Combined indicator of vitamin B12 status: modification for missing biomarkers and folate status and...

Clin Chem Lab Med 2015; aop

*Corresponding author: Sergey N. Fedosov, Department of

Clinical Chemistry, Aarhus University, Science Park, Gustav

Wieds Vej 10C, 8000, Aarhus C, Denmark, E-mail: [email protected] ;

[email protected]

Alex Brito and Lindsay H. Allen: USDA, ARS, Western Human

Nutrition Research Center, University of California, Davis, CA, USA

Joshua W. Miller: Department of Nutritional Sciences, Rutgers

University, New Brunswick, NJ, USA ; and Department of Pathology

and Laboratory Medicine, University of California, Davis, CA, USA

Ralph Green: Department of Pathology and Laboratory Medicine,

University of California, Davis, CA, USA

Sergey N. Fedosov * , Alex Brito , Joshua W. Miller , Ralph Green and Lindsay H. Allen

Combined indicator of vitamin B 12 status: modification for missing biomarkers and folate status and recommendations for revised cut-points DOI 10.1515/cclm-2014-0818

Received August 14 , 2014 ; accepted December 7 , 2014

Abstract

Background: A novel approach to determine vitamin B 12

status is to combine four blood markers: total B 12

(B 12

),

holotranscobalamin (holoTC), methylmalonic acid (MMA)

and total homocysteine (tHcy). This combined indica-

tor of B 12

status is expressed as cB 12

= log 10

[(holoTC · B 12

)/

(MMA · Hcy)] – (age factor). Here we calculate cB 12

in data-

sets with missing biomarkers, examine the influence of

folate status, and revise diagnostic cut-points.

Methods: We used a database with all four markers

(n = 5211) plus folate measurements (n = 972). A biomarker Z

(assumed missing) was plotted versus X (a combination of

other markers) and Y (age). Each chart was approximated

by a function Z theor

, which predicted the potentially absent

value(s). Statistical distributions of cB 12

were aligned with

physiological indicators of deficiency and used to deter-

mine cut-offs.

Results: The predictive functions Z theor

allowed assess-

ment of the “ incomplete ” indicators, 3cB 12

(three markers

known) and 2cB 12

(two markers known). Predictions con-

tained a systematic deviation associated with disper-

sion along two axes Z and X (and unaccounted by the

least squares fit). Increase in tHcy at low serum folate

was corrected (cB 12

+ Δ folate

) based on the function of

Δ folate

= log 10

(Hcy real

/Hcy theor

) versus folate. Statistical distri-

butions of cB 12

revealed the boundaries of groups with B 12

deficiency, i.e., cB 12

< – 0.5.

Conclusions: We provide equations that combine two,

three or four biomarkers into one diagnostic indicator,

thereby rescaling unmatched data into the same coordi-

nate system. Adjustment of this indicator is required if

serum folate is < 10 nmol/L and tHcy is measured. Revised

cut-points and guidelines for using this approach are

provided.

Keywords: cobalamin; deficiency; diagnostics; folate;

markers; vitamin B12

.

Introduction Vitamin B

12 (cobalamin, Cbl) and folate are essential

micronutrients for humans [1, 2] . Dietary B 12

mainly comes

from animal source foods, whereas folate is primarily of

plant origin. In many countries folic acid is added to food

staples for neural tube defect prevention. The absorption

of dietary B 12

may be insufficient among older adults due

to reduced production of gastric intrinsic factor and low

gastric acid secretion [3] . Deficiency in either of the two

B-vitamins is associated with hematological disorders and

neurological alterations of varying severity [4] .

Vitamin B 12

status is assessed predominantly by total

serum B 12

, due to the simplicity of the method and its low

cost [5] . However the sensitivity of this method in early

stages of deficiency is questionable [5 – 8] , and the speci-

ficity of low serum B 12

for diagnosing diminished B 12

status

is also poor. It is also possible to measure the substrates

of Cbl-dependent reactions, namely methylmalonic acid

(MMA) and total homocysteine (tHcy), whose levels in

serum are inversely associated with total B 12

concentra-

tions [6 – 8] . The interpretation of metabolite measure-

ments is, however, not straightforward because tHcy also

increases with low folate status, and both tHcy and MMA

are elevated as a result of impaired renal function [9, 10] .

Brought to you by | Aarhus University Library / StatsbiblioteketAuthenticated | [email protected] author's copy

Download Date | 1/16/15 9:42 AM

2 Fedesov et al.: Combined indicator of B 12

status

More recently the Cbl-saturated form of the specific trans-

port protein, holoTC, has been used as an indicator of B 12

status with somewhat better predictive accuracy than total

B 12

[11, 12] . In general, B 12

deficiency is expected to result in

low concentrations of total B 12

and holoTC, accompanied

by high MMA and tHcy. The aforementioned biomarkers

are considered to be early (pre-clinical) indicators of defi-

ciency. However, diagnosis is often contradictory when

based on the four tests independently [5, 8, 9, 11, 13 – 17] .

For this reason efforts have been made to improve

the biochemical assessment of B 12

status. Most methods

use algorithms with a sequential selection (if → then)

based on the conventional cut-point for each marker [18,

19] . The ratio of holotranscobalamin (holoTC)/B 12

(a two-

component marker) showed a higher predictive potential

than each single marker separately [20] . Most recently, we

developed a novel approach to combine all four biomark-

ers and thereby increase precision of diagnosis [21] . The

metabolic fingerprint of each subject was presented as

either a point on a “ diagnostic surface ” or a single param-

eter, called w [21] . The distribution of points on the “ diag-

nostic surface ” was characterized by several frequency

peaks ( Figure 1 A). These selections contained combina-

tions of the four markers transformed into a single vari-

able dependent on age, see sketch in Figure 1 B [22, 23] .

The vertical distance between the tested measurements

and the reference level (the most frequent metabolite com-

bination at mature age) was called the combined indica-

tor of vitamin B 12

status (cB 12

). Its values were previously

interpreted as high-normal ( + 0.4), normal (0), low-normal

( – 0.5), deficient ( – 1.5) and severely deficient ( – 2.5). The

effect of age on the combined variable was assessed [22,

23] , and the reference level was described by an equa-

tion [23] instead of a fixed value [21] . This approach was

validated using two physiological characteristics mod-

erately affected by B 12

status: hemoglobin and cognitive

score [23] .

Despite a higher reliability of the four-marker analy-

sis, the routine measurements usually include only two or

three tests due to the lower overall costs. This restriction

apparently precludes calculation of the four-component

indicator and complicates comparison of the diagnostic

results based on different markers.

This article addresses the above issues. First, we

suggest a method for calculating the combined indica-

tor of B 12

status (cB 12

) when one or two biomarkers are

missing. Second, we examine the effect of folate status on

cB 12

since folate affects tHcy (one of the four components

of cB 12

). Finally, the gradation of cB 12

status is revised, and

useful cut-points are proposed.

Materials and methods Theory

The four-component indicator w was previously called the “ wellness

parameter ” [21, 23] or “ wellness score ” [22] . Now we have replaced

this term by a more descriptive defi nition, “ combined indicator of

vitamin B 12

status ” or “ cB 12

” . It can be expressed by Equation 1 below:

12 12

12 10 10 2.6

Test Ref

holoTC B holoTC B 3.79cB log log Test

MMA Hcy MMA Hcy age1

230

⎛ ⎞ ⎛ ⎞⋅ ⋅= − = −⎜ ⎟ ⎜ ⎟⋅ ⋅⎝ ⎠ ⎝ ⎠ ⎛ ⎞

+⎜ ⎟⎝ ⎠

(1)

The fi rst element of Equation 1 represents the logarithmic ratio of test

measurements, and the second element is a reference combination at

the stipulated age. Further details of the original formula have been

presented elsewhere [21, 23] .

Figure 1 Basic description of the method.

(A) Distribution of the main peaks on the diagnostic surface. (B) Schematic sketch, where the four-component variable from peaks 1 – 5

(panel A) is plotted versus age. The vertical distance between these selections and the reference level no. 2 corresponds to the combined

indicator of vitamin B 12

status (cB 12

).



Fedesov et al.: Combined indicator of B 12

status 3

Study populations

A pooled database (n = 5211) with plasma or serum concentrations of

B 12

, MMA, tHcy and holoTC (the list of methods in the Supplemen-

tary Material, Table S1 that accompanies the article http://www.

degruyter.com/view/j/cclm.2015.53.issue-8/cclm-2014-0818/cclm-

2014-0818.xml?format = INT ) was provided by the authors of origi-

nal papers [24 – 32] or obtained from the literature [33] . The database

included apparently healthy volunteers (n = 76) from Aarhus, Den-

mark [24] ; subjects from Aarhus (n = 647) suspected to be B 12

defi cient

[25] ; healthy vegans from Britain (nv144) without any serious medical

disorders [26] ; elderly people (n = 946) from Banbury (Oxfordshire,

England) [27] ; Hispanic elderly (n = 665) participants in the Sacra-

mento Area Latino Study on Aging (SALSA) who were exposed to

mandatory folic acid fortifi cation and in some cases to folic acid sup-

plementation [28] ; community-dwelling elderly (n = 320) from San-

tiago, Chile, classifi ed as having adequate, marginal and defi cient B 12

status based on serum B 12

measurement who were consuming folic

acid fortifi ed bread [29, 30] ; young adults (n = 2439) recruited at the

University of Dublin, Ireland [31] ; community living Mexican adult

women (n = 128) who were exposed to mandatory folic acid fortifi ca-

tion [32] ; and fi nally US subjects with folate (n = 19) or B 12

defi ciency

(n = 72) confi rmed both biochemically and clinically [33] . The adjust-

ment for serum folate was examined in 972 individuals who had data

on this variable.

Eligibility criteria

Data were excluded from subjects with: 1) plasma creatinine > 100

μ mol/L; 2) holoTC concentrations beyond the reliable range of 1 – 500

pmol/L; and 3) total B 12

> 1000 pmol/L. If interventions were conducted,

only baseline values were included. All the subjects were apparently

healthy, except for those with confi rmed folate or B 12

defi ciency.

Standardization of units

Concentrations of biomarkers were standardized to: 1) pmol/L for B 12

and holoTC; 2) μ mol/L for tHcy and MMA; and 3) nmol/L for serum

folate.

Calculation of combined B 12 status in the absence of one or two biomarkers

We used the “ pooled database ” (containing the results of all four

tests) to predict each biomarker considered to be missing. Its loga-

rithmic concentration ( Z ) was plotted versus the logarithmic combi-

nation of the three other markers ( X ) and age ( Y ) ( Figure 2 A). The

points obtained were initially approximated by the linear surface

3

2

AB

C D

1

0100

8060

X=log10 (B12

/(MMA·Hcy))

Z=

log 1

0 (h

oloT

C)

Log 1

0 (h

oloT

C) re

alR

esid

uals

(tr

ue –

pre

dict

.)

3cB

12 (

–hol

oTC

)

Log10 (holoTC)theor

1 1.5

Deviation

Deviation

2 2.5Y=Age

4020

0 -20

24

2.5

1.5

0.5

3

2

1

0

2.5

1.5

0.5

1

0

1.5

0.5

2

-2

x-ax

is (

fit 0

)

x-axis (fit 1)

Fit method

calc. method 012

021

-4

-6-5 -4 -3 -2 -1

4cB12

0 1 2 -5 -4 -3 -2 -1

4cB12

0 1 2

0

-0.5

-1

Figure 2 Prognosis of “ incomplete ” indicator at missing holoTC.

(A) Predictive surface, where holoTC is expressed via B 12

, tHcy, MMA and age. The best fit ( Z theor

function) is shown in Table 1 , entry 2. (B)

Validation of this function by plotting real holoTC versus theoretical holoTC. The linear slope (solid line) is equal to 1.0000000003. Fit

by smoothing splines is shown as a dashed line. Upper and lower insets illustrate the cases with purely vertical and vertical – horizontal

dispersion of points. (C) Alignment of the predicted “ incomplete ” indicator 3cB 12

(without holoTC) with the complete four-variable indicator

4cB 12

. Assumption of 3cB 12

or 4cB 12

as x -axis gives two different fits: “ 0 ” (dashed line) and “ 1 ” (dotted line). Fit “ 2 ” shows the intermedi-

ate situation (solid line). Three respective sets of linear coefficients were used to correct the “ incomplete ” indicator cB 12

. (D) Residual

plot (4cB 12

– cB 12

) versus 4cB 12

. Residuals correspond to cB 12

calculated according to corrections “ 0 ” (red), “ 1 ” (green) and “ 2 ” (blue). Solid

lines show the smoothing splines of the respective datasets. The average deviations σ cB12

are equal to 0.181 (set “ 0 ” ), 0.189 (set “ 1 ” ) and

0.183 (set “ 2 ” ).




status

Z theor

= A 1 + A

2 · X + A

3 · Y . It was then supplemented by trial elements

(e.g., A 4 · X · Y , A

4 · X 2 , etc.) that aff ected the goodness of fi t. The best

additional expression (if any) was included into the fi nal equation.

Refi nement was continued until the linear slope of the verifi cation

plot Z real

versus Z theor

(e.g., Figure 2 B) deviated from unity by < 5 · 10 6 .

The value of Z theor

fi lled a “ gap ” in Equation 1 and allowed assessment

of the “ incomplete ” three-component indicator 3cB 12

calculated from

the three known measurements and the prediction (inserted into

Equation 1 instead of the absent value). Analogous logic was applied

to calculation of 2cB 12

based on two known markers and prediction

for the other two missing values expressed, e.g., as log 10

(holoTC/

MMA).

The initial estimates of 3cB 12

and 2cB 12

were subjected to addi-

tional correction to compensate for a systematic deviation. The

problem lies in the least squares estimate of Z theor

in Figure 2 A,

which does not consider an error of the “ independent ” variable

X . This introduces a small systematic deviation to Z theor

( Figure 2 B)

and to the corresponding values of 3cB 12

or 2cB 12

( Figure 2 C). The

deviation was corrected using dependence of 3cB 12

(or 2cB 12

) on 4cB 12

( Figure 2 C). This linear chart has no pre-defi ned x -axis, and both

3cB 12

and 4cB 12

can be assigned as the apparent “ independent ” vari-

able x . If x = 3cB 12

and y = 4cB 12

the line y = A 1 + A

2 · x has by defi nition the

intercept A 1 = 0 and the slope A

2 = 1 (fi tting method 0). Swapping of

coordinates ( x = 4cB 12

, y = 3cB 12

) changes the best fi t to, e.g., A 1 = 0.14,

A 2 = 0.96 (fi tting method 1). The intermediate line between the two

above fi ttings represents total least squares fi t and has the interme-

diate linear parameters of A 1 = 0.07, A

2 = 0.98 (fi tting method 2). Three

sets (0, 1 and 2) of the correction coeffi cients A 1 and A

2 were used to

adjust cB 12

:

12 1

12

2

3cBcB

AA

−= (2)

The eff ect of each correction was examined on the residual plot (see

an example in Figure 2 D).

Correction of the combined indicator for folate status

Low folate status is expected to increase tHcy thereby shift ing cB 12

downward. The folate eff ect can be expressed as Δ folate

= log 10

(Hcy real

/

Hcy theor

), where Hcy real

refers to an observed measurement at any

folate status and Hcy theor

is a theoretical prediction expected at the

normal plasma folate concentration (see entries 3 and 4 in Table 1 ).

One of the folate datasets had no measurements of holoTC [33] , and

its Hcy theor

was calculated from MMA and B 12

(entry 4 in Table 1 ).

Successful correction, however, requires independence of the other

markers of folate.

Cut-points of the combined indicator of vitamin B 12 status

Five categories of B 12

status ( Figure 1 A) were used to calculate the val-

ues of cB 12

and plot them as probability density functions with the

same normalized area. The intersections show the optimal cut-points

[34] . Selections of cB 12

above and below – 0.5 ( ± 0.1, ± 0.5) were com-

pared in terms of hemoglobin (Hb) and cognitive score 0% – 100%

(mini mental state examination [27] and global cognitive score [28,

29] ) as surrogates for hematological status and cognitive function,

respectively.

Results

Assessment of combined vitamin B 12 status in the absence of individual biomarkers

HoloTC is one of the most frequently absent biomarkers;

therefore its prediction is described in the most detail. The

logarithmic concentrations of holoTC were taken from the

four-marker database and plotted versus the three other

markers X and age Y ( Figure 2 A). The points were approxi-

mated by a theoretical surface Z theor

as explained in the

Methods. The final approximating equation is presented

in Table 1 (entry 2). Figure 2 B validates accuracy of this

function by showing the linear chart of real holoTC versus

theoretical holoTC (slope of 1.0000000003). It appears,

however, that this chart has dispersion along both axes,

as in the sketch shown in the lower inset of Figure 2 B. Dis-

persion of theoretical holoTC in Figure 2 B was encoded

during the least squares fit in Figure 2 A, where the error

associated with combination of three markers could not

be considered. This caused a skewed dispersion of points

in Figure 2 B (main chart and lower inset) leading to a sys-

tematic deviation of the fitting line. The contrary situa-

tion (with purely vertical dispersion and a “ perfect ” fit) is

depicted in the upper sketch of Figure 2 B for a comparison.

To correct deviation, the three-component indicator

3cB 12

(derived from observed B 12

, MMA, tHcy and theo-

retical holoTC) was aligned with the true four component

4cB 12

in Figure 2 C. Depending on the assumed coordinate

x the best linear fit changes its position: fit “ 0 ” for x = 3cB 12

(dashed line); fit “ 1 ” for x = 4cB 12

(dotted line); fit “ 2 ” for

the average situation of total least squares (solid line).

Assuming one or another set of linear fitting coefficients,

we have calculated three variants of the corrected indicator

cB 12

(Equation 2). Deviation between the real 4cB 12

and the

predicted value of cB 12

was examined on the residual plot

( Figure 2 D). The average dispersion around zero ( σ cB12

) of

green points (fit “ 1 ” ) and blue points (fit “ 2 ” ) increased by

4% and 0.9%, respectively, when compared with red points

(fit “ 0 ” ). Blue and green points showed the lowest system-

atic shifts based on spline approximations. Based on these

observations, procedure “ 2 ” was chosen as the preferable

method to adjust prediction for the “ incomplete ” cB 12

.

The same methodology was repeated for the other

three markers, see the Supplementary Material (Figures

S1 – S10 and Table S2). The suggested predictive functions

for the markers and the corresponding correction equa-

tions of 3cB 12

and 2cB 12

are shown in Table 1 . The low

average dispersion of points around zero in the residual

charts ( σ cB12

column in Table 1 ) reveals the most reliable




status 5

Table 1 Calculation of the “ incomplete ” indicator cB 12

.

№

Missing marker(s), Z

Measured markers, X

Equation Z theor == F ( X , Y ) , Y == age (slope of Z real versus Z theor )

cB 12 corrected a

12

12

bcB

cB

best

σ

σ

σ

⎛ ⎞⎜ ⎟⎜ ⎟⎝ ⎠

1 log (B 12

) holoTC

logHcy MMA

⎛ ⎞⎜ ⎟⋅⎝ ⎠

1.904 0.3708 0.002487

0.001935

X YX Y

+ ⋅ + ⋅− ⋅ ⋅

Linear:

A 1 = 0.0052

A 2 = 0.965

0.176

(1.008)

(slope 1.0000000003)

2 log (holoTC)

12B

logHcy MMA

⎛ ⎞⎜ ⎟⎝ ⋅ ⎠

6.136 0.8914 0.003920

2.6143 4

X YX

+ ⋅ + ⋅

− ⋅ +Linear:

A 1 = 0.0056

A 2 = 0.962

0.183

(1.009)

(slope 1.0000000002)

3 log (Hcy) 12

B holoTClog

MMA

⎛ ⎞⋅⎜ ⎟⎝ ⎠

0.57060.07971 2.321 ( 0.5)

0.002757

XY

−− + ⋅ ++ ⋅

Linear:

A 1 = 0.0025

0.123

(1.004)

(slope 1.0000000003) A 2 = 0.993

4 log (Hcy)

12B

logMMA

⎛ ⎞⎜ ⎟⎝ ⎠

0.003774168.2 166.5 ( 0.5)

0.00236

XY

− ⋅ ++ ⋅

n.d. n.d.

(slope 1.0000000003)

5 log (MMA)

12B holoTC

logHcy

⎛ ⎞⋅⎜ ⎟⎝ ⎠

5 2

745.49 ( 2)743.97

2.0058276.127 10 0.001251

XXY X Y−

⋅ +−+

+ ⋅ ⋅ − ⋅ ⋅

Linear: A

1 = 0.0063

A 2 = 0.958

0.198(1.012)

(slope 1.000005)

6

⎛ ⎞⎜ ⎟⎝ ⎠

holoTClog

MMA 12

Blog

Hcy

⎛ ⎞⎜ ⎟⎝ ⎠

1.476 0.6034 0.005416

0.005331

X YX Y

+ ⋅ − ⋅+ ⋅ ⋅

Linear:

A 1 = 0.016

A 2 = 0.892

0.317

(1.027)

(slope 1.000000003)

7 holoTC

logHcy

⎛ ⎞⎜ ⎟⎝ ⎠

12B

logMMA

⎛ ⎞⎜ ⎟⎝ ⎠

− + ⋅+ ⋅

1.1962 0.6133

0.0007543

XY

Linear:

A 1 = 0.0094

A 2 = 0.937

0.240

(1.016)

(slope 1.00000009)

8 12

Blog

Hcy

⎛ ⎞⎜ ⎟⎝ ⎠

holoTClog

MMA

⎛ ⎞⎜ ⎟⎝ ⎠

0.61.849 1.6539 ( 1)

0.002138 0.002475

XY X Y

− + ⋅ ++ ⋅ − ⋅ ⋅

Linear:

A 1 = 0.0091

A 2 = 0.939

0.234

(1.015)

(slope 1.000000004)

a Correction of “ incomplete ” indicators 3cB 12

or 2cB 12

compensates for a systematic shift at the border values ( Figure 2 B – D and the Sup-

plementary Figures S1B,C,D – S9B,C,D). This procedure slightly increases the average deviation σ cB 12

if compared to the smallest deviation

σ best

(usually observed at absent correction “ 0 ” ). Linear correction of cB 12

uses Equation 2 and coefficients A 1 and A

2 ; b The average deviation

of the predicted “ incomplete ” cB 12

from the true values obtained in the complete four-variable test. The span 0 ± σ cB 12

covers 68% of all the

points. Low σ cB 12

deviation means better correspondence with the full test.

marker combination in the “ incomplete ” tests (see the

Discussion).

Correction of combined vitamin B 12 status for folate status

Assessment of the folate-associated shift in cB 12

was

done as explained in Methods. The amplitude of shift

Δ folate

= log 10

(Hcy real

/Hcy theor

) showed a steep increase at

folate concentrations < 10 nmol/L ( Figure 3 A). The com-

bination of the three other markers (excluding tHcy)

was, in contrast, essentially independent of serum folate

( Figure 3 B). This means that only tHcy requires correction.

The data in Figure 3 A were used to obtain the empirical

correction (Equation 3) for the folate-associated shift:

folate

3folate

1.1 eΔ−

= ⋅ (3)

Here the folate concentration is given in nmol/L. The

B 12

-related fraction of the combined indicator (at plasma

folate < 10 nmol/L) can be estimated by Equation 4:

Hcy shift

12 12 folatecB cB Δ= + (4)

Similar correction using red blood cell (RBC) folate failed,

because changes in RBC folate considerably affected the




status

Figure 3 Correction of combined vitamin B 12

status for serum folate.

(A) tHcy shift. Logarithmic ratios of Hcy real

(at any folate) and Hcy theor

(predicted for normal folate and arbitrary B 12

statuses) are plotted

versus serum/plasma folate. Solid line shows the best fit by Equation 3. (B) Absent shift in B 12

, holoTC and MMA at different folate. Clinically

identified patients with B 12

deficiency from [33] were excluded. Dataset from [29, 30] included only individuals after B 12

treatment (baseline

data were excluded, because B 12

deficiency was suspected in many cases). Solid line shows fit by smoothing splines.

ratio of the three other markers making correction impos-

sible (data not shown). This is not surprising, since red

blood cell folate is known to be influenced by B 12

status [5] .

Cut-points of combined vitamin B 12 status

Five areas in Figure 1 A covered the major Gaussian

peaks and their age-related shifts: area 1 with borders of

(holoTC · B 12

) 0.5 from 120 to 170 and ½ log(MMA · Hcy) from

– 0.05 to 0.3; area 2 (85 – 120) and (0 – 0.35); area 3 (45 – 85)

and (0.05 – 0.85); area 4 (25 – 45) and (0.3 – 0.8); area 5 (5 –

25) and (0.5 – 1.6). The values of cB 12

were extracted from

these regions and presented as probability distributions

( Figure 4 A). The points of their intersection indicate

the approximate borderlines. The main distributions in

Figure 4 A cover cB 12

values from – 5 to 1. A small distribu-

tion above 1.5 (n = 44, not shown) was classified as “ ele-

vated vitamin B 12

” (supra-physiological B 12

), see also Table

2 . The two dominating peaks, 1 and 2 (prevalent in young

and adult cohorts) were pooled together and regarded

as the “ adequate vitamin B 12

status ” category (cB 12

from

– 0.5 to + 1.5). Peak 3 has intermediate characteristics and

noticeably overlaps peak 2. We decided to treat the inter-

val of cB 12

from approximately – 1.5 to – 0.5 as indicating

“ low vitamin B 12

” . The main part of distribution 4 was

classified as a group with “ possible B 12

deficiency ” (cB 12

from – 2.5 to – 1.5). Finally, the broad peak 5 (cB 12

< – 2.5) was

categorized as the group with “ probable B 12

deficiency ” .

The gray zone in Figure 4 A represented a transi-

tional region around the general cut-point of – 0.5, where

identification of B 12

status might be ambiguous. We

considered a broad zone (cB 12

= – 0.5 ± 0.5) and a narrow

zone ( – 0.5 ± 0.1). The measurements of hemoglobin and

cognitive score were selected to the left and to the right

of these diffuse separators and presented as the corre-

sponding distributions. The selections with adequate

status (above separator) had better physiological char-

acteristics ( Figure 4 B, C red lines) if compared with the

selection having possible/probable B 12

deficiency (below

separator, green lines). Differences were highly signifi-

cant according to Kolmogorov-Smirnov and Kruskal-Wal-

lis non-parametric tests. A sharp separator at cB 12

= – 0.5

gave a decreased level of significance for hemoglobin

and ambiguous results for cognitive function, balancing

at the border of significance.

Table 2 provides a summary description of suggested

classifications of B 12

status together with guidelines for

further application of this approach.

Discussion

Assessment of combined vitamin B 12 status with missing markers

The method presented enables calculation of the “ incom-

plete ” three-component indicator 3cB 12

and the two-

component indicator 2cB 12

. Both types of analysis are

economically advantageous due to their lower laboratory

costs (compared with the four-marker procedure). Table 1

shows the equations necessary for calculation of 3cB 12

and 2cB 12

, and we also provide spreadsheets in the Sup-

plemental Materials to simplify calculations. It should be

understood that measurement of all four markers gives

more reliable determination of B 12

status because of the

larger amount of information encoded. Nevertheless the

truncated procedure (based on three or two markers)




status 7

A B

C

Freq

uenc

yFr

eque

ncy

Freq

uenc

y

0.8 0.35

0.3

0.25

0.2

0.15

0.1

0.05

06 7 8 9 10 11

Hb, mM

0.6

0.6

0.5

0.4

0.3

0.2

0.1

0

0.4

0.2

0-5 -4

0 20 40 60 80 100

-3 -2 -1 0-0.5

1

Probablydeficient

Possiblydeficient

LowAdequate

1.5

1

2

3

4

5

cB12

cB12

Cognitive score, %

>0

>-0.4<-1

<-0.6

>0

>-0.4

<-1

<-0.6

cB12

Figure 4 Analysis of frequency distributions.

(A) Distributions of the combined indicator cB 12

extracted from the representative areas ( № 1 – 5) in Figure 1 A. Approximate gradation of

status is notated at the top. Gray shading indicates a zone with potentially ambiguous diagnostics. General cut-point of – 0.5 is proposed.

(B) Distributions of hemoglobin from the selection with cB 12

> 0 or –0.4 (red line) and cB 12

< – 1 or –0.6 (green line). Significance of difference

was judged by Kolmogorov-Smirnov (**) and Kruskal-Wallis (***) tests, see also [23] . (C) Distributions of cognitive score. Selections and

significance (*** by K-S test, ** by K-W test) were obtained as in panel B, see also [23] .

still represents an improvement on one biomarker tests.

Table 1 classifies the “ incomplete ” tests according to their

reliability (lower σ cB12

= greater reliability). The best simu-

lation of the full-scale procedure was obtained with B 12

,

holoTC and MMA measured (tHcy excluded), see entry

3. The next best simulation was shared by two tests of

almost equal precision: without B 12

(entry 1) or without

holoTC (entry 2). Two-variable indicators expectedly

showed a higher deviation ( σ cB12

). The smallest error was

observed for the test based on holoTC and MMA (entry 8).

The errors of ± 0.25 are apparently acceptable considering

the span of status groups in Figure 4 A, e.g., “ adequate

B 12

” from – 0.5 to 1.5.

Correction of the combined indicator of vitamin B 12 status by folate status

The combined indicator cB 12

needs adjustment when

serum (plasma) folate is below 10 nmol/L and calculations

include tHcy. Correction by RBC folate was impossible

because of changes in all markers. B 12

deficiency causes

accumulation of methyl folate with increased plasma

folate levels and reduced red cell folate [35] . Correction

for low folate should therefore be omitted if interactions

between folate and B 12

status are at issue. The influences

of molecular genetic determinants, such as polymor-

phisms, were not included in our models, but represent a

potential topic to be explored in the future.

Cut-points and guidelines for using the combined indicator of vitamin B 12 status

In the present study, we define cut-points from statistical

criteria and physiological indicators based on hemoglobin

and cognitive score. However, it is important to acknowl-

edge that both indicators depend on many factors apart

from B12

status. Yet, even this rough approach indicated

changes in the selections above and below the “ gray

zones ” of cB 12

( Figure 4 ). The diffused separators were used

to highlight the artificial character of the sharp cut-point

(cB 12

= – 0.5) and to estimate its possible error (assessed as a

value between ± 0.1 and ± 0.2). Further research is planned

to connect the combined indicator with other pathophysi-

ological manifestations of B 12

deficiency.

For researchers and clinicians we suggest classifica-

tion of B 12

status as “ elevated vitamin B 12

” , “ vitamin B 12




status

Tabl

e 2

De

fin

itio

n o

f vi

tam

in B

12 s

tatu

s b

y co

mb

ine

d i

nd

ica

tor

cB 12

.

For r

esea

rche

rs a

nd cl

inic

ians

Fo

r epi

dem

iolo

gica

l pur

pose

s

Clas

sific

atio

n Eq

uiva

lenc

e to

si

ngle

cut-p

oint

s a Bi

olog

ical

inte

rpre

tatio

n Gu

idel

ines

Cl

assi

ficat

ion

Equi

vale

nce

to

sing

le cu

t-poi

nts a

Guid

elin

es

Ele

vate

d B

12

B 1

2 > 6

50

The

pa

tho

ge

ne

sis

of

hig

h B

12 i

s n

ot

full

y u

nd

ers

too

d

Co

ns

ide

r p

ote

nti

al

cau

se

s o

f h

igh

B 1

2

leve

ls s

uch

as

liv

er

dis

ea

se

or

curr

en

t o

r

rece

nt

su

pp

lem

en

tati

on

or

tre

atm

en

t

B 1

2 a

de

qu

acy

B 1

2 > 1

86

No

act

ion

ad

vis

ed

un

les

s

sig

ns

/sym

pto

ms

pre

se

nt

> 1.5

ho

loTC

> 19

0 > –

0.5

ho

loTC

> 37

tHcy

< 8.0

tHcy

< 13

.6

MM

A < 0

.11

MM

A < 0

.35

B 1

2 a

de

qu

acy

18

6 < B

12 < 6

50

Ex

pe

cte

d t

o a

cco

mp

lis

h a

ll B

12

sta

tus

de

pe

nd

en

t fu

nct

ion

s

No

act

ion

ad

vis

ed

un

les

s s

ign

s/

sym

pto

ms

pre

se

nt

– 0

.5 –

1.5

37

< ho

loTC

< 19

0

13

.6 > t

Hcy

> 8.0

0.3

5 > M

MA

> 0.1

1

Low

B 1

2

11

9 < B

12 < 1

86

Po

ten

tia

l su

b-c

lin

ica

l ma

nif

es

tati

on

s

of

B 1

2 d

efi

cie

ncy

. i.

e.,

ab

se

nce

of

he

ma

tolo

gic

al

cha

ng

es

, b

ut

su

bcl

inic

al

ne

uro

log

ica

l im

pa

irm

en

t

Co

ns

ide

r re

com

me

nd

ing

ora

l

su

pp

lem

en

ts

Tra

ns

itio

na

l B

12

sta

tus

– 0

.5 –

– 2

.5

11

9 < B

12 < 1

86

Ne

ed

ora

l s

up

ple

me

nta

tio

n

– 1

.5 –

– 0

.52

0 < h

olo

TC < 3

72

0 < h

olo

TC < 3

7

19

.2 > t

Hcy

> 13

.61

9.2

> tH

cy > 1

3.6

0.8

4 > M

MA

> 0.3

51

.7 > M

MA

> 0.3

5

Po

ss

ible

B 1

2

de

fici

en

cy

– 2

.5 –

– 1

.5

11

6 < B

12 < 1

19

Po

ten

tia

l m

an

ife

sta

tio

ns

of

B 1

2

de

fici

en

cy

Po

ten

tia

lly

pre

scr

ibe

ora

l s

up

ple

me

nts

,

as

se

ss

ag

ain

in

3 –

6 m

on

ths

Low

B 1

2 s

tatu

sB

12 < 1

19

Ne

ed

im

me

dia

te i

nte

rve

nti

on

wit

h I

M i

nje

ctio

ns

or

hig

h d

os

e

ora

l s

up

ple

me

nts

, m

on

ito

rin

g

of

sta

tus

ove

r ti

me

8.4

< ho

loTC

< 20

< – 2

.5h

olo

TC < 2

0

51

> tH

cy > 1

9.2

tHcy

> 51

1.7

> MM

A > 0

.84

MM

A > 1

.7

Pro

ba

ble

B 1

2

de

fici

en

cy

< – 2

.5

B 1

2 < 1

16

It i

s p

os

sib

le t

o o

bs

erv

e c

lin

ica

l

ma

nif

es

tati

on

s o

f B

12 d

efi

cie

ncy

.

Cli

nic

al

ou

tco

me

s a

re n

ee

de

d t

o

con

firm

po

ten

tia

l cl

inic

al

de

fici

en

cy

Co

ns

ide

r im

me

dia

te t

rea

tme

nt

wit

h

IM i

nje

ctio

ns

, d

ete

rmin

e c

au

se

wit

h

pri

ma

ry c

on

sid

era

tio

n f

or

the

po

ss

ibil

ity

of

pe

rnic

iou

s a

ne

mia

ho

loTC

< 8.4

tHcy

> 51

M

MA

> 1.7

a A

pp

roxi

ma

te e

qu

iva

len

ce t

o s

ing

le b

iom

ark

er

cut-

po

ints

wa

s b

as

ed

on

th

e c

alc

ula

tio

n o

f e

ach

sin

gle

bio

ma

rke

r co

ns

titu

tin

g t

he

co

rre

sp

on

din

g v

alu

e o

f th

e c

om

bin

ed

in

dic

ato

r o

f vi

tam

in B

12

sta

tus

. B

12 a

nd

ho

loTC

are

ex

pre

ss

ed

as

pm

ol/

L, t

Hcy

an

d M

MA

as

μ m

ol/

L. N

ote

: C

ut-

po

ints

we

re e

sti

ma

ted

ba

se

d o

n s

tati

sti

cal

crit

eri

a a

nd

its

va

lid

ati

on

ba

se

d o

n h

em

og

lob

in a

nd

co

gn

itiv

e

ind

ica

tors

. Fu

rth

er

vali

da

tio

ns

are

cu

rre

ntl

y u

nd

er

as

se

ss

me

nt

in m

ult

iple

stu

die

s.

The

refo

re,

the

se

cu

t-p

oin

ts a

nd

gu

ide

lin

es

re

pre

se

nt

a r

efe

ren

ce a

nd

it

is p

os

sib

le t

he

y m

ay

be

mo

dif

ied

ba

se

d o

n t

he

re

su

lts

of

futu

re v

ali

da

tio

ns

. Fo

r fu

rth

er

info

rma

tio

n r

eg

ard

ing

th

e c

alc

ula

tio

n o

f th

is n

ew

in

dic

ato

r p

lea

se

co

nta

ct S

erg

ey

Fed

os

ov

or

Ale

x B

rito

.




status 9

adequacy ” , “ low vitamin B 12

” , “ possible vitamin B 12

defi-

ciency ” and “ probable vitamin B 12

deficiency ” ( Table 2 ).

Definition of “ elevated vitamin B 12

status ” is not clear-

cut, and covers the relatively small group (n = 44 out of

5211) with the highest values of cB 12

(equivalent to serum/

plasma vitamin B 12

> 650 pmol/L). We separate this group

to attract attention to potential causes of high B 12

levels,

such as liver disease, autoimmune disorders, presence of

solid tumors [36] , inaccurate analytical measurements or

current/recent supplementation or treatment.

Classification of B 12

status suggested by the World

Health Organization (WHO) uses only three categories and

does not correct for age. The most frequently used classi-

fication includes “ vitamin B 12

adequacy ” ( > 221 pmol/L),

“ low B 12

” (between 148 and 221 pmol/L) and “ B 12

deficiency ”

( < 148 pmol/L) [37] . The approximate correspondence of

these categories to our novel approach can be achieved at

cB 12

> – 0.5; – 2.5 < cB 12

< – 0.5; and cB 12

< – 2.5, respectively.

The individual markers (constituting cB 12

in Table 2 )

are not identical with the combined indicator itself. For

example, the students of Trinity college [31] with low B 12

status included 13.7% and 6.7% based on B 12

< 186 pmol/L

and cB 12

< – 0.5, respectively. Here the result of the com-

bined analysis seems to be more realistic.

Potential uses of the combined indicator of vitamin B 12 status

One combined indicator (calculated form several meas-

urements) allows for more reliable diagnostics compared

with any single biomarker. In this way detection of sub-

clinical B 12

deficiency can be improved. Calculation of

“ incomplete ” cB 12

rescales data, based on unmatched

marker combinations, into a universal coordinate system,

thereby simplifying comparison of the results. A “ higher

weight ” of cB 12

in terms of accumulated information might

help to validate the conventional cut-points of the indi-

vidual markers and expose functional consequences of

low B 12

status. At the epidemiological level, there is the

potential to better define: 1) the consequences of clinically

undetected B 12

deficiency; and 2) the magnitude of B 12

defi-

ciency as a public health problem.

Limitations

The databases used here and earlier [21, 23] are heteroge-

neous in both participants and methods. Therefore, the

statistical estimates obtained from our pooled database

are potentially affected by multiple sources of variation.

The definition of cut-points needs further revision based

on clinical and functional manifestations of B 12

deficiency.

As there were no children in any of datasets, we currently

limit application of our approach to individuals above

18 years of age. Since plasma B 12

, tHcy, and MMA concen-

trations are altered during pregnancy [38, 39] this should

also be taken into account.

Future perspectives

The next step should be validation of the cut-points in

the context of functional B 12

deficiency. Further experi-

ence is required to establish whether this approach can

become a standard tool for more accurate determination

of B 12

status, as well as evaluating the response to B 12

supplementation.

Conclusions We provide equations and spreadsheets that combine

two, three or four biomarkers into one diagnostic param-

eter called “ combined indicator of vitamin B 12

status ” .

Adjustment of this indicator is required at serum or

plasma folate < 10 nmol/L because of increase in tHcy.

Revised cut-points and guidelines for the novel approach

are suggested.

Acknowledgments: The work of L.H. Allen and

A. Brito was supported in part by USDA, ARS Project

# 5306-51000-003-00D.

Author contributions: All the authors have accepted

responsibility for the entire content of this submitted

manuscript and approved submission.

Financial support: None declared.

Employment or leadership: None declared.

Honorarium: None declared.

Competing interests: The funding organization(s) played

no role in the study design; in the collection, analysis, and

interpretation of data; in the writing of the report; or in the

decision to submit the report for publication.

References 1. Allen LH. Vitamin B

12 . Adv Nutr 2012;3:54 – 5.

2. Herrmann W, Obeid R. Cobalamin deficiency. Subcell Biochem

2012;56:301 – 22.




status

3. Andr è s E, Loukili NH, Noel E, Kaltenbach G, Abdelgheni MB,

Perrin AR, et al. Vitamin B 12

(cobalamin) deficiency in elderly

patients. Can Med Assoc J 2004;171:251 – 9.

4. Lachner C, Steinle NI, Regenold WT. The neuropsychiatry of

vitamin B 12

deficiency in elderly patients. J Neuropsychiatry Clin

Neurosci 2012;24:5 – 15.

5. Green R. Indicators for assessing folate and vitamin B 12

status

and for monitoring the efficacy of intervention strategies. Food

Nutr Bull 2008;29:S52 – 63.

6. Elin RJ, Winter WE. Methylmalonic acid: a test whose time has

come ? Arch Pathol Lab Med 2001;125:824 – 7.

7. Scott JM. Folate and vitamin B 12

. Proc Nutr Soc 1999;58:441 – 8.

8. Carmel R, Green R, Rosenblatt S, Watkins D. Update on cobala-

min, folate, and homocysteine. Hematology Am Soc Hematol

Educ Program 2003:62 – 81.

9. Obeid R, Schorr H, Eckert R, Herrmann W. Vitamin B 12

status in

the elderly as judged by available biochemical markers. Clin

Chem 2004;50:238 – 41.

10. Lindgren A. Elevated serum methylmalonic acid. How much

comes from cobalamin deficiency and how much comes from

the kidneys ? Scand J Clin Lab Invest 2002;62:15 – 9.

11. Nexo E, Hoffmann-L ü cke E. Holotranscobalamin, a marker of

vitamin B 12

status: analytical aspects and clinical utility. Am J

Clin Nutr 2011;94:359S – 65S.

12. Herbert V, Fong W, Gulle V, Stopler T. Low holotranscobala-

min II is the earliest serum marker for subnormal vitamin B 12

(cobalamin) absorption in patients with AIDS. Am J Hematol

1990;34:132 – 9.

13. Carmel R. Biomarkers of cobalamin (vitamin B 12

) status in the

epidemiologic setting: a critical overview of context, applica-

tions, and performance characteristics of cobalamin, meth-

ylmalonic acid, and holotranscobalamin II. Am J Clin Nutr

2011;94:348S – 58S.

14. Chatthanawaree W. Biomarkers of cobalamin (vitamin B 12

)

deficiency and its application. J Nutr Health Aging 2011;15:

227 – 31.

15. Wuerges J, Garau G, Geremia S, Fedosov SN, Petersen

TE, Randaccio L. Structural basis for mammalian vitamin

B 12

transport by transcobalamin. Proc Natl Acad Sci

2006;103:4386 – 91.

16. Fedosov SN. Physiological and molecular aspects of cobalamin

transport. Subcell Biochem 2012;56:347 – 67.

17. Solomon LR. Cobalamin-responsive disorders in the ambulatory

care setting: unreliability of cobalamin, methylmalonic acid,

and homocysteine testing. Blood 2005;105:978 – 85.

18. Remacha AF, Sard à MP, Canals C, Queralt ò JM, Zapico E,

Remacha J, et al. Role of serum holotranscobalamin (holoTC) in

the diagnosis of patients with low serum cobalamin. Compari-

son with methylmalonic acid and homocysteine. Ann Hematol

2014;93:565 – 9.

19. Palacios G, Sola R, Barrios L, Pietrzik K, Castillo MJ, Gonz á lez M.

Algorithm for the early diagnosis of vitamin B 12

deficiency in

elderly people. Nutr Hosp 2013;28:1447 – 52.

20. Miller JW, Garrod MG, Rockwood AL, Kushnir MM, Allen LH,

Haan MN, et al. Measurement of total vitamin B 12

and holotransco-

balamin, singly and in combination, in screening for metabolic

vitamin B 12

deficiency. Clin Chem 2006;52:278 – 85.

21. Fedosov SN. Metabolic signs of vitamin B 12

deficiency in

humans: computational model and its implications for diagnos-

tics. Metabolism 2010;59:1124 – 38.

22. Lildballe DL, Fedosov S, Sherliker P, Hin H, Clarke R, Nexo E.

Association of cognitive impairment with combinations of

vitamin B 12

-related parameters. Clin Chem 2011;57:1436 – 43.

23. Fedosov SN. Biochemical markers of vitamin B 12

deficiency

combined in one diagnostic parameter: the age-dependence

and association with cognitive function and blood haemoglobin.

Clin Chim Acta 2013;422:47 – 53.

24. Hvas AM, M ø rkbak AL, Nex ø E. Plasma holotranscobalamin

compared with plasma cobalamins for assessment of vitamin

B 12

absorption; optimization of a non-radioactive vitamin B 12

absorption test (CobaSorb). Clin Chim Acta 2007;376:150 – 4.

25. Hvas AM, Nex ø E. Holotranscobalamin – a first choice assay

for diagnosing early vitamin B 12

deficiency ? J Intern Med

2005;257:289 – 98.

26. Wright ZL, Hvas AM, M ø ller J, Sanders TA, Nex ø E. Holotransco-

balamin as an indicator of dietary vitamin B 12

deficiency. Clin

Chem 2003;49:2076 – 8.

27. Hin H, Clarke R, Sherliker P, Sherliker P, Atoyebi W, Emmens K,

et al. Clinical relevance of low serum vitamin B 12

concentra-

tions in older people: the Banbury B 12

study. Age Ageing

2006;35:416 – 22.

28. Campbell AK, Miller JW, Green R, Haan MN, Allen LH. Plasma

vitamin B 12

concentrations in an elderly Latino population are

predicted by serum gastrin concentrations and crystalline vita-

min B 12

intake. J Nutr 2003;133:2770 – 6.

29. S á nchez H, Albala C, Lera L, Castillo JL, Verdugo R, Lavados M,

et al. Comparison of two modes of vitamin B 12

supplementation

on neuroconduction and cognitive function among older people

living in Santiago, Chile: a cluster randomized controlled trial, a

study protocol. Nutr J 2011;10:100.

30. Brito A, Allen LH, Verdugo R, S á nchez H, Hertrampf E, Albala C,

et al. Cyanocobalamin treatment improves vitamin B 12

status

and peripheral neuroconduction in deficient Chilean elderly.

[Abstract]. FASEB J 2012;26:126.2.

31. Stone N, Pangilinan F, Molloy AM, Shane B, Scott J, Ueland PM,

et al. Bioinformatic and genetic association analysis of micro-

RNA target sites in one-carbon metabolism genes. PLoS One

2011;6:e21851.

32. Shahab-Ferdows S, Anaya-Loyola MA, Vergara-Casta ñ eda H,

Rosado JL, Keyes WR, Newman JW. Vitamin B-12 supplementa-

tion of rural Mexican women changes biochemical vitamin B- 12

status indicators but does not affect hematology or a bone

turnover marker. J Nutr 2012;142:1881 – 7.

33. Stabler SP, Marcell PD, Podell ER, Allen RH, Savage DG,

Lindenbaum J. Elevation of total homocysteine in the serum

of patients with cobalamin or folate deficiency detected by

capillary gas chromatography-mass spectrometry. J Clin Invest

1988;81:466 – 74.

34. S ø reide K. Receiver-operating characteristic curve analysis in

diagnostic, prognostic and predictive biomarker research. J Clin

Pathol 2009;62:1 – 5.

35. Hansj ö rg S, Wilmanns W. Cobalamin dependent methionine syn-

thesis and methyl-folate-trap in human vitamin B 12

deficiency. Br

J Haematol 1977;36:189 – 98.

36. Arendt J, Nexo E. Unexpected high plasma cobalamin/Proposal

for a diagnostic strategy. Clin Chem Lab Med 2013;51:489 – 96.

37. Allen LH. Guidelines on food fortification with micronutrients.

World Health Organization. Geneva: Department of Nutrition for

Health and Development, 2006.




status 11

38. Murphy MM, Molloy AM, Ueland PM, Fernandez-Ballart JD,

Schneede J, Arija V, et al. Longitudinal study of the effect

of pregnancy on maternal and fetal cobalamin status in

healthy women and their offspring. J Nutr 2007;137:1863 – 7.

39. Greibe E, Andreasen BH, Lildballe DL, Morkbak AL, Hvas AM,

Nexo E. Uptake of cobalamin and markers of cobalamin status:

a longitudinal study of healthy pregnant women. Clin Chem Lab

Med 2011;49:1877 – 82.

Supplemental Material: The online version of this article

(DOI: 10.1515/cclm-2014-0818) offers supplementary material,

available to authorized users.



Combined indicator of vitamin B12 status: modification for missing biomarkers and folate status and...

Documents

Transcript of Combined indicator of vitamin B12 status: modification for missing biomarkers and folate status and...