Circulating beta-2 microglobulin and risk of cancer: the Atherosclerosis Risk in Communities Study (ARIC)

Abstract

Background

Serum β-2 microglobulin (Β2M), a major histocompatibility complex class I molecule that is a biomarker of kidney filtration and increased cell turnover, is elevated at the time of diagnosis in hematological and some solid cancers. However, serum Β2M was not examined prospectively as a marker for cancer risk. We hypothesized that in a population without a prior cancer diagnosis, serum Β2M is associated with risk of cancer (n=2,436), including colorectal (n=255), lung (n=298), breast (n=424), prostate (n=524) cancers, and hematological (n=176) malignancies.

Methods

The analytical cohort (n=12,300) was followed for incident cancers from 1990 through 2006. Β2M (range: 0.9–57.8 mg/L) was measured in stored serum collected in 1990–1992. Cox proportional hazards models were used to estimate hazard ratios (HR) and 95% confidence intervals for cancer incidence and mortality in relation to quartiles of Β2M.

Results

Adjusting for age, sex, race, center, education, body mass index, smoking, aspirin and hormone therapy (in women) and comparing highest to lowest Β2M quartiles, HRs were 1.25 (1.06–1.47; P-trend=0.002) for total cancer risk and 2.21 (1.32–3.70; P-trend=0.001) for colorectal cancer risk, with similar HRs for colon and rectal cancers. These associations remained after adjustment for an inflammatory biomarker, C-reactive protein, and after excluding the first three years of follow-up. Significant associations were also observed for mortality from total, lung and hematological cancers.

Conclusions

These findings provide the first evidence that higher serum Β2M is associated with increased colorectal cancer risk.

Impact

This study supports B2M as a potential biomarker for colorectal cancer risk.

Introduction

Beta-2-microglobulin (B2M) is a subunit of the major histocompatibility complex (MHC) class I molecule; it is found on the surface of all nucleated cells and is especially abundant in lymphocytes and monocytes (1). Under normal physiological conditions, some amount of Β2M may be secreted into circulation from cell surfaces or as a result of intracellular release (2–4) and is removed from blood predominantly by the kidneys (5). Thus, the net concentration of serum Β2M is determined by its generation and secretion into serum, as well as its elimination by the kidneys, so that serum B2M concentration increases with elevated turnover of cells and/or reduced kidney function (6). Therefore, in people without kidney disease, elevated Β2M serves as a marker of altered cell proliferation. Elevated serum Β2M concentrations have been observed in many pathological conditions, including renal disease, immunodeficiency, and autoimmune diseases (1, 7–9). Further, Β2M concentrations have been reported as elevated at the time of diagnosis in many solid (3, 10–14) and hematological cancers (15, 16).

There is also evidence to suggest that elevated serum Β2M concentrations are associated with risk of all-cause death. In a prospective cohort study of nondisabled people 65 years and older, Β2M concentrations better predicted death than high sensitivity C-reactive protein (CRP), an established inflammatory marker (1). In addition, several cohort studies, including the Atherosclerosis Risk in Communities Study (ARIC) (7), have reported positive associations between Β2M and aging, various cardiovascular conditions, and mortality from diabetes, CVD and all causes, suggesting that serum B2M may also serve as an inflammatory biomarker (7, 17, 18).

However, despite the evidence showing increased Β2M concentrations in cancer patients, there have been no published studies that examined serum Β2M in relation to the risk of subsequent cancer in the general population. Using ARIC data, we examined the risk of all cancers combined, the four most common solid cancers (colorectal, lung, prostate, and breast cancers) individually, and hematological cancers combined. We hypothesized that the strongest associations would be observed for colorectal and hematological cancers, but the mechanisms underlying these associations would be different. Given that colorectal cancer has been associated with immune response and inflammation (19–22), we reasoned that higher serum Β2M would be an early biomarker associated with increased risk for future colorectal cancer. While this hypothesis might also be relevant for breast and prostate cancers (23, 24), associations between circulating immune and inflammatory markers and these cancers have been less consistently reported (23, 25) than for colorectal cancer. In contrast, we hypothesized that B2M may serve as a marker of early detection for hematological cancers, since elevated B2M concentrations are consistently observed in multiple myeloma, malignant lymphomas, and leukemia and associated with progression and worse prognosis (4,5,15,16). Similarly, B2M could be a potential early detection biomarker for lung cancer, which is often diagnosed at an advanced stage, given that B2M was reported to be higher and occur more frequently in lung cancer cases compared to unaffected controls (26).

Materials and Methods

ARIC, a multicenter population-based prospective cohort study of atherosclerosis, was initiated in 1987–1989 when 15,792 volunteers ages 45–64 years were recruited from 4 communities: suburban Minneapolis (MN), Forsyth County (NC), Jackson (MS), and Washington County (MD) (27, 28). Baseline and four follow-up visits in 1990–1992, 1993–1995, 1996–1998, and 2011–2013 included interviews, laboratory measurements, and clinic examinations (27, 28). At each visit, participants were asked about their demographic characteristics, education, lifestyle behaviors, and reproductive and medical histories. Trained personnel collected anthropometric measures, blood (serum and plasma), and urine samples.

Β2M was measured in serum samples collected at Visit 2 (93% of the living cohort was re-examined), therefore most of the covariates shown in Table 1 and tested in the multivariable regression models were also collected at Visit 2, i.e., body mass index (BMI), waist-to-hip ratio (WHR), smoking status, pack-years, alcohol intake (g/week), use of hormone replacement therapy (HRT), and aspirin use during the 2 weeks before the examination. Education, consumption of red meat, and dietary fiber were collected at Visit 1. The analytical cohort was followed from Visit 2 until the end of 2006, for a median of 14.8 years of follow-up (maximum was 16.9 years).

Table 1.

Characteristics across quartiles of serum β-2-microglobulin (B2M) in 12,300 participants; ARIC, 1990–1992.

Characteristics (mean (SD) or prevalence %)	Quartiles of B2M (mg/L)
	≤1.66 N=3063	1.67–1.88 N=3140	1.89–2.15 N=3046	≥2.16 N=3051
Age at Visit 2 (years)	54.3 (5.1)	56.2 (5.5)	57.6 (5.6)	59.1 (5.5)
Female, %	59.0	53.2	53.3	56.1
White, %	63.7	75.4	80.2	78.5
Body mass index, kg/m2	26.7 (4.8)	27.3 (5.0)	28.0 (5.3)	28.9 (5.9)
Education, %
< high school	18.3	19.6	21.9	27.7
high school	31.4	33.1	34.3	33.8
> high school	50.4	47.3	43.8	38.6
Pack-years of smoking	12.7 (18.6)	14.4 (19.9)	15.6 (20.9)	16.7 (22.3)
Alcohol intake, g/week	6.1 (13.4)	5.6 (13.4)	5.1 (13.8)	4.4 (13.6)
Aspirin use, % yes	44.5	45.3	47.7	48.2
Diabetes, % yes	11.4	8.8	9.5	14.3
HRT use, % ever, for females	38.8	34.1	29.6	25.0
eGFR, mL/min/1.73 m2 (SD)	110 (11)	100 (10)	92 (11)	78 (16)
Cystatin C, mg/dL (SD)	0.7 (0.1)	0.8 (0.1)	0.9 (0.1)	1.1 (0.5)
Creatinine, mg/dL (SD)	1.1 (0.2)	1.1 (0.2)	1.2 (0.2)	1.3 (0.7)
hs-CRP*, mg/L (SD)	1.8 (1.8)	2.1 (2.1)	2.6 (2.6)	3.7 (4.0)

^*Geometric mean for hs-CRP was calculated because of the non-normal distribution.

Local institutional review boards approved the ARIC protocol and all participants provided their informed consent, which included permission for follow-up for the occurrence of non-cardiovascular diseases.

Ascertainment of cancer incidence and mortality

Incident cancers in 1990–2006 were identified by linkage to the state cancer registries of MD, MN, MS, and NC and were supplemented by active surveillance of the cohort, which included recording of hospital discharges and hospitalizations that were self-reported during annual follow-up telephone calls (29, 30). Additional cancer cases (not reported by cancer registries) were included after hospital records were obtained, reviewed, and verified for cancer. Primary site, date of cancer diagnosis, and source of diagnostic information (e.g., pathology report) were recorded (29, 30). Data on cancer stage were not uniformly collected.

Deaths from cancer as the underlying cause were obtained from death certificates in 1990–2011. The follow-up for death was five years longer than for incidence (1990–2006), which explains the higher number of deaths from lung cancer (n=352) compared to the number of incident lung cancers (n=298). Also, although the inclusion of only first primary incident cancers did not affect cancer deaths, it resulted in a decrease of incident cancer cases.

Measurements of biological markers

WBC count was measured in local centers at the time of the Visit 2 exams (1990–1992). Serum concentrations of Β2M, high-sensitivity CRP (hs-CRP) as well as cystatin (marker of kidney function), were measured in 2012–2013 using stored serum from Visit 2. Β2M and CRP were measured using the Roche β2M reagent and Roche latex-particle enhanced immunoturbidimetric assay kits, respectively, and read on the Roche Modular P Chemistry analyzer (Roche Diagnostics, Indianapolis, IN). Cystatin C was measured using the Gentian immunoassay (Gentian Technology AS, Moss, Norway). Another marker of kidney function, creatinine, was measured by the modified kinetic Jaffé method (31) in 1990–1992. Estimated glomerular filtration rate (eGFR), an index of kidney function, was calculated in ARIC using standardized cystatin C and standardized creatinine (32).

The analytical sample for this analysis (n=12,300) excluded the participants who did not consent to participate in studies of non-cardiovascular diseases (n=187); were not black or white (n=48); were blacks from Minneapolis and Washington County (n=55); did not participate in Visit 2 (n=1323), self-reported history of cancer at baseline (n=882); had cancer diagnosed before Visit 2 (n=214), or had missing information about B2M (n=786) (some individuals were excluded for more than one reason).

Statistical analysis

Cox proportional hazards models were used to estimate hazard ratios (HR) and 95% confidence intervals (CI) for incident cancers in relation to Β2M modeled as quartiles. Test for trends were calculated by fitting quartile numbers as an ordinal variable into the Cox regression model. The proportional hazards assumption was tested by adding an interaction term between follow-up time and Β2M concentration, and the assumption was not violated for any individual cancer or total cancer.

For each cancer we analyzed two multivariable-adjusted models. Model 1 was adjusted for age (continuous), sex, and a 5-level variable combining race and ARIC field center. Model 2 was additionally adjusted for education (less than high school, high school, more than high school), body mass index (BMI; <25, 25–29.9, ≥30 kg/m²), smoking status (yes, no) and pack-years of smoking (continuous), use of aspirin (yes, no), sex-HRT variable (men, women never taking HRT, and women who were former or current HRT users), and eGFR (continuous). We adjusted for eGFR for two reasons: (1) serum Β2M concentration depends on its elimination by the kidneys (Spearman correlation coefficient between Β2M and eGFR ρ=−0.72) and (2) previous studies found that reduced kidney function and chronic kidney diseases were associated with poor prognosis for several cancers, including cancers of the lung (33), bladder (33), kidney (34, 35), and several others (36–39). Therefore, we aimed to understand whether the Β2M-cancer associations were independent of reduced kidney function. Additional covariates that were tested for confounding but were not included in final models included consumption of alcohol, red meat, and dietary fiber.

The fully-adjusted breast cancer models also included age at menarche (continuous), number of live births (0, 1–2, 3–4, and ≥5), and menopausal status (yes, no). The colorectal cancer analysis was also run separately by subsite (colon, rectum). In addition, we conducted several supplementary analyses for all cancer types. We further adjusted for WHR, in addition to BMI, to decrease the likelihood of residual confounding by body fatness (the Spearman correlations were: between Β2M and BMI ρ=0.15; B2M and WHR ρ=0.18; BMI and WHR ρ=0.46). We also assessed the independence of the B2M and colorectal cancer association from hs-CRP, which we previously found to be associated with colorectal cancer in ARIC (30). To do so, we first determined the Spearman correlation between Β2M and hs-CRP (ρ=0.27). In addition, we conducted several sensitivity analyses: we excluded participants (~10% of participants) with evidence of acute inflammation (hs-CRP >10 mg/L) and those with reduced kidney function measured as eGFR ≤ 90 mL/min/1.73 m² (~33% of participants), i.e., the cut-point recommended by the National Kidney Foundation (40). To minimize the impact of undiagnosed disease, we excluded people with less than 3 years of follow-up (4% of participants). Although we were not able to study associations across cancer stage, we examined associations between B2M and risk of death from cancer.

Given that each biomarker is measured with some degree of error, we sought to determine whether combining B2M and hs-CRP would better capture the association with colorectal cancer than either biomarker measurement alone; we restricted this analysis to colorectal cancer due its stronger association with B2M. To do so, we created a combined score, by summing hs-CRP and B2M as categorical variables (where the quartiles for each variable were coded 0–3). The resulting combined score, which ranged from 0 to 6, was categorized into 4 groups (0–1, 2–3, 4–5, and 6). In addition, we tested the risk of colorectal cancer in relation to two other combined scores based on hs-CRP and B2M that were created as (1) a sum of the normalized log-transformed continuous biomarkers; (2) a linear combination of normalized log-transformed continuous biomarkers with weights equal to beta coefficients derived from Cox regression when hs-CRP and B2M were included simultaneously.

We also tested whether the multivariable models that included both B2M and hs-CRP (normalized log-transformed), provided better fit than the model with Β2M alone; we tested for differences in model fit using the likelihood ratio chi-square test. Analyses were performed using SAS v9.3 (SAS Institute Inc., Cary, NC); all statistical tests were two-sided.

Results

The analytical cohort (n=12,300; 55% women, 26% black and 74% white) was followed for incident cancer from 1990–1992 through 2006. During follow-up, 2,436 incident cancers were ascertained, including 255 colorectal, 298 lung, 424 breast, 524 prostate, and 176 hematological cancers. Serum Β2M concentrations at the start of follow-up ranged from 0.9 to 57.8 mg/L, with a mean (SD) of 2.0 (1.5) mg/L. B2M was unrelated to sex, but higher Β2M quartiles were associated with older age, higher proportions of whites (versus blacks), less than high school education, higher BMI, greater pack-years of smoking, recent aspirin use, and less ever use of hormone therapy (Table 1). These findings were mostly consistent with previous studies (41–43).

In the minimally adjusted Model 1, for the highest versus lowest Β2M quartile, there were significant positive associations of B2M with the risk of total (HR =1.20, 95% CI, 1.07–1.36; P-trend=0.0003) and colorectal cancers (HR=1.55, 95% CI, 1.04–2.30; P-trend=0.02) and a possible association with hematological cancers (HR=1.42; 95% CI, 0.92–2.21; P-trend=0.07) (Table 2). In adjusted Model 2, compared to the lowest quartile, people in the highest quartile were more likely to develop total (HR=1.25, 95% CI, 1.06–1.47; P-trend=0.002) and colorectal cancers (HR= 2.21; 95% CI, 1.32–3.70, P-trend=0.001). When the analysis for colorectal cancer was stratified by subsite, HRs were similar for colon (n=206) and rectal (n=67) cancers: 1.72 and 1.98 for the highest versus lowest Β2M quartile (not presented in the tables). After multivariable adjustment (Model 2), the HRs for hematological cancers did not change substantively. For lung cancer, there was a statistically significant association in the fourth versus first quartile only: HR=1.64 (95% CI 1.03–2.61) in Model 2 (Table 2). B2M was not associated with risk of breast or prostate cancers in either Model 1 or 2.

Table 2.

Hazard ratios (95% CI) for the incidence of total cancer and cancer types by quartiles of B2M; ARIC, 1990–2006.

Cancer	No. of cases	Quartiles of B2M (mg/L)
		≤1.66	1.67–1.88	1.89–2.15	≥2.16	P-trend
Total cancer	2,436	512	583	659	682
Person-years	162,617	42,356	42,688	40,069	37,503
Incidence ratesa	14.98	12.09	13.66	16.45	18.18
Model 1b		1 (reference)	1.01 (0.89–1.14)	1.14 (1.02–1.29)	1.20 (1.07–1.36)	0.0003
Model 2c		1 (reference)	1.02 (0.89–1.17)	1.20 (1.05–1.38)	1.25 (1.06–1.47)	0.002
Colorectal cancer	255	44	64	77	70
Person-years	162,617	42,356	42,688	40,069	37,503
Incidence rates	1.57	1.04	1.50	1.92	1.87
Model 1		1 (reference)	1.35 (0.91–1.99)	1.66 (1.13–2.43)	1.55 (1.04–2.30)	0.02
Model 2		1 (reference)	1.37 (0.89–2.09)	1.93 (1.25–2.99)	2.21 (1.32–3.70)	0.001
Lung cancer	298	59	72	66	101
Person-years	162,617	42,356	42,688	40,069	37,503
Incidence rates	1.83	1.39	1.69	1.65	2.69
Model 1		1 (reference)	0.99 (0.70–1.40)	0.88 (0.61–1.26)	1.31 (0.93–1.84)	0.12
Model 2		1 (reference)	1.09 (0.74–1.61)	0.98 (0.64–1.50)	1.64 (1.03–2.61)	0.06
Breast cancer (women only)	424	103	91	103	127
Person-years	92,554	25,434	23,206	22,114	21,801
Incidence rates	4.58	4.05	3.92	4.66	5.83
Model 1		1 (reference)	0.88 (0.66–1.17)	0.99 (0.75–1.31)	1.19 (0.90–1.56)	0.14
Model 2*d		1 (reference)	0.84 (0.58–1.23)	1.13 (0.78–1.64)	1.01 (0.64–1.59)	0.62
Prostate cancer	524	119	153	149	103
Person-years	70,062	16,922	19,482	17,955	15,703
Incidence rates	7.48	7.03	7.85	8.30	6.56
Model 1		1 (reference)	1.12 (0.87–1.42)	1.15 (0.90–1.48)	0.82 (0.62–1.08)	0.21
Model 2		1 (reference)	1.14 (0.88–1.48)	1.20 (0.89–1.62)	0.95 (0.66–1.36)	0.98
Hematological cancers	176	35	38	43	60
Person-years	162,617	42,356	42,688	40,069	37,503
Incidence rates	1.08	0.83	0.89	1.07	1.60
Model 1		1 (reference)	0.92 (0.58–1.47)	1.02 (0.65–1.62)	1.42 (0.92–2.21)	0.07
Model 2		1 (reference)	1.06 (0.62–1.79)	1.17 (0.67–2.00)	1.50 (0.80–2.81)	0.18

^aPer 1,000 person-years

^bModel 1 – adjusted for age, sex, race-study center.

^cModel 2 – all variables in Model 1 + education, BMI, smoking status and pack-years of smoking, use of aspirin, hormone replacement therapy (in women) and eGFR.

^dModel 2*– all variables in Model 2 + menopausal status, age at menarche, and number of live births.

Additional adjustment of Β2M-cancer associations for WHR or log transformed hs-CRP did not markedly change any of the observed associations (Supplementary table 1, Models 3 and 4, respectively). After excluding participants with a possible acute inflammatory response (CRP >10 mg/L), associations for colorectal and total cancer remained statistically significant (Supplementary table 1, Model 5). Likewise, the associations for total and colorectal cancer also did not appreciably change after excluding those with eGFR≤90 mL/min/1.73 m²: for the highest versus lowest B2M quartile, the HRs were 1.21 (95% CI, 0.96–1.51; P-trend=0.004) for total cancer and 2.18 (95% CI, 1.18–4.04; P-trend=0.004) for colorectal cancer (Supplementary table 1, Model 6). After excluding people with less than 3 years of follow-up, the positive associations with total or colorectal cancer did not change, whereas the hazard ratios for lung and hematological cancers for the fourth quartiles (versus the first quartiles) diminished (Supplementary table 1, Model 7). Further, there was no evidence of multiplicative interactions of serum Β2M with age, BMI, smoking, HRT or aspirin use in relation to any cancer (P-values for all interactions were > 0.1). Stratified associations for colorectal and lung cancers are presented in Supplementary tables 2a and 2b, respectively.

Further, we analyzed serum B2M and risk of death from cancer during follow-up through 2011 and showed that the patterns of associations were similar to the associations for incident total cancer and each cancer type, respectively (Table 3). The association for death from hematological cancers was stronger than for incidence and was statistically significant, while death from colorectal cancer was not significantly associated with B2M (with only 85 deaths observed). The association for death from lung cancer was of the same magnitude as for incidence and was significant.

Table 3.

Hazard ratios (95% CI) of death from total cancer and cancer types by quartiles of serum B2M; ARIC, 1990–2011.

Cause of death	No. of deaths	Quartiles of B2M (mg/L)
		≤1.66	1.67–1.88	1.89–2.15	≥2.16	P-trend
Total cancer	1160	237	261	322	344
Person-years	219,393	57,374	57,868	54,563	49,588
Mortality ratea	5.29	4.13	4.51	5.90	6.94
Model 2b		1 (reference)	0.92 (0.76–1.12)	1.09 (0.89–1.34)	1.30 (1.02–1.64)	0.01
Colorectal cancer	85	13	27	24	21
Person-years	219,393	57,374	57,868	54,563	49,588
Mortality rate	0.39	0.23	0.47	0.44	0.42
Model 2		1 (reference)	1.60 (0.77–3.32)	1.50 (0.68–3.33)	1.73 (0.68–4.44)	0.34
Lung cancer	352	70	84	84	114
Person-years	219,393	57,374	57,868	54,563	49,588
Mortality rate	1.60	1.22	1.45	1.54	2.30
Model 2		1 (reference)	1.05 (0.73–1.50)	1.05 (0.72–1.55)	1.69 (1.10–2.59)	0.02
Breast cancer (women only)	77	19	18	23	17
Person-years	124,783	34,480	31,434	29,888	28,982
Mortality rate	0.62	0.55	0.57	0.77	0.59
Model 2		1 (reference)	1.12 (0.49–2.53)	1.23 (0.51–2.95)	0.80 (0.26–2.47)	0.83
Prostate cancer	58	14	15	20	9
Person-years	94,609	22,894	26,434	24,676	20,605
Mortality rate	0.61	0.61	0.57	0.81	0.44
Model 2		1 (reference)	0.78 (0.36–1.68)	0.93 (0.42–1.68)	0.42 (0.14–2.07)	0.25
Hematological cancers	138	21	29	34	54
Person-years	219,393	57,374	57,868	54,563	49,588
Mortality rate	0.63	0.37	0.50	0.82	1.09
Model 2		1 (reference)	1.21 (0.64–2.28)	1.48 (0.78–2.82)	2.44 (1.20–4.95)	0.001

^aPer 1,000 person-years

^bMultivariable model (Model 2, Table 2) – adjusted for age, sex, race, study center, education, BMI, smoking status and pack-years of smoking, use of aspirin, hormone replacement therapy (in women) and eGFR.

Finally, we examined whether the combined score based on hs-CRP and B2M was more strongly associated with colorectal cancer incidence than was Β2M alone. As shown in Table 4, the associations for Combined Score 1, which included categorical B2M and hs-CRP, were very similar in magnitude and behaved in a parallel fashion to the Β2M-colorectal cancer association: the HR for highest versus lowest quartile was 2.49 (95% CI: 1.42–4.38, P-trend=0.0002). The associations with two other combined scores based on continuous log-transformed hs-CRP and B2M were also similar to those with B2M and to each other (Supplementary Table 3). However, the multivariable model with continuous hs-CRP and B2M included simultaneously in the model showed improvement in model fit compared to the multivariable model that included Β2M only (P=0.02).

Table 4.

Hazard ratios (95% CIs) for colorectal cancer incidence by quartiles of biomarker (hs-CRP or B2M) or inflammatory score; ARIC, 1990–2006.

Biomarker	Quartile 1	Quartile 2	Quartile 3	Quartile 4	p-value
B2M (mg/L)	≤1.66	1.67–1.88	1.89–2.15	≥2.16
Number of cases/participants	44/3,063	64/3,140	77/3,046	70/3,051
HR(95% CI) a	1 (reference)	1.37 (0.89–2.09)	1.93 (1.25–2.99)	2.21 (1.32–3.70)	0.001
hs-CRP (mg/L)	≤1.07	1.07–2.26	2.27–4.87	≥4.88
Number of cases/participants	42/3,069	59/3,091	76/3,069	78/3,066
HR (95% CI)	1 (reference)	1.36 (0.89–2.09)	1.84 (1.21–2.81)	2.17 (1.39–3.37)	0.0002
Combined Score 1b	0–1	2–3	4–5	6
Number of cases/participants	44/2,813	82/4,621	97/3,660	32/1,201
HR (95% CI)	1 (reference)	1.19 (0.75–1.67)	1.84 (1.20–2.82)	2.49 (1.42–4.38)	0.0002

^aMultivariable model (Model 2, Table 2) – adjusted for age, sex, race, study center, education, BMI, smoking status and pack-years of smoking, use of aspirin, hormone replacement therapy (in women) and eGFR.

^bSum of hs-CRP and B2M (categorized in quartiles that ranged from to 0 to 3)

Discussion

In this population-based cohort study of adults followed for a maximum of 17 years, participants with a serum Β2M concentration in the highest versus lowest quartiles had a 25% higher risk of total cancer incidence and 121% higher risk of colorectal cancer incidence. These findings support our a priori hypothesis that B2M, a biomarker of increased cell turnover, is associated with CRC risk.

The robust association between B2M and colorectal cancer risk was independent of the inflammatory marker hs-CRP and persisted in all sensitivity analyses: it remained after excluding those with eGFR ≤90 mL/min/1.73m² (potentially abnormal kidney function) or those with CRP <10 mg/L (acute inflammation). It also did not markedly change after excluding the first three years of follow-up, which implies that higher Β2M concentrations do not reflect the subclinical presence of malignancy. The magnitude of the Β2M-colorectal associations was similar to the strength of the association between hs-CRP and colorectal cancer in this study: for the highest versus lowest hs-CRP quartile, the HR was 2.17 (95% CI, 1.39–3.37; P-trend=0.0002). This association is unlikely to be explained by residual confounding by obesity, because Β2M concentration was only weakly correlated with BMI or WHR (Spearman ρ=0.15 and ρ=0.18, respectively). This characteristic distinguishes B2M from CRP, since CRP concentrations are moderately associated with obesity, both in our study (Spearman r=0.39) and other studies (44, 45). Also, we showed that linear combinations of two biomarkers, Β2M and CRP, were positively associated with the risk of developing colorectal cancer. Importantly, the linear combinations of Β2M and CRP did not yield a higher HR than either biomarker put into the model alone. However, the inclusion of CRP and B2M simultaneously in the model improved the fit of the model suggesting that the combined score may be a better marker for the development of colorectal cancer than Β2M alone.

The associations for incidence and death associated with B2M behaved in parallel fashion for total cancer and each cancer type. The trends for deaths were significant for total, lung and hematological cancers. For the lung cancer death, the association was driven by the increased hazard ratio in the highest versus lowest B2M quartile. For death from hematological cancers, the HRs in all quartiles were higher than the corresponding HRs for risk of hematological cancers, which is consistent with a strong correlation between B2M concentration and the progression of hematological cancer (4, 5). Additional studies are needed to better understand the link between B2M and risk of hematological and lung cancers.

The observed associations of elevated B2M with the total cancer incidence and death, colorectal cancer incidence and the death from lung and hematological cancers are consistent with experimental evidence that links Β2M expression with carcinogenesis. In mouse studies, Josson et al. reported that Β2M overexpression in cancer cells promoted the growth and progression of human prostate, breast, lung, and renal cancer cells and led to metastases and lethality (46). In addition, several human studies have shown that serum Β2M is elevated at the time of cancer diagnosis or thereafter and increases with progression in many solid cancers, including those of the prostate (11, 12), breast (10, 13, 14), lung (26, 47), and kidneys (48), and, to a greater extent, in hematological cancers, such as lymphomas and multiple myeloma (15, 16, 49). Further, serum Β2M was reported to be a strong indicator of poor prognosis and reduced survival of patients with solid and hematological cancers, including kidney, prostate and breast tumors, multiple myeloma, lymphomas and leukemia (11, 14, 16, 50–52).

Mechanisms explaining the positive associations between serum Β2M concentrations and carcinogenesis have not been precisely established, but, most likely, Β2M plays several roles: (1) pro-angiogenic; (2) pro-tumorigenic; (3) it may drive innate pro-inflammatory cytokines (e.g., TNF-α, interleukin (IL)-1, IL-6 and IL-8) and growth promoting factors (e.g., vascular endothelial growth factor, epidermal growth factor receptor, fatty acid synthase, insulin and insulin-like growth factor 1 receptor) in both cancer patients and unaffected people (3, 4, 12, 53); and (4) may drive epithelial–mesenchymal transition (46).

Thus, Β2M may have a direct pathogenic role in immune response and inflammation or may serve as an early biomarker of increased cell turnover. Its role as an inflammation marker is corroborated by studies reporting positive associations of B2M with aging, CVD, diabetes and mortality from those diseases and all causes (1,7–9). In a prospective cohort study of 1034 healthy people 65 years and older followed for 8 years, the risk of all-cause death was significantly increased by 102% in the second and by184% in the third tertile versus the first B2M tertile, independent of lifestyle factors, renal dysfunction and preexisting comorbidities (1). Of note, B2M was a better predictor of mortality than CRP or cystatin.

The function of B2M as a marker of increased cell turnover may be particularly relevant to colorectal cancer, since the tissues in colon and rectum are characterized by rapidly-dividing epithelia cells and immune response elicited by gut microbiota (22, 54). However, we do not know exactly, whether B2M is secreted by immune or tumor cells, i.e. whether the elevated B2M concentrations reflect local or systemic response. Finally, B2M could be a marker of undiagnosed cancer, especially for hematological malignancies, since B2M is an established marker of progression and worse prognosis in those cancers (4, 5,16).

A limitation of our study is that Β2M was measured only once at Visit 2 (1990–92). However, a recent study using Third National Health and Nutrition Examination Survey (NHANES III) data on 795 people examined two B2M measurements taken 18 days apart and concluded that Β2M has a low-short-term variability (coefficient of variation was 8.4%) (55). Another limitation of our study is the low number of cases ascertained in each cancer group, especially hematological cancers, which precluded both the study of leukemia and lymphoma separately and the conduct of subgroup analyses. The main strength of this study is that, to our knowledge, it is the first to examine serum Β2M concentrations in relation to future cancer risk. We studied this association in a large prospective population-based study that has rigorously collected data on multiple risk factors related to Β2M and cancer, including information on other markers of kidney function and inflammation. This allowed us to evaluate the association of Β2M with cancer development independently of other risk factors.

In conclusion, our study provides the first evidence that higher serum Β2M in people without a prior cancer diagnosis is associated with increased colorectal cancer risk. Significant associations were also observed for mortality from total, lung and hematological cancers. Further research is needed to determine whether for these associations, circulating Β2M is marking increased turnover of tumor or immune cells.