Kaplan-Meier estimates of survival according to tertiles of β2-microglobulin (A and B), cystatin C (C and D), and C-reactive protein (E and F) concentrations in men (A, C, and E) and women (B, D, and F). Tertiles 1 (lowest), 2 (middle), and 3 (highest) represent 1.5 mg/L or less, 1.6 to 1.8 mg/L, and 1.9 mg/L or greater for β2-microglobulin; 0.80 mg/L or less, 0.81 to 0.93 mg/L, and 0.94 mg/L or greater for cystatin C; and 0.33 mg/L or less, 0.34 to 0.85 mg/L, and 0.86 mg/L or greater for C-reactive protein (to convert to nanomoles per liter, multiply by 9.524), respectively. Cumulative mortality rates during 8-year follow-up are given in parentheses.
Crude (A) and adjusted (B) hazard ratios of β2-microglobulin (β2-M), the renal function measures estimated glomerular filtration rate (GFR) and cystatin C, and the inflammation marker C-reactive protein (CRP) for 8-year mortality using the Cox proportional hazards model. Error bars represent 95% confidence intervals.
Receiver operating characteristic curves of 3 risk markers for 8-year mortality. The areas under the receiver operating characteristic curves are 0.70 (95% confidence interval [CI], 0.66-0.74), 0.66 (95% CI, 0.62-0.70), and 0.57 (95% CI, 0.53-0.61) for β2-microglobulin, cystatin C, and C-reactive protein (CRP), respectively.
Shinkai S, Chaves PHM, Fujiwara Y, Watanabe S, Shibata H, Yoshida H, Suzuki T. β2-Microglobulin for Risk Stratification of Total Mortality in the Elderly PopulationComparison With Cystatin C and C-Reactive Protein. Arch Intern Med. 2008;168(2):200-206. doi:10.1001/archinternmed.2007.64
Copyright 2008 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2008
The clinicoepidemiologic relevance of moderately elevated concentrations of circulating β2-microglobulin (β2-M) has not been established.
We examined whether serum β2-M concentration independently predicts total mortality in community-dwelling older populations and compared its predictive value with that of cystatin C and C-reactive protein (CRP) using a prospective cohort study of 1034 initially nondisabled persons 65 years and older as part of the Tokyo Metropolitan Institute of Gerontology Longitudinal Interdisciplinary Study on Aging. Cox proportional hazards models were used to examine independent associations between baseline β2-M levels and total mortality.
During a median follow-up of 7.9 years, 223 persons died. A strong dose-response relationship was found between baseline serum β2-M concentration and mortality risk, even after multiple adjustments. Compared with individuals in the lowest tertile of serum β2-M concentration, those in the middle (hazard ratio, 2.02; 95% confidence interval [CI], 1.35-3.04) and highest (hazard ratio, 2.84; 95% CI, 1.92-4.20) tertiles had a substantially increased mortality risk. Respective values were 1.28 (95% CI, 0.86-1.90) and 1.95 (95% CI, 1.31-2.89) for cystatin C and 1.39 (95% CI, 0.98-1.98) and 1.44 (95% CI, 1.00-2.06) for CRP; only the highest tertiles showed significantly higher mortality risks. The area under the receiver operating characteristic curve for 8-year mortality was greatest for β2-M (0.70; 95% CI, 0.66-0.74), followed by cystatin C (0.66; 95% CI, 0.62-0.70) and CRP (0.57; 95% CI, 0.53-0.61). Additional adjustment for renal function measures, inflammation markers, or both only partially reduced the association between β2-M and mortality.
Serum β2-M is an independent predictor of total mortality in a general population of older adults and may be a better predictor than cystatin C or CRP.
Knowledge about biological risk markers for mortality in the elderly population remains limited.1- 4 Such knowledge is relevant because it may provide important clues about the characterization of causal mechanisms for mortality or the development of predictive screening algorithms for identifying high-risk individuals. In this context, a search for novel mortality risk markers is warranted.
B2-microglobulin (β2-M) constitutes a light chain of the class I major histocompatibility antigens. Widely distributed in nucleated cells in the body, it is especially rich in immunocompetent cells, such as lymphocytes and monocytes. Various stimuli cause substantial amounts of the molecule to be shed into the circulation.5 Owing to a small molecular mass (11.8 kDa), circulating β2-M passes freely through the glomeruli and is reabsorbed and metabolized in the proximal tubules of the kidneys.5 As such, β2-M concentration in the blood is largely affected by the glomerular filtration rate (GFR) of the kidneys.6 In healthy individuals, β2-M concentration is fairly constant. On the other hand, blood levels of β2-M have been documented to increase in disease states such as renal dysfunction (owing to reduced catabolism) and in certain malignancies, autoimmune diseases, and infections (owing to increased production).7- 10 In the clinical setting, serum β2-M has been particularly useful as a marker of chronic kidney disease–related dysfunction.
Advanced aging also seems to affect the circulating β2-M concentration. Moderately elevated circulating β2-M levels are often found in older people. Whether this elevation merely reflects decreased renal function or a presence of low-grade inflammation in older adults11- 13 or also has its own clinical prognostic relevance in this population remains unclear.
The objectives of this study are 2-fold: (1) to test the hypothesis that increased serum β2-M concentration is a risk factor for total mortality in community-dwelling older adults independent of the impact of renal dysfunction, inflammation, and other major potential confounders and (2) to compare the predictive value of β2-M for total mortality with that of the renal function measures estimated GFR and cystatin C14 and the inflammation marker CRP, all of which are major mortality predictors.15- 17
The Tokyo Metropolitan Institute of Gerontology Longitudinal Interdisciplinary Study on Aging (TMIG-LISA) is a long-term prospective study on aging and health in Japanese older people who reside in Koganei City, a suburb of Tokyo, and Nangai Village, a rural area in northern Japan.18 Details of the cohort selection process were previously published.19 Briefly, in Koganei City, a random sample consisting of one-tenth of the population aged 65 to 84 years (n = 996) was recruited. Of those, 814 persons (81.7%) responded to the initial home visit interview survey in 1991, and 405 of those (49.8%) further participated in the baseline medical examination conducted at a community hall in 1991. In Nangai Village, of all community-dwelling residents 65 years and older (n = 940), 852 ambulatory persons (90.6%) were invited to participate in the baseline survey conducted at a community hall in 1992. Of those, 748 (87.8%) were interviewed, and 735 (86.3%) also underwent medical examination.
Overall, 1140 men and women aged 65 to 89 years participated in the baseline interview and medical examination in the TMIG-LISA. Of those, 1091 persons reported no dependency in any of 5 basic activities of daily living (moving, dressing, eating, toileting, and bathing); this subset constituted the cohort of our study. Study participants were followed up through January 31, 1999 (Koganei), and August 31, 2000 (Nangai) (the end of the first term of the TMIG-LISA). Of the original 1091 in the cohort, 11 had missing β2-M values, 18 had missing creatinine values, 26 had missing cystatin C values, 22 had missing CRP values, and 40 had missing data on one of the other covariates. The final sample size for this analysis was 1034. The Tokyo Metropolitan Institute of Gerontology review committee approved the study protocol, and informed consent was obtained from all the participants.
The baseline interviews and medical examinations included demographic questions, self-rated health status, functional status in instrumental and basic activities of daily living, medical history, smoking habits, alcohol intake, anthropometry, standard blood pressure measurement, a resting 12-lead electrocardiogram, collection of blood specimens, and performance-based physical function.
Blood samples were centrifuged at the examination sites, and the resulting serum samples were kept at 4°C until analysis or storage at −80°C within 24 hours. Serum β2-M concentrations at baseline were determined at a laboratory (SRL, Inc, Tokyo) in 1991 or 1992 using the latex immunoprecipitation method and an autoanalyzer (JCA-BM12; Nippon Denshi, Tokyo). The calibration and quality control data were as follows: the intrasubject coefficient of variation range was 0.5% to 1.5%, and the interday coefficient of variation range was 1.1% to 5.5% across a wide range of β2-M concentrations.
Other laboratory measurements (total cholesterol, high-density lipoprotein cholesterol, triglycerides, glucose, hemoglobin A1c, and hematologic values) at baseline were conducted using standardized procedures. The presence or absence of proteinuria was assessed using a casual dipstick, and its grade was categorized as “none,” “trace,” or “greater than trace.” Serum creatinine, cystatin C, and high-sensitivity CRP levels were determined at SRL, Inc, in 2006 using serum samples collected in 1991-1992 and stored at −80°C. Creatinine was measured using the enzyme colorimetric method and an autoanalyzer (Hitachi7170; Hitachi Ltd, Tokyo). Cystatin C and CRP were measured using particle-enhanced immunonephelometric assays (N Latex Cystatin C and N Latex CRP II, respectively; Dade Behring Inc, Deerfield, Illinois) and a nephelometer (BN II; Dade Behring). Intrasubject and interday coefficient of variation ranges were, respectively, 1.7% to 1.9% and 1.6% to 2.3% for cystatin C and 0.9% to 1.7% and 2.3% to 3.0% for CRP across a wide range of cystatin C and CRP levels.
Body mass index was calculated as weight in kilograms divided by height in meters squared. Medical history included the self-report of physician-diagnosed stroke, heart disease, type 2 diabetes mellitus, and hypertension. Performance-based measures of physical function included a 5-m measured walk at a usual pace (timed to the 0.1 second) and 2 measures of maximal grip strength in the dominant hand (to the nearest kilogram) using a Smedley-type dynamometer (Yagami Co, Tokyo). Mortality data were obtained through a comprehensive surveillance system that has been used successfully since the beginning of the TMIG-LISA. All deaths were ascertained by checking the local registries.
Cumulative survival through the end of follow-up by tertiles of baseline β2-M, cystatin C, and CRP concentrations were calculated using the Kaplan-Meier method. Differences in survival curves were evaluated using the log-rank test. The independent role of β2-M as a predictor of total mortality was evaluated using Cox proportional hazards regression models, with adjustments for age, sex, preexisting diseases, and other major potential confounding factors selected on the basis of previous literature findings. These factors included body mass index, systolic blood pressure, total and high-density lipoprotein cholesterol levels, albumin level, hemoglobin A1c level, smoking and alcohol drinking status (current, former, or never), self-rated health (excellent or good vs fair or poor), and usual walking speed (meters per second), which were entered into the models as continuous variables, except for smoking and alcohol status and self-rated health. Preexisting diseases included physician-diagnosed stroke, heart diseases (ischemic heart disease or others), hypertension, and type 2 diabetes mellitus, and each was entered into the models as dichotomous data (present vs absent). In these analyses, predictive values of β2-M for total mortality were evaluated using adjusted relative risks in each of the middle and highest tertiles of baseline β2-M concentration compared with the lowest tertile.
To further examine whether the association of β2-M and mortality was independent of renal function and inflammation, additional adjustment was made for renal function measures (degree of proteinuria, estimated GFR, and cystatin C level), inflammation markers (white blood cell [WBC] count and CRP level), or both using Cox regression models. We estimated the GFR using the 6-variable version of the Modification of Diet in Renal Disease equation20; we further multiplied the calculated figure by 0.741 to correct for the Japanese build.21 Estimated GFR, cystatin C level, WBC count, and log-transformed CRP level were entered into the models as continuous variables, and degree of proteinuria was entered as a categorical variable.
Receiver operating characteristic (ROC) curves were plotted for 3 risk markers (β2-M, cystatin C, and CRP). We used occurrence compared with nonoccurrence of events in 8 years as the outcome measure for this analysis. For statistical significance, 2-tailed P < .05 was used throughout the analysis. All data analyses were performed using a statistical software program (SPSS version 14.0 for Windows; SPSS Inc, Chicago, Illinois).
Serum β2-M concentrations at baseline in 1034 nondisabled older participants ranged from 0.8 to 6.6 mg/L, with a mean (SD) of 1.77 (0.55) mg/L and a leftward-skewed distribution. The distributions for men and women were similar, and the mean values were not significantly different (P = .12, t test). Thus, we categorized the participants into 3 tertile groups based on the distribution of β2-M concentrations in the whole population: lowest tertile, 1.5 mg/L or less; middle tertile, 1.6 to 1.8 mg/L; and highest tertile, 1.9 mg/L or greater.
Characteristics of the study population at baseline by β2-M categories are presented in Table 1. We observed strong and significant associations of increased serum β2-M levels with increased age and renal dysfunction, as assessed by proteinuria, serum creatinine level, estimated GFR, and cystacin C level. In addition, β2-M concentration was positively associated with preexisting medical conditions such as stroke, ischemic heart disease, and hypertension; CRP level; and former alcohol drinking and smoking habits and was negatively associated with albumin level, total and high-density lipoprotein cholesterol level, self-rated health, and usual walking speed.
During a median follow-up of 7.9 years, 121 men (28.1%) and 102 women (16.9%) died. We observed a graded relationship between serum β2-M level and total mortality (Figure 1A and B). The cumulative mortality rates during 8-year follow-up were 13.4% (20/149), 24.1% (32/133), and 46.3% (69/149) for men and 7.6% (20/264), 16.7% (27/162), and 29.9% (53/177) for women in the lowest, middle, and highest tertiles of β2-M concentration. The respective 1000 person-year mortality rates were 18.0, 34.8, and 74.2 for men and 10.1, 23.1, and 44.4 for women. There were clear differences in survival curves among the tertiles in men and women. The 3 cystatin C categories exhibited a similar pattern to those of β2-M in women, but the Kaplan-Meier estimates of mortality did not differ between the lowest (tertile 1) and middle (tertile 2) tertiles in men (P = .43, log-rank test). The 3 CRP categories did not grade mortality risk in women (P = .52, log-rank test).
These results were upheld in the Cox regression analysis (Figure 2). Multiple adjustments somewhat weakened the β2-M relationships, but they remained significant in subgroups with higher β2-M levels. When participants in the middle and highest tertiles of β2-M concentration were compared with those in the lowest tertile, the adjusted hazard ratios were 2.02 (95% confidence interval [CI], 1.35-3.04) and 2.84 (95% CI, 1.92-4.20).
Estimated GFR, cystatin C, and CRP also predicted mortality after multiple adjustments, but the prognostic ability of these indexes was inferior to that of β2-M. For example, although the intermediate and highest tertiles of β2-M concentration were associated with increased mortality risk compared with the lowest tertile, only the highest tertiles of estimated GFR, cystatin C, and CRP levels were significantly associated with higher mortality risks; the adjusted hazard ratios of the intermediate and highest tertiles were 1.09 (95% CI, 0.76-1.58) and 1.55 (95% CI, 1.07-2.23) for estimated GFR, 1.28 (95% CI, 0.86-1.90) and 1.95 (95% CI, 1.31-2.89) for cystatin C, and 1.39 (95% CI, 0.98-1.98) and 1.44 (95% CI, 1.00-2.06) for CRP, respectively (Figure 2).
The ROC curves of 3 risk markers (β2-M, cystatin C, and CRP) were generated for 8-year mortality (Figure 3). The area under the ROC curve was greatest for β2-M (0.70; 95% CI, 0.66-0.74), followed by cystatin C (0.66; 95% CI, 0.62-0.70) and CRP (0.57; 95% CI, 0.53-0.61). Additional adjustment for renal function measures, inflammation markers, or both partially reduced the association of β2-M and mortality (models 3A, 3B, and 4 in Table 2).
Exclusion of the deaths that occurred during the first 2 years of follow-up (n = 36) did not materially alter the results (adjusted hazard ratios for the middle and highest tertiles of β2-M, 1.66 [95% CI, 1.07-2.55] and 2.56 [95% CI, 1.70-3.85], respectively). Likewise, exclusion of individuals who had abnormal (within the worst 3%) values at baseline (n = 294) in body mass index, walking speed, handgrip strength, systolic blood pressure, hemoglobin, albumin, hemoglobin A1c, CRP, WBC count, creatinine, or glutamic pyruvic transaminase, plus those who rated their health as poor (n = 34), did not affect the association of β2-M and mortality (adjusted hazard ratios for the middle and highest tertiles, 2.14 [95% CI, 1.29-3.54] and 3.12 [95% CI, 1.89-5.13], respectively).
We observed a strong and independent association between β2-M concentration and total mortality during 8-year-follow-up in 65- to 89-year-old men and women enrolled in the population-based TMIG-LISA in 1991-1992. To our knowledge, this is the first study to document the clinical prognostic value of circulating β2-M concentration as an independent predictor of total mortality in a general population of older adults. We found evidence that the predictive value of circulating β2-M concentration may surpass that provided by established prognostic factors for mortality, such as estimated GFR and cystatin C (chronic kidney disease markers) and CRP (inflammation marker). In terms of risk stratification, serum β2-M levels significantly discriminated mortality risk in a dose-response manner across 3 tertile groups, whereas only the highest tertiles of estimated GFR, cystatin C level, and CRP concentration showed a significantly increased risk. Also, the area under the ROC curve for β2-M was greater than that for cystatin C or CRP. Compared with previously reported risk markers in terms of strength of association, high β2-M levels seem to be a better predictor of mortality than CRP,16 interleukin 6,17,22 fibrinogen,23 and total homocysteine levels.24,25
Cystatin C, a cysteine protease inhibitor, has been proposed to represent a superior marker for the detection of renal impairment compared with creatinine or a creatinine-based estimate of GFR and was recently documented to have significant prognostic value for total mortality and cardiovascular morbidity and mortality.15 C-reactive protein is a strong mortality predictor in older persons as well as an inflammation marker.16,17 The present study builds on the literature not only by replicating previous findings on the association of cystatin C and CRP in community-dwelling older Japanese persons but principally by documenting the independent value of β2-M as a predictor of mortality and by comparing its predictive value with that of cystatin C and CRP. These results are consistent with the previous finding that serum β2-M predicted the onset of functional decline in the Nangai cohort of the TMIG-LISA during 6 years of follow-up.26 Taken together, these findings suggest that serum β2-M could be a useful prognostic tool in the evaluation of elderly persons.
What potential explanations are there for the close link between β2-M and mortality? There is the possibility that β2-M is a better marker of reduced renal function than conventional measures, such as creatinine or estimated GFR, and even than cystatin C. Recent studies15,27 have documented that individuals with moderately reduced renal function are at increased risk for total mortality and cardiovascular morbidity and mortality. Older persons with higher β2-M concentrations may well be at increased risk for mortality via a renal dysfunction mechanism. However, as evidenced by the fact that additional adjustment for renal function measures, including estimated GFR and cystatin C, only moderately reduces the association of β2-M and mortality, such a renal mechanism alone could not fully explain the close link between β2-M and total mortality in this healthy older cohort.
Another possibility is that β2-M is an inflammation marker. Considering the relevant role of β2-M in immune responses, higher circulating β2-M concentrations may derive from increased systemic or local inflammation, possibly associated with subclinical cardiovascular diseases or malignancies. In addition to the low expected rate of β2-M–related malignancies (eg, multiple myeloma) in the 2 study sites of the TMIG-LISA, the close relationships between β2-M and preexisting cardiovascular diseases at baseline suggest the involvement of cardiovascular diseases. Also, there was a significant relationship between serum β2-M and log CRP. Low-grade inflammation characterized by increased levels of cytokines and acute-phase proteins as well as elevated WBC counts have been documented to predict all-cause mortality and cardiovascular mortality in the elderly population.16,17,22,23,28 In summary, the hypothesis that circulating β2-M concentrations reflect low-grade inflammation is attractive. However, as evidenced by the fact that additional adjustments for WBC counts and CRP levels only slightly reduced the association of β2-M and mortality, inflammation does not likely explain the association between β2-M and total mortality. Nonetheless, CRP and WBC count do not reflect the total inflammation burden, and the possibility of residual confounding by prevalent chronic disease burden, explaining (at least in part) the association observed herein, cannot be excluded.
In the clinical setting, β2-M is known to relate to several diseases with high mortality rates. To examine whether serious underlying illnesses would have confounded the association of β2-M with mortality, we analyzed data after exclusion of the deaths in the first 2 years of follow-up or after exclusion of those performing worst in a variety of variables at baseline. Neither analysis altered the association of β2-M and mortality, which speaks against the notion that the observed association may be totally explained by residual confounding from serious illnesses.
Several other unresolved issues remain. It remains to be established whether high levels of circulating β2-M have a direct pathogenic effect, analogously to what has been speculated for high CRP levels. In addition, knowledge about clinical factors that may affect β2-M concentrations beyond age, sex, chronic conditions, renal function, and inflammation remains limited. It also remains to be determined whether circulating β2-M levels are potentially modifiable and, if so, what the impact of such changes would be on β2-M levels.
This study has several strengths that should be acknowledged, particularly the use of a population-based sample of the TMIG-LISA cohort and 8-year follow-up. Also, the availability of comprehensive information in the TMIG-LISA database allowed us to adjust for important factors that might have confounded the relationship between elevated β2-M levels and total mortality. The study participants were all initially nondisabled older persons, enabling us to extend the results to general community-dwelling elderly persons.
In conclusion, serum β2-M concentration is an independent predictor of total mortality in a general population of older adults, and it may be a better predictor than cystatin C or CRP levels. Knowledge about the mechanisms of this association remains limited, but because of the strong association with mortality, investigation of potential mechanisms is warranted. Information about cause of death could provide clues about potential mechanisms (eg, atherosclerotic events and increased risk of infection).
Correspondence: Shoji Shinkai, MD, PhD, MPH, Department of Community Health, Tokyo Metropolitan Institute of Gerontology, 35-2 Sakae-cho, Itabashi-ku, Tokyo 173-0015, Japan (email@example.com).
Accepted for Publication: September 2, 2007.
Author Contributions: Dr Shinkai had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Shinkai, Chaves, Shibata, and Suzuki. Acquisition of data: Shinkai, Fujiwara, Watanabe, Shibata, Yoshida, and Suzuki. Analysis and interpretation of data: Shinkai, Chaves, and Watanabe. Drafting of the manuscript: Shinkai and Chaves. Critical revision of the manuscript for important intellectual content: Chaves, Fujiwara, Watanabe, Shibata, Yoshida, and Suzuki. Statistical analysis: Shinkai and Watanabe. Obtained funding: Shinkai, Shibata, and Suzuki.
Financial Disclosure: None reported.
Funding/Support: This study was supported by grants from the Tokyo Metropolitan Government, by the Japan Arteriosclerosis Prevention Fund, and by Grant-in-Aid for Scientific Research (B)(2)14370150 and Grant-in-Aid for Exploratory Research 17659192 from the Ministry of Education, Culture, Sports, Science, and Technology, Japan.
Role of the Sponsors: None of the funding sources had any role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript.
Previous Presentation: This study was presented at the 59th Annual Scientific Meeting of the Gerontological Society of America; November 17, 2006; Dallas, Texas.
Additional Information: This study was conducted as part of the TMIG-LISA (1991-present).
Additional Contributions: We thank the participants and municipal officers in Koganei City and Nangai Village as well as other staff members of the TMIG-LISA for their cooperation with this study.