralateral BC, and death; competing risks: other malignancy; censoring points: emigration and end-of-follow-up) and Kaplan–Meier estimates on OS (events: death; censoring points: emigration and end-of-follow-up).

Ethical approval

The study was conducted in accordance with relevant guidelines and legislation. All projects applying for samples at the RBGB need approval from the Danish Data Protection Agency and the Danish Council on Ethics. The conduction of the study is approved by the Danish Council on Ethics (no. 1-10-72-192-20) and registered as a scientific project at Region Midtjylland, Denmark (no. 1-16-02-299-20). Informed consent was obtained from all participants included in the study.

Results

Our cohort included 2673 patients with a median age of 62 years at BC diagnosis (see Table 1). The median BMI was lowest in CRP-Q1 (BMI = 22.49 kg/m2) and highest in CRP-Q4 (BMI = 28.33 kg/m2). In CRP-Q4, more patients were postmenopausal, had higher CCI scores, and larger tumors compared to CRP-Q1. Chemotherapy was more often administered to patients in the CRP-Q1 compared with CRP-Q4. In total, 64 patients (2.39%) had underweight, 1265 (47.33%) had normal-weight, 818 (30.60%) had overweight, and 486 (18.18%) had obesity. The characteristics of the patients in the adjusted analyses (N = 2485) were similar to the characteristics of the 2673 patients in the crude analyses.

Table 1 Descriptive characteristics of women with breast cancer stage I-III referred for surgery at Aarhus University Hospital, Denmark for primary breast cancer. Patients were diagnosed with breast cancer and donated blood samples for future research between 2010 and 2020.
Full size table

In DFS analyses, 368 clinical events occurred over 14,962 person-years (median follow-up time 5.55 years). In the mortality analyses, 298 deaths were recorded during 15,803 person-years (median follow-up time = 6.02 years).

Table 2 presents the estimated DFS hazard ratios across CRP quartiles. In total, 70 events occurred in 3980 person-years in CRP-Q1, 82 events in 3724 person-years in CRP-Q2, 100 events in 3744 person-years in CRP-Q3, and 116 events occurred during 3513 person-years in CRP-Q4. In the adjusted analyses, we found a positive association between CRP-Q4 and the risk of clinical events compared to CRP-Q1 (CRP-Q4, HRadj: 1.62 [95% CI = 1.14–2.28]). Supplementary Tables 1 and 2 present the estimated DFS hazard ratios across CRP quartiles in more adjusted models, and with other malignancy treated as event in Supplementary Table 2.

Table 2 C-reactive protein quartiles in relation to disease-free survival and overall survival in breast cancer patients.
Full size table

Table 2 shows the estimated mortality hazard ratios across CRP quartiles. In total, 39 deaths were recorded during 4157 person-years in CRP-Q1, 66 deaths in CRP-Q2 during 3932 person-years, 88 deaths during 3949 person-years in CRP-Q3, and 105 deaths in CRP-Q4 during 3764 person-years. In the adjusted analyses, CRP-Q4 was associated with higher mortality risk compared to CRP-Q1 (CRP-Q4, HRadj: 2.47 [95% CI = 1.62–3.76]). Supplementary Table 3 presents the estimated mortality hazard ratios across CRP quartiles in more adjusted models.

Figure 2 displays the cumulative incidences of clinical events (BC recurrence, contralateral BC, and death) across BMI groups. We saw an evident increase of incidences in CRP-Q4 compared to the other quartiles in BC patients with normal-weight or overweight, but not in patients with obesity.

Figure 2
figure 2

Cumulative incidences (Aalen–Johansen estimator) of clinical events (recurrence, contralateral breast cancer, and death) in women with breast cancer stage I–III according to C-reactive protein quartiles in patients with (a) normal-weight (body mass index 18.5- < 25 kg/m2) (CRP-Q1: ≤ 0.41 mg/L, CRP-Q2: 0.42–0.81 mg/L, CRP-Q3: 0.82–1.82 mg/L, CRP-Q4: ≥ 1.83 mg/L), (b) overweight (body mass index 25- < 30 kg/m2) (CRP-Q1: ≤ 0.85 mg/L, CRP-Q2: 0.86–1.62 mg/L, CRP-Q3: 1.63–3.26 mg/L, CRP-Q4: ≥ 3.28 mg/L), and (c) obesity (body mass index ≥ 30 kg/m2) (CRP-Q1: ≤ 1.75 mg/L, CRP-Q2: 1.77–3.55 mg/L, CRP-Q3: 3.56–7.02 mg/L, CRP-Q4: ≥ 7.07 mg/L). CRP-Q1 C-reactive protein quartile 1 (lowest quartile); CRP-Q4 C-reactive protein quartile 4 (highest quartile). The figures were made using Stata® version 18.

Full size image

Table 3 displays the estimated DFS hazard ratios according to CRP quartiles in BMI groups. We demonstrated an increased risk of an event among patients with CRP-Q4 as compared with CRP-Q1 in patients with normal-weight (CRP-Q4, HRadj: 1.70 [95% CI = 1.09–2.66]) and overweight (CRP-Q4, HRadj: 1.75 [95% CI = 1.08–2.86]). In patients with obesity, we found an increased risk of a clinical event in CRP-Q4 (CRP-Q4, HRadj: 1.73 [95% CI = 0.78–3.83]), though the precision of the estimate was less precise. It should be noted that the number of patients was lowest in the obesity group.

Table 3 C-reactive protein quartiles in relation to disease-free survival in breast cancer patients across body mass index groups.
Full size table

Figure 3 shows the cumulative incidences of death across BMI groups. We observed a higher number of deaths in CRP-Q4 compared with the other quartiles among patients with normal-weight or overweight. In patients with obesity, we observed an increase in deaths in CRP-Q4 as well, becoming evident after eight years of follow-up.

Figure 3
figure 3

Cumulative incidences (Kaplan–Meier estimator) of deaths in women with breast cancer stage I–III according to C-reactive protein quartiles in patients with (a) normal-weight (body mass index 18.5- < 25 kg/m2) (CRP-Q1: ≤ 0.41 mg/L, CRP-Q2: 0.42–0.81 mg/L, CRP-Q3: 0.82–1.82 mg/L, CRP-Q4: ≥ 1.83 mg/L), (b) overweight (body mass index 25- < 30 kg/m2) (CRP-Q1: ≤ 0.85 mg/L, CRP-Q2: 0.86–1.62 mg/L, CRP-Q3: 1.63–3.26 mg/L, CRP-Q4: ≥ 3.28 mg/L), and (c) obesity (body mass index ≥ 30 kg/m2) (CRP-Q1: ≤ 1.75 mg/L, CRP-Q2: 1.77–3.55 mg/L, CRP-Q3: 3.56–7.02 mg/L, CRP-Q4: ≥ 7.07 mg/L). Abbreviations: CRP-Q1 C-reactive protein quartile 1 (lowest quartile); CRP-Q4 C-reactive protein quartile 4 (highest quartile). The figures were made using Stata® version 18.

Full size image

Table 4 displays the estimated mortality hazard ratios according to CRP quartiles in BMI groups. In patients with normal-weight, being in the highest CRP quartile compared to the lowest CRP quartile was associated with an increased risk of death (CRP-Q4, HRadj: 3.66 [95% CI = 1.95–6.87]). In patients with overweight, an association was observed between CRP-Q4 and an increased risk of death compared to CRP-Q1 (CRP-Q4, HRadj: 1.92 [95% CI = 1.06–3.46]). In patients with obesity, we also observed an increased risk of death for patients in CRP-Q4 (CRP-Q4, HRadj: 1.40 [95% CI = 0.64–3.09]) compared with CRP-Q1, however, the precision of the estimate was weaker than in the other BMI groups.

Table 4 C-reactive protein quartiles in relation to overall survival in breast cancer patients across body mass index groups.
Full size table

Discussion

This study demonstrated an association between high CRP levels and inferior outcomes in both DFS and OS analyses. In the BMI stratified analyses, we observed an association between high CRP and inferior DFS in patients with normal-weight, overweight, and obesity, although less evident among patients with obesity, which may be explained by low numbers of patients with obesity. In the OS analyses, we saw an over three-fold increased risk of death in patients with normal-weight and high CRP compared with low CRP. In patients with overweight, the increased risk of death was nearly two-fold, whereas a 40% increased risk of death was seen in patients with high CRP and obesity, but the precision of the estimate was lower than in normal-weight and overweight. We did not present BMI-stratified analyses for patients with underweight (BMI < 18.5 kg/m2), as the results were too imprecise due to the low number of patients with underweight.

Prior literature has shown inconsistent results regarding the prognostic value of CRP in BC. The meta-analysis by Mikkelsen et al. suggested a limited potential for CRP as a prognostic marker in non-metastatic settings19. In studies using CRP as a categorical variable, high CRP was associated with lower DFS and OS, but the estimates were imprecise. A Danish study of 2910 patients showed that the highest CRP tertile was associated with reduced OS and DFS37. The patients had BC stage I-IV diagnosed between 2002 and 2009, and blood was drawn at the time of diagnosis. In a cohort of BC patients with stage I-III disease in the United States (N = 2919), similar results were reported, in which blood was drawn at least 12 months after no evidence of disease (median 21.7 months)38. Higher all-cause mortality was observed in patients with CRP levels \(\ge\) 10 mg/L compared to patients with CRP levels < 1 mg/L. A decrease in OS in pre-diagnostic CRP levels \(\ge\) 10 mg/L compared to < 10 mg/L was found in a study by Wulaningsih et al.39 (N = 6606). Contrary to these studies, Frydenberg et al. found a decreased risk of all-cause death in the highest pre-diagnostic CRP tertile, and a similar correlation was found in DFS analyses (N = 192)40. Our findings are consistent with results from the larger studies37,38,39.

To our knowledge, only three studies have investigated the association between CRP and BC prognosis stratified by BMI 20,21,22. Our study is the first to explore the association with DFS across BMI groups. The latest study, a Chinese prospective multicenter cohort study with BC patients stage I-IV (N = 514) by Ruan et al.20, found a strong correlation between CRP > 10 mg/L and all-cause mortality. CRP > 10 mg/L was associated with lower OS in patients with BMI \(\ge\) 24 kg/m2 and BMI < 24 kg/m2, however, the precision of the estimate was weaker in patients with BMI < 24 kg/m2 compared to patients with BMI \(\ge\) 24 kg/m2. In 1114 BC patients (stage in situ to IV), Nelson et al. reported that only patients with higher CRP levels and normal-weight had an increased risk of death (HR: 1.39 [95% CI = 1.03–1.89]) for every 1 standard deviation increase in logCRP21. Patients with BMI > 25 kg/m2 and higher CRP levels had a slightly lower risk of death, but the precision of the estimate was low. In the NHANES III cohort, Wulaningsih et al. included 7780 females aged ≥ 20 without a cancer history at baseline22. A total of 44 BC deaths were reported. The risk of BC death per log CRP increase was higher in BMI < 30 kg/m2 compared to BMI \(\ge\) 30 kg/m2 (HR 1.94 [95% CI = 0.51–7.29] vs. HR 1.40 [95% CI = 0.52–3.77]), however, the precision of the estimates was low.

Since our findings suggest that increased CRP across all BMI groups may be linked to worse BC prognosis, our results are similar to most of the results from the previous studies cited above. However, variations in study designs make a direct comparison of results difficult. Comparing our results with Ruan et al. is problematic since a BMI \(\ge\) 24 kg/m2 includes both normal-weight, overweight, and obesity, and they used CRP as a binary exposure 20. Also, stage IV patients were included, the cohort was younger (mean 53.7 years), and treatment choice differed (i.e. only 5.3% received radiotherapy). It is not clear when Ruan et al. collected their blood samples. Our study cohort is larger, but based on a single-center cohort study. In the study by Nelson et al., the patients were older (mean age from 70.3 to 71.5 years in CRP quartiles), and they included patients with in situ and stage IV disease21. Blood samples were collected, on average 7.8 years before BC diagnosis which constitutes a major difference to our study21. Also, patients with underweight were included in the stratified analyses, and patients with CRP > 10 mg/L were excluded. Like Nelson et al., the CRP levels were measured before BC diagnosis by Wulaningsih et al.22. The authors had no information on BC incidence and disease characteristics22.

Our findings could have clinical implications. CRP at the time of diagnosis may be used by clinicians to identify BC patients with an increased risk of inferior outcomes. The precision of the estimate in patients with obesity could improve with a larger sample size, as our results in patients with obesity could be due to a type 2 error. However, many other factors are involved in the link between obesity and BC, such as adipokines and estrogens, as we previously reviewed41, and could potentially be of more significant importance than CRP for patients with obesity. Also, CRP is a surrogate marker for systemic inflammation and many factors (e.g. smoking and blood pressure) influence CRP levels42, and we were not able to take all these factors into consideration. Furthermore, CRP is not an appropriate marker for the local inflammatory environment in the breast.

Our study has limitations. First, it is a single-institutional study and the results may not apply to other institutions and countries. Second, we adjusted for potential confounders, but we cannot rule out the possibility of residual confounding, such as smoking status and alcohol consumption. Third, not all BC patients referred for surgery at AUH agreed to donate a blood sample, and we do not have information on the non-participants, which could lead to selection bias. Fourth, BMI is an indicator of general obesity but does not reveal information on body composition, which is important information as the inflammatory signatures differ between abdominal and gynoid obesity43. Fifth, we only measured CRP levels at a single time point, so we were unable to evaluate the impact of fluctuations in CRP, for example, due to acute infection or lifestyle factors.

Conclusions

High circulating CRP levels at the time of BC diagnosis were associated with an inferior BC prognosis in this large Danish cohort. Furthermore, our study suggests that CRP may be a clinically relevant prognostic marker for BC prognosis across BMI groups. Future studies should investigate the relationship between CRP and BC in patients with obesity on a larger scale. Also, we encourage the investigation of other obesity-associated biomarkers in mapping the link between obesity and BC prognosis to identify patients in need of additional intervention.