Abstract
Background and Aims
Recent research has suggested that high vitamin B12 levels may be associated with increased mortality after ICU admission. However, it is known that impaired liver function may lead to elevated B12 since B12 is metabolized through the liver, and therefore high B12 levels may serve as a proxy for poor liver function. The aim of this study is to assess the impact that liver function and liver disease have on the relationship between high vitamin B12 levels and mortality in the ICU.
Methods
We performed an observational cohort study using ICU data that were collected from patients admitted to four ICU types (medical, surgical, cardiac care and cardiac surgery recovery) in one large urban hospital from 2001 to 2008. We analyzed the medical records of 1,684 adult patients (age ≥ 18 years) who had vitamin B12 and liver function measurements up to 14 days prior to ICU admission or within 24 hours after admission.
Results
While we found an association between high B12 and mortality when we did not control for any potential confounders, after we adjusted for liver function and liver disease, no significant association existed between B12 and mortality using multivariable logistic regression (30-day mortality: OR=1.18, 95% CI 0.81 to 1.72, p=0.3890; 90-day mortality: OR=1.20, 95% CI 0.84 to 1.71, p=0.3077).
Conclusions
Elevated B12 levels are not a significant predictor of mortality after ICU admission when liver function is controlled for, and may instead be a proxy for poor liver function.
Keywords: Vitamin B12, critical care, liver function, mortality, intensive care unit, liver disease
1. Introduction
Much of the clinical research on vitamin B12 (cobalamin) has focused on B12 deficiency, but less is known about the effects of elevated B12 levels. Several studies have looked at B12 and mortality in elderly populations: some have found an association between high B12 levels and mortality1–6 and others have not7, 8. When studying B12, it is important to account for liver function and liver disease, as impaired liver function is known to affect B12 levels, primarily because B12 is metabolized by the liver9, 10. Although some of the above studies did adjust for albumin (ALB)1, 4, 5, none of them explicitly adjusted for a full liver function panel in a multivariable model. Lin et al11 found that elevated B12 levels were associated with increased mortality in hepatocellular carcinoma patients (N=90). Dou et al12 studied N=149 patients with liver failure and found that B12 levels were positively correlated with mortality after adjusting for the Model for End-stage Liver Disease (MELD) score13. However, neither study focused on ICU patients.
Few studies have looked at elevated vitamin B12 levels and mortality for ICU patients. Olivieri et al14 conducted a study of 250 elderly patients with acute myocardial infarction admitted to a cardiac care unit (CCU) and found that vitamin B12 was not a significant predictor for cardiovascular-related mortality. Corcoran et al15, 16 studied on the effect of vitamin B12 levels on mortality in the ICU and found that there was no association between B12 deficiency and mortality, and that weak correlations existed between B12 levels, CRP, and Sequential Organ Failure Assessment (SOFA) score, a measure of acuity in the ICU17. Manzanares and Hardy18 conducted a review of studies on the relationship between B12 and mortality and noted that the data on B12 and mortality for critical care patients was scarce.
In a recent study, Sviri et al19 examined B12 in the medical ICU (MICU), N=663, and found that log B12 levels were linearly positively correlated with 90-day mortality risk after adjusting for several other factors including cirrhosis and chronic liver disease but not measures of liver function (OR 1.7, 95% CI 0.9 to 3.1, p<0.05). We hypothesize that poor liver function may be the underlying cause of both the high B12 levels and increased mortality and, furthermore that B12 levels are not linearly associated with mortality risk, i.e., mortality risk may be elevated for patients with either deficient or high B12 levels. In this study, we investigate whether there is a relationship between high B12 and mortality after adjusting for liver function and liver disease.
2. Materials and methods
We performed an observational cohort study based on electronic medical records that were collected during the course of clinical ICU care and subsequently de-identified. Our study population consisted of all adult patients that had B12 values and liver function values (ALB, ALT, AST, ALP, PT, and TB) recorded in version 2.6 of the Multiparameter Intelligent Monitoring in Intensive Care II (MIMIC-II)20 database. MIMIC-II contains a wide variety of clinical and demographic information on approximately 24,000 adult patients 18 years or older that were admitted to the ICU at Boston’s Beth Israel Deaconess Medical Center from 2001 to 2008. The database is freely available; any researcher who accepts the data use agreement and has completed human subjects training can apply for permission to access the data. We did not need patient consent as all of the data are de-identified. All of the authors have completed human subjects training, and we conducted the study under National Institutes of Health Institutional Review Board exemption number 4193.
We used lab values, including B12 values, recorded from 14 days before ICU admission up to 24 hours after admission. Other variables, such as comorbidities, were based on combinations of International Classification of Diseases, ninth revision (ICD-9) diagnosis codes, which are assigned at the end of the hospital stay.
There is no consensus on the definition of high or elevated B12. Previous studies have chosen a variety of levels to represent high or very high B12 (pmol/l) levels. Some examples include: 664.219; 5131; 700 (high) and 1500 (very high)11; and 601 (high) and 1000 (very high)10. Excess B12 is either stored in the liver, present in the blood without any adverse effects, or excreted. There are no studies investigating the potential clinical effects of high B12 because, in general, there are no known adverse effects caused by elevated B12. It is probably for this reason that there is no cutoff for high B12 based on clinical outcomes. We chose a level of 1000 pmol/l because we wanted to be sure of identifying a cohort with abnormally high levels of B12 that may potentially have adverse clinical outcomes. With regard to B12 deficiency, the cut point for clinical B12 deficiency is ≤ 148 pmol/l21, although even the definition of deficiency is an area of debate22, 23. We chose to categorize the B12 levels into deficient (≤ 148 pmol/l), mid-range (> 148 pmol/l to ≤ 1000 pmol/l), and high (> 1000 pmol/l) groups. By doing so, we were able to estimate the mortality risk for elevated B12 compared to mid-range or deficient levels, and we did not need to make the assumption that mortality risk increases linearly with B12 levels.
We modeled two outcomes: death 30 days and 90 days after admission to the ICU. We used four groups of independent variables in our analysis: vitamin B12 category, demographics, other potential confounders, and liver disease and function. Demographic information included age at admission, gender, race, and insurance coverage. We included the following variables to adjust for liver function: ALT, AST, TB, ALP, ALB, PT, liver cancer, and non-cancerous liver disease. In general, we categorized laboratory tests using quartiles (except B12) because the laboratory values did not appear to have a linear relationship with mortality. The diagnosis of non-cancerous liver disease was defined using the following ICD-9 codes: 570 (liver necrosis), 571.5–9 (chronic liver disease and cirrhosis, excluding the 571.0–571.3 codes explicitly mentioning alcoholic liver disease as we have added alcoholism to the analysis separately), 572* (liver abscess and sequelae of chronic liver disease), and 573* (other disorders of the liver, including hepatitis and hepatopulmonary syndrome), where * indicates the group of codes that begin with that sequence. Liver cancer was defined using ICD-9 codes 155* (primary liver and intrahepatic bile duct cancer) and 197.7 (secondary liver metastases). Elevated liver function test values were not part of the definition of liver disease; liver function tests were added to the analysis separately.
We considered several other potentially confounding factors for inclusion in the final model. We used the Simplified Acute Physiology Score (SAPS)24 – with the age component removed so we could observe the effect of age separately – in order to account for the patient’s acuity during the first 24 hours of the ICU stay. We considered other ICU-related factors such as first service unit – CCU, CSRU (cardiac surgery recovery unit), MICU, and SICU (surgical ICU) – and non-liver lab tests such as serum creatinine, hemoglobin, and mean corpuscular volume (MCV). We also collected information on B12 supplementation and other interventions (dialysis, total parenteral nutrition (TPN), and inotropic support). Finally, we considered multiple disease conditions for inclusion in the final model: alcoholism, sepsis, respiratory disease, myeloproliferative disease, stroke, coronary artery disease (CAD), hypertension, other cardiovascular disease conditions not captured by any other comorbidities (Other CVD), immunocompromised status, gastrointestinal (GI) hemorrhage, and neuropsychiatric disorders. These potentially confounding variables were considered as candidates in the final model for a variety of reasons, either because they affect or are related to B12 levels (myeloproliferative disease, B12 supplementation, TPN, creatinine, hemoglobin, MCV), are known to affect liver function (alcoholism), or identify common subpopulations in the ICU (first service unit, CAD, stroke, hypertension, other CVD conditions not already captured by other information). We also considered for inclusion variables associated with chronic conditions that are known to be associated with higher mortality in the ICU, namely, dialysis, immunocompromised status, and respiratory disease. These chronic conditions are used in the APACHE-II (acute physiology and chronic health evaluation) score25 which is an alternative to the SAPS score for measuring acuity in the ICU but, unlike the SAPS score, includes variables for pre-existing conditions. Most comorbidities were captured using ICD-9 codes, except immunocompromised status and alcoholism, which were supplemented with a manual review of the notes. Finally, we also collected variables included in the study by Sviri et al19 (inotropic support, sepsis, GI hemorrhage, neuropsychiatric disorders).
With regard to B12 supplementation, four patients were known to have received a B12 supplement prior to B12 measurement during their hospital stay. However, we did not have information about B12 supplementation prior to ICU admission for any patient and therefore it is possible that other patients also received B12 supplements prior to measurement. For this reason, we chose to keep the four patients who received B12 supplements in the ICU prior to B12 measurement in the cohort.
2.1 Statistical analysis
For the descriptive statistics comparing the independent variables across B12 levels in Tables 1 and 2, we used nonparametric versions of standard statistical tests in order to accommodate small sample sizes in some subgroups and non-normal data. We used nonparametric ANOVA (Kruskal-Wallis test26) to compare continuous variables, such as laboratory values, and Fisher’s exact test to compare categorical variables, such as gender and race. Where appropriate, we used the Jonckheere-Terpstra test of trend for continuous predictors27 and the Cochran-Armitage chi-squared test for trend of proportions to investigate whether the predictors had an increasing or decreasing trend with increasing B12 levels28. A p-value <0.05 was considered significant for all hypothesis tests.
Table 1.
Univariate descriptive table of B12, demographics, and liver function.
| Covariate | Deficient B12 ≤148 pmol/l N=37 | Mid-range B12 148–1000 pmol/l N=1347 | High B12 >1000 pmol/l N=300 | Total N=1684 | P-value Test of association | P-value Test of trend |
|---|---|---|---|---|---|---|
| Mortality | ||||||
| 30-day | 3 (8.1) | 197 (14.6) | 98 (32.7) | 298 (17.7) | <0.0001 | <0.0001 |
| 90-day | 4 (10.8) | 268 (19.9) | 121 (40.3) | 393 (23.3) | <0.0001 | <0.0001 |
| B12, pmol/l | 118 (112.9, 135) | 411 (274.1, 586.6) | 1383 (1148.8, 1475.6) | 465.2 (290.7, 797.7) | <0.0001 | <0.0001 |
| Age, years | 0.2367 | - | ||||
| <45 | 3 (8.1) | 131 (9.7) | 35 (11.7) | 169 (10) | 0.6179 | 0.2781 |
| 45–60 | 14 (37.8) | 457 (33.9) | 118 (39.3) | 589 (35) | 0.4515 | 0.1431 |
| 60–80 | 16 (43.2) | 481 (35.7) | 91 (30.3) | 588 (34.9) | 0.3434 | 0.0401 |
| ≥80 | 4 (10.8) | 278 (20.6) | 56 (18.7) | 338 (20.1) | 0.4410 | 0.9090 |
| Age, continuous | 66.8 (58.5, 72.7) | 68.2 (55.9, 78.3) | 64.4 (52.8, 76.7) | 67.7 (55.4, 78.1) | 0.0822 | 0.0587 |
| Female | 14 (37.8) | 525 (39) | 138 (46) | 677 (40.2) | 0.0785 | 0.0303 |
| Race | 0.4232 | - | ||||
| White | 31 (83.8) | 966 (71.7) | 214 (71.3) | 1211 (71.9) | 0.8187 | 0.4279 |
| Black | 2 (5.4) | 123 (9.1) | 34 (11.3) | 159 (9.4) | 0.4827 | 0.1542 |
| Other | 4 (10.8) | 258 (19.2) | 52 (17.3) | 314 (18.6) | 0.5409 | 0.8767 |
| Insurance | 0.0237 | |||||
| Insured | 22 (59.5) | 434 (32.2) | 93 (31) | 549 (32.6) | 0.0762 | 0.0675 |
| Medicaid/Self-pay | 1 (2.7) | 187 (13.9) | 40 (13.3) | 228 (13.5) | 0.1946 | 0.5644 |
| Medicare | 14 (37.8) | 702 (52.1) | 158 (52.7) | 874 (51.9) | 0.6168 | 0.3827 |
| Other | 0 (0) | 24 (1.8) | 9 (3) | 33 (2) | 0.3316 | 0.1067 |
| Liver function | ||||||
| ALT, IU/l | 17 (11, 27) | 23 (15, 38) | 38 (18, 82) | 24 (15, 43) | <0.0001 | <0.0001 |
| AST, IU/l | 26 (18, 43) | 27 (19, 49) | 50 (26, 125.2) | 29 (20, 58.2) | <0.0001 | <0.0001 |
| TB, mg/dl | 0.6 (0.4, 0.8) | 0.5 (0.3, 0.9) | 1.1 (0.5, 3.5) | 0.6 (0.4, 1.1) | <0.0001 | <0.0001 |
| ALP, IU/l | 66 (53, 85) | 82 (64, 110) | 118.5 (79.5, 194.2) | 85 (66, 121.2) | <0.0001 | <0.0001 |
| ALB, g/dl | 3.6 (3.3, 3.8) | 3.4 (2.9, 3.8) | 2.9 (2.3, 3.3) | 3.3 (2.8, 3.8) | <0.0001 | <0.0001 |
| PT, seconds | 14.5 (13.5, 15.5) | 14.4 (13.2, 16.3) | 15.8 (14.1, 19) | 14.6 (13.3, 16.7) | <0.0001 | <0.0001 |
| Liver disease | 0 (0) | 126 (9.4) | 104 (34.7) | 230 (13.7) | <0.0001 | <0.0001 |
| Liver cancer | 0 (0) | 19 (1.4) | 16 (5.3) | 35 (2.1) | 0.0004 | <0.0001 |
Data presented as number (%) or median (inter-quartile range).
Table 2.
Univariate descriptive table of features of the ICU stay, procedures, and comorbidities.
| Covariate | Deficient B12 ≤148 pmol/l N=37 | Mid-range B12 148–1000 pmol/l N=1347 | High B12 >1000 pmol/l N=300 | Total N=1684 | P-value Test of association | P-value Test of trend |
|---|---|---|---|---|---|---|
| SAPS | 14 (9, 17) | 12 (8, 15) | 12 (9, 16) | 12 (8, 16) | 0.1260 | 0.3991 |
| First service unit | <0.0001 | - | ||||
| CCU | 6 (16.2) | 188 (14) | 27 (9) | 221 (13.1) | 0.0929 | 0.0200 |
| CSRU | 18 (48.6) | 362 (26.9) | 21 (7) | 401 (23.8) | <0.0001 | <0.0001 |
| MICU | 13 (35.1) | 698 (51.8) | 225 (75) | 936 (55.6) | 0.0005 | <0.0001 |
| SICU | 0 (0) | 99 (7.3) | 27 (9) | 126 (7.5) | 0.1496 | 0.1057 |
| Admission typea | <0.0001 | - | ||||
| Elective | 9 (24.3) | 147 (10.9) | 6 (2) | 162 (9.6) | <0.0001 | <0.0001 |
| Emergency | 28 (75.7) | 1101 (81.7) | 271 (90.3) | 1400 (83.1) | 0.5316 | 0.0002 |
| Urgent | 0 (0) | 35 (2.6) | 7 (2.3) | 42 (2.5) | 1 | 0.8695 |
| LOS, days | ||||||
| ICU | 2.4 (1.2, 3.7) | 2.2 (1.2, 4.6) | 2.8 (1.5, 7) | 2.3 (1.2, 4.8) | 0.0090 | 0.0025 |
| Hospital | 7 (5, 11) | 9 (5, 16) | 13.5 (6, 24) | 9 (5, 17) | <0.0001 | <0.0001 |
| Lab tests | ||||||
| Creatinine, mg/dl | 0.9 (0.7, 1.1) | 1 (0.7, 1.6) | 1.3 (0.8, 2.4) | 1 (0.7, 1.7) | <0.0001 | <0.0001 |
| Hemoglobin, g/dl | 10.4 (9.3, 11.2) | 10.3 (9, 11.7) | 10 (8.8, 11.2) | 10.2 (9, 11.6) | 0.0365 | 0.0131 |
| Folate, ng/mla | 8.9 (6.3, 12.9) | 13.1 (9.5, 19.4) | 14.8 (10.3, 20) | 13.4 (9.6, 19.9) | 0.0008 | <0.0001 |
| MCV, fLa | 89 (83, 93) | 90 (86, 95) | 93 (88, 100) | 91 (86, 96) | <0.0001 | <0.0001 |
| Procedures | ||||||
| B12 supplement | 8 (21.6) | 51 (3.8) | 15 (5) | 74 (4.4) | 0.0002 | 0.1962 |
| Dialysis | 0 (0) | 124 (9.2) | 50 (16.7) | 174 (10.3) | <0.0001 | <0.0001 |
| TPN | 0 (0) | 27 (2) | 9 (3) | 36 (2.1) | 0.4080 | 0.1746 |
| Inotropic support | 9 (24.3) | 283 (21) | 74 (24.7) | 366 (21.7) | 0.3350 | 0.2692 |
| Comobidities | ||||||
| Alcoholism | 4 (10.8) | 218 (16.2) | 78 (26) | 300 (17.8) | 0.0003 | <0.0001 |
| Sepsis | 4 (10.8) | 190 (14.1) | 82 (27.3) | 276 (16.4) | <0.0001 | <0.0001 |
| Respiratory disease | 2 (5.4) | 179 (13.3) | 39 (13) | 220 (13.1) | 0.4288 | 0.6487 |
| Myeloproliferative disease | 0 (0) | 6 (0.4) | 5 (1.7) | 11 (0.7) | 0.0675 | 0.0179 |
| Stroke | 2 (5.4) | 108 (8) | 24 (8) | 134 (8) | 0.9559 | 0.8178 |
| CAD | 26 (70.3) | 544 (40.4) | 61 (20.3) | 631 (37.5) | <0.0001 | <0.0001 |
| Other CVD | 20 (54.1) | 807 (59.9) | 151 (50.3) | 978 (58.1) | 0.0085 | 0.0104 |
| Hypertension | 26 (70.3) | 716 (53.2) | 120 (40) | 862 (51.2) | <0.0001 | <0.0001 |
| Immunocomprimized | 6 (16.2) | 358 (26.6) | 110 (36.7) | 474 (28.1) | 0.0007 | 0.0001 |
| Neuropsychiatric | 8 (21.6) | 420 (31.2) | 105 (35) | 533 (31.7) | 0.1852 | 0.0855 |
| GI hemorrhage | 0 (0) | 71 (5.3) | 22 (7.3) | 93 (5.5) | 0.1332 | 0.0571 |
Data presented as number (%) or median (inter-quartile range).
Admission type had 80 (4.8%) missing values, folate had 503 (29.9%) missing values, and MCV had 1 (0.1%) missing value.
We investigated the relationship between B12, mortality, and liver disease and function by fitting a multivariable logistic regression model that adjusted B12 measurement and liver disease and liver function, as well as other potential confounders. This model is referred to as the “fully-adjusted” model. We attempted to account for potential confounders, as well as factors adjusted for in the multivariable model of Sviri et al19. In order to create the fully-adjusted model, we first fitted a model with all B12, demographic, and potentially confounding variables, and then we removed from the model all variables with p<0.05 (except B12 and liver function and liver disease variables, as they relate to our primary hypothesis).We performed a global likelihood ratio test to confirm that these variables did not significantly contribute to the model as a group.
In addition, in order to illustrate the attenuation effect on mortality risk due to B12 level, we also showed the results from the intermediate model building steps (from consecutively adding B12, demographic variables, other potential clinical confounders, and finally liver disease and liver function variables). We label these intermediate models, Models 1–3, with “Model 4” referring to our fully-adjusted model. We repeated this analysis for 30-day and 90-day mortality. Collinearity of the independent variables was assessed by using the variance inflation factor (VIF) and generalized variance inflation factor (GVIF)29 for variables with >2 degrees of freedom (df ) (i.e., more than 2 categories); no VIF or GVIF1/(2df) was >2. All data analysis was performed using the R statistical software30.
We conducted two sensitivity analyses. The first, was an attempt to replicate the results of the Sviri19 study. The Sviri study19 was comprised of MICU patients only, and in that study only the 90-day mortality model found a significant relationship between B12 and mortality. For this reason we built an additional model for 90-day mortality based only on MICU patients in order to be able to make more direct comparison to the Sviri study. We applied the same model building algorithm that we used to build the fully-adjusted models (Model 4) to the MICU subset. In the second sensitivity analysis we used different B12 cutoffs for low and high B12, in order to observe if this affected the results. We used 200 pmol/l (the threshold for subclinical deficiency22) for low B12 and 664 pmol/l (an upper threshold used by a previous study19).
Finally, we also performed a univariate analysis to compare the patients in our study to the rest of the patients in the MIMIC-II database (i.e., the patients without B12 values or values for the liver function tests) in order to investigate possible sources of selection bias (see Table 5). Compared to the rest of the adult patients in the MIMIC-II database, the patients in the B12 study tended to have more comorbidities, but typically there was no more than a 5% difference in prevalence between the B12 and non-B12 groups, with the exception of alcoholism (17.8% compared to 3.5%), sepsis (15.1% compared to 9.7%), liver disease (13.7% compared to 6%), CVD conditions not already captured by other variables (58.2% compared to 50.6%), immunocompromised status (28.1% compared to 6.5%) and neuropsychiatric disorders (32.4% compared to 25.4%). There were also a higher proportion of patients in the MICU in our study than in the rest of the MIMIC-II database (55.6% for the B12 study group compared to 35.8% for the non-B12 group) and less SICU patients (7.5% for the B12 group compared to 27.7% for the non-B12 group). Both sets of patients tended to have similar liver function measures and SAPS scores (a median of 12 for the B12 study patients compared to 10 for the rest of the ICU population).
Table 5.
Comparing the ICU patients in the B12 study and ICU patients who did not have a B12 measurement at baseline.
| Variable | B12 study (N=1684) | Not in B12 study (N=22,814) | P-value |
|---|---|---|---|
| 30-day mortality | 298 (17.7) | 3049 (13.4) | <0.0001 |
| 90-day mortality | 393 (23.3) | 3845 (16.9) | <0.0001 |
| Age, years | 67.7 (55.4, 78.1) | 65.1 (51, 77.7) | <0.0001 |
| Female | 677 (40.2) | 9918 (43.5) | 0.0087 |
| Race | <0.0001 | ||
| White | 1211 (71.9) | 15713 (68.9) | 0.0094 |
| Black | 159 (9.4) | 1601 (7) | 0.0003 |
| Other | 314 (18.6) | 5500 (24.1) | <0.0001 |
| Insurance type | <0.0001 | ||
| Insured | 549 (32.6) | 9457 (42.1) | <0.0001 |
| Medicaid/Free/Self-pay | 228 (13.5) | 2583 (11.5) | 0.0131 |
| Medicare | 874 (51.9) | 9833 (43.8) | <0.0001 |
| Other | 33 (2) | 578 (2.6) | 0.1267 |
| Liver function: | |||
| ALT, IU/L | 24 (15, 43) | 25 (16, 46) | 0.0013 |
| AST, IU/L | 29 (20, 58.2) | 30 (20, 63) | 0.1296 |
| TB, mg/dL | 0.6 (0.4, 1.1) | 0.6 (0.4, 1) | 0.9987 |
| ALP, IU/L | 85 (66, 121.2) | 80 (61, 111) | <0.0001 |
| ALB, g/dL | 3.3 (2.8, 3.8) | 3.4 (2.9, 3.9) | <0.0001 |
| PT, seconds | 14.6 (13.3, 16.7) | 13.8 (12.9, 15.5) | <0.0001 |
| Liver disease | 230 (13.7) | 1360 (6) | <0.0001 |
| Liver cancer | 35 (2.1) | 495 (2.2) | 0.8625 |
| SAPS | 12 (8, 16) | 10 (6, 14) | <0.0001 |
| First service Unit | <0.0001 | ||
| CCU | 221 (13.1) | 3659 (16) | 0.0015 |
| CSRU | 401 (23.8) | 4653 (20.4) | 0.0010 |
| MICU | 936 (55.6) | 8175 (35.8) | <0.0001 |
| SICU | 126 (7.5) | 6327 (27.7) | <0.0001 |
| Admission type | <0.0001 | ||
| Elective | 162 (10.1) | 3474 (15.8) | <0.0001 |
| Emergency | 1400 (87.3) | 17581 (79.9) | <0.0001 |
| Urgent | 42 (2.6) | 945 (4.3) | 0.0008 |
| ICU LOS | 2.3 (1.2, 4.8) | 2 (1.1, 4.1) | <0.0001 |
| Hospital LOS | 9 (5, 17) | 7 (4, 12) | <0.0001 |
| Non-liver lab values: | |||
| Creatinine, mg/dL | 1 (0.7, 1.7) | 0.9 (0.7, 1.3) | <0.0001 |
| Hemoglobin, g/dL | 10.2 (9, 11.6) | 11.4 (9.9, 13.1) | <0.0001 |
| Folate, ng/mL | 13.4 (9.6, 19.9) | 15 (10.4, 20) | 0.0053 |
| MCV, ng/mL | 91 (86, 96) | 89 (86, 93) | <0.0001 |
| Procedures: | |||
| B12 supplement | 74 (4.4) | 298 (1.3) | <0.0001 |
| Dialysis | 174 (10.3) | 808 (3.5) | <0.0001 |
| TPN | 36 (2.1) | 467 (2) | 0.7892 |
| Inotropic support | 366 (21.7) | 3557 (15.6) | <0.0001 |
| Comorbidities: | |||
| Alcoholism | 300 (17.8) | 791 (3.5) | <0.0001 |
| Sepsis | 276 (16.4) | 2217 (9.7) | <0.0001 |
| Respiratory disease | 220 (13.1) | 2641 (11.6) | 0.0704 |
| Myeloproliferative disease | 11 (0.7) | 85 (0.4) | 0.1002 |
| Stroke | 134 (8) | 2569 (11.3) | <0.0001 |
| CAD | 631 (37.5) | 7553 (33.1) | 0.0003 |
| Other CVD | 978 (58.1) | 11610 (50.9) | <0.0001 |
| Hypertension | 862 (51.2) | 10767 (47.2) | 0.0016 |
| Immunocompromised | 474 (28.1) | 1474 (6.5) | <0.0001 |
| Neuro-psychiatric disorder | 533 (31.7) | 5847 (25.6) | <0.0001 |
| GI hemmorrhage | 93 (5.5) | 766 (3.4) | <0.0001 |
Data presented as number (%) or median (inter-quartile range).
3. Results
3.1 Univariate analysis
Univariate descriptions of the relationship between B12 levels, the mortality outcomes, demographics, and liver function are given in Table 1. The high B12 group had higher 30-day mortality than other groups and the morality rate increased with increasing B12 (mortality rate for the high B12 group was 32.7% compared to 8.1% and 14.6% for the deficient and mid-range B12 groups, p<0.0001). Similar results were found for 90-day mortality (mortality rate for high B12 group was 40.3% compared to 10.8% and 19.9% for the deficient and mid-range groups, p<0.0001). The differences in age across the B12 groups were statistically significant, but not clinically important (median age 64.4 years for high B12 compared to 66.8 and 68.2 years for the deficient and mid-range groups, p<0.0001). The B12 groups had a similar proportion of females (46.0% compared to 37.8% and 39% for deficient and mid-range groups, p=0.0785). All the liver function measures and rates of liver disease and cancer generally increased with increasing B12 levels with the exception (as might be expected) of albumin which decreased with increasing B12 levels (p<0.0001 for all). These results are in agreement with Dou et al12, Lin et al11, Ermens et al9 and others who found that poor liver function is associated with elevated B12 levels.
The univariate examination of the relationships between B12 and the remaining potentially confounding factors is given in Table 2. SAPS were similar for all B12 groups (for example, median SAPS was 14 for the deficient B12 group and 12 for both the mid-range and high B12 groups, p=0.1260). However, in general, high B12 levels appeared to be associated with conditions and factors usually associated with poorer health compared to deficient and mid-range B12 groups. The high B12 group had proportionately more patients in the MICU (75.0% versus 35.1% and 51.8% for the deficient and midrange groups, respectively, p<0.0001) and less patients in the CCU and CSRU than other B12 groups (9.0% in CCU compared to 16.2% and 14.0% in the deficient and mid-range groups, p=0.0200, and 7.0% in the CSRU compared to 48.6% and 26.9% in the deficient and mid-range B12 groups). The proportion of elective admissions tended to be lower in the high B12 group (2% for high B12 compared to 24.3% and 10.9% for deficient and mid-range groups, p<0.0001), and the proportion of emergency admissions increased with increasing B12 (90.3% for high B12 compared to 75.7% and 81.7% for deficient and midrange, p=0.0002). Higher B12 levels were associated with longer lengths of stay (ICU LOS: median 2.8 days for high B12 compared to 2.4 and 2.2 for deficient and mid-range groups, p=0.0025; Hospital LOS: median 13.5 days for high B12 compared to 7 and 9 days for deficient and mid-range groups, p<0.0001).
There were statistically significant, but not necessarily clinically important, trends for all non-liver lab values across the B12 groups (higher B12 levels were associated with higher creatinine, folate, and MCV levels and lower hemoglobin levels). As might be expected, the deficient B12 group had the highest rate of B12 supplementation compared to the other B12 levels (21.6% compared to 3.8% and 5% for mid-range and high B12 groups, p=0.0002). With regard to other procedures, only rates of dialysis tended to increase as B12 increased (16.7% for high B12 compared to 9.2% and 0% for mid-range and deficient B12, p<0.0001). Finally, the high B12 group tended to have higher rates of alcoholism (p<0.0001), sepsis (p<0.0001), myeloproliferative disease (p<0.0001), and immunocompromised status (p=0.0001) and lower rates of CAD (p<0.0001), Other CVD (p=0.0104), and hypertension (p<0.0001).
3.2 Multivariable models
A summary of Models 1 to 4 are given in Table 3. Model 1 gives the unadjusted estimate of the mortality risk for high and deficient B12 levels compared to mid-range levels, repeated for mortality at 30 and 90 days after ICU admission. Deficient B12 was not significantly different to the mid-range B12 group for any of the models. For the high B12 group, the unadjusted risk was approximately 2.7–2.8 times that of mid-range B12 group (for 30-day: OR 2.83, 95% CI 2.13 to 3.76, p<0.0001; for 90-day: OR 2.72, 95% CI 2.08 to 3.55, p<0.0001). Model 2 represents the risk of B12 adjusted for demographic variables (age, gender, race, and insurance type). The mortality risks for the high B12 group in Model 2 were similar to the unadjusted risks (for 30-day: OR 3.12, 95% CI 2.31 to 4.20, p<0.0001; for 90-day: OR=3.09, 95% CI 2.33 to 4.09, p<0.0001). When we added other potential non-liver related clinical confounders to Model 2 to create Model 3, the mortality risk for the high B12 patients compared to midrange patients dropped to an approximately 100% increased risk but this difference was still significant compared to mid-range B12 patients (for 30-day: OR 1.93, 95% CI 1.38 to 2.72, p=0.0001; for 90-day: OR=1.97, 95% CI 1.43 to 2.71, p<0.0001).
Table 3.
Results from the four models for mortality 30 and 90 days after ICU admission.
| Models: 30 days | Description | Deficient B12 OR (95% CI) | P-value | High B12 OR (95% CI) | P-value |
|---|---|---|---|---|---|
| Model 1 | B12 only | 0.52 (0.16, 1.69) | 0.2743 | 2.83 (2.13, 3.76) | <0.0001 |
| Model 2 | +Demographics | 0.65 (0.19, 2.16) | 0.4776 | 3.12 (2.31, 4.20) | <0.0001 |
| Model 3 | +Other potential confounders | 0.89 (0.22, 3.57) | 0.8651 | 1.93 (1.38, 2.72) | 0.0001 |
| Model 4 | +Liver function and diseases | 1.32 (0.32, 5.52) | 0.7033 | 1.18 (0.81, 1.72) | 0.3890 |
|
| |||||
| Models: 90 days | |||||
|
| |||||
| Model 1 | B12 only | 0.49 (0.17, 1.39) | 0.1790 | 2.72 (2.08, 3.55) | <0.0001 |
| Model 2 | +Demographics | 0.58 (0.20, 1.67) | 0.3098 | 3.09 (2.33, 4.09) | <0.0001 |
| Model 3 | +Other potential confounders | 0.99 (0.30, 3.30) | 0.9919 | 1.97 (1.43, 2.71) | <0.0001 |
| Model 4 | +Liver function and diseases | 1.30 (0.37, 4.53) | 0.6774 | 1.20 (0.84, 1.71) | 0.3077 |
Table 3 shows the model building procedure. In order to demonstrate the attenuation of the effect of B12 on mortality, we added variables to the B12 predictor sequentially to create the final, fully-adjusted model (“model 4”).Model 1 includes B12 only; model 2 includes B12 and demographic variables; model 3 includes B12, demographic, and other potentially confounding factors; and model 4 includes B12, demographic variables, other potential confounders, and liver function and disease (see Table 4 for details). The odds ratio compares the risk of the deficient or high B12 group to the mid-range B12 group.
However, after we adjusted for liver disease and liver function (Model 4) in addition to all the other factors, the mortality risk for the high B12 group decreased to an approximately 20% increased risk compared to mid-range patients, and the difference between high B12 and mid-range B12 was no longer significant (for 30-day: OR 1.18, 95% CI 0.81 to 1.72, p=0.3890; for 90-day: OR=1.20, 95% CI 0.84 to 1.71, p=0.3077). The details of the fully-adjusted model (Model 4) are given in Table 4. Significant predictors for the 30-day mortality model were age (p<0.0001), insurance (p=0.0128), SAPS (p<0.0001), first service unit (p<0.0001), stroke (p<0.0001), immunocompromised status (p<0.0001), alcoholism (p=0.0244), inotropic support (p=0.0331), TB (p=0.0335), ALP (0.0156), and ALB (p<0.0001). As mentioned previously, high B12 was not a significant predictor in the model (p=0.3886) and neither was B12 overall (p=0.6626).
Table 4.
Results from final model for mortality for 30 and 90 days after ICU admission.
| Covariate | 30-day mortality OR (95% CI) | P-value | 90-day mortality OR (95% CI) | P-value |
|---|---|---|---|---|
| B12, pmol/l (Ref=Mid- range) | 0.6626 | 0.5719 | ||
| Deficient B12 | 1.28 (0.31, 5.32) | 0.7363 | 1.27 (0.37, 4.41) | 0.7060 |
| High B12 | 1.18 (0.81, 1.72) | 0.3886 | 1.20 (0.84, 1.70) | 0.3148 |
| Age, years (Ref=45–65) | <0.0001 | <0.0001 | ||
| <45 | 0.46 (0.24, 0.90) | 0.0224 | 0.48 (0.26, 0.88) | 0.0181 |
| 65–80 | 1.97 (1.22, 3.15) | 0.0051 | 2.25 (1.47, 3.44) | 0.0002 |
| ≥80 | 6.44 (3.82, 10.85) | <0.0001 | 7.22 (4.48, 11.63) | <0.0001 |
| Gender (Ref=Female) | 0.94 (0.68, 1.30) | 0.7200 | 1.18 (0.88, 1.58) | 0.2725 |
| Race (Ref=White) | 0.4647 | 0.4164 | ||
| Black | 0.71 (0.41, 1.25) | 0.2363 | 0.73 (0.44, 1.20) | 0.2141 |
| Other | 1.02 (0.67, 1.54) | 0.9307 | 0.89 (0.60, 1.31) | 0.5490 |
| Insurance type (Ref=Insured) | 0.0128 | 0.2368 | ||
| Medicaid/Self-pay | 1.67 (0.94, 2.96) | 0.0819 | 1.25 (0.74, 2.12) | 0.4017 |
| Medicare | 1.42 (0.94, 2.16) | 0.0982 | 1.17 (0.81, 1.70) | 0.4081 |
| Other | 4.66 (1.72, 12.61) | 0.0025 | 2.65 (1.01, 6.95) | 0.0469 |
| Dialysis | - | - | 1.82 (1.19, 2.80) | 0.0062 |
| SAPS | 1.09 (1.06, 1.13) | <0.0001 | 1.08 (1.05, 1.11) | <0.0001 |
| First service unit (Ref=CCU) | <0.0001 | <0.0001 | ||
| CSRU | 0.32 (0.15, 0.68) | 0.0031 | 0.37 (0.19, 0.72) | 0.0032 |
| MICU | 1.14 (0.67, 1.94) | 0.6254 | 1.05 (0.66, 1.67) | 0.8416 |
| SICU | 0.19 (0.08, 0.47) | 0.0004 | 0.38 (0.18, 0.77) | 0.0072 |
| B12 supplement | 0.49 (0.21, 1.13) | 0.0941 | - | - |
| Sepsis | 1.39 (0.96, 2.03) | 0.0839 | 1.46 (1.03, 2.07) | 0.0330 |
| TPN | - | - | 2.11 (0.92, 4.85) | 0.0780 |
| Respiratory disease | - | - | 1.72 (1.17, 2.52) | 0.0060 |
| Stroke | 3.77 (2.25, 6.32) | <0.0001 | 3.21 (1.98, 5.21) | <0.0001 |
| CAD | 0.77 (0.51, 1.15) | 0.2015 | 0.58 (0.40, 0.84) | 0.0037 |
| Hypertension | - | - | 0.79 (0.58, 1.07) | 0.1310 |
| Other CVD | - | - | 1.42 (1.02, 1.98) | 0.0354 |
| Immunocompromised | 2.30 (1.62, 3.26) | <0.0001 | 2.43 (1.76, 3.36) | <0.0001 |
| Alcoholism | 1.66 (1.07, 2.59) | 0.0244 | - | - |
| MCV, fL (Ref≤86) | 0.0529 | 0.0455 | ||
| Q2 (86, 91] | 0.91 (0.57, 1.45) | 0.6832 | 0.82 (0.54, 1.24) | 0.3439 |
| Q3 (91, 96] | 1.05 (0.64, 1.70) | 0.8583 | 0.93 (0.60, 1.44) | 0.7403 |
| Q4 (96, 135] | 1.57 (1.00, 2.47) | 0.0487 | 1.39 (0.93, 2.09) | 0.1063 |
| Inotropic support | 1.53 (1.03, 2.27) | 0.0331 | - | - |
| Neuropsychiatric disorder | 0.70 (0.49, 1.01) | 0.0561 | - | - |
| ALT, IU/l (Ref≤15) | 0.3293 | 0.6625 | ||
| Q2 (15, 24] | 1.21 (0.74, 1.99) | 0.4535 | 1.18 (0.76, 1.83) | 0.4598 |
| Q3 (24, 43] | 1.62 (0.96, 2.73) | 0.0683 | 1.35 (0.84, 2.16) | 0.2179 |
| Q4 (43, 4170] | 1.40 (0.76, 2.58) | 0.2828 | 1.19 (0.68, 2.08) | 0.5407 |
| AST, IU/l (Ref≤20) | 0.8722 | 0.7431 | ||
| Q2 (20, 29] | 1.20 (0.70, 2.05) | 0.5086 | 1.13 (0.71, 1.80) | 0.6009 |
| Q3 (29, 58.2] | 1.03 (0.59, 1.79) | 0.9273 | 0.90 (0.55, 1.48) | 0.6861 |
| Q4 (58.2, 4253] | 1.14 (0.58, 2.24) | 0.6997 | 1.04 (0.57, 1.91) | 0.8996 |
| TB, mg/dl (Ref≤0.4) | 0.0335 | 0.0165 | ||
| Q2 (0.4, 0.6] | 1.49 (0.93, 2.38) | 0.0966 | 1.32 (0.86, 2.02) | 0.2075 |
| Q3 (0.6, 1.1] | 0.99 (0.61, 1.60) | 0.9629 | 1.33 (0.87, 2.03) | 0.1888 |
| Q4 (1.1, 54.2] | 1.81 (1.12, 2.94) | 0.0160 | 2.06 (1.32, 3.21) | 0.0014 |
| ALP, IU/L (Ref≤66) | 0.0156 | 0.0430 | ||
| Q2 (66, 85] | 1.29 (0.77, 2.16) | 0.3367 | 1.48 (0.94, 2.34) | 0.0896 |
| Q3 (85, 121] | 1.85 (1.13, 3.03) | 0.0140 | 1.66 (1.07, 2.58) | 0.0248 |
| Q4 (121, 4585] | 2.02 (1.23, 3.29) | 0.0052 | 1.85 (1.19, 2.89) | 0.0065 |
| ALB, g/dl (Ref≤2.8) | <0.0001 | <0.0001 | ||
| Q2 (2.8, 3.3] | 0.92 (0.63, 1.32) | 0.6427 | 0.81 (0.57, 1.13) | 0.2156 |
| Q3 (3.3, 3.8] | 0.45 (0.28, 0.71) | 0.0007 | 0.38 (0.25, 0.58) | <0.0001 |
| Q4 (3.8, 5.6] | 0.28 (0.14, 0.53) | 0.0001 | 0.24 (0.14, 0.43) | <0.0001 |
| PT, second (Ref≤13.3) | 0.4065 | 0.8901 | ||
| Q2 (13.3, 14.6] | 1.09 (0.68, 1.74) | 0.7338 | 1.02 (0.67, 1.56) | 0.9164 |
| Q3 (14.6, 16.7] | 0.95 (0.58, 1.58) | 0.8555 | 1.00 (0.64, 1.56) | 0.9989 |
| Q4 (16.7, 150] | 1.34 (0.84, 2.13) | 0.2221 | 1.15 (0.75, 1.76) | 0.5305 |
| Liver disease | 1.61 (0.99, 2.61) | 0.0531 | 1.71 (1.09, 2.67) | 0.0197 |
| Liver cancer | 2.31 (0.97, 5.46) | 0.0576 | 3.54 (1.52, 8.26) | 0.0035 |
For 90-day mortality we saw broadly similar results, with some difference in the selection of the confounding variables. Again, high B12, deficient B12, and B12 as a whole were not significant (p=0.3148, p=0.6774, and p=0.5719, respectively). Significant predictors in the 90-day mortality model were: age (p<0.0001), dialysis (p=0.0062), first service unit (p<0.0001), sepsis (p=0.0330), respiratory disease (p=0.0060), stroke (p<0.0001), CAD (p=0.0037), CVD (p=0.0354), immunocompromised status (p<0.0001), MCV (p=0.0455), TB (p=0.0165), ALP (p=0.0430), ALB (p<0.0001), liver disease (p=0.0197), and liver cancer (p=0.0035). In general, the 90-day model tended to select chronic conditions such as dialysis or heart disease, and tended to exclude acute predictors that might be expected to have more of a short-term effect, such as inotropic support.
3.2 Sensitivity analyses
We also built a corresponding Model 4 for 90-day mortality based only on the MICU patients (N=936), and again we found that high B12 was not a significant predictor (p=0.8063) of mortality (results not shown). Note that the 90-day, fully-adjusted mortality model based on the MICU patients is the model that most closely mirrors the Sviri et al19 multivariable model that found that high B12 was a significant predictor of mortality, except that our model is adjusted for liver function. In general, most of the significant predictors in the 90-day MICU model were also significant in the corresponding 90-day model based on the whole ICU sample (age, SAPS, stroke, Other CVD, immunocompromised status, MCV, TB, ALB, and liver cancer). There were some differences; the MICU model dropped dialysis, sepsis, TPN, respiratory disease, CAD, and hypertension. Furthermore, gender was significant in the MICU 90-day model but not in the general 90-day, fully adjusted model; conversely, ALP and liver disease were not significant in the MICU 90-day model but were significant in the general 90-day, fully-adjusted model.
We also tesed the fully-adjusted 30-day and 90-day models in the paper with different cutoff levels for B12 of 200 pmol/l for low B12 and 664 pmol/l for high B12, as described in the methods section. We found similar results, and in fact the mortality risk for high B12 levels was actually smaller with these thresholds than with fully-adjusted model presented in the paper. The risk for high B12 compared to mid-range B12 was not significant for 30-day mortality (OR=1.07; 95%CI 0.76–1.52; p=0.68) or for 90-day mortality (OR=1.11; 95%CI 0.81–1.52; p=0.70).
4. Discussion
We showed that after adjusting for liver function and disease high B12 levels were not associated with higher mortality for ICU patients. It is possible that poor liver function or liver disease cause both elevated B12 levels and negative clinical outcomes, and high B12 levels serve as a proxy for poor liver function. While it is always difficult to definitively exclude an association, to our knowledge, our study is the largest study so far to look at the relationship between elevated B12 and mortality for the ICU population. Also, as far as we are aware, our study is the only one to consider a non-linear association between B12 levels and mortality, contain a comprehensive set of liver function measurements, and include patients from the SICU and CSRU.
It is important to consider liver function when studying B12 as the liver is directly involved in B12 metabolism. The liver typically stores several years’ supply of B12 and also synthesizes transcobalamin II (TCII), which is a B12 transport protein. After dietary B12 is absorbed from the terminal ileum into the enterohepatic circulation, TCII promotes uptake of B12 into the liver. TCII also carries B12 in the serum and binds to receptors on non-hepatocyte cell surfaces to facilitate B12 transport into cells9. Although TCII carries some B12 and is responsible for promoting cellular B12 uptake, the majority of serum B12 is bound to transcobalamin I/III (also called haptocorrin, or HC), which is synthesized in salivary gland, gastric, and myeloid cells (precursors to red blood cells, platelets, and non-lymphocyte white blood cells). HC does not play a role in cellular B12 uptake; however, B12 bound to HC (HC-B12) is taken up by the liver and is one mechanism by which circulating B12 is cleared9.
Liver damage can lead to high vitamin B12 levels by several mechanisms: hepatocyte lysis (break- down of the hepatic cells releases stored B12), decreased liver production of TCII (which decreases uptake by tissues and increases the amount of circulating B12), and decreased liver uptake of HC-B12 (which also increases the amount of circulating B12)9. Therefore it is plausible that poor liver function is the underlying cause of both higher mortality rates and elevated B12 levels. High B12 levels can also be the result of increased HC production in certain hematologic conditions due to the increased number of myeloid cells (which increases serum HC-B12 levels) and increased ingestion or therapeutic B12 administration9.
Furthermore, there are several issues with using elevated B12 levels as a biomarker for mortality for ICU patients. First, in contrast to liver function, vitamin B12 is not routinely measured in the ICU in the United States. Second, again unlike liver function, there is no known mechanism to link elevated B12 with mortality. Third, there are already a number of scores that have been developed and validated to measure acuity of illness in the ICU, such as SAPS24, APACHE-II25, and SOFA17, so B12 measurements are not necessary for this purpose. In contrast to elevated B12, B12 deficiency is a known problem for elderly populations21—23.
The primary limitation of our study is that it is a retrospective, observational study and therefore, as with all observational studies, we were unable to rule out bias due to the non-randomization of unknown confounders or selection bias due to only being able to study patients who had B12 levels taken. However, we attempted to assess this possible source of bias by comparing the 1684 patients included in our study with the 22,814 patients in the MIMIC-II database but not included in this study, and while we found some clinically significant differences (our study group had a higher incidence of alcoholism, immuno-compromised status patients, and patients admitted to the MICU), none seemed to suggest a reason as to why B12 and not liver function was responsible for higher mortality. A prospective study on the association between high B12 and mortality where B12 values are measured on all participants would be valuable in ruling out possible selection bias, but to date no such study has been conducted. We were also not able to collect information on B12 therapy prior to admission to the hospital. It is possible that this may create some bias as some of the patients on B12 therapy may have appeared in the “high” B12 group that otherwise would have appeared in the “deficient” category, for example. However, in general, it is unlikely that a person could have achieved a B12 level over 1000 pmol/l (our cutoff for high B12) due to B12 therapy alone.
5. Conclusions
After accounting for liver function and liver disease, elevated B12 levels were not a significant predictor of mortality for ICU patients, for either 30 or 90 days after admission. B12 is known to be elevated in patients with liver disease and it is likely that high B12 levels are often a proxy for poor liver function. Moreover, other composite measures of acuity for the ICU population are already in use have good predictive power for mortality. We conclude that elevated B12 is not a good predictor of mortality for ICU patients when liver function is known.
Acknowledgments
The authors would like to thank Roger Mark, MD, PhD and the MIMIC team (MIT Laboratory for Computational Physiology) for collecting and making the MIMIC data publicly available and in providing additional information for our analysis, and Samantha Fletcher and Marnica Johnson for manually annotating some of the clinical notes. This research was supported by the Intramural Research Program of the National Institutes of Health (NIH), National Library of Medicine (NLM), under NIH IRB exemption.
Funding source
Support for this study was provided by the Intramural Research Program of the National Institutes of Health (NIH), National Library of Medicine (NLM), under an NIH Institutional Review Board exemption.
Abbreviations
- ALB
albumin
- ALP
alkaline phosphatase
- ALT
alanine aminotransferase
- AST
aspartate aminotransferase
- CCU
coronary care unit
- CSRU
cardiac surgery recovery unit
- GVIF
generalized variance inflation factor
- HC
haptocorrin
- ICD-9
International Classification of Disease codes, ninth revision
- IQR
inter-quartile range
- MCV
mean corpuscular volume
- MELD
Model for End-stage Liver Disease
- MICU
medical intensive care unit
- MIMIC-II
Multiparameter Intelligent Monitoring in Intensive Care II
- MMA
methylmalonic acid
- PT
prothrombin time
- SICU
surgical intensive care unit
- TB
total bilirubin
- TCII
transcobalamin II
- tHcy
total homocysteine
- TPN
total parenteral nutrition
- VIF
variance inflation factor
Footnotes
Statement of authorship
FC drafted the manuscript, contributed to the study design, collected the data, analyzed and interpreted the data, and provided critical revisions to the manuscript. KL conceived of the study, contributed to the study design, provided data interpretation, and contributed to critical revisions of the manuscript. SA and DD contributed to the study design, data acquisition, data interpretation, and critical revisions of the manuscript. CJM contributed to the in study conception and design, data interpretation, critical revisions of the manuscript, and also supervised the study. All authors read and approved the final manuscript.
Conflict of interest
The authors have no conflicts of interest to disclose.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Contributor Information
Fiona M Callaghan, Email: Fiona.Callaghan@nih.gov, (Co-first author) Lister Hill National Center for Biomedical Communications, National Library of Medicine, National Institutes of Health, 8600 Rockville Pike, Building 38A/9N911, Bethesda, MD 20894, USA, 301-435-3199 (phone) 301 496 0673 (fax).
Kira Leishear, Email: Kira.Leishear@nih.gov, (Co-first author) Division of Epidemiology, Statistics, and Prevention Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, 6100 Executive Boulevard, Rockville, MD 20892, USA.
Swapna Abhyankar, Email: Swapna.Abhyankar@nih.gov, Lister Hill National Center for Biomedical Communications, National Library of Medicine, National Institutes of Health, 8600 Rockville Pike, Bethesda, MD 20894, USA.
Dina Demner-Fushman, Email: Dina.Demner-Fushman@nih.gov, Lister Hill National Center for Biomedical Communications, National Library of Medicine, National Institutes of Health, 8600 Rockville Pike, Bethesda, MD 20894, USA.
Clement J. McDonald, Email: Clement.McDonald@nih.gov, Lister Hill National Center for Biomedical Communications, National Library of Medicine, National Institutes of Health, 8600 Rockville Pike, Bethesda, MD 20894, USA.
References
- 1.Tal S, Guller V, Shavit Y, Stern F, Malnick S. Mortality predictors in hospitalized elderly patients. J Am Ger Soc. 2011;104:933–938. doi: 10.1093/qjmed/hcr093. [DOI] [PubMed] [Google Scholar]
- 2.Salles N, Herrmann F, Sieber C, Rapin C. High vitamin B12 level and mortality in elderly inpatients. J Nutr Health Aging. 2008;12:219–221. doi: 10.1007/BF02982624. [DOI] [PubMed] [Google Scholar]
- 3.Shahar A, Feiglin L, Shahar DR, Levy S, Seligsohn U. High prevalence and impact of subnormal serum vitamin B12 levels in Israeli elders admitted to a geriatric hospital. J Nutr Health Aging. 2001;5:124–127. [PubMed] [Google Scholar]
- 4.Hemmersbach-Miller M, Conde-Martel A, Betancor-Leon P. Vitamin B-12 as a predictor of mortality in elderly patients. J Am Geriatr Soc. 2005;53:2035–2036. doi: 10.1111/j.1532-5415.2005.00479_2.x. [DOI] [PubMed] [Google Scholar]
- 5.Baztan JJ, Gavidia JJ, Gomez-Pavon J, Esteve A, Ruiperez I. High vitamin B12 levels and in-hospital mortality. J Am Geriatr Soc. 2010;58:2237–2238. doi: 10.1111/j.1532-5415.2010.03116.x. [DOI] [PubMed] [Google Scholar]
- 6.Zeitlin A, Frishman WH, Chang CJ. The association of vitamin B-12 and folate blood levels with mortality and cardiovascular morbidity incidence in the old. The Bronx Aging Study. Am J Ther. 1997;4:275–281. doi: 10.1097/00045391-199707000-00008. [DOI] [PubMed] [Google Scholar]
- 7.Gonzalez S, Huerta JM, Fernandez S, Patterson AM, Lasheras C. Homocysteine increases the risk of mortality in elderly individuals. Br J Nutr. 2007;97:1138–1143. doi: 10.1017/S0007114507691958. [DOI] [PubMed] [Google Scholar]
- 8.Dangour AD, Breeze E, Clarke R, Shetty PS, Uauy R, Fletcher AE. Plasma homocysteine, but not folate or vitamin B-12, predicts mortality in older people in the United Kingdom. J Nutr. 2008;138:1121–1128. doi: 10.1093/jn/138.6.1121. [DOI] [PubMed] [Google Scholar]
- 9.Ermens AAM, Vlasveld LT, Lindemans J. Significance of elevated cobalamin (vitamin B12) levels in blood. Clin Biochem. 2003;36:585–590. doi: 10.1016/j.clinbiochem.2003.08.004. [DOI] [PubMed] [Google Scholar]
- 10.Arendt JFB, Nexo E. Cobalamin related parameters and disease patterns in patients with increased serum cobalamin levels. PLoS ONE. 2012 doi: 10.1371/journal.pone.0045979. published online September 21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Lin CY, Kuo CS, Lu CL, Wu MY, Huang RFS. Elevated serum vitamin B12 levels in association with tumor markers as the prognostic factors predictive for poor survival in patients with hepatocellular carcinoma. Nutr Cancer. 2010;62:190–197. doi: 10.1080/01635580903305334. [DOI] [PubMed] [Google Scholar]
- 12.Dou JF, Xu WY, Ye B, Zhang Y, Mao WL. Serum vitamin B12 levels as indicators of disease severity and mortality of patients with acute-on-chronic liver failure. Clin Chim Acta. 2012;413:1809–1812. doi: 10.1016/j.cca.2012.07.008. [DOI] [PubMed] [Google Scholar]
- 13.Kamath PS, Wiesner RH, Malinchoc M, Kremers W, Therneau TM, Kosberg CL, D’Amico G, Dickson ER, Kim WR. A model to predict survival in patients with end-stage liver disease. Hepatol. 2001;33:464–470. doi: 10.1053/jhep.2001.22172. [DOI] [PubMed] [Google Scholar]
- 14.Olivieri F, Spazzafumo L, Antonicelli R, Marchegiani F, Cardelli M, Sirolla C, Galeazzi R, Giovagnetti S, Mocchegiani E, Franceschi C. Combination of biomarkers to predict mortality in elderly patients with myocardial infarction. Mech Ageing Dev. 2008;129:231–237. doi: 10.1016/j.mad.2008.01.002. [DOI] [PubMed] [Google Scholar]
- 15.Corcoran TB, O’Neill MA, Webb SAR, Ho KM. Prevalence of vitamin deficiencies on admission: relationship to hospital mortality in critically ill patients. Anaesth Intensive Care. 2009;37:254–260. doi: 10.1177/0310057X0903700215. [DOI] [PubMed] [Google Scholar]
- 16.Corcoran TB, O’Neill MA, Webb SAR, Ho KM. Inflammation, vitamin deficiencies and organ failure in critically ill patients. Anaesth Intensive Care. 2009;37:740–747. doi: 10.1177/0310057X0903700510. [DOI] [PubMed] [Google Scholar]
- 17.Vincent J, Moreno R, Takala J, Willatts S, DeMendonca A, Bruining H, Reinhart CK, Suter PM, Thijs LG. The SOFA (sepsis-related organ failure assessment) score to describe organ dysfunction/failure. Intensive Care Med. 1996;22:707–710. doi: 10.1007/BF01709751. [DOI] [PubMed] [Google Scholar]
- 18.Manzanares W, Hardy G. Vitamin B12: the forgotten micronutrient for critical care. Curr Opin Clin Nutr Metab Care. 2010;13:662–668. doi: 10.1097/MCO.0b013e32833dfaec. [DOI] [PubMed] [Google Scholar]
- 19.Sviri S, Khalaila R, Daher S, Bayya A, Linton DM, Stav I, van Heerden PV. Increased vitamin B12 levels are associated with mortality in critically ill medical patients. Clin Nutr. 2012;31:53–59. doi: 10.1016/j.clnu.2011.08.010. [DOI] [PubMed] [Google Scholar]
- 20.Saeed M, Villarroel M, Reisner AT, Clifford G, Lehman L-W, Moody G, Heldt T, Kyaw TH, Moody B, Mark RG. Multiparameter intelligent monitoring in intensive care II: a public-access intensive care unit database. Crit Care. 2011;39:952–960. doi: 10.1097/CCM.0b013e31820a92c6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Natural History and Prevalence of Vitamin B12 Deficiency. Atlanta, GA: Centers for Disease Control and Prevention; 2009. [Accessed July 1, 2013]. at http://www.cdc.gov/ncbddd/b12/history.html. [Google Scholar]
- 22.Carmel R. Subclinical cobalamin deficiency. Curr Opin Gastroenterol. 2012;28:151–158. doi: 10.1097/MOG.0b013e3283505852. [DOI] [PubMed] [Google Scholar]
- 23.Leishear K, Boudreau R, Studenski S, Ferrucci L, Rosano C, Harris TB, Newman AB, Strotmeyer ES. Vitamin B12 and central and peripheral neurological function: is there a threshold level? Gerontologist. 2012;52:508. [Google Scholar]
- 24.Le Gall JR, Loirat P, Alperovitch A, Glaser P, Granthil C, Mathieu D, Mercier P, Thomas R, Villers D. A simplified acute physiology score for ICU patients. Crit Care Med. 1984;12:975–977. doi: 10.1097/00003246-198411000-00012. [DOI] [PubMed] [Google Scholar]
- 25.Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: A severity of disease classification system. Crit Care. 1985;13:818–829. [PubMed] [Google Scholar]
- 26.Kruskal WH, Wallis WA. Use of ranks in one-criterion variance analysis. JASA. 1952;47:583–621. [Google Scholar]
- 27.Jonckheere AR. A distribution-free k-sample test against ordered alternatives. Biometrika. 1954;41:133–145. [Google Scholar]
- 28.Cochran WG. Some methods for strengthening the common tests. Biometrics. 1954;10:417–451. [Google Scholar]
- 29.Fox J, Monette G. Generalized collinearity diagnostics. JASA. 1992;87:178–183. [Google Scholar]
- 30.R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2012. Available at: http://www.R-project.org/ [Google Scholar]