Olfactory dysfunction in diabetes mellitus

Physiology&Behavior,Vol.53, pp. 17-21, 1993

0031-9384/93 $6.00+ .00 Copyright© 1993PergamonPressLtd.

Printed in the USA.

Olfactory Dysfunction in Diabetes Mellitus R U T H S. WEINSTOCK,*~f I H E R B E R T N. W R I G H T t 2 A N D D O U G L A S U. S M I T H t

*Department of Veterans Affairs Medical Center, ?~SUNY Health Science Center, Syracuse, N Y 13210 Received 14 April 1992 WEINSTOCK, R. S., H. N. WRIGHT AND D. U. SMITH. Olfactorydysfunction in diabetesmellitus. PHYSIOL BEHAV 53(1) 17-21, 1993.--Olfactory dysfunction has been reported in individuals with diabetes mellitus, but the etiology is unknown. Diabetes is often complicated by serious medical conditions which could be related to the development of decreased olfactory ability. Overall,our 111 subjectswith diabetes showed deficienciesin their ability to identify odorants measured with the Odorant Confusion Matrix (mean = 67.8% correct). The presence of macrovascular disease was found to be associated with olfactory dysfunction. Glycemic control as well as the type and duration of diabetes were not related to olfactory ability. Also, there was no distinct association with the presence of neuropathy, retinopathy, nephropathy, hypertension, or impotence. Consistent with previous studies utilizing measures of odorant identification, performance decreased with increased age, females were somewhat superior to males, and smoking had a deleterious effect. Other nondiabetes-relatedmedical conditions and medications had no apparent effect on the olfactory ability of our subjects. These results suggest that the sequelae associated with macrovascular disease, such as perhaps, ischemia, to the olfactory area, impact negativelyon olfactory ability. Olfaction

Diabetesmellitus

Vasculardisease

METHOD

THE study of individuals with diabetes mellitus permits the investigation of physiological mechanisms related to the processing of olfactory information. Recent studies have shown that the olfactory region in rat brain is enriched in insulin receptors and insulin receptor mRNA (7,13). Given that the pathophysiology of diabetes mellitus involves decreased responsivity to insulin and that decreased olfactory ability has been found in some patients with diabetes (1,9,16), it is reasonable to hypothesize that abnormalities of receptor function in the olfactory area, if they occur in the human condition, could be expressed as disturbances in the sense of smell. It is known that insulin receptor number can decrease and function can be abnormal in the liver, muscle, and adipose tissues of subjects with diabetes (3,10). Insulin receptor number increases and insulin action can improve with the lowering of blood glucose levels by dietary, sulfonylurea, or insulin therapy (5,8,12). It is, therefore, possible that elevated blood glucose levels reflecting poor glycemic control are related to impaired olfactory ability. Diabetes mellitus is frequently accompanied by the development of several other medical conditions which also could be related to decreased olfactory ability. Notable among these conditions are neuropathy and vascular disease. Also to be considered, and controlled for, are the effects of age and sex (4), as well as smoking status (6). To examine the extent to which any of these factors contribute to impaired olfaction, we performed detailed olfactory testing in 111 subjects with diabetes mellitus in whom the glycemic control and the presence of diabetesrelated complications were documented.

Subjects Patients enrolled in the Diabetes Clinics of the Department of Veterans Affairs Medical Center and SUNY Health Science Center at Syracuse were recruited to participate in this study. Olfaction was tested at the time of routine outpatient appointments when the patients were feeling well. Excluded were patients with nasal congestion, as well as patients in whom there was concern about their nasal patency. For each subject the following information was recorded: age, sex, duration of diabetes, type of diabetes, presence of diabetes-related complications (neuropathy, retinopathy, nephropathy, impotence, and macrovascular complications such as coronary artery disease or peripheral vascular disease), presence of hypertension, thyroid disease, and other medical problems. A list of each patient's medications, a recent hemoglobin AIC measurement (an indicator ofglycemic control), and smoking status were also recorded. This information was obtained by patient history, physical examination, and chart review. Informed consent was obtained from all patients. The study was approved by the Research Subcommittees on Human Studies of the Department of Veterans Affairs Medical Center and SUNY Health Science Center at Syracuse.

Olfaction Evaluation Olfaction was quantitatively tested using the Odorant Confusion Matrix (OCM) as previously described (19). In this test procedure, ten pure chemicals diluted to represent their percep-

This work was supported in part by NIH Grant P01 DC00220 and the Department of Veterans Affairs. Requests for reprints should be addressed to Ruth S. Weinstock, M.D., Ph.D., Department of Veterans Affairs Medical Center, Endocrinology (l I 1), 800 Irving Avenue, Syracuse, NY 13210. 2 Deceased March 26, 1992.

17

18

WEINSTOCK. W R I G H I

tual attributes in e v e ~ d a y life (such as ammonia, cinnamon, etc.) are randomly presented to each observer, in one session in 10 blocks of 10 each with the constraint that when moving from one block of ten to the next, no one odorant follows itself. In addition, an initial block of 10 is presented to the observers to familiarize them with the procedure. The observers were provided an alphabetized list of the odors. Their task was to identify' the odorant presented. The observers were not made aware of the familiarization procedure, nor was any feedback provided on the accuracy of their responses. The sniff bottle technique was used. Responses by the observers were limited to one of the ten alternatives and all responses, whether correct or not, were recorded. In this study, we consider the possible impact of individual characteristics and medical conditions on the percent correct identifications (out of 100 possible) derived from the main-diagonal of the confusion matrix.

I. 2. 3. 4. 5. 6. 7.

age, macrovascular complications. sex, smoking status, age × sex interaction, age × smoking status interaction, and sex × smoking status interaction.

The resultant best model had an r 2 of 0,41, and can be expressed as:

In OR - 2.940 + 0.034X1 - 0.759X2 - 0.023X3 -- 7.186X4 + 5.824X3X4 - 0.725X~X3 + 0.432XtX4 where: In O R X~

Statistical Analyses Data analysis was divided into two stages: variable selection and linear modeling (14). Stage one consisted of determining which variables (from among the original candidate set of 17 variables) best contributed to the prediction of performance on the OCM as measured by the log-odds transformation of the proportion of correct identifications on the OCM task. Beginning with an initial set of 17 candidate variables (regressors), a backward elimination strategy was employed in which the regressor with the lowest evidence of partial correlation (F -_<2) was eliminated. This process was then continued for the remaining variables until evidence for the lowest partial correlation among the remaining (k) regressors equalled or exceeded the criterion, F >-2. Of the original 17 candidate variables, five remained after this selection process. They were: 1. 2. 3. 4. 5.

age, macrovascular complications, treatment, impotence status unknown, and smoking status.

)(2 -¥3 )(4

natural log odds ratio of correct to incorrect O C M responses age code (1 = 20-50 years, 2 = 51-60 years, 3 = 6170 years, 4 >_ 70 years) macrovaseular complications (0 = absent, 1 = present) sex (0 = female, 1 = male) smoking status (0 = not a current smoker, 1 ~ current smoker). RESULTS

The clinical characteristics of the 111 subjects who participated in this study are shown in Table 1. The majority of the subjects were from the Department of Veterans Affairs Medical Center which accounts for the large number of male subjects. Twenty-seven percent of subjects had insulin-dependent diabetes mellitus (IDDM). All of these subjects, by definition, had an absolute deficiency in insulin production, were ketosis-prone TABLE 1 CHARACTERISTICSOF SUBJECTS Percentage Sex

Given this candidate set of predictor variables, first-order interaction variables were formed and subjected to the same procedure described in the preceding paragraph. Higher interaction terms were then formed from those obtained in this step, and the procedure repeated. In addition to the aforementioned set of five favored variables, two first-order interactions among this set satisfied the inclusion criterion. They were:

Age (years)

1. age and impotence status unknown, and 2. impotence status unknown and smoking status.

Duration of diabetes (years)

The coefficient of determination for these seven regressors was r 2 = 0.37. This model served as our basic regression model which was then subjected to further testing. Given the basic model, estimates of regression coefficients and their standard errors were then obtained by adding each of the eliminated regressors to the model described in the preceding paragraph. Estimates were also obtained for the seven variables included in the basic regression model. Identification of strong predictors was based on those regressors whose 95% confidence intervals about the estimated regression coefficient excluded zero. The resulting enhanced model included seven variables. They were:

,kN1) S M I I H

Smokers

Type of diabetes

Insulin treatment Glycemic control (hemoglobin A ~ C )

Male Female 20-50 51-60 61-70 >70 Yes No Unknown IDDM* NIDDMt < 1- 10 11- 19 >20 Yes No~c Excellent (~6.1%) Suboptimal (6.2%-9.0%) Poor (>9.0%)

86.5 13.5 25.2 26.1 39.6 9.0 18.0 36.9 45.0 27.0 73.0 34.2 44.7 18.0 77.5 22.5 9.9 57.7 32.4

n = 111. * IDDM: Insulin-dependent diabetes meUitus. t NIDDM: Noninsulin-dependent diabetes. :~ Noninsulin-dependent diabetes mellitus subjects: Treated with oral hypoglycemic agents (sulfonylurea drugs): 20.7%; treated with diet alone: 1.8%; total: 22.5%.

OLFACTION IN DIABETES MELLITUS and required insulin therapy. The rest of the subjects had noninsulin-dependent diabetes mellitus (NIDDM) and, therefore, had varying degrees of insulin deficiency and insulin resistance. Many of the NIDDM patients were treated with insulin. Measurements of hemoglobin A~C were obtained on all subjects. This assay is a well-established index of glycemic control and reflects the patient's mean blood glucose levels over the previous couple of months (11,18). Ten percent of subjects had normal hemoglobin A~C measurements reflecting excellent glycemic control, whereas 58% had suboptimal and 32% had poor glycemic control (Table 1). Of the 61 subjects for whom smoking history was available, approximately one-third were smokers. Patients with diabetes mellitus are at risk for the development of several serious medical conditions. The microvascular complications of diabetes include peripheral and autonomic neuropathy, retinopathy, and nephropathy. The macrovascular complications include coronary artery disease and peripheral vascular disease. Fifty-one percent of our subjects had clinical evidence of neuropathy, and 60% had retinopathy or nephropathy (Table 2). Evidence of macrovascular disease was found in 39% of subjects. Hypertension was present in 39% of subjects as well (Table 2). Eighty-two percent of the subjects also had a history of nondiabetes-related medical conditions and 74% were taking other medications in addition to their insulin or oral hypoglycemic agent (Table 2). Olfactory ability was assessed using the odorant confusion matrix as previously described (19). Overall, olfactory ability was impaired in our study group (mean percent correct identifications = 67.8% as compared to >80% in normal subjects (19)). When results of olfaction testing were analyzed by subject characteristics and associated medical conditions, olfactory ability most strongly correlated with age and macrovascular disease (Table 3). The oldest subjects (>70 years old) had a correct odorant identification response rate of 52%, whereas the correct response rate observed in younger subjects (20-50 years old) was 81% (Table 3). Subjects with macrovascular disease also had impaired olfaction compared to subjects without coronary artery or peripheral vascular disease (59% vs. 74% correct responses, respectively).

19 Natural log odds ratio analyses confirmed that age and presence of macrovascular disease were distinct indicators of olfactory deficiencies. Smoking status was a weakly distinct indicator, whereas insulin treatment was an indistinct indicator. Glycemic control (as reflected by the hemoglobin AjC level), presence of neuropathy or other microvascular complications, and the presence of nondiabetes-related medical problems and medications were not found to be indicators of olfactory ability. DISCUSSION Many patients with diabetes mellitus have impaired olfaction, the physiological basis of which is unknown (9). Individuals with diabetes are at increased risk for the development of multiple medical complications, including macrovascular disease. In this series of 111 subjects with diabetes, the presence of coronary artery or peripheral vascular disease correlated with olfactory dysfunction. Vascular disease with ischemia to the olfactory area could account, at least in part, for the disturbances in the sense of smell reported in our group of patients. The olfactory area in the rat is particularly rich in insulin receptors and insulin receptor mRNA (7,13). The presence of these receptors suggests that insulin may play a role in olfactory processes. In our series, the use of insulin in the treatment of diabetes was an indistinct indicator of olfactory ability, and the use of oral sulfonylurea drugs, which can augment insulin secretion, was a poor indicator. The adequacy of the insulin or sulfonylurea therapy, however, is best assessed by examining the subject's glycemic control. Normalization of blood glucose profiles improves insulin receptor number and function in peripheral tissues (5,8,12). If insulin receptors in the olfactory region play a role in olfactory transduction and are abnormal in inadequately controlled diabetes, poor glycemic control would be expected to be associated with olfactory dysfunction. We did not, however, find a correlation between glycemic control and olfactory ability. We assessed glycemic control by the use of the hemoglobin A jC assay. This is a well-established index of mean blood glucose levels, and is the major measure used in trials to evaluate the

TABLE 2 MEDICALCONDITIONSPRESENT IN STUDY SUBJECTS Percentage MedicalConditions

Present

Diabetes-Related Conditions Neuropathy Retinopathy or nephropathy Coronary artery or peripheral vascular disease Impotence Hypertension Other Medical conditions other than diabetes mellitus Prescribed medications other than insulin and oral hypoglycemic agents n = 111. * Number of medications prescribed: percentage:

1-2 42.3

Absent

51.4 60.4

48.6 39.6

38.7 25.2 38.7

61.3 65.8 61.3

82.0

18.0

73.8*

26.2

3-4 17.1

5-6 12.6

Unknown

9.0

>6 1.8

total 73.8

20

WEINSTOCK. W R I G H I

.\NI) SMII'Ii

TABLE 3 RESULIS OF OLFACTORY 7FESTING Percent Correct Subject Characteristic Sex Male Female Age (years) 20-50 51-60 61-70 71+ Duration of diabetes (years) 9.0'~.~

Mean ± SD

66.0 ± 19.7 78.5 ± 23.8 80.5 71.9 60.4 51.5

± 18.7 _+ 14.5 _+ 20.1 ± 18.9

67.7 ± 25.8 69.6 _+ 16.6 62.8 ± 18.6 67.4 ± 21.9 67.9 _+ 20.3 67.7 ± 20.8 67.9 ± 20.5 65.3 +_ 25.6 66.0 ± 20.7 70.8 ± 18.5

association of glycemic control with diabetic complications (2,11). This assay reflects mean glucose concentrations over the previous couple of months (18). If metabolic control is important in the development of diabetes-related medical complications, an assay which reflects glucose levels over longer periods of time might also be useful, but such an assay is not available. Other investigators, however, have found hemoglobin A~C levels useful in the study of diabetes of long duration. For example, Orchard and co-workers (15) studied patients with I D D M of greater than 25 years duration and found that lower hemoglobin A~C levels correlated with a lower prevalence of diabetes-related complications. In our series, the hemoglobin AjC values did not correlate with olfactory ability. Our patients with the longest duration of diabetes (and presumably the longest duration of hyperglycemia) also did not have a more marked decrease in their sense of smell than those in whom the diabetes was of relatively recent onset. These findings, again, do not support a direct association between glycemic control and olfactory function. The severity and rate of progression of the microvascular complications of diabetes have been related, at least in part, to the degree of hyperglycemia [for review see (17)]. In our series,

Percellt COlTeC{

Medical Conditions Diabetes Related Conditions Neuropathy Present Absent Retinopathy or nephropathy Present Absent Coronary artery or peripheral vascular disease Present Absent Impotence Present Absent Unknown Hypertension Present Absent Nondiabetes-Related Conditions Smoking Present Absent Unknown Other medical problems Present Absent No. of medications None I-2 3-4 5-6 >6

Mean + SD

65.1 + 20.3 70.6 ± 20.8 67.9 _+ 18.7 67.6 + 20.2

58.6 _+ 18.6 73.7 _+ 19.9 66.8 _+ 17.3 68.0 _+ 22.0 67.1 ± 19.2 65.0 ± 20.7 69.5 + 20.6

63.5 _+ 17.9 70.0 _+ 20.8 67.4 ± 21.4 66.9 ± 20.7 71.7 _+ 20.5 73.1 67.0 65.5 62.9 64.7

± 21.8 + 19.1 _+ 23.8 ± 17.0 ± 19.6

the presence of microvascular disease in the retina and kidney also did not correlate with olfactory ability. This is consistent with the lack of association between glycemic control, duration of diabetes, and results of olfaction testing. It has been postulated that defects in olfactory nerve function are responsible for the impaired sense of smell in subjects with diabetes (9). We found no relationship between the presence of clinically apparent neuropathy (another microvascular complication of diabetes) and olfactory dysfunction. We cannot exclude the possibility, however, that subjects with diabetes have abnormalities in their olfactory nerves that are not associated with the presence of generalized neuropathy. A decrease in the ability to identify odorants with increasing age has been well documented. Such a decrease has been found to occur more rapidly for males than females (4). Cigarette smoking is also associated with decrements in the ability to identify odorants (6). We should note that the measure of odorant identification in these studies was the University of Pennsylvania Smell Identification Test (UPSIT) in which the observers are presented with a 40-item four-alternative forced choice task. Our confusion matrix test (OCM) procedure for odorant identification is a replicated ten-alternative forced choice task, permitting

O L F A C T I O N IN DIABETES M E L L I T U S

21

within-observer multidimensional analysis of the off-diagonal responses. Nevertheless, the results from the O C M and UPSIT, at least with respect to percent correct identification, correlate well (19). Thus, the results from the present study with the O C M showing decrements in olfactory ability with an increase in age and with smoking and with males performing less well than females, are consistent with previous studies with the UPSIT. In summary, we found that impaired olfactory ability in our subjects with diabetes mellitus was most closely associated with increasing age and the presence of macrovascular disease. Greater impairment was also seen in males and in current smokers. After taking the joint influence of age, macrovascular disease, sex, and

smoking status into account, glycemic control, duration of diabetes, and the presence of neuropathy did not correlate additionally with the results of olfaction testing. Our findings suggest that the sequelae associated with vascular disease contribute to the olfactory dysfunction observed in patients with diabetes mellitus. Whether such disease contributes to olfactory disabilities in individuals without diabetes remains speculative and is an area under active investigation. ACKNOWLEDGEMENTS We thank Paul Sheehe, Sc.D. for his assistance in the preparation of this manuscript and Denise Fryer for her technical assistance.

REFERENCES 1. Borgogna, E.; Moniaci, D.; Barioglio, M. G.; Benzi, M.; Cantino, G.; Pettiti, G. Esame olfattometrico per odori olfattogustativi in pazienti diabetici. Min. Stom. 32:39-41; 1983. 2. DCCT Research Group. Diabetes control and complications trial (DCCT) update. Diabetes Care 13:427-433; 1990. 3. Defronzo, R. A.; Ferrannine, E.; Koivisto, V. New concepts in the pathogenesis and treatment of non-insulin-dependent diabetes mellitus. Am. J. Med. 74:52-81; 1983. 4. Doty, R. L.; Shaman, P.; Applebaum, S. L.; Giberson, R.; Siksorski, L.; Rosenberg, L. Smell identification ability: Changes with age. Science 226:1441-1443; 1984. 5. Foley, J. E.; Kashiwagi, A.; Verso, M. A.; Reaven, G.; Andrews, J. Improvement in in vitro insulin action after one month of insulin therapy in obese noninsulin-dependent diabetics. Measurements of glucose transport and metabolism, insulin binding, and lipolysis in isolated adipocytes. J. Clin. Invest. 72:1901-1909; 1983. 6. Frye, R. E.; Schwartz, B. S.; Doty, R. L. Dose-related effects of cigarette smoking on olfactory function. JAMA 263:1233-1236; 1990. 7. Hill, J. M.; Lesniak, M. A.; Pert, C. B.; Roth, J. Autoradiographic localization of insulin receptors in rat brain: Prominence in olfactory and limbic areas. Neuroscience 17:1127-1138; 1986. 8. Hjollund, E.; Pedersen, O.; Richelson, B.; Beck-Nielsen, H.; Sorensen, N. S. Increased insulin binding to adipocytes and monocytes and increased insulin sensitivity of glucose transport and metabolism in adipocytes from non-insulin-dependent diabetics after low-fat/ high-starch/high-fiber diet. Metabolism 32:1067-1075; 1983. 9. Jorgensen, M. B.; Buch, N. H. Studies on the sense of smell and taste in diabetics. Acta Otolaryngol. 53:539-545; 1961.

10. Kahn, C. R.; White, M. F. The insulin receptor and the molecular mechanism of insulin action. J. Clin. Invest. 82:115 l - l 156; 1988. I I. Koenig, R. J.; Peterson, C. M.; Jones, R. L.; Saunder, C.; Lehrman, M.; Cerami, A. Correlation of glucose regulation and hemoglobin AIC in diabetes mellitus. N. Engl. J. Med. 295:417-420; 1976. 12. Kolterman, O. G.; Olefsky, J. M. The impact of sulfonylurea treatment upon the mechanisms responsible for the insulin resistance in type II diabetes. Diabetes Care 7:81-88; 1984. 13. Marks, J. L.; Porte, D., Jr.; Stahl, W. L.; Baskin, D. G. Localization of insulin receptor mRNA in rat brain by in situ hybridization. Endocrinology 127:3234-3236; 1990. 14. Mozell, M. M.; Sheehe, P. R.; Swieck, S. W., Jr.; Kurtz, D. B.; Hornung, D. E. A parametric study of the stimulation variables affecting the magnitude of the olfactory nerve response. J. Gen. Physiol. 83:233-267, 1984. 15. Orchard, T. J.; Dorman, J. S.; Maser, R. E.; Becker, D. J.; Ellis, D.; LaPorte, R. E.; Kuller, L. H.; Wolfson, S. K., Jr.; Drash, A. L. Factors associated with avoidance of severe complications after 25 years of IDDM. Diabetes Care 13:741-747; 1990. 16. Patterson, D. S.; Turner, P.; Smart, J. V. Smell threshold in diabetes mellitus. Nature 5023:625; 1966. 17. Rosenstock, J.; Raskin, P. Diabetes and its complications: Blood glucose control versus genetic susceptibility. Diabetes Metab. Rev. 4:417-435; 1988. 18. Svendsen, P. A.; Lauritzen, T.; Soegaard, V.; Nerup, J. Glycosylated haemoglobin and steady-state mean blood glucose concentration in Type l (insulin-dependent)diabetes. Diabetologla 23:403--405; 1982. 19. Wright, H. N. Characterization of olfactory dysfunction. Arch. Otolaryngol. Head Neck Surg. 113:163-168; 1987.