Upsala J Med Sci 79: 97-102, 1974 

Choice of Variables for Predicting the Heart Volume 

LARS BACKLUND, KJELL BERGSTROM, PER ERICSSON, U N O  ERIKSON, 
BERNHARD HUITFELDT AND HANS RENCK 

From the D e p a r t m e n t s  of A n a e s i h e s i o l o g y ,  Clinical Physiology, Internnl Medicine a n d  Diagnostic 
R a d i o l o g y ,  University H o s p i t a l ,  U p p s a l u ,  S w e d e n  

ABSTRACT 
Roentgenological heart volume determinations in the 
supine posture with calculation according to the formulae 
of both Jonsell & Kjellberg et al., and measurements of 
the height, weight, total haemoglobin, blood volume and 
physical work capacity, were made on 45 men of ages 
64-86 years, previously prostatectomized and with no 
subjective symptoms of cardiopulmonary disease, and 
on 22 healthy men and 17 healthy women of ages 58-71 
years from a health survey. Statistical analysis of these 
data comprised simple correlation calculations and step- 
wise regression. The body weight showed, as a rule, a 
stronger correlation than the body surface area with the 
heart volume. On testing a combination of variables, 
measures of body size in combination with blood vol- 
ume appeared to have a stronger explanatory capacity 
than other combinations. Measures of physical work 
capacity seemed to have a weaker predictive capacity 
in the age groups concerned. All correlations between 
heart volume and the other variables were weaker in a 
group with ECG anomalies possibly expressing heart 
disease than in the groups with a normal ECG. 

INTRODUCTION 

Roentgenological measurement of the heart volume 
is a very common procedure in clinical practice 
and is used in the diagnosis of heart diseases, in 
evaluating the course of a disease and in assess- 
ing therapeutic effects. In certain cases, especially 
for diagnostic purposes, attempts are made to esti- 
mate the size of the individual heart chambers, but 
in general a measurement of the total heart volume 
is considered adequate. In sports physiology and 
clinical work physiology heart volume determination 
is performed with the aim of correlating the size 
of the heart with the cardiac function. Heart volume 
determination is also used in health surveys as a 
screening method for heart disease. With this ex- 
tensive application of the procedure it is of great 
importance to have access to a method with good 
precision and t o  use uniform and adequate models 
for prediction of normal values. A recent question- 

naire study (3) showed, however, that within 
Sweden alone several different methods are used, 
with variations both in the examination technique 
and in the formula for calculation of the volume. 
In normal persons the heart volume is related, 
among other things to body size and circulatory 
functional capacity. In calculating normal values, 
therefore, consideration must be taken of one or 
more such factors. 

The methods usually used in Sweden for heart 
volume determination have been developed by 
Jonsell (10) and Kjellberg et al. ( 1  I ,  12, 13). The 
volume according to the method of Jonsell (Vol,) 
is calculated from the formula: 

71 ( f - v r y  (f-vl) 

6 .P 
Val,=- a b c  

where a ,  b and c are the diameters illustrated 
in Fig. 1, f is the film-focus distance (=125cm), 
v f r  is the distance from the centre of the heart 
to the film in the frontal plane and v' is the 
corresponding distance to the film in the lateral 
plane. The formula corrects for degree of magnifi- 
cation. 

The volume according to Kjellberg et al. (Vol,) 
is calculated from the formula: 

where c 1  and e l  are c and e reduced for the 
degree of magnification and d ,  e and c are diameters 
as illustrated in Fig. 1. The other notation is the 
same as in the previous formula. 

The body position and projection are also of 
importance. Bergstrom et al. ( 1 ,  2 )  found that the 
differences in mean values obtained with the differ- 
ent projections were often considerable and larger 
in calculation according to Jonsell's formula than 

7 - 742855 Upsala J Med Sci 79 



98 L .  Backlund et al. 

in calculation by the formula of Kjellberg. In their 
material of healthy young men, Kjellberg’s formula 
gave, as a rule, somewhat higher values than 
Jonsell’s formula. The range of variation was 
smaller with exposure in the recumbent than in 
t h e  sitting position. Angled projections did not 
result in smaller ranges of variation than straight 
projections. 

T h e  heart volume is usually expressed in relation 
t o  calculated body surface area as a measure of 
body size. Reindell e t  al. (14) have pointed o u t ,  
however, that body weight has a stronger and more 
constant correlation than the calculated body sur- 
face area with the heart volume. Other methods 
of predicting the heart volume a r e  to use rneas- 
ures of t h e  circulatory functional capacity, e.g. 
physical work capacity, o r  measures of t h e  volume 
of t h e  circulatory system, e.g. total haemoglobin 
or blood volume (6, 7, 14). To increase the possi- 
bilities of accurate prediction it is important to 
ascertain which variables have the strongest corre- 
lation with t h e  heart volume, but also to know 
how these correlations vary with age. Reindell e t  
al. (14) have shown that both age and sex as 
well a s  degree of physical fitness influence t h e  
correlations between physical work capacity and 
heart volume. At a given heart volume the oxygen 
pulse a t  submaximal work is lower in women than 
in men a n d  higher in well-trained athletes than in 

Fig. 1 .  Schematic drawing showing the 
three diameters according to Jonsell, (u, b 
and c )  and the three diameters according 
to Kjellberg et al. (d, e and c ) .  

normal men. Further, there is a weaker correla- 
tion between oxygen pulse and heart volume in 
higher than in younger ages. 

T h e  aim of the present study was t o  find o u t  
which single variable and which combination of 2, 
3 and 4 relevant variables, respectively, have the 
greatest explanatory value in prediction of t h e  
heart volume in elderly persons free from heart 
disease. 

M A T E R I A L  
The study comprised both a group of men with no 
subjective symptoms of cardiopulmonary disease, who 
5-7 months previously had undergone prostatectomy 
with no postoperative complications, and a group of 
men and women from a health survey on a selection 
from the population of the urban district of Uppsala. 
This latter group had undergone a comprehensive clin- 
ical examination, including static and dynamic spiro- 
metry and a submaximal work test, at which they had 
exhibited no signs indicative of respiratory or circula- 
tory disease or other abnormality. None of them suf- 
fered from any disease which could be assumed to be 
limiting for the physical work capacity or to affect the 
blood formation. In the group of prostatectomized 
patients all were clinicully free of cardiopulmonary dis- 
ease. This group was divided into two sub-groups, one 
with a normal ECG both at rest and during exercise 
and the other with a pathological ECG at rest and/or 
during exercise. The sub-groups are referred to in the 
following as ECG-normal and ECG-abnormal. The ECG 

Table I .  Anthropometric data f o r  the material 

Height, cm Weight, kg 
- - - 

Age, yrs 
- 

Group n X Range X Range X Range 

A. Prostatectomized, 

B .  Prostatectomized, 
ECG-a bnormal 22 73.2 61-86 172.1 165-184 76.8 53-97 

C. Healthsurvey, men 22 64.6 59-71 176.0 164-184 75.3 59-95 
D. Health survey, women 17 63.0 5 9 4 9  162.9 150-178 63.5 47-79 

ECG-normal 23 72.5 64-82 173.3 166-183 75.5 58-95 

Upsala J Med Sci 79 



Heart volume prediction 99 

formula of both Jonsell (Val,) and Kjellberg (Vol,). 
In the calculations correction was made for the fact 
that the formula of KjeUberg et al. is intended for 30” 
angulation in the frontal projection, while in the present 
study the central beam was directed perpendicularly to 
the frontal plane ( I ,  2). 

Table 11. Meon values, standard deviations and 
rangesfor m e a s i m s  of’ heart volume, physical work 
cupucity und total hartnoglobin 

Group 

A. Prostatectomized, 
ECG-normal 

B .  Prostatectomized, 
ECG-a bnormal 

C. Health survey, 
men 

D. Health survey, 
women 

Vari- 
able ,f S.D. Range n 

Volj 985 226 6 6 6 1 3 7 7  23 
VolK 843 206 523-1280 23 
Wls,) 733 145 498-1 145 21 
THb 695 127 509-1 140 21 
V01j 967 219 663-1341 22 
VOlK 810 132 623-1096 22 
Wise 713 154 34&l000 19 
THb 672 142 442- 907 22 
V01j 890 177 667-1336 22 
VolK 649 94 487- 827 22 
WIso 798 170 497-1 108 20 
THb 673 103 525- 998 22 
Volj 641 110 426- 904 17 
VOIK 505 90 389- 712 17 
Wise 484 158 200- 810 16 
THb 520 88 395- 734 17 

changes consisted in bundle branch block, low-grade 
A-V block or unspecific ST-T changes. 

Table I gives the age distribution and certain anthropo- 
metric data for the different groups. 

PHYSIOLOGICAL METHODS 
The physical work capacity was determined by bicycle 
ergometry (graded loads of 6 min duration). W,,, and 
W,,o, defined as the work intensity at heart rates of 
130 and 150 b e a t s h i n ,  respectively, were calculated by 
inter- or extrapolation (7). At the end of the work test 
capillary blood was taken for lactate determination, 
whereafter the variable W,,,,, defined a s  the work load 
in kpmlmin performed at a lactate concentration of 
5 mEq/l, was calculated for the group of prostatecto- 
mized patients (15). Total haemoglobin (THb) was de- 
termined by the alveolar CO method, wherewith dupli- 
cate determinations were performed on the “health 
survey” group (5) and single determinations on the 
group of prostatectomized patients (15). Determinations 
of the haemoglobin concentration (Hb) was performed 
on capillary blood and from the value of THb and Hb 
the total blood volume (TBV) was calculated. 

ROENTGENOLOGICAL METHODS 
All patients were examined in the supine posture with 
antero-posterior and lateral projections and with the 
central beam directed at right angles to the frontal 
plane and lateral plane. The film-focus distance was 
125 cm and frontal and lateral films were exposed 
simultaneously without ECG triggering. The subject 
breathed calmly and shallowly during the exposure. 
Measurement on the films was performed by two of 
the authors in collaboration ( K .  B. and U. E.) and in 
all cases the volume was calculated according to the 

COMPUTER ANALYSIS AND STATISTICAL 
METHODS 
The body surface area (BSA), THb, TBV and heart 
volume were calculated from measured primary data by 

Table 111. Certain regression equations, giving re- 
sidual variances and correlation coefficients ( r )  

Regr. Residual 
coeff. Const. variance r n 

A. Prostatectomized, ECG-normal 
Val,= 13.5 BW - 36.7 171 
VolK= 9 . 5 B W  + 128.0 183 
Volj= 1068.2 BSA- 1033.8 176 
Vol,= 748.9BSA- 572.6 185 
Val,= 0.6W1so- 585.1 233 
V O ~ K =  0.2W,sn+ 742.4 219 
Volj= 54.7 TBV+ 627.4 201 
VolK= 80.9TBV+ 351.1 186 

B. Prostatectomized, ECG-abnormal 
Val,= 8.7 BW + 301.7 208 
VolK= 4.4 BW + 472.1 129 
Volj= 563.3BSA- 101.1 215 
VolK= 313.3BSA+ 215.5 131 
Val,= 0.7W150+ 434.2 207 
VolK= 0.2W1,50+ 632.2 115 
Volj= -16.7 TBV + 1055.5 229 
VolK= 7.0TBV+ 772.3 138 

C .  Health survey, men 

Val,= 1.5 BW + 
Vol,= 13.5 BW - 

Volj- 907.7BSA- 
VolK= 139.2BSA+ 
Vol j = 0.5 Wisnf 

Val,= 115.4TBv+ 
VolK= 66.9 TBV+ 

D. Health survey, women 
Vol j = 8 . 9 B W  + 
VolK= 6 . 9 B W  + 

vOlK= 0. I WlSo,Of 

Volj= 564.4BSA - 
VolK= 471.9BSA- 
Val,= 0.4W,,o+ 
volK= 0.3 W,sof 
Val,= 44.5 TBv+ 
VolK= 34.6 TBV+ 

129.8 138 
537.1 98 
845.4 137 
382.5 97 
490.2 170 
550.0 101 
249.8 155 
277.6 79 

83.2 77 
69.0 66 

307.2 81 
287.5 65 
453.8 102 
344.8 80 
438.9 111 
347.9 92 

0.69 23 
0.53 23 
0.67 23 
0.51 23 
0.29 20 
0.10 20 
0.29 20 
0.44 20 

0.43 22 
0.36 22 
0.36 22 
0.33 22 
0.41 17 
0.23 17 

-0.08 22 
0.06 22 

0.67 22 
0.14 22 
0.68 22 
0.20 22 
0.48 20 
0.21 20 
0.56 22 
0.61 22 

0.76 17 
0.73 17 
0.72 17 
0.74 17 
0.53 16 
0.56 16 
0.31 17 
0.29 17 

Upsala J Med Sci 79 



100 L .  Backlund et al. 

Table IV. Some resrrlts f r o m  s t e p w i s e  regression, 
gi\>ing the mirltiple correlation coefficient ( R )  a s  (in 
index of the explanatory value of the regression 

~ 

Depend- Regression coefficients 
ent vari- 
able Const. BW TBV W,30 W,,,, BSA R 

A. Prostatectomized, ECG-normal, n =20 
Vol, 127.3 11.0 
VOIJ -263.0 11.4 62.8 

V0lJ -413.9 11.4 69.3 0.2 
VolJ -322.6 12.3 62.3 0 . 4  -0.5 
VOIJ -958.6 44.5 -0.5 

Vol, 133.5 11.0 -0.01 

B. Prostatectomized, ECG-abnormal, n =22 
Vol, 301.7 8.7 

Vol, 1.9 7.7 0.7 
VOIJ 448.7 9.8 -43.9 

VOIJ 147.0 8 . 9  -52.1 0.7 
VOIJ 899.7 10.4 -134.9 1 . 1  -1.2 
Vol, 412.4 -88.0 -0.8 

C .  Health survey, men, n=22 
VolJ -129.8 13.5 
VolJ -389.9 1 1 . 1  80.0 

VOIJ -399.1 11.5 87.6 -0.1 
VolJ -148.4 12.8 0.1 

VOIJ -827.9 0.1 

Vol J 52.5 9 . 4  

Vol, 60.2 8.8 0.1 

VOIJ -313.1 0.1 

VOIJ -816.9 60.8 

D. Health survey, women, n = I6 

VolJ -76.0 9 . 0  34.7 

Vol J 82.5 9.1 36.0 -0.02 
VOIJ -420.7 25.4 

0.63 
0.71 
0.63 
0.72 
0.76 

1003.7 0.73 

0.43 
0.47 
0.51 
0.56 
0.67 

737.0 0.46 

0.67 
0.77 
0.68 
0.77 

716.3 0.73 
866.9 0.69 

0.78 
0.81 
0.79 
0.81 

566.8 0.76 
546.9 0.75 

means of a computer. The statistical analysis was also 
performed by a computer; in this analysis the regression 
line, residual variance and coefficient of correlation 
were calculated for each individual variable against 
each of the other variables. Stepwise linear regression 
analysis (4) was performed in order to determine which 
variable and which combination of variables had the 
greatest value for predicting the heart volume. Only 
persons for whom complete data had been obtained 
were included in these analyses. 

RESULTS 

The mean values and ranges for heart volume, 
W,,, and T H b  in the different groups are pre- 
sented in Table 11. It can be seen in the table 
that VOI, was higher than Vol, in all groups 
(P,<0.05, P,<O.Ol, P,<O.OOl, P,<O.OOl). It is 
also evident from the table that the heart volume 

was somewhat higher in the two groups of pro- 
statectomized patients than in the healthy men 
from the health survey. This difference was sta- 
tistically significant for Vol,(P<O.OOl). There was 
no significant differences in heart volume, how- 
ever, between ECG-normal and ECG-abnormal 
prostatectomized patients. 

Table I11 A-D presents certain regression equa- 
tions, giving residual variance and coefficient of 
correlation as measures of the predictive value 
of the regression. It is evident from the table 
that the residual variance was generally somewhat 
lower and the coefficient of correlation somewhat 
higher when body weight (BW) was the independ- 
ent variable than when this variable was body 
surface area (BSA). The explanatory value for BW 
and BSA was, in general, higher than for W,,, and 
TBV. The coefficient of correlation was usually 
higher in regressions with Vol, than with Vol, 
as the dependent variable. 

Table IV A-D gives certain results from the 
stepwise regression analysis with the multiple 
correlation coefficient ( R )  as a measure of the 
prediction value of the regression. The starting 
point of the calculations was the finding that in 
the simple regression analysis VOlJ and BW had 
shown a higher degree of correlation to different 
variables than VOlJ and B S A ,  respectively. The 
table shows that all regressions had a considerably 
lower prediction value in the ECG-abnormal group 
of prostatectomized patients than in the other 
groups. It can also be seen in the table that BW 
alone has a relatively high explanatory value and 
that BW in combination with blood volume seems 
t o  have a higher explanatory value than B S A  in 
combination with blood volume. WIs, as a 
further variable did not increase the predictive 
value of the regression noteworthily except in the 
group of ECG-abnormal prostatectomized patients. 

DISCUSSION 

Groups C and D were taken from a statistically 
selected sample of the population in the age 
section concerned. The persons included comprise 
a relatively stringently selected group among those 
who fulfilled the criterion of having no manifest 
disease of importance for the ventilatory capacity, 
central circulatory capacity or blood formation. 
These groups (C and D) can thus be said t o  re- 

Zlpsala J Med Sci 79 



present healthy people in this age section. Group A 
(prostatectomized patients with no manifest cardio- 
pulmonary disease) reasonably can be considered 
to be representative of healthy men of the age 
range in question. Groups A ,  C and D exhibited 
essentially the same pattern as regards the correla- 
tion between heart volume and body dimensions 
and between circulatory functional capacity and 
blood volume. I n  group B (ECG-abnormal pro- 
statectomized patients) these correlations were 
weaker despite the absence of a history of or 
other clinical signs of heart disease. The im- 
portance of ECG changes for correlations with 
the heart volume has been demonstrated previ- 
ously by Strandell (16). 

Reindell et al. (14), among others, have found 
the heart volume to be more strongly correlated 
to body weight than to body surface area. This 
observation is also corroborated by the present 
investigation. Thus there appear to be convincing 
arguments favouring the view that the total heart 
volume shall be related to body weight rather 
than to body surface area. Situations with rapid 
changes in weight or other abnormal weight con- 
ditions should probably comprise exceptions. In 
these extreme cases, however, it is doubtful 
whether any measure of body size at all can be 
of value for predicting the heart volume. 

A further improvement in prediction of the heart 
volume and establishment of normal values is 
obtained if the blood volume is added to the body 
weight. This is in agreement with previous results 
(16). The findings in the present study indicate, 
on the other hand, that the predictive value of 
physical work capacity is considerably smaller 
than has been reported earlier for younger age 
groups (8, 9, 16). In our material the correlation 
between physical work capacity, as a measure of 
the central circulatory capacity, and the heart 
volume, as an expression of the volume of the 
circulatory system, was relatively weak. THb and 
TBV can also be said to be measures of the 
volume of the circulatory system. In a larger 
sample of men and women of corresponding ages 
from the same health survey material, the corre- 
lations between physical work capacity and THb 
and TBV, respectively, were studied and found to 
be weaker than those reported earlier for younger 
persons (5). Both of these observations may in- 
dicate that in a population such as this the 
physical work capacity is limited by other factors 

Heart volume prediction I0 1 

than the volume of the circulatory system and the 
oxygen transport capacity. 

In a comparison between the calculation proce- 
dures of Jonsell and Kjellberg et al., the correlations 
between heart volume and other variables were 
found to be stronger throughout with the Jonsell 
formula despite the fact that the variance was, as 
a rule, lower in calculations according to Kjell- 
berg et al. The implication of this observations 
cannot be stated with certainty, and further analy- 
sis would seem to be motivated. A possible ex- 
planation can be that determinations according to 
Kjellberg for special body types and/or in the 
case of obesity have a poorer volume discrimina- 
tory capacity than Jonsell’s method. This might 
possibly explain the observation that the residual 
variances are lower for Kjellberg’s method while 
the correlation coefficients are higher for measure- 
ments according to Jonsell. 

REFERENCES 
I .  

2 .  

3. 

4. 

5 .  

6 .  

7. 

8. 

9. 

10. 

Bergstrom, K . ,  Backlund, L., Erikson, U .  & Gustafs- 
son, B.: Heart volume and its relation to measures 
of circulatory function in healthy young men. Acta 
Med Scand 185: 471, 1969. 
Bergstrom, K . ,  Erikson, U. & Gustafsson, B.: 
Roentgenological determination of the heart volume. 
Acta Soc Med Upsal74: 81, 1969. 
Bergstrom, K .  & Erikson, U.: Rontgenologisk h j P t -  
volymsbestamning. En enkat-undersokning. Lakar- 
tidningen 68; 2599, 1971, 
Efroymson, M. A.: Multiple regression analysis. 
In Mathematical Methods for Digital Computers, 
Part V (17) (ed. A. Ralston & S .  Wilf). John Wiley 
and Sons, 1960. 
Ericsson, P.: The effect of iron supplementation on 
the physical work capacity in the elderly. Acta 
Med Scand 188: 36 I ,  1970. 
Ericsson, P.: Total haemoglobin and physical work 
capacity in elderly people. Acta Med Scand 188: 
15, 1970. 
Ericsson, P. & Irnell, L.: Physical work capacity 
and static lung volumes in elderly people. Acta 
Med Scand 185: 185, 1969. 
Holmgren, A., Jonson, B., Levander, M., Linder- 
holm, H., Sjostrand, T. & Strom, G.: Low physical 
working capacity in suspected heart cases due to 
inadequate adjustment of peripheral blood flow 
(vasoregulatory asthenia). Acta Med Scand 158: 
413, 1957. 
Holmgren, A. & Strandell, T.: The relationship 
between heart volume, total hemoglobin and physi- 
cal working capacity in former athletes. Acta Med 
Scand 163: 149, 1959. 
Jonsell, S . :  A method for the determination of the 
heart size by teleroentgenography ( a  heart volume 
index). Acta Radio1 (Stockh) 20: 325, 1939. 

Upsala J Med Sci 79 



102 L .  Backlund et al. 

11. Kjellberg, S. R., Rudhe, U .  & Sjostrand, T.: The 
relations of the cardiac volume to the weight and 
surface area of the body, the blood volume and the 
physical capacity for work. Acta Radiol (Stockh) 
31: 113, 1949. 

12. Kjellberg, S .  R., Lonroth, H. & Rudhe, U.: The 
effect of various factors on the roentgenological 
determination of the cardiac volume. Acta Radiol 
(Stockh)35: 413, 1951. 

13. Larsson, H. & Kjellberg, S .  R.: Roentgenological 
heart volume determination with special regard to 
pulse rate and the position of the body. Acta Radiol 
(Stockh) 29: 159, 1948. 

14. Reindell, H., Konig, K. & Roskamm, H.: Funktions- 
diagnostik des gesunden und kranken Herzens. 
George Thieme Verlag, Stuttgart, 1967. 

15. Renck, H.: The elderly patient after anaesthesia 
and surgery. With special regard to certain respira- 
tory, circulatory, metabolic and muscular functions. 
Acta Anaesthesiol Scand, Suppl. X X X I V .  1969. 

16. Strandell, T.: Heart volume and its relation to some 
anthropometric data in old men compared with 
young men. Acta Med Scand 176: 205, 1964. 

Received November 15. 1973 

Address for reprints: 

Lars Backlund, M .  D. 
Dept. of Clinical Physiology 
University Hospital 
S-750 14 Uppsala 
Sweden 

Upsala J Med Sci 79