THE JOURNAL OF INFECTIOUS DISEASES. VOL. 137, NO.2.

FEBRUARY 1978
© 1978 by the University of Chicago. All rights reserved. 0022-1899178/3702-0003$00.88

A Method for Testing for Synergy with Any Number of Agents

M. C. Berenbaum From the Wellcome Laboratories of

Experimental Pathology, Variety
Club Research Wing, St. Mary's
Hospital Medical School, London,
United Kingdom

The standard checkerboard titration for detecting synergy between antibiotics is

practicable for combinations of two antibiotics, laborious for combinations of three,
and not feasible for combinations of four or more. Nevertheless, methods for testing
of combinations of several antibiotics are urgently needed because some combina-
tions might be superior to those in use and enable the successful treatment of infec-
tions resistant to current therapy. A simple method for measurement of synergy (or

Downloaded from http://jid.oxfordjournals.org/ at Syracuse University Library on April 6, 2014

antagonism) with combinations of any number of agents has been developed which
requires less effort than the standard checkerboard titration of two agents. With this
method, the concentrations of each of n agents producing some specified effect (such
as minimal inhibitory concentration or minimal bactericidal concentration) are de-
termined. A reference combination made up of lin of each of these concentrations is
titrated to find a dilution that produces the specified effect. The degree of dilution
required is equal to the sum of the fractional inhibitory concentrations (concentration
of each agent in combination/concentration of each agent alone) as conventionally
determined by checkerboard titrations; sums of < I, I, and> I indicate synergy, addi-
tivity, and antagonism, respectively.

The use of synergistic combinations of antibiotics
is often needed in the treatment of serious infec-
tions. Testing of combinations for synergy is
therefore widely practiced, and there is much
literature on the subject. The general procedure is
to conduct a so-called checkerboard titration, in
which two antibiotics are tested in serial dilutions
and in all combinations of these dilutions to-
gether to find the concentrations of each anti-
biotic, both alone and in combination, that pro-
duce some specified, easily determined effect.
The nature of the interaction between the two
antibiotics is then determined either algebraically
or geometrically. In the former method, the con-
centration of each antibiotic in the combination
that produces the specified effect is expressed as a
fraction of the concentration that produces the
same effect when the antibiotic is used alone,
i.e., its fractional inhibitory concentration. When
Received for publication March 18, 1977, and in revised
form July 12, 1977.
This study was supported by the Medical Research Coun-
cil and the Cancer Research Campaign.
I thank Dr. H. Gaya for helpful discussion.
Please address requests for reprints to Dr. M. C. Beren-
baum, Department of Experimental Pathology, St. Mary's
Hospital Medical School, London W2, United Kingdom.
the sum of these fractions is I, the combination
is additive; when the sum is <I, the combination
is synergistic; and when the sum is > I, the
combination is antagonistic. In the second
method, which is the geometric counterpart of the
first, a graph is constructed with the axes represent-
ing antibiotic concentrations on linear scales.
When the combination is additive, the isobole
(i.e., the line joining the points that represent all
combinations with the same effect, including the
equally effective concentrations of the antibiotics
used alone) is straight. Synergistic combinations
give concave isoboles, and antagonistic combina-
tions give convex isoboles [I].
However, these procedures are rarely, if ever,
carried out with combinations of three antibiotics,
and there does not appear to have been any at-
tempt to apply them to combinations of four or
more. The reason is obvious. 1£, for instance, each
antibiotic is tested in 10 dilutions (including a
control), 100 tests are required for a combination
of two antibiotics, 1,000 tests for a combination
of three, 10,000 tests for a combination of four,
and so on. Testing of combinations of three anti-
biotics is therefore laborious, and testing of four
or more antibiotics appears to be unfeasible. Fur-
thermore, although it is easy to plot isoboles for

122
Synergy with Any Number of Agents 123

combinations of two antibiotics, plotting them for

combinations of three antibiotics entails a three- 8
dimensional construction (see below), and it is i3
0 7
0
physically impossible to plot isoboles for com- 0
6
0
binations of more than three agents. II:
W
I- 5
However, there is unquestionably a need for Z
w
4
methods for testing for synergy with combinations ::::l
u,
of any number of agents, because such testing 0 3
;!
might lead to the discovery of antibiotic combina- Cl 2
0
tions considerably more potent than those now in ..J

use and effective against infections resistant to cur-

0
rent therapy. The method described herein allows 0 100 200 400 1000
synergy or antagonism to be measured for combi- STREPTOMYCIN llJ9/mll

Downloaded from http://jid.oxfordjournals.org/ at Syracuse University Library on April 6, 2014

nations of several antibiotics with less effort than is
involved in a conventional checkerboard titration
of two agents. The principles are not restricted
to the use of antibiotics (2).
Methods
Combinations of one agent. The simplest ap-
proach to this problem is first to define additivity
and then to see how synergy and antagonism dif-
fer from additivity. Agents that act additively are
no more and no less effective in combination
than they are separately; in other words, there is
no advantage in combining them. There is one
sort of combination that must always behave in
this way, i.e., the spurious combination of any
agent with itself. Such "combinations" are there-
fore convenient models for additivity, and, by
examining their behavior, we can see how addi-
tive combinations of different agents should be-
have. Furthermore, if synergy is broadly defined
as a condition in which the effectiveness of agents
is increased when they are combined and antago-
nism as the reverse, this additive model provides
a frame of reference by which we can see how syn-
ergistic and antagonistic combinations should be-
have.
For illustration, let us consider the effect of
streptomycin on enterococci as illustrated in fig-
ure I from the data of Moellering et al. [3]. Sup-
pose this antibiotic were placed in two containers,
A and B, and we conducted an experiment to find
the effects of A and B alone and in combination.
To reduce loglo cfu of the enterococci from 8.2 to
2.0, we need a concentration of either 400 J-Lg of A/
ml or 400 J-Lg of B/m1. This same effect is pro-
duced by a combination of 200 J-Lg of A/ml and
Figure 1. Relation between concentrations of strep-
tomycin and growth of enterococci (data from [3]).
200 J-Lg of B/m!. Let the concentrations of A and
B that each produces the same specified effect
(equally effective concentrations) when the anti-
biotics are used alone be called A e and Be' and let
A e and Be be their concentrations when used in
combination. Then A e = Be= 400 J-Lg/ml and
=
A e/ A e + Be/Be 200/400 + 200/400 1. =
We also know (because A and B are in fact the
same agent) that other combinations having this
effect are, for instance, 100 J-Lg of A/ml + 300 J-Lg
of B/ml or 350 J-Lg of A/ml + 50 J-Lg of B/ml:
= =
Ae/A e + Be/Be 100/400 + 300/400 I and 350/
400 + 50/400 = 1. Alternatively, if we specify that
the effect we wish to produce is a reduction in IOglO
cfu of the enterococci from 8.2 to 7.0, then A e = Be
= 200 J-Lg/m1. This effect will be produced by a
combination of, for example, 100 J-Lg of A/ml and
100 J-Lg of Byrnl or 150 J-Lg of A/ml and 50 J-Lg of
B/ml: Ae/A e + Be/Be = 100/200 + 100/200 =
I and 150/200 + 50/200 = 1. In other words, with
this "pair" of antibiotics, which we know can only
behave additively, equation I (Ae/ A e + Be/Be =
I) always holds.
Equation I can also be expressed geometrically,
because it is an equation of a straight line. Figure
2 is an isobologram, showing the effects of the
various combinations mentioned above and illus-
trating the fact that in this spurious case in which
the two constituents of the combination are in
fact identical, the points representing all combi-
nations with an equal effect lie on the straight line
joining the concentrations of the two constituents
that are equally effective when used alone.
124
-,
400 2
B
300 2
-,
200
-, 7 2 B
100 8
"7 ~2
0 8.2 8
"7 "2
o 100 200 300 400
A
Figure 2. Isobologram showing effects of combina-
tions of two identical preparations, A and B, of strep-
tomycin on the growth of enterococci (data from fig-
ure I). Values are loglO cfu of the organisms after cul-
ture for 24 hr with the antibiotic preparation. Points
representing combinations with equal effects lie on the
straight line joining the concentrations of A and B
alone that produce the same effect (equally effective
concentrations}.
Combination of two agents. If the antibiotic
in container B is mixed with an equal part of an
inert material that has no effect on the antibiotic,
the antibiotic's actions, or the bacteria, the equally
effective concentrations of A and B that reduce
the 10glO cfu of enterococci from 8.2 to 7.0 are 200
and 400 JLg/ml, respectively. Thus combinations
producing this effect are, for instance, 100 JLg of
A/ml + 200 JLg of B/ml or 50 JLg of A/ml + 300
=
JLg of B/ml: Ae/Ae + Be/Be 100/200 + 200/400
= I and 50/200 + 300/400 = I.
It appears, therefore, that when a combination
of two agents is simply additive, equation I holds,
irrespective of the effect specified or whether the
equally effective concentrations of the two agents
are the same or different, and that the equa-
tion does not depend on any particular rela-
tion between concentration and effect. This case
may also be expressed geometrically. It is clear
from figure 3 that all additive combinations pro-
ducing the same effect as A e and Be lie on the
straight line between A e and Be whether A, and
Beare the same or different.
Berenbaum
400 7
300
\ 7
\
200 8 7
Downloaded from http://jid.oxfordjournals.org/ at Syracuse University Library on April 6, 2014
100
o 8.2 8 7
o 100 200
A
Figure 3. Isobologram showing effects of combina-
tions of streptomycin (A) and streptomycin mixed
with an equal part of inert material (B). Values are
IOglO cfu of the organisms after culture for 24 hr
with the antibiotic preparation. Points representing
combinations with equal effects lie on the straight
line joining the concentrations of A and B that are
equally effective when used alone.
An additive combination of two agents may
therefore be defined as one that satisfies equation
I and that therefore, in an isobologram, lies on
the straight line joining the concentrations of the
two agents that, when the agents are used alone,
produce the same effect as the combination (the
equally effective concentrations).
Unequivocal and precise definitions of synergy
and antagonism follow from this definition of
additivity. Synergistic agents are more effective in
combination than they are separately, i.e., less is
required to produce a given effect when the anti-
biotics are used together than when they are used
separately, and the sum of the fractions in equa-
tion I is therefore <I. Conversely, for antagonistic
agents, more is required to produce a given effect
Synergy with Any Number of Agents 125

Figure 4. Isobolograms showing

the relation between sums of
fractional inhibitory concentrations
and isobolar shapes for combina-
tions of two agents. In this example
equal effects are produced by 8
units of agent A and 12 units of
12 1.0 12 1.0
agent B. The sum of fractional in-
hibitory concentrations for each
combination is readily calculated
10
by reference to the dose axes of
A and B and is shown to the right B B
of the point (e) representing that
8
combination. When the sum of the
fractional inhibitory concentrations
is <1 (left), the point lies below 6

Downloaded from http://jid.oxfordjournals.org/ at Syracuse University Library on April 6, 2014

the additive line, the isobole is
concave, and the combination is
synergistic. When the sum is >1
(right), the point lies above the
additive line, the isobole is convex,
and the combination is antagonistic:
Note that, if 0 is the origin of the
isobologram, C the point repre-
senting any combination (e), and C'
2 4 6
the intersect (0) of OC or its extra- A A
polate with the additive line, then
the sum of fractional inhibitory
concentrations for C equals oct
OC'. Synergy and antagonism are
usually most marked along OX' or
its extrapolate, where X' is the mid-
point of the additive line.

when the antibiotics are used together than when
they are used separately, and the sum of the frac-
tions is> I.
Figure 4 shows the connection between equa-
tion I and the shape of the isobole for combina-
tions of two agents. When for any given combina-
tion the sum of the fractions is <I, the point rep-
resenting that combination lies below the addi-
tive line, and when all combinations with the
same effect are synergistic, the isobole for that
effect is concave. Conversely, when the sum of
the fractions for a combination is > I, the point
representing it lies above the additive line, and
when all combinations with the same effect are
antagonistic, the isobole for that effect is convex.
Combination of three agents. Let us now add
to our containers of streptomycin (A and B) a
third, C, in which the antibiotic is mixed with
twice its weight of inert material. The concentra-
tions of A, B, and C required to reduce the loglo
c£u of enterococci from 8.2 to 7.0 are now 200,
400, and 600 p,g/ml, respectively. Possible com-
binations of the three that will result in a final
concentration of 200 p,g of streptomycin/rnl are
100 p,g of A/ml + 100 p,g of B/ml + 150 p,g of
C/ml and 50 p,g of A/ml + 100 p,g of B/ml + 300
p,g of C/ml: 100/200 + 100/400 + 150/600 = I
and 50/200 + 100/400 + 300/600 = 1. Thus equa-
tion I can be extended for use with combinations
of three agents by adding a term for the third, as
long as the sum of the fractions is I (equation 2):
A c / A e + Be/Be + Ce/C e = 1.
Equation 2 describes a plane in three dimen-
sions, and figure 5 accordingly shows that, when
the sum of the fractions in this equation is I, the
points representing the combinations lie in the
flat plane joining A e, Be' and Ceo
When three agents are synergistic, the sum of
the fractions in equation 2 is <I, and when they
are antagonistic, the sum is > 1. In the former case,
the isobolar surface is concave, and in the latter,
convex (figure 6).
Combinations of any number of agents. It is
evident that the argument used to extend equa-
126
C
600
• 600
C
tion 1 to equation 2 can be used to extend it for
use with any number of agents. For instance, if we
had six different mixtures of an antibiotic with
an inert material (A-F) such that the concentra-
tions of each required to produce a given speci-
fied effect were 200, 400, 800, 900, 1,000, and
1,200 JLg/ml, respectively, or six different anti-
biotics (A-F) with these equally effective concen-
trations, we could form a combination of all six
that would itself have this specified effect if we
C • 600
•
C
600
Berenbaum
Figure 5. Three-dimensional iso-
bolograms showing that the points
representing additive combinations
of antibiotics lie in the flat plane
that joins the concentrations of the
three agents that are equally ef-
fective when used alone, and that
the sum offractional inhibitory con-
centrations for such combinations
is 1. In this example, equally
effective concentrations of A, B,
and Care 200, 400, and 600
/1g/ml, respectively. Combination
Downloaded from http://jid.oxfordjournals.org/ at Syracuse University Library on April 6, 2014
X consists of 100 /1g of Ajml, 100
/1g of B/ml, and 150 /1g of C/m!.
Combination Y consists of 50
/1g of A/ml, 100 /1g of B/ml, and
300 /1g of C/m!.
ensured that A c/200 + B c/400 + C c/800 + D c/900
+ Ec/I,OOO + Fc/I,200 = 1.
Therefore, we can write a general equation (3)
for use with any number of antibiotics: A c / A. +
Bc/B. + Cc/C.+ ... + Xc/X. = <I for synergy,
I for additivity, or > I for antagonism.
Experimental design. Equation 3 indicates
how to measure the degree of synergy or antago-
nism for any combination of any number of
agents, but not which of the infinite number of
Figure. 6. Three-dimensional iso-
bolograms showing that points re-
presenting synergistic combinations
of three agents lie on concave
surfaces, and antagonistic combina-
tions on convex surfaces, that
join the concentrations of the three
agents that are equally effective
when used alone. In this example,
equally effective concentrations of
A, B, and C are as in figure 5.
Combination X consists of 80
/1g of A/ml, 50 /1g of B/ml, and 100
/1ft of C/ml, and the sum of the
fractional inhibitory concentrations
is 0.69. Combination Y consists
of 150 /1g of A/ml, 150 /1g of B/
ml, and 200 /1g of C/ml and the
sum is 1.46.
Synergy with Any Number of Agents
...
Figure 7. In three-agent isobolo-
c C
grams, synergy and antagonism are
usually most marked along OX'
or its extrapolate, where 0 is the
origin of the isobologram and
X' the midpoint of the additive
plane. The sum of fractional inhi-
bitory concentrations for each
combination of antibiotics A, B,
and C is shown below the point
representing that combination, and,
as in figure 4, it equals the ratio
OC/OC', where C is the point
representing the combination (e)
and C' the intersect of OC or
its extrapolate with the additive
plane (0).
combinations possible with any given number of
agents would detect interactions (synergy or an-
tagonism) most efficiently. If there were no inter-
action, equation 3 would be satisfied by a linear
isobole or isobolar surface (the term "surface"
here includes hypersurfaces in more than three
dimensions). When there is an interaction, the
isoboles are deflected from the linear condition.
The problem is therefore one of selecting the
point on a linear isobole or isobolar surface likely
to suffer the greatest deflection if synergy or an-
tagonism is present.
Intuitively, one would expect isobolar lines and
surfaces to be more or less symmetric about the
line from the origin of the isobologram to the
midpoint of the additive line or plane, and that
synergy or antagonism would therefore usually be
most marked along this line. Examination of pub-
lished isoboles for combinations of two antibiotics
shows that this is in fact usually the case; as noted
by O'Grady [4], synergy with such combinations is
generally greatest when combinations are made up
in the ratios of the respective MICs. Figures 4 and
7 illustrate this principle. The combination at the
midpoint of a two-agent additive line (point X' in
figure 4) consists of one-half of the equally effec-
tive concentration of each agent, and the combi-
nation at the midpoint of the three-agent additive
plane (point X' in figure 7) consists of one-third
of the equally effective concentrations of each.
In general, therefore, for a combination of n
agents, synergy and antagonism are most efficiently
127
+
Downloaded from http://jid.oxfordjournals.org/ at Syracuse University Library on April 6, 2014
detected by testing of the reference combination
made up of I In of the equally effective concentra-
tions of each of the agents The degree of synergy
(or antagonism) in each case is measured by test-
ing of serial dilutions (or increasing concentra-
tions) of this reference combination (i.e., by
titrating along line X'O or the extrapolate of OX'
in figures 4 and 7) until a combination X is found
that has the same effect as the equally effective
concentrations of any of the antibiotics used alone.
Dilution or concentration of the reference com-
bination X' by factor f to arrive at combination X
dilutes or concentrates each of its constituents by
the same factor; therefore, the sum of the frac-
tions in equation 3, which is I for combination
X', becomes f for combination X. In other words,
the sum of the fractions in equation 3 is given
simply by the degree to which the reference com-
bination has been diluted or concentrated.
The testing of any number of agents for synergy,
additivity, or antagonism may therefore be re-
duced to the following procedure. (1) Find the
concentration (or dose) of each agent that pro-
duces the specified effect when the agents are used
on their own. (2) Make up a reference combina-
tion of n agents consisting of lin of the above
concentrations (or doses) of each agent. (3) Ti-
trate serial dilutions or concentrations of the
reference combination to find a combination
producing the specified effect, repeating step I in
parallel with this step (see below). This procedure
gives the equally effective concentrations of the
128
agents used alone and in combination, and the
degree of synergy or antagonism is found simply
by substituting these values in equation 3. (In
the case of antagonism, one would in practice
start with a combination, perhaps, two or four
times as concentrated as the reference combina-
tion and would test serial dilutions of this com-
bination.)
Experience with this method emphasizes two
practical points (H. Gaya, personal communica-
tion). First, the assay of equally effective concen-
trations may be affected by factors such as inocu-
lum size, the constancy of which cannot beguaran-
teed from one experiment to another. Therefore,
to ensure that equally effective concentrations of
agents used alone and together are determined un-
der the same conditions, it is necessary to repeat
step 1, at least over the relevant concentration
ranges, in parallel with step 3. Second, in a check-
erboard titration it is usually easy to detect and
allow for discrepancies due to technical errors that
give anomalous results in one or a few tests out of
the whole array. As no such internal check is em-
bodied in the method described here, it demands
more than ordinarily careful technique, and
replicate assays are desirable.
Accordingly, when combinations of only two
antibiotics are tested, this procedure may save lit-
tle labor as compared with the standard checker-
board titration, but there is a clear advantage for
combinations of three agents, and the advantage
increases rapidly with the number of agents com-
bined. In testing n antibiotics in d dilutions of
each (including controls), a checkerboard titra-
tion needs d» tests; the method described here re-
quires (2n + l)(d - 1) + 2 tests, assuming a full
repeat of step 1 in parallel with step 3, and an-
other (n + l)(d - 1) + 1 tests for each subsequent
replicate assay For example, testing of five anti-
biotics together in eight dilutions (including con-
trols) would involve 32, 768 tests with the checker-
board method. With the procedure described
here, only 79 tests would be needed, with another
43 tests for each subsequent replication.
Naturally, this method is most efficient in de-
tecting synergy or antagonism when these are most
marked along the line joining the origin of the
isobologram to the point representing the refer-
ence combination. Its efficiency is not greatly im-
paired when this is not the case because, even with
Berenbaum
markedly skewed isoboles, titration along the line
joining the midpoint of the additive line or plane
to the origin will, in the vast majority of cases,
reveal the nature of the interaction unambigu-
ously.
However, a minority of isoboles are anomalous,
showing synergy with some combinations and an-
tagonism with others. A good example is illus-
trated by McAllister [5]. If the exploration of such
anomalies is required, they may be located by test-
ing combinations in the additive line or plane
other than the reference combination specified
above. One economical and symmetric arrange-
Downloaded from http://jid.oxfordjournals.org/ at Syracuse University Library on April 6, 2014
ment would be to make up combinations in which
each agent in turn is at (n + 1)/2n and each of
its partners at 1/2n of their respective equally ef-
fective concentrations. Titrations from these
points toward or away from the origin could be
made if, for instance, one or another of these com-
binations was found to be more effective than the
reference combination or showed an interaction
opposite to that of the reference combination.
Combinations including ineffective agents. For
effective antibiotics, the concentrations used in
combinations are based on the concentrations that
produce specified effects when used alone (the
equally effective concentrations as defined above).
For agents that alone are without effect on the
target organism, the concentrations used in com-
binations cannot be so based, for their equally
effective concentrations do not exist, i.e., they are
infinite. Nevertheless, such combinations may re-
quire examination, for agents that are ineffective
themselves may alter the effectiveness of other
agents. Concentrations of such agents to be used
in combinations must be based on criteria other
than equally effective concentrations. Clinical
criteria, such as the average or maximal level
found in tissues, seem appropriate, and it would
be reasonable, at least in the first instance, to
titrate the effective agents in the presence of con-
centrations of the ineffective agents fixed at these
selected levels.
For example, if three effective agents, A, B, and
C, are combined with an ineffective one, D, the
appropriate equation would be AelA e + Be/Be
+ CelC e + D.v/oo = <I for synergy, I for additiv-
ity, and > I for antagonism, where D", is a con-
centration based on clinicalcriteria and is kept
constant in titration of the c~mbination. As the
Synergy with Any Number of Agents
fraction for D equals 0, it follows from the pro-
cedure outlined in the experimental design that
the reference combination consists of one-third
(not one-fourth) of the equally effective concen-
trations of A, B, and C.
Discussion
Patients with infections that are life-threatening
or resistant to treatment with single antibiotics
tend to do better when treated with antibiotic
combinations shown in vitro to act synergistically
against strains of the offending species [6-9].
Moreover, patients treated with combinations
shown in vitro to act synergistically against the
infecting organisms isolated from their tissues
tend to do better than patients treated with com-
binations shown not to be synergistic [10-13].
There is thus no question of the advantage of
treating serious infections with antibiotic com-
binations that act synergistically against the in-
fecting organism(s). However, the clinical use of
any combination requires the prior in vitro dem-
onstration that it is synergistic, for randomly
chosen combinations of antibiotics mayor may
not be so. Indeed, some combinations are antago-
nistic, and the consequences of using these com-
binations clinically may sometimes be serious
[14, 15]. The most reliable method for demon-
strating synergy is the checkerboard titration, and,
since this method is readily carried out only with
combinations of two antibiotics, clinical practice
has consequently been almost entirely restricted
to such combinations. Empirical therapy with
combinations of four or five antibiotics has been
used successfully [16, 17], and as many as 11
agents have been used (unsuccessfully) together
[17]. However, no practicable method existed for
showing whether such combinations were syner-
gistic or not, and such modes of treatment were
therefore essentially blind.
It is inherently unlikely that the maximal de-
gree of synergy attainable with antibiotics is pres-
ent with combinations of only two, and it is possi-
ble that much higher degrees of synergy may be
present in combinations with multiple compo-
nents. The method of testing described here
should enable large numbers of such combina-
tions to be tested rapidly, reliably, and econom-
ically.
129
Much confusion exists on the subject of synergy
between antibiotics because of the Widespread
use of incorrect criteria for determining the na-
ture of antibiotic interactions. If these criteria
were to be used in examining combinations with
multiple components for synergy, incorrect con.
elusions would certainly be reached. According
to the most generally used of these incorrect cri-
teria, which appears to have been adopted by most
workers in this field, a combination is said to be
synergistic if its effect markedly exceeds the effect
of any of its constituents or, better still, the sum
of these individual effects.
Downloaded from http://jid.oxfordjournals.org/ at Syracuse University Library on April 6, 2014
Figure 1, constructed from the results of Moel-
lering et al. [3], shows the fallacy in this approach.
It was asserted by the authors [3] that some
combinations of streptomycin and penicillin were
synergistic because their effects were 100 times
greater than that of the most effective of these
agents used alone. Now suppose that 200 /Lg of
antibiotic Alml alone and 200 /Lg of antibiotic BI
ml alone each reduces the loglo cfu of a culture by
1.2 and that the combination of the two reduces
the 10glO cfu by 6.2. The effect of this combination
is therefore 105 times greater than that of either of
its constituents and> 1,000 times greater than the
sum of their effects. According to the criterion de-
scribed above, marked synergy is undoubtedly
present between A and B, but figure 1 shows that
this sort of result may be obtained when A and
B are in fact the same antibiotic, circumstances in
which synergy must by definition be absent. This
criterion cannot therefore be used to decide
whether synergy is present. In fact, as shown else-
where [2], even an antagonistic combination may
have an effect exceeding the sum of the effects of
its constituents. The possible clinical conse-
quences of using combinations alleged to be syn-
ergistic according to this criterion when they are
in fact antagonistic require no elaboration.
Another source of confusion arises from the
temptation to introduce into the definition of
synergy considerations of clinical usefulness or
statistical reliability. For instance, it is often said
that a combination of two agents is synergistic
only if the MIC of each agent in the combination
is one-fourth or less of the MIC of each agent used
alone; when the MICs of each in the combination
are one-half of their MICs used alone, the com-
bination is said to be additive. This criterion may
130
have been introduced because a mere halving of
the MIC in one test may be due to the inherent
error of titration with doubling dilutions, or be-
cause of the belief that synergy shown in vitro may
be of little use in vivo unless it is marked. Ques-
tions as to what effect should be measured (i.e.,
MIC, MBC, growth rate) and what length of ex-
posure to an antibiotic is most appropriate also
exercise some workers, because some measure-
ments may be clinically more relevant than others.
Furthermore, the fact that a combination is
synergistic in vitro does not, in itself, guarantee its
usefulness in vivo, for the combination may be
too toxic to the host or the required concentra-
tions may not be achievable. However, all of
these issues must be carefully separated from the
primary question as to whether synergy is present.
Synergy is a quantitative relation between
amounts of agents required to produce a specified
effect when used alone and in combination, and
is defined unequivocally by equation 3 above.
Whether the degree of synergy so measured is sta-
tistically significant or clinically relevant are mat-
ters for investigation by appropriate statistical,
experimental, and clinical methods, and such con-
siderations should not be allowed to obscure the
issue as to whether or not a combination is syner-
gistic.
References
I. Loewe, S., Muischnek, H. Uber Kombinationswirk-
ungen. I. Mitteilung: Hilfsmittel der Fragestellung.
Arch. Exp. PathoI. Pharmako!. 114:313-326, 1926.
2. Berenbaum, M. C. Synergy, additivism and antagonism
in immunosuppression: a critical review. Clin, Exp.
ImmunoI. 28:1-18,1977.
3. Moellering, R. C., Jr., Wennersten, C., Weinberg, A. N.
Studies on antibiotic synergism against enterococci.
I. Bacteriologic studies. J. Lab. Clin. Med. 77:821-
828,1971.
4. O'Grady, F. Co-trimoxazole: properties and thera-
peutic uses. In J. Klastersky [ed], Clinical use of com-
binations of antibiotics. Hodder and Stoughton,
London, 1975, p. 165-178.
5. McAllister, T. A. Treatment of osteomyelitis. Br. J.
Hosp. Med. 12:535-545, 1974.
Berenbaum
6. Sabath, L. D., Elder, H. A., McCall, C. E., Finland, M.
Synergistic combinations of penicillins in the treat-
ment of bacteriuria. N. Eng!. J. Med. 277:232-238,
1967.
7. Klastersky, J., Cappel, R., Daneau, D. Therapy with
carbenicillin and gentamycin for patients with can-
cer and severe infections caused by gram-negative
rods. Cancer 31:331-336,1973.
8. Klastersky, J., Daneau, D., Henri, A., Cappel, R., Hens-
gens, C. Antibiotic combinations for gram-negative
infections in patients with cancer. Eur, J. Cancer
9:407-415,1973.
9. Caya, H. Antimicrobial therapy in neutropenic pa-
tients with malignant disease. In J. Klastersky [ed.],
Clinical use of combinations of antibiotics. Hodder
and Stoughton, London, 1975, p. 117-125.
Downloaded from http://jid.oxfordjournals.org/ at Syracuse University Library on April 6, 2014
10. McCabe, W. R., Jackson, G. G. Treatment of pyelone-
phritis: bacterial, drug and host factors in success or
failure among 252 patients. N. Eng!. J. Med. 272:
1037-1044, 1965.
II. Klastersky, J., Cappel, R., Daneau, D. Clinical signifi-
cance of in vitro synergism between antibiotics in
gram-negative infections. Antimicrob. Agents Chern-
other. 2:470-475, 1972.
12. Klastersky, J., Hensgens, C., Debusscher, L. Empiric
therapy for cancer patients: comparative study of
ticarcillin-tobramycin, ticarcillin-cephalothin, and
cephalothin-tobramycin. Antimicrob. Agents Chemo-
ther. 7:640-645, 1975.
13. Klastcrskv, J., Hensgens, C., Meunier-Carpenter, F.
Comparative effectiveness of combinations of ami-
kacin with penicillin G and amikacin with carbeni-
cillin in gram-negative septicemia: double-blind
clinical tria!' J. Infect. Dis. 134 (SuppI.)S433-S440,
1976.
14. Lepper, M. H., Dowling, H. F. Treatment of pneumo-
coccic meningitis with penicillin compared with pen-
icillin plus aureomycin: studies including observa-
tions on an apparent antagonism between penicillin
and aureomycin. Arch. Intern. Med. 88:489-494, 1951.
15. Mathies, A. W., Jr., Leedom, J. M., Ivler, D., Wehrle,
P. F., Portnoy, B. Antibiotic antagonism in bacterial
meningitis. Antimicrob. Agents Chernother. 7:218-
224,1967.
16. Tattersall, M. H. N., Hutchinson, R. M., Gaya, H.,
Spiers, A. S. D. Empirical antibiotic therapy in febrile
patients with neutropenia and malignant disease.
Eur. J. Cancer 9:417-423,1973.
17. Tattersall, M. H. N., Spiers, A. S. D., Darrell,J. H. Ini-
tial therapy with combination of five antibiotics in
febrile patients with leukaemia and neutropenia.
Lancet 1:162-165, 1972.