- Email: [email protected]
The effect of phenformin on “prediabetes”
The Effect of Phenformin on “Prediahetes” By DOUGLAS L.
WILANSKY, INGE HAHN AND REUBEN SCHUCHJZR
Observations are presented on the cortisone glucose tolerance test in 568 relatives of diabetic patients. Twentyone per cent showed a positive cortisone glucose tolerance curve, including 4.4 per cent who were regarded overtly diabetic from the conventional glucose tolerance curve. A double-blind study was carried out, on the effect of a 6 week course of phenformin on cortisone-induced glucose intolerance in 60 nondiabetic relations of diabetic persons. Following placebo administration, there was a progressive decline in the number of subjects whose cortisone glucose tolerance curve was rendered normal, to 42 per cent 1 year after discontinuing the agent. With phenformin there was a progressive rise to 71 per cent. The phenformin effect was statistically significantly greater than placebo (0.05 > p > 0.02). These subjects were comprised of 2 groups. In subjects designated “prediabetics,” the conventional glucose tolerance curves were entirely normal. In “probable diabetics,” the conventional glucose tolerance curve showed some abnormality, but not sufficient to be regarded as overt diabetes.
In the probable diabetes group, 1 year following placebo marked deterioration of cortisone glucose tolerance occurred, while in the phenformin treated there was improvement. The differences were highly statistically significant (0.005 > p > 0.001). Not 1 of 7 phenformin treated persons became overtly diabetic, whereas 3 of 5 placebo treated showed overt diabetes after 1 year. Following 1 year of observation, no change occurred in the mean values of the cortisone glucose tolerance curve in a group of prediabetics. (p > .0.9). One year after placebo administration a slight fall occurred, (0.05 > p > 0.02), but following phenformin a marked and highly significant drop was observed (p < 0.001). Placebo appears to exert a significant but progressively wanning influence on the glucose intolerance of prediabetes, and no effect on probable diabetes. Phenformin results in marked amelioration of the cortisone-induced glucose intolerance of both prediabetes and probable diabetes, and appears to exert a retarding effect on the development of the diabetic syndrome.
F
AJANS, Conn and others have demonstrated that 25 per cent of close nondiabetic relatives of diabetic patients have abnormal glucose tolerance curves following administration of 2 doses of cortisone prior to the test.1-4 In long-term follow-up studies of cortisone positive and negative relatives, overt diabetes was found to emerge almost exclusively from the cortisone positive group. It was decided, therefore, to study the effect of an oral hypoglycemic agent given for a short period, upon this phenomenon. Rather than use an agent acting upon beta cell production cf insulin with the possibility of inducing an exhaustion phenomenon, we decided to use an agent acting on peripheral utilization of glucose and, thus, phenformin was chosen. The term prediabetes has recently been used in varying manner. Some would reserve drate
it for that period abnormality
in the genesis
is demonstrable
of diabetes
mellitus
before
carbohy-
even by the stress of preadministration
From the Department of Medicine, Jewish Received fog publication Feb. 23, 1965.
General
Hospital,
Montreal,
of
Canada
793 METABOLISM,VOL. 14, No. 7
(JULY), 1963
794
WILANSKY,
HAHN AND SCHUCHER
adrenal steroids, others at the time that steroid-induced glucose intolerance is demonstrable. In the light of our present incomplete knowledge it would appear that both categories may represent phases of the “prediabetic state.” Without engaging in a semantic discussion, for purposes of this publication the term “prediabetes” will be applied to relatives of diabetic patients with cortisone-induced glucose intolerance and normal conventional glucose tolerance curve. A previous preliminary report on small numbers of patients suggested that a g-week course of phenformin was capable of restoring the abnormal cortisone glucose tolerance to normal 3 months after its discontinuation.” The present paper details an annual follow-up on a larger series. Since several years must elapse before overt diabetes emerges from the prediabetic state the present data must be regarded as preliminary. METHODS Cortisone
glucose
tolerance
curves
were
performed
on relatives
of
diabetic
patients
according to Fajans and Conn’s method.2 Cortisone in doses of 50 or 62.5 mg. was given orally 8% and 2 hours prior to an oral glucose tolerance test using 100 Gm. of glucose. Family history and body weight were recorded. Any person taking a thiazidr drug or other agent known to influence blood sugar was excluded from the study. The blood sugar method used was the Auto Analyzer@ (ferricyanide reduction) method. in order to compare figures obtained with those of the Nelson-Somogyi method, duplicate determination on 35 specimens were performed. h1ean blood sugar values were found to be 9.8 mg./lOO ml. higher than by the Nelson-Somogyi method. The upper limits of normal for steroid glucose tolerance curve according to Conn (Nelson-Somogyi) are 160, 150, 140 mg./lOO ml. at 1, 11/2, and 2 hours, respectively. Values of 170, 160, and 150 were therefore used for purposes of our study using the Auto Analyzer. Fajans and Conn have emphasized the importance of the 1 and 2 hour
blood
sugar
values in the interpretation of the positive curve. 2 The lr/ hour value is helpful in doubtful cases. For the present, a cortisone glucose tolerance curve was considered positive when 2 of the 3 values, that is 170, 160, 150, mg./lOO ml., at 1, l%, and 2 hours, respectively, were exceeded. Individuals with a positive tolerance
curve
and those
steroid
whose
curve
values
were
exceeded
subjected
to a conventional
170,
and
150,
130,
at 1, 1%
glucose and 2
hours, respectively, were considered frankly diabetic. Cortisone positive diabetic relatives were then given phenformin (DBI), time disintegration 50 mg. capsules, 1 daily for 6 weeks or placebo, in double-blind manner. The cortisone glucose tolerance curve was repeated 1 week after therapy, and again 3 months and 1 year after discontinuation of therapy. On each occasion body weight was recorded and any subject showing weight change was excluded from the study. RESULTS
Table 1 indicates that since the outset of the study 568 relatives of diabetic patients have been subjected to the cortisone glucose curve. One hundred and twenty or 21 per cent showed a positive steroid curve and of this group 25 or 4.4 per cent were diabetic as defined by the conventional glucose tolerance curve. The incidence of prediabetes, then, in nondiabetic relatives of diabetic patients was 16.7 per cent. Table 2 shows that 21 per cent (25 subjects) of the individuals with abnormal cortisone glucose tolerance had overt diabetes. The ratio, therefore, of
PHJSNFORMIN
Table
EFFECT
795
ON PBEDIABETFS
l.--Incidence
of Diabetes
and Prediabetes
in a Study of Relatives Percentage
No.
Subjects studied
568
Normal cortisone glucose tolerance curve Abnormal cortisone glucose tolerance curve Diabetes Prediabetes
100
448
78.9
120
21.1
25
4.4
95
16.7
of Prediabetes and Diabetes in 120 Subjects with Positive Cortisone Glucose Tolerance Curve Percentage No.
Table 2.-Zncidence
Subjects with positive cortisone glucose tolerance curve Prediabetes Diabetes Note the ratio of diabetes in cortisone
120 95 25 positive relatives,
100 79 21 is similar to the ratio of corti-
sone positive to total diabetic relatives shown in table 1.
Table 3.-Eflect
of Phenformin on Cortisone Glucose Tolerance Curves Normal at End Therapy
Normal 3 mo. Later
NOj,,,l 1 yr. Later
Placebo
Number Percentage
16/26 61
13/26 50
10/24 42
Phenformin
Number Percentage
16/35 47
22/35 63
20/28 71
The results are at the end of therapy,
3 months
and 1 year later,
using double-blind
technic.
cortisone positive relatives to the total population of diabetic relatives is equal to the ratio of overt diabetes to cortisone positive relatives. Table 3 shows the effect of phenformin and placebo on cortisone glucose tolerance curves. The number and percentage of those becoming normal at the end of therapy, 3 months and 1 year later are shown. Placebo initially successful in causing 61 per cent (16 of 26) to revert to normal was effective in only 42 per cent (10 of 24) at 1 year. With phenformin a progressive rise in rate of success from 47 per cent (16 of 35) to 71 per cent (20 of 28) occurred. There was no significant difference in the incidence of persons with a normal curve in placebo and phenformin groups at the end of therapy (0.30 > p > 0.20) by chi square testing or 3 months later (0.50 > p > 0.30). However, 1 year following the end of therapy there was a significantly higher incidence of normals in the phenformin group (0.05 > p > 0.02). As the study proceeded we recognized that we were dealing with 2 groups of cortisone positive relatives, other than the overt diabetic. In one group, designated “probable diabetes,” some hyperglycemic values were present in the conventional glucose tolerance curves, but they were not sufficient to be
796
WILANSKY,
Table 4.-Effect
of Phenformin on Cortisone Glucose Curve in the Probable Diabetic Initial
No. of Subjects Placebo
5
HAHN
Values
1
1 ‘/
221 k x*
175 26
Values 2
1 Yr. after
1
118 2
12
AND
Tolerance Therapy
1 ‘/
2
_.___-
‘2.53 i
20
SCHUCHER
239 2 26
p
__
191 2 26
v&let
0.005 > p > U.UOl
Phenformin *Values
7 are
+Analysis Note
mean
of
that
241 i k
variance,
the
values
14
199 Ifi 9
standard 3 factor 1 year
experiments,
after
Table 5.-Cortisone
therapy
Observation
196 i
repeated are
Glucose Initial
No.of Subiects
141 2 5
II
with
Tolerance
placebo.
Values
1
1%
2
1 ‘,c, 171 -c 14
134 r
12
p >
155 t
120 t
11
0.06 > p >
0.02
109 t
6
p <
O.UUl
2O’i t
Phenformin
20
214 i
6
18X r
3
144 t
fi
188kX
11
11
142 2 8
0.90
error.
of variance. differences
drop
” valuet
210 -c 14
11
nificant
._
2
7
134 t
with
Therapy
127 t
6
1 year
phenformin.
177 -f 4 190 t
at
with
216 -t lO* 6
Significant
lower
1 Yr. after
1
213 t
standard
hut
in the Prediubetic
Curves
Values
16
f
14
~~
9
tAnslysis
104 2
measures.’
higher
Placebo
*Mean
143 i- lti
error.
3 factor
observation.
w&8 noted
experiments,
in the 3 groups With
following
repeated
are present
placebo
some
phenform:n
fall
measures.’
(0.05
>
p >
occurred
treatment
(p
0.02).
(0.05 >
>
No P
significant >
O.OZ),
change but
the
occurred most
sig-
0.001).
regarded as overt diabetes. (The 2 hour value was usually normal and the 1 and 11/2 hour values usually elevated.) In the other group designated “prediabetic,” the conventional glucose tolerance curves were entirely normal. These 2 groups were then separated for individual examination. The blood glucose values at 1, 1% and 2 hours on the cortisone glucose tolerance curves in “probable diabetes” is shown in table 4. In the placebo treated, the values 1 year after therapy are much higher at 1, 1% and 2 hours. The reverse is the case in the phenformin-treated patients, where marked lowering of values occurred at each point (table 4). The statistical analysis of these results was performed by analysis of variance adapted for use with 3 factor experiments with repeated measures as described by Wirier.“” This difference between placebo and phenformin effect is highly significant (0.005 > p > 0.001). In the placebo group none of the 5 subjects had normal cortisone glucose tolerance curves at 3 months or 1 year after administration. On the contrary, with phenformin, 6 of 9 and 3 of 7 subjects had normal curves at 3 months and subjects had 1 year, respectively. Furthermore, 3 of the 5 placebo-treated acquired overt diabetes, whereas none of the 7 phenformin-treated subjects were diabetic after 1 year. The data on groups of prediabetics handled in 3 different ways, i.e., observation (no treatment), phenformin and placebo treatment, are shown in table 5. Analysis of variance ,8 showed that there was a statistically significant decline in blood glucose levels after 1 year in the data as a whole (p < O.OOOS). When the data were examined in individual categories (i.e., when treatment x year interaction was tested) statistically significant differences were also
PHENFORMIN
EFFECT ON PREDIABETES
797
found (0.05 > p > 0.02). Subsequent t tests, showed that there was no significant difference in the observation group (p > OQO), that a significant decline occurred in placebo group (0.05 > p > 0.02) and a highly significant decline in the phenformin group (p < 0.001). DISCUSSION
The incidence of prediabetes of 16.7 per cent in this series is lower than the 25 per cent reported by Corm and Fajans, as is the incidence of diabetes of 4.4 per cent. This discrepancy may be partly explained by the fact that these authors studied close relatives, whereas the present study concerned itself with any diabetic relative. Discrepancy in age or sex incidence in the 2 series may play a part in the difference. At the moment of study, 21 per cent of the subjects showed positive cortisone glucose tolerance curves, and of these 21 per cent were diabetic. The exact significance of this observation is not clear, but the symmetry of ratios, may reflect 2 stages in time, of the genetic aspect of the dynamically evolving diabetic syndrome. While significant differences between phenformin and placebo success rates were not demonstrated by chi square testing at the end of therapy nor 3 months later, they were evident 1 year later (table 3). Furthermore, the trend with time suggests that initially the phenformin may have had an inhibitory effect on the rate of conversion of abnormal to normal cortisone glucose tolerance curve relative to placebo. It has been shown that the initially high blood insulin level in certain obese subjects may be lowered during phenformin therapy, presumably due to diminished need for pancreatic insulin release, consequent upon increased peripheral utilization of glucose induced by the phenformin. gslo One might speculate that the lower success rate with phenformin immediately after therapy in our prediabetic subjects might then reflect the effect of lowered blood insulin. The increase in numbers of normal responses by the phenformin with time cannot be fully explained. It would appear, however, that phenformin set in motion a process which resulted in continuing improvement in glucose tolerance, after its discontinuation. With placebo, the diminishing success reflected the diminishing effect of the psychologic stimulus with the passage of time. In the prediabetic a marked fall in levels of blood glucose occurred following phenformin therapy. A smaller drop occurred with placebo, and no change in the group on observation (table 5). While mean values for a group of prediabetic subjects on observation alone, showed no significant change after 1 year, the situation was different in terms of individuals. Seven of 9 subjects’ have remained continually abnormal, 2 have shown spontaneous remission and exacerbation. This does not cast doubt on the validity of this test as much as suggest, that the carbohydrate abnormality of the prediabetic may be intermittently abnormal in some subjects. That placebo has shown some ameliorative effect in the prediabetic indicates that in the early stages of the genesis of diabetes, carbohydrate intoler-
798
WILANSKY,
HAHX
AND SCHUCHER
ante may be influenced beneficially by psychologic factors. One might argue, contrariwise, that adverse psychologic stimuli may increase the carbohydrate intolerance, and constitute one factor that might accelerate carbohvdratr decompensation and the emergence of overt diabetes. At a later stage. when probable diabetes was apparent, no placebo effect was observed and deterioration of carbohydrate tolerance occurred. This suggests that at this time the% metabolic disturbance is too well established to be influenced by the relative]! small stimulus of placebo. Dohan and Lukens have demonstrated that in the experimental animal ;I short course of phloridzin or insulin, given early in the course of hypophyseal diabetes in partially pancreatectomized cats, led to a permanent remission of the diabetic state.“,’ If the above agents were given after the diabetes was long established, they were ineffective in leading to remission of the diabetes. These observations taken with the data herein reported add weight to the need for studying short-term therapy of the earliest recognizable phases of diabetes and prediabetes. The fact that phenformin may induce remission of the carbohydrate abnormality in the prediabetic for at least 1 year suggests that single or multiplt short courses of this agent may be capable of retarding the emergence and onset of clinical diabetes mellitus. The present group of subjects are being followed further to evaluate this possibility. Preliminary results from the 2 year follow-up studies indicate further divergence between phenformin and placebo effectiveness in the prediabetic. This is largely due to further loss of carbohydrate tolerance in the placebo-treated group. ACKNOWLEDGMENT We gratefully acknowledge the support of Dr. J. L. Brandt, Physician in Chief. Jewish General Hospital, and the assistance of Dr. N. Kalant, Associate Director of Research. This study was supported by grants from the Arlington Funk Laboratories Division, U. 5’. Vitamin and Pharmaceutical Corporation, New York, Montreal, thanks to Robert Wygant and Dr. Harvey Sadow. In addition, support was received from a Dominion Provincial Orant to Dr. Kalant. The cortisone was provided by Merck-Sharpe ratories, Montreal, courtesy of Dr. T. Airlie &own.
and Dohme
Statistical analyses McGill University.
hlemorial
were
performed
by
Dr.
Bernard
Grad,
Allan
Labo-
Institute,
We are also grateful to Dr. M. M. Hoffman, Associate Professor of Medicine, University, for reviewing the manuscript and for his positive suggestions.
hlcCil1
REFERENCES 1. Fajans, S. S., and Conn, J. W.: An approach to the prediction of diabetes mellitus by modification of the glucose tolerance test with cortisone. Diabetes 3:296-304, 1954. 2. -, and -: The early recognition of diabetes mellitus. N. Y. Acad. Sci. 82:208-218, 1959. 3. -: In Williams, R. H. (Ed.): Diabetes.
New York. P. B. Hoeber.
Inc..
1960.
p. 389. -1. Conn. J. W., and Fajans, S. S.: The prediabetic state. Amer. J. Med. 31: 839-850, 1961. 3. Wilansky, D. L., Hahn, I.. and Schucher. R.: The influence of phenformin on prediabetes. Appl. Therap. 4:471-175. 1962.
PHENFORMIN 6. Dohan,
F.
799
EFFECT ON PREDIABETES
C.,
Chambers,
A. H.,
and
Fish, C. A.: The metabolism of dogs with permanent diabetes produced by anterior pituitary extract. Endocrinology 28:566-579,
1941.
7. Lukens, F. D. W., Dohan, Wolcott, M. W.: Pituitary
F. C., and diabetes in
the cat: Recovery following phlorhidzin treatment. Endocrinology 32: 475487, 1943. 8. Winer, B. J.: Statistical Procedures in Experimental Design. New York, Mc-
Graw-Hill
Co.
Inc.,
1962,
pp.
9. Grodsky,
G. M., Karam,
J. H., Pavlatos,
F. C., and Forsham,
P. H.: Reduction
by
excessive
phenformin
levels
after
and diabetic 278, 10. Sadow,
of
glucose subjects.
loading
Metabolism
H. 8:
A fundamental
to hypoglycemic 12:4:333-345,
therapy.
12:
approach Metabolism
1963.
Inge Hahn, R.N., Nurse in charge, The Diagnostic Unit, Jewish General Hospital, Montreal, Canada. Schucher,
insulin in obese
1963.
Douglas L. Wilumky, M.D., F.R.C.P. (C), Assistant Physician, Department of Medicine, Jewish General Hospital, Montreal, Cam&.
Reuben
319-
330.
Ph.D., Hospital Biochemist, Hospital, Montreal, Canada.
Jewish General
WILANSKY, INGE HAHN AND REUBEN SCHUCHJZR
Observations are presented on the cortisone glucose tolerance test in 568 relatives of diabetic patients. Twentyone per cent showed a positive cortisone glucose tolerance curve, including 4.4 per cent who were regarded overtly diabetic from the conventional glucose tolerance curve. A double-blind study was carried out, on the effect of a 6 week course of phenformin on cortisone-induced glucose intolerance in 60 nondiabetic relations of diabetic persons. Following placebo administration, there was a progressive decline in the number of subjects whose cortisone glucose tolerance curve was rendered normal, to 42 per cent 1 year after discontinuing the agent. With phenformin there was a progressive rise to 71 per cent. The phenformin effect was statistically significantly greater than placebo (0.05 > p > 0.02). These subjects were comprised of 2 groups. In subjects designated “prediabetics,” the conventional glucose tolerance curves were entirely normal. In “probable diabetics,” the conventional glucose tolerance curve showed some abnormality, but not sufficient to be regarded as overt diabetes.
In the probable diabetes group, 1 year following placebo marked deterioration of cortisone glucose tolerance occurred, while in the phenformin treated there was improvement. The differences were highly statistically significant (0.005 > p > 0.001). Not 1 of 7 phenformin treated persons became overtly diabetic, whereas 3 of 5 placebo treated showed overt diabetes after 1 year. Following 1 year of observation, no change occurred in the mean values of the cortisone glucose tolerance curve in a group of prediabetics. (p > .0.9). One year after placebo administration a slight fall occurred, (0.05 > p > 0.02), but following phenformin a marked and highly significant drop was observed (p < 0.001). Placebo appears to exert a significant but progressively wanning influence on the glucose intolerance of prediabetes, and no effect on probable diabetes. Phenformin results in marked amelioration of the cortisone-induced glucose intolerance of both prediabetes and probable diabetes, and appears to exert a retarding effect on the development of the diabetic syndrome.
F
AJANS, Conn and others have demonstrated that 25 per cent of close nondiabetic relatives of diabetic patients have abnormal glucose tolerance curves following administration of 2 doses of cortisone prior to the test.1-4 In long-term follow-up studies of cortisone positive and negative relatives, overt diabetes was found to emerge almost exclusively from the cortisone positive group. It was decided, therefore, to study the effect of an oral hypoglycemic agent given for a short period, upon this phenomenon. Rather than use an agent acting upon beta cell production cf insulin with the possibility of inducing an exhaustion phenomenon, we decided to use an agent acting on peripheral utilization of glucose and, thus, phenformin was chosen. The term prediabetes has recently been used in varying manner. Some would reserve drate
it for that period abnormality
in the genesis
is demonstrable
of diabetes
mellitus
before
carbohy-
even by the stress of preadministration
From the Department of Medicine, Jewish Received fog publication Feb. 23, 1965.
General
Hospital,
Montreal,
of
Canada
793 METABOLISM,VOL. 14, No. 7
(JULY), 1963
794
WILANSKY,
HAHN AND SCHUCHER
adrenal steroids, others at the time that steroid-induced glucose intolerance is demonstrable. In the light of our present incomplete knowledge it would appear that both categories may represent phases of the “prediabetic state.” Without engaging in a semantic discussion, for purposes of this publication the term “prediabetes” will be applied to relatives of diabetic patients with cortisone-induced glucose intolerance and normal conventional glucose tolerance curve. A previous preliminary report on small numbers of patients suggested that a g-week course of phenformin was capable of restoring the abnormal cortisone glucose tolerance to normal 3 months after its discontinuation.” The present paper details an annual follow-up on a larger series. Since several years must elapse before overt diabetes emerges from the prediabetic state the present data must be regarded as preliminary. METHODS Cortisone
glucose
tolerance
curves
were
performed
on relatives
of
diabetic
patients
according to Fajans and Conn’s method.2 Cortisone in doses of 50 or 62.5 mg. was given orally 8% and 2 hours prior to an oral glucose tolerance test using 100 Gm. of glucose. Family history and body weight were recorded. Any person taking a thiazidr drug or other agent known to influence blood sugar was excluded from the study. The blood sugar method used was the Auto Analyzer@ (ferricyanide reduction) method. in order to compare figures obtained with those of the Nelson-Somogyi method, duplicate determination on 35 specimens were performed. h1ean blood sugar values were found to be 9.8 mg./lOO ml. higher than by the Nelson-Somogyi method. The upper limits of normal for steroid glucose tolerance curve according to Conn (Nelson-Somogyi) are 160, 150, 140 mg./lOO ml. at 1, 11/2, and 2 hours, respectively. Values of 170, 160, and 150 were therefore used for purposes of our study using the Auto Analyzer. Fajans and Conn have emphasized the importance of the 1 and 2 hour
blood
sugar
values in the interpretation of the positive curve. 2 The lr/ hour value is helpful in doubtful cases. For the present, a cortisone glucose tolerance curve was considered positive when 2 of the 3 values, that is 170, 160, 150, mg./lOO ml., at 1, l%, and 2 hours, respectively, were exceeded. Individuals with a positive tolerance
curve
and those
steroid
whose
curve
values
were
exceeded
subjected
to a conventional
170,
and
150,
130,
at 1, 1%
glucose and 2
hours, respectively, were considered frankly diabetic. Cortisone positive diabetic relatives were then given phenformin (DBI), time disintegration 50 mg. capsules, 1 daily for 6 weeks or placebo, in double-blind manner. The cortisone glucose tolerance curve was repeated 1 week after therapy, and again 3 months and 1 year after discontinuation of therapy. On each occasion body weight was recorded and any subject showing weight change was excluded from the study. RESULTS
Table 1 indicates that since the outset of the study 568 relatives of diabetic patients have been subjected to the cortisone glucose curve. One hundred and twenty or 21 per cent showed a positive steroid curve and of this group 25 or 4.4 per cent were diabetic as defined by the conventional glucose tolerance curve. The incidence of prediabetes, then, in nondiabetic relatives of diabetic patients was 16.7 per cent. Table 2 shows that 21 per cent (25 subjects) of the individuals with abnormal cortisone glucose tolerance had overt diabetes. The ratio, therefore, of
PHJSNFORMIN
Table
EFFECT
795
ON PBEDIABETFS
l.--Incidence
of Diabetes
and Prediabetes
in a Study of Relatives Percentage
No.
Subjects studied
568
Normal cortisone glucose tolerance curve Abnormal cortisone glucose tolerance curve Diabetes Prediabetes
100
448
78.9
120
21.1
25
4.4
95
16.7
of Prediabetes and Diabetes in 120 Subjects with Positive Cortisone Glucose Tolerance Curve Percentage No.
Table 2.-Zncidence
Subjects with positive cortisone glucose tolerance curve Prediabetes Diabetes Note the ratio of diabetes in cortisone
120 95 25 positive relatives,
100 79 21 is similar to the ratio of corti-
sone positive to total diabetic relatives shown in table 1.
Table 3.-Eflect
of Phenformin on Cortisone Glucose Tolerance Curves Normal at End Therapy
Normal 3 mo. Later
NOj,,,l 1 yr. Later
Placebo
Number Percentage
16/26 61
13/26 50
10/24 42
Phenformin
Number Percentage
16/35 47
22/35 63
20/28 71
The results are at the end of therapy,
3 months
and 1 year later,
using double-blind
technic.
cortisone positive relatives to the total population of diabetic relatives is equal to the ratio of overt diabetes to cortisone positive relatives. Table 3 shows the effect of phenformin and placebo on cortisone glucose tolerance curves. The number and percentage of those becoming normal at the end of therapy, 3 months and 1 year later are shown. Placebo initially successful in causing 61 per cent (16 of 26) to revert to normal was effective in only 42 per cent (10 of 24) at 1 year. With phenformin a progressive rise in rate of success from 47 per cent (16 of 35) to 71 per cent (20 of 28) occurred. There was no significant difference in the incidence of persons with a normal curve in placebo and phenformin groups at the end of therapy (0.30 > p > 0.20) by chi square testing or 3 months later (0.50 > p > 0.30). However, 1 year following the end of therapy there was a significantly higher incidence of normals in the phenformin group (0.05 > p > 0.02). As the study proceeded we recognized that we were dealing with 2 groups of cortisone positive relatives, other than the overt diabetic. In one group, designated “probable diabetes,” some hyperglycemic values were present in the conventional glucose tolerance curves, but they were not sufficient to be
796
WILANSKY,
Table 4.-Effect
of Phenformin on Cortisone Glucose Curve in the Probable Diabetic Initial
No. of Subjects Placebo
5
HAHN
Values
1
1 ‘/
221 k x*
175 26
Values 2
1 Yr. after
1
118 2
12
AND
Tolerance Therapy
1 ‘/
2
_.___-
‘2.53 i
20
SCHUCHER
239 2 26
p
__
191 2 26
v&let
0.005 > p > U.UOl
Phenformin *Values
7 are
+Analysis Note
mean
of
that
241 i k
variance,
the
values
14
199 Ifi 9
standard 3 factor 1 year
experiments,
after
Table 5.-Cortisone
therapy
Observation
196 i
repeated are
Glucose Initial
No.of Subiects
141 2 5
II
with
Tolerance
placebo.
Values
1
1%
2
1 ‘,c, 171 -c 14
134 r
12
p >
155 t
120 t
11
0.06 > p >
0.02
109 t
6
p <
O.UUl
2O’i t
Phenformin
20
214 i
6
18X r
3
144 t
fi
188kX
11
11
142 2 8
0.90
error.
of variance. differences
drop
” valuet
210 -c 14
11
nificant
._
2
7
134 t
with
Therapy
127 t
6
1 year
phenformin.
177 -f 4 190 t
at
with
216 -t lO* 6
Significant
lower
1 Yr. after
1
213 t
standard
hut
in the Prediubetic
Curves
Values
16
f
14
~~
9
tAnslysis
104 2
measures.’
higher
Placebo
*Mean
143 i- lti
error.
3 factor
observation.
w&8 noted
experiments,
in the 3 groups With
following
repeated
are present
placebo
some
phenform:n
fall
measures.’
(0.05
>
p >
occurred
treatment
(p
0.02).
(0.05 >
>
No P
significant >
O.OZ),
change but
the
occurred most
sig-
0.001).
regarded as overt diabetes. (The 2 hour value was usually normal and the 1 and 11/2 hour values usually elevated.) In the other group designated “prediabetic,” the conventional glucose tolerance curves were entirely normal. These 2 groups were then separated for individual examination. The blood glucose values at 1, 1% and 2 hours on the cortisone glucose tolerance curves in “probable diabetes” is shown in table 4. In the placebo treated, the values 1 year after therapy are much higher at 1, 1% and 2 hours. The reverse is the case in the phenformin-treated patients, where marked lowering of values occurred at each point (table 4). The statistical analysis of these results was performed by analysis of variance adapted for use with 3 factor experiments with repeated measures as described by Wirier.“” This difference between placebo and phenformin effect is highly significant (0.005 > p > 0.001). In the placebo group none of the 5 subjects had normal cortisone glucose tolerance curves at 3 months or 1 year after administration. On the contrary, with phenformin, 6 of 9 and 3 of 7 subjects had normal curves at 3 months and subjects had 1 year, respectively. Furthermore, 3 of the 5 placebo-treated acquired overt diabetes, whereas none of the 7 phenformin-treated subjects were diabetic after 1 year. The data on groups of prediabetics handled in 3 different ways, i.e., observation (no treatment), phenformin and placebo treatment, are shown in table 5. Analysis of variance ,8 showed that there was a statistically significant decline in blood glucose levels after 1 year in the data as a whole (p < O.OOOS). When the data were examined in individual categories (i.e., when treatment x year interaction was tested) statistically significant differences were also
PHENFORMIN
EFFECT ON PREDIABETES
797
found (0.05 > p > 0.02). Subsequent t tests, showed that there was no significant difference in the observation group (p > OQO), that a significant decline occurred in placebo group (0.05 > p > 0.02) and a highly significant decline in the phenformin group (p < 0.001). DISCUSSION
The incidence of prediabetes of 16.7 per cent in this series is lower than the 25 per cent reported by Corm and Fajans, as is the incidence of diabetes of 4.4 per cent. This discrepancy may be partly explained by the fact that these authors studied close relatives, whereas the present study concerned itself with any diabetic relative. Discrepancy in age or sex incidence in the 2 series may play a part in the difference. At the moment of study, 21 per cent of the subjects showed positive cortisone glucose tolerance curves, and of these 21 per cent were diabetic. The exact significance of this observation is not clear, but the symmetry of ratios, may reflect 2 stages in time, of the genetic aspect of the dynamically evolving diabetic syndrome. While significant differences between phenformin and placebo success rates were not demonstrated by chi square testing at the end of therapy nor 3 months later, they were evident 1 year later (table 3). Furthermore, the trend with time suggests that initially the phenformin may have had an inhibitory effect on the rate of conversion of abnormal to normal cortisone glucose tolerance curve relative to placebo. It has been shown that the initially high blood insulin level in certain obese subjects may be lowered during phenformin therapy, presumably due to diminished need for pancreatic insulin release, consequent upon increased peripheral utilization of glucose induced by the phenformin. gslo One might speculate that the lower success rate with phenformin immediately after therapy in our prediabetic subjects might then reflect the effect of lowered blood insulin. The increase in numbers of normal responses by the phenformin with time cannot be fully explained. It would appear, however, that phenformin set in motion a process which resulted in continuing improvement in glucose tolerance, after its discontinuation. With placebo, the diminishing success reflected the diminishing effect of the psychologic stimulus with the passage of time. In the prediabetic a marked fall in levels of blood glucose occurred following phenformin therapy. A smaller drop occurred with placebo, and no change in the group on observation (table 5). While mean values for a group of prediabetic subjects on observation alone, showed no significant change after 1 year, the situation was different in terms of individuals. Seven of 9 subjects’ have remained continually abnormal, 2 have shown spontaneous remission and exacerbation. This does not cast doubt on the validity of this test as much as suggest, that the carbohydrate abnormality of the prediabetic may be intermittently abnormal in some subjects. That placebo has shown some ameliorative effect in the prediabetic indicates that in the early stages of the genesis of diabetes, carbohydrate intoler-
798
WILANSKY,
HAHX
AND SCHUCHER
ante may be influenced beneficially by psychologic factors. One might argue, contrariwise, that adverse psychologic stimuli may increase the carbohydrate intolerance, and constitute one factor that might accelerate carbohvdratr decompensation and the emergence of overt diabetes. At a later stage. when probable diabetes was apparent, no placebo effect was observed and deterioration of carbohydrate tolerance occurred. This suggests that at this time the% metabolic disturbance is too well established to be influenced by the relative]! small stimulus of placebo. Dohan and Lukens have demonstrated that in the experimental animal ;I short course of phloridzin or insulin, given early in the course of hypophyseal diabetes in partially pancreatectomized cats, led to a permanent remission of the diabetic state.“,’ If the above agents were given after the diabetes was long established, they were ineffective in leading to remission of the diabetes. These observations taken with the data herein reported add weight to the need for studying short-term therapy of the earliest recognizable phases of diabetes and prediabetes. The fact that phenformin may induce remission of the carbohydrate abnormality in the prediabetic for at least 1 year suggests that single or multiplt short courses of this agent may be capable of retarding the emergence and onset of clinical diabetes mellitus. The present group of subjects are being followed further to evaluate this possibility. Preliminary results from the 2 year follow-up studies indicate further divergence between phenformin and placebo effectiveness in the prediabetic. This is largely due to further loss of carbohydrate tolerance in the placebo-treated group. ACKNOWLEDGMENT We gratefully acknowledge the support of Dr. J. L. Brandt, Physician in Chief. Jewish General Hospital, and the assistance of Dr. N. Kalant, Associate Director of Research. This study was supported by grants from the Arlington Funk Laboratories Division, U. 5’. Vitamin and Pharmaceutical Corporation, New York, Montreal, thanks to Robert Wygant and Dr. Harvey Sadow. In addition, support was received from a Dominion Provincial Orant to Dr. Kalant. The cortisone was provided by Merck-Sharpe ratories, Montreal, courtesy of Dr. T. Airlie &own.
and Dohme
Statistical analyses McGill University.
hlemorial
were
performed
by
Dr.
Bernard
Grad,
Allan
Labo-
Institute,
We are also grateful to Dr. M. M. Hoffman, Associate Professor of Medicine, University, for reviewing the manuscript and for his positive suggestions.
hlcCil1
REFERENCES 1. Fajans, S. S., and Conn, J. W.: An approach to the prediction of diabetes mellitus by modification of the glucose tolerance test with cortisone. Diabetes 3:296-304, 1954. 2. -, and -: The early recognition of diabetes mellitus. N. Y. Acad. Sci. 82:208-218, 1959. 3. -: In Williams, R. H. (Ed.): Diabetes.
New York. P. B. Hoeber.
Inc..
1960.
p. 389. -1. Conn. J. W., and Fajans, S. S.: The prediabetic state. Amer. J. Med. 31: 839-850, 1961. 3. Wilansky, D. L., Hahn, I.. and Schucher. R.: The influence of phenformin on prediabetes. Appl. Therap. 4:471-175. 1962.
PHENFORMIN 6. Dohan,
F.
799
EFFECT ON PREDIABETES
C.,
Chambers,
A. H.,
and
Fish, C. A.: The metabolism of dogs with permanent diabetes produced by anterior pituitary extract. Endocrinology 28:566-579,
1941.
7. Lukens, F. D. W., Dohan, Wolcott, M. W.: Pituitary
F. C., and diabetes in
the cat: Recovery following phlorhidzin treatment. Endocrinology 32: 475487, 1943. 8. Winer, B. J.: Statistical Procedures in Experimental Design. New York, Mc-
Graw-Hill
Co.
Inc.,
1962,
pp.
9. Grodsky,
G. M., Karam,
J. H., Pavlatos,
F. C., and Forsham,
P. H.: Reduction
by
excessive
phenformin
levels
after
and diabetic 278, 10. Sadow,
of
glucose subjects.
loading
Metabolism
H. 8:
A fundamental
to hypoglycemic 12:4:333-345,
therapy.
12:
approach Metabolism
1963.
Inge Hahn, R.N., Nurse in charge, The Diagnostic Unit, Jewish General Hospital, Montreal, Canada. Schucher,
insulin in obese
1963.
Douglas L. Wilumky, M.D., F.R.C.P. (C), Assistant Physician, Department of Medicine, Jewish General Hospital, Montreal, Cam&.
Reuben
319-
330.
Ph.D., Hospital Biochemist, Hospital, Montreal, Canada.
Jewish General