Intact taste and smell function is important in HIV-infected individuals because chemosensory abnormalities may contribute to inadequate food intake leading to malnutrition, weight loss, and ultimately wasting. Caloric intake and subsequent nutritional status may be affected by impaired taste and smell functioning [1–3]. The perceptions of taste and smell normally lead to the initiation of cephalic phase reflexes, a series of physiologic events that prepare the gastrointestinal tract to digest food. Cephalic phase reflexes result in increased salivary secretion, increased gastric secretion of pepsinogen and hydrochloric acid, augmented pancreatic endocrine and exocrine function, and improved glucose metabolism [4,5]. Thus, the senses of smell and taste can potentially play a role in the utilization of nutrients, and the chemosensory losses and distortions that occur in HIV-infected patients may contribute to their impaired nutritional status.
Chemosensory losses in HIV-infected individuals may result from a variety of pathophysiologic processes that affect taste and smell receptors or the peripheral and central nervous systems. These processes may include oral pathology, opportunistic infections, neoplasms, and HIV-associated neurological disease. Medications also play a major role in taste losses because the majority of the drugs prescribed to HIV-infected patients have been associated clinically with taste disorders. Antimicrobial drugs, including amphotericin B, ampicillin, metronidazole, and tetracyclines can cause a loss or distortion of taste [1,6]. Nebulized pentamidine, which is commonly used as prophylaxis against Pneumocystis carinii pneumonia, can cause a metallic taste .
There are scattered reports of taste and smell dysfunction in the HIV-infected population, but the prevalence of these abnormalities is unknown. Brody et al.  measured odor identification in 42 HIV-infected patients along a continuum of progressive immunodeficiency and 37 healthy age- and sex-matched individuals. The subjects were administered the University of Pennsylvania Smell Identification Test (UPSIT), which consists of common odorants and evaluates the ability to identify each odor from a four-item word list. Patients (both asymptomatic and symptomatic) scored significantly lower than control subjects. Hornung et al.  also used the UPSIT to test smell identification in 50 HIV-infected patients and concluded that 32% (16 out of 50) were either hyposmic (had reduced sense of smell) or anosmic (had no sense of smell) according to the norms for the test. The reduced olfactory perception was not related to age, presence of symptoms, presence of fatigue, HIV risk factors or CD4+ lymphocyte count. Razani et al.  found an elevated odor threshold for butanol in 15 patients with AIDS dementia complex when compared with 15 HIV-infected patients without AIDS dementia complex and 19 HIV negative subjects. Lehrner et al.  performed threshold testing with butanol, odor identification testing and odor memory testing on 18 HIV-infected subjects with CD4 cell counts of 240–700 × 106/l, 19 HIV-infected subjects with CD4 cell counts below 170 × 106/l, and 18 controls. He found mild olfactory impairment in HIV-infected persons that seemed to increase with progression of disease. In a study of 25 healthy HIV-infected subjects with CD4+ lymphocyte counts below 100 × 106/l by Mattes et al. , 72% of subjects reported chemosensory alterations including increased sensitivity in 32%, decreased sensitivity in 32% and distortions in 32%. Seventy per cent of patients with taste disorders indicated that medication exacerbated their problem. However, no significant differences in taste identification ability or intensity ratings were observed between patients and controls. Three HIV-infected subjects had markedly decreased odor identification scores, but the group mean was similar to controls.
In a recent study in our laboratory , we found abnormalities of both taste and smell function in HIV-infected subjects. Taste and smell function was measured in 40 HIV-infected individuals and 40 healthy control subjects matched for age, sex, race, smoking behavior, and number of years of education. Detection thresholds for glutamic acid, quinine hydrochloride, and menthol were increased in the HIV-infected subjects. A significant correlation was not established between taste and smell function and Centers for Disease Control and Prevention (CDC) classification of HIV infection , although trends were observed suggesting worsening function with progression of disease.
This study was performed to delineate the scope of taste and smell complaints in HIV-infected patients at the Duke University Infectious Diseases Clinic (DUIDC), to evaluate the factors associated with taste and smell complaints, and to determine the impact on quality of life.
The survey was conducted at DUIDC between 27 February 1997 and 22 May 1997, and consisted of four parts: (i) a self-administered questionnaire about the senses of taste and smell; (ii) a self-administered quality of life questionnaire, the 30-item Medical Outcomes Study HIV Health Survey (MOS-HIV-30) [15,16], (iii) a targeted patient interview designed to elicit a brief medical history and the patient's current medical regimen; and (iv) a brief chart review. Typically, the clinic receptionist would introduce each HIV-infected patient to the investigator (A.E.H.) while the patient was waiting to see his/her physician. The investigator would explain the survey and obtain written informed consent. Patients who were willing to complete the survey then underwent the 5 min targeted interview and were given the taste and smell questionnaire and quality of life questionnaire to complete prior to leaving the clinic. Assistance in completing the questionnaires was offered to patients who had difficulty reading or writing. Chart review was performed later, after laboratory values had returned.
The taste and smell questionnaire was developed by the investigators based on their clinical experience. It was pilot-tested on a group of 10–15 patients. Vague questions were reworded, and the questionnaire was administered to the study population. The taste section of the taste and smell questionnaire consisted of nine questions addressing changes in the sense of taste, changes in the way a food tastes, presence and quality of a bad taste in the mouth, changes in specific taste qualities (salt, sweet, sour, bitter), effect of medications on the sense of taste, and a rating of the overall severity of taste abnormalities. A taste complaint score, ranging from 0 (no taste complaints) to 10 (many taste complaints), was calculated by adding one point for each taste complaint and two points for a rating of ‘severe’ on the overall rating of taste abnormalities. Patients were also asked which medications tasted bad, but this question was not included in the taste complaint score, because individuals with normal taste function would also find that some medications taste bad.
The smell section of the taste and smell questionnaire consisted of five questions addressing changes in the sense of smell, changes in the way food smells, the effect of medications on the sense of smell, changes in the strength of odors, and a rating of the overall severity of smell abnormalities. A smell complaint score, ranging from 0 (no complaints) to 6 (many smell complaints), was calculated by adding one point for each smell complaint and two points for a rating of ‘severe’ on the overall rating of smell complaints. Patients were also asked which medications smelled bad, but this question was not included in the smell complaint score, because individuals with normal smell function would find that some medications smell bad. The chemosensory score was calculated by adding the taste complaint score and the smell complaint score.
The quality of life questionnaire consisted of the MOSHIV-30. This instrument was adapted from the health status instruments used in the Medical Outcomes Study  and has been extensively validated in HIV-infected persons [15,16]. It measures quality of life on a scale of 0–100 in multiple domains including general health perception, physical function, role function, social function, cognitive function, pain, mental health, and health distress. Summary scores of physical and mental health were also calculated .
The targeted patient interview was conducted to elicit the patient's current medication regimen, history of HIV-related problems, and history of general medical problems such as nasal polyps, chronic sinusitis, seasonal allergies, frequent upper respiratory tract infections, diabetes, hypertension, severe head injury, occupational exposure, and extensive dental work that might predispose to taste/smell abnormalities. A brief chart review was performed to confirm the facts elicited on patient interview and to obtain the most recent CD4 cell count and HIV-1 viral load.
Taste and smell testing
Formal taste and smell testing was performed on a subset of 50 patients. These subjects were recruited from the 207 patients who completed the taste and smell survey. Surveyed subjects were invited to return to the taste and smell laboratory on another day for formal taste and smell testing after completing the targeted interview. Priority was given to patients who lived in the area and had more free time. Subjects were paid US$ 20 to defray the cost of travel and parking.
Taste detection and recognition thresholds were measured using a two alternative forced choice ascending series method. Subjects were presented with a pair of stimuli, one being the tastant (sucrose, sodium chloride, citric acid, or quinine hydrochloride) dissolved in deionized water (the test stimulus), and the other being the deionized water alone (the control). The position of the test stimulus was randomized across all pairs. Subjects started at the lowest concentration and progressed to increasingly stronger concentrations. The subject's task was to determine which of the two stimuli tasted stronger, and to describe the quality of the stronger stimulus if possible. Fifteen milliliters of each taste stimulus were presented in half-filled 30 ml plastic medicine cups. After tasting each pair, subjects rinsed their mouths with deionized distilled water and expectorated. The subject's detection threshold was considered to be the first in a series of five consecutive correctly identified test stimuli. Similarly, the subject's taste recognition threshold was the first in a series of five consecutive correctly recognized test stimuli. The taste threshold stimuli were separated by a dilution factor of 2. The concentration ranges for the target stimuli were as follows: sucrose, 3.125 × 10−4 M up to 1.6 × 10−1 M; sodium chloride, 3.125 × 10−4 M up to 1.6 × 10−1 M; citric acid, 6.10 × 10−6 M up to 3.125 × 10−3 M; quinine hydrochloride, 6.25 × 10−7 M up to 3.2 × 10−4 M.
Olfactory function was measured using the odor triad test. The subject was presented odor stimuli in triads in which two of the smells were identical and a third smell was different. Fifteen milliliters of each stimulus were presented in 30 ml amber bottles with a cotton ball and an odor-tight lid. The smell stimuli were presented to the subject at the level of the nostrils, and the instruction was given to the subject to give all smell stimuli a ‘whiff’ of equal magnitude. The subject was asked to determine which of the smells was different from the other two. The subject was also asked to describe or identify, if possible, the ‘different’ smell as well as the other two smells. Fifteen triads were presented to each patient. The odor triad score could range from 0 to 15 and was the number of triads in which the subject correctly identified the different smell. The concentrations of the odor stimuli used in the odor triad test were as follows: menthol, 3.16 × 10−2 g/ml; benzaldehyde, 1.06 × 10−2 ml/ml; n-butanol, 4.6 × 10−3 ml/ml; caproic acid, 1.26 × 10−2 ml/ml; citronellal, 5.4 × 10−3 ml/ml; guaiacol, 2.14 × 10−2 ml/ml; geraniol, 2 × 10−2 ml/ml; citral, 2 × 10−2 ml/ml; methyl salicylate, 1.48 × 10−2 ml/ml. All of the concentrations listed above for the odorants were found to be isointense after extensive pretesting.
Simple descriptive statistics were used to summarize the prevalence, character, and severity of taste and smell abnormalities in the study population. χ2 analysis was used to compare categorical variables such as gender, race, and CDC clinical stage. The Wilcoxon rank-sum test was used to compare variables such as age, body mass index, CD4 cell count, HIV-1 viral loads, detection and recognition thresholds, and odor triads.
Multivariate stepwise analysis was performed using SAS software version 6.10  to determine which factors predicted the overall chemosensory score. Dummy variables were used to code for gender, race, and tobacco use. An interval scale was used to summarize the CDC clinical stage. Age, number of AIDS diagnoses, number of medications, number of protease inhibitors, body mass index, CD4 cell count, and log-transformed viral load were entered as continuous variables. A term for an interaction between the number of medications and the CDC clinical stage was also included. The model was validated by repeating the analysis on 100 bootstrap samples .
Quality of life scores were compared between subjects with and without chemosensory complaints using the Wilcoxon rank-sum test. Further analysis of the effect of chemosensory complaints on quality of life was performed by using a regression model to control for number of AIDS diagnoses, number of medications taken CD4 cell count and HIV-1 viral load.
During the study period, 790 patients were being followed at DUIDC. Two-hundred and eighteen patients were approached in the waiting room about completing the survey. Seven patients refused. Four patients agreed, but did not fill out the survey completely enough to be included in the analysis. The demographics of the 207 patients who completed the survey were compared with the 583 patients who were not surveyed (Table 1). Patients surveyed were similar to those who were not surveyed with respect to gender, race, age and viral load. Subjects who were surveyed were more likely to have a more advanced stage of disease and a lower CD4 cell count than those who were not surveyed. The most likely explanation for this discrepancy is that patients with more advanced stages of disease and lower CD4 cell counts are seen more frequently in the clinic, and were therefore more likely to be approached about completing the survey during the 3-month survey period.
One hundred and forty-four patients [70%; 95% confidence interval (CI), 64–76] surveyed reported chemosensory complaints. Ninety-one (44%; 95% CI, 37–51) had both taste and smell complaints, 47 (23%; 95% CI, 17–28) reported only taste complaints, and six (3%; 95% CI, 1–5) reported only smell complaints. The remaining 63 (30%) reported no taste or smell complaints. Taste complaint scores ranged from 0 to 10 (Fig. 1A). The most common taste complaints were food tasting different from usual for 61 patients (30%), an overall change in the sense of taste for 57 patients (28%), and alterations in the taste of salt for 72 patients (35%), alterations in the taste of sweet for 52 patients (25%), alterations in the taste of sour for 52 patients (25%), and alterations in the taste of bitter for 49 patients (24%). Patients rated their abnormal sense of taste as moderate in 55 (27%) of cases, and severe in seven (3%) cases.
Smell complaint scores ranged from 0 to 6 (Fig. 1B). The most common smell complaints were a change in the ability to detect odors for 65 patients (31%), an overall change in the sense of smell for 33 patients (16%), and food smelling different from usual for 29 patients (14%). Patients rated their abnormal sense of smell as moderate in 46 (22%) cases, and severe in 11 (5%) cases.
Forty-three patients (21%) reported that medications interfered with their sense of taste, and five (2%) reported that medications interfered with their sense of smell. Seventy-nine (38%) reported that medications taste bad, and 43 (21%) reported that medications smelled bad. Patients could report problems with medications they were currently taking, or medications they had previously taken. The medications which were most commonly listed as problematic were indinavir (Crixivan), ritonavir (Norvir), clarithromycin (Biaxin), zidovudine (Retrovir), didanosine (Videx), lamivudine (Epivir) and trimethoprim–sulfamethoxazole (Bactrim or Septra). Although we did not have information about how many patients had ever taken each medication, it appeared that the number of complaints was disproportionately high for protease inhibitors.
Table 2 compares the characteristics of the 144 patients with chemosensory complaints with the 63 patients without chemosensory complaints. Bivariate analysis showed that chemosensory complaints were significantly associated with taking a greater number of medications, more advanced CDC clinical stage, increased number of AIDS diagnoses, lower CD4 cell count, female gender, a history of oral candidiasis, and higher HIV-1 viral load.
Multivariate analysis was used to generate a model to predict the overall chemosensory complaint score (Fig. 1C). The factors summarized in Table 3 were entered into a multivariate linear regression model to predict the overall chemosensory complaint score. The dummy variables for tobacco use were considered in a block for the model, as were the dummy variables for race. Stepwise elimination of variables using P = 0.05 for entry into the model and P = 0.15 for retention in the model revealed that the number of medications taken, hay fever, and tobacco use provided the best predictive model.
The multivariate model was validated using bootstrapping techniques. One hundred bootstrap samples were generated, and the above-mentioned stepwise regression analysis was performed on each new set (Table 3). The number of medications appeared in the model 85% of the time, hay fever 75% of the time, and tobacco use 65% of the time. The chemosensory score increased with the number of medications taken (mean β = 0.35), the presence of hay fever (mean β = 1.51) and tobacco use (mean β = 1.17 for current smokers and mean β = 2.21 for former smokers).
Quality of life scores for patients with taste/smell complaints and for patients without complaints are summarized in Table 4. Chemosensory complaints were associated with decreased quality of life in all domains. After controlling for CD4 cell count, HIV-1 viral load, number of AIDS diagnoses, and number of medications taken, high chemosensory score was still significantly associated with poor quality of life as measured in all domains as well as by mental health summary (P = 0.0001) and physical health summary (P = 0.0001).
Fifty subjects also underwent chemosensory testing. These subjects were similar to the 157 subjects who did not undergo chemosensory testing with regard to demographic factors, CDC clinical stage, number of medications taken, CD4 cell count and HIV-1 viral load (data not shown). Similar proportions of subjects had chemosensory complaints; 35 patients (70%) who underwent chemosensory testing and 109 patients (69%) who did not undergo chemosensory testing had chemosensory complaints. Quality of life scores were similar in the two groups.
Of the 50 subjects who underwent chemosensory testing, 32 (64%) had taste complaints. The median taste complaint score for subjects who had complaints was 4. Detection and recognition thresholds for the four primary senses (sweet, salty, sour, bitter) are summarized in Table 5. There were no significant differences between subjects with and without taste complaints in the thresholds for any of the tastants using the Wilcoxon rank-sum test. Threshold levels for both groups were similar to previously published values for healthy subjects . Twenty six (52%) patients had smell complaints. The median smell complaint score was for subjects who had complaints was 1. No statistically significant difference in the odor triad score was found between patients with and without smell complaints; the median score for both groups was 12 out of a maximum score of 15.
This study indicates that a substantial fraction of HIV-infected patients seen at DUIDC have chemosensory complaints. Almost three-quarters of patients surveyed reported chemosensory complaints, and almost 50% of patients reported both taste and smell complaints. Previous reports have demonstrated taste and smell abnormalities in HIV-infected persons [8–13], but this study is the first to demonstrate the pervasiveness of chemosensory complaints. Chemosensory abnormalities are important because of their potential influence on quality of life, nutrition, and medication compliance.
The association between chemosensory complaints and quality of life in this study was very striking. Subjects with chemosensory complaints scored lower on all domains of the quality of life instrument, the MOSHIV. The association of chemosensory complaints with poor quality of life persisted after controlling for number of AIDS diagnoses, number of medications taken, CD4 cell count and HIV-1 viral load. It is impossible to tell whether chemosensory difficulties result directly in decreased quality of life, or whether chemosensory difficulties are another marker of poor quality of life. In either case, strategies aimed at reducing chemosensory complaints could have a positive influence on quality of life.
Chemosensory abnormalities may also have a potential influence on nutrition. Taste and smell abnormalities may impair initiation of cephalic phase reflexes that optimize nutrient absorption. This may contribute to inadequate oral intake, leading to malnutrition, weight loss and ultimately wasting. Our study failed to show a relationship between chemosensory complaints and body mass index, but as a cross-sectional study, it would not be able to detect changes in body mass index over time, which may be a more sensitive indicator of nutritional status.
Multivariate analysis demonstrated that the severity of the chemosensory complaint score was associated with the number of medications taken, hay fever and tobacco use. Of note, there were many subjective complaints about medications; many subjects complained that medications tasted or smelled bad, or interfered with their senses of taste and smell. A more rigorous evaluation of the effect of each medication on the senses of taste and smell could have been performed by asking each patient about each drug individually, but the questionnaire would have become very long, and respondent burden would have become prohibitively high.
In the era of potent antiretroviral therapy, adherence to medical therapy is critical in the prevention of drug resistance. Erratic compliance with medications results in intermittent or subtherapeutic levels of antiretrovirals, which promotes the natural selection of resistant strains. If a medication tastes or smells bad, or alters the way food tastes or smells, patients are less likely to be compliant with taking the medication. The chemosensory effect of medications should be considered a side-effect to be taken into account when clinicians decide which combination of medications to prescribe. Elucidation of the molecular mechanism of the chemosensory effects of drugs may allow development of effective medications with less impairment of chemosensory function.
Measures aimed at decreasing chemosensory complaints may improve patients’ quality of life, nutritional status, and compliance with medications. Reversible causes of chemosensory dysfunction, such as oral candidiasis, should be treated. Chemosensory side-effects should be taken into account when prescribing medications. Taste supplementation using flavor enhancers, commercially available preparations of odorous molecules that enhance the natural of flavor of food, may be added to food to increase oral intake in anorectic individuals who complain that food tastes bland. These patients could also experience an improvement in quality of life if the pleasure associated with eating is increased.
We believe that the survey respondents are representative of the HIV-infected patients followed at DUIDC, but the results may have been affected by selection bias or recall bias. Selection bias was minimized by inviting patients to complete the survey before ascertaining whether they had chemosensory complaints. Recall bias, leading to over-reporting of chemosensory complaints, may have resulted from the directed nature of the survey. This could increase the number of subjects reporting minor taste and smell complaints. However, 30% of patients reported moderate-to-severe impairment of their sense of taste, and 27% of patients reported moderate-to-severe impairment of their sense of smell.
Chemosensory testing of a subset of 50 subjects failed to demonstrate a difference in taste thresholds between those with and without taste complaints. Similarly, no difference in smell discrimination was demonstrated between subjects with and without smell complaints. Other investigators have noted difficulty in correlating taste and smell complaints with formal taste and smell testing [12,22]. There are several possible reasons for this discrepancy. The most likely explanation is that most subjects complained of dysgeusia, or taste distortions, which taste detection and recognition thresholds would not evaluate well. Future testing should be directed towards better diagnosis of taste distortions. Second, because the senses of taste and smell are so closely linked, abnormalities of taste can be confused with abnormalities of smell. Perhaps a directed interview could better distinguish taste from smell complaints than a questionnaire.
This study has demonstrated that there is a high prevalence of self-reported chemosensory complaints in HIV-infected persons followed at the DUIDC. Chemosensory difficulties are important because of their striking association with poor quality of life, and their possible influence on nutrition and medical compliance. Further investigation of the molecular mechanism of taste and smell abnormalities associated with medications, and potential interventions such as taste supplementation is warranted.