Measuring Taste Impairment in Epidemiologic Studies The Beaver Dam Offspring Study

Measuring Taste Impairment in

Epidemiologic Studies

The Beaver Dam Offspring Study

L. M. Bartoshuk,cG. H. Huang,d B. E. K. Klein,a R. Klein,aF. J. Nieto,aJ. S. Pankow,eT. S. Tweed,a

E. M. Krantz,aand G. S. Moya a_{University of Wisconsin, Madison, Wisconsin, USA}

b_{Yale University, New Haven, Connecticut, USA} c_{University of Florida, Gainesville, Florida, USA}

d_{National Chiao Tung University, Hsinchu, Taiwan, Republic of China} e_{University of Minnesota, Minneapolis, Minnesota, USA}

Taste or gustatory function may play an important role in determining diet and nutri-tional status and therefore indirectly impact health. Yet there have been few attempts to study the spectrum of taste function and dysfunction in human populations. Epidemi-ologic studies are needed to understand the impact of taste function and dysfunction on public health, to identify modifiable risk factors, and to develop and test strategies to prevent clinically significant dysfunction. However, measuring taste function in epi-demiologic studies is challenging and requires repeatable, efficient methods that can measure change over time. Insights gained from translating laboratory-based methods to a population-based study, the Beaver Dam Offspring Study (BOSS) will be shared. In this study, a generalized labeled magnitude scale (gLMS) method was used to measure taste intensity of filter paper disks saturated with salt, sucrose, citric acid, quinine, or 6-n-propylthiouracil, and a gLMS measure of taste preferences was administered. In addition, a portable, inexpensive camera system to capture digital images of fungiform papillae and a masked grading system to measure the density of fungiform papillae were developed. Adult children of participants in the population-based Epidemiology of Hearing Loss Study in Beaver Dam, Wisconsin, are eligible for this ongoing study. The parents were residents of Beaver Dam and 43–84 years of age in 1987–1988; offspring ranged in age from 21–84 years in 2005–2008. Methods will be described in detail and preliminary results about the distributions of taste function in the BOSS cohort will be presented.

Key words: taste; epidemiology; methods

Introduction

Taste or gustatory function may be an im-portant determinant of health through

possi-Address for correspondence: Karen J. Cruickshanks, PhD, Department of Ophthalmology, University of Wisconsin, 610 N. Walnut Street, 1038 Warf, Madison, WI 53726-2336. Voice: 608-262-4032; fax: 608-2625-2148. cruickshanks@episense.wisc.edu

ble affects on food choice and consumption. However, there have been few epidemiologic studies measuring taste function in adults.1_We

do not know if or how often taste intensity or flavor recognition changes with aging in the general population, nor if taste dysfunction is an important contributor to the risk of cer-tain chronic diseases, frailty, or inability to re-cover from serious infections, surgeries, or other

International Symposium on Olfaction and Taste: Ann. N.Y. Acad. Sci. 1170: 543–552 (2009). doi: 10.1111/j.1749-6632.2009.04103.x c 2009 New York Academy of Sciences.

is optimal for health, but data from epidemio-logic studies are needed to quantify the preva-lence of taste disorders and the impact of taste dysfunction on health, and to identify the de-terminants of taste dysfunction.

In order to conduct epidemiologic studies of taste function there must be reliable, inexpen-sive, portable tests with low respondent burden and agreed-upon definitions of taste function and dysfunction outcomes. As yet, these are not readily available, although efforts are un-derway as part of the NIH Toolbox Initiative to reach consensus on recommended measure-ment methods and definitions. The purpose of this paper is to report the methods used to mea-sure taste as part of an ongoing epidemiologic study of human aging sensory systems in Beaver Dam, Wisconsin.

Study Population

The Beaver Dam Offspring Study (BOSS) is a study of age-related hearing, vision, and olfactory impairments, and ocular disorders among the adult children of participants in the population-based Epidemiology of Hearing Loss Study (EHLS), a longitudinal study of ag-ing that began in 1993.2,3In 1987–1988 a pri-vate census of the city and township of Beaver Dam, Wisconsin, was conducted to identify all residents ages 43–84 years.4 These 5924 indi-viduals were invited to participate in an exten-sive examination for a study of age-related oc-ular disorders, the Beaver Dam Eye Study. The EHLS was timed to coincide with the 5-year follow-up for the eye study cohort, and 3753 (82.6%) of the 4541 surviving, eligible subjects

of the prevalence of taste dysfunction or distri-butions await the complete data set. Previously, it has been shown that the EHLS cohort is simi-lar in age, gender, and education to residents of mid-sized cities in the United States, although the cohort is primarily non-Hispanic white.2

Although taste testing was not part of the original scope of the study, we were approached by the National Institute on Deafness and Other Communication Disorders to include taste testing and charged with developing, stan-dardizing, and testing methods and protocols for measuring the sense of taste in field stud-ies, obtaining digital images of the tongue, and measuring the density of fungiform papillae. We were to use these methods to measure the distributions of taste intensity and fungiform papilla density, associations of taste intensity with fungiform papilla density, and associations of taste with social, medical, and lifestyle factors and health.

Measurement Methods

Taste Intensity Measurement A generalized labeled magnitude scale (gLMS) was used to quantify the perceived taste intensity of filter paper disks (Fig. 1A).5 _Disk

were impregnated with 1.0 M sodium chlo-ride (salt), 1.8 M sucrose (sweet), 0.1 M citric acid (sour), 0.001 M quinine (bitter), or 6-n-propylthiouracil (PROP). Whatman #1 filter paper was soaked in room-temperature solu-tions of sodium chloride, sucrose, citric acid, and quinine and dried; disk diameters were 3 cm. Filter paper was soaked in a saturated

Figure 1. (A) gLMS used to measure intensity of remembered sensations and tastes. (B) gLMS used to measure intensity of food likes and dislikes.

solution of PROP near boiling and dried; each paper disk contained 1.2–1.6 mg PROP. These disks were produced in the laboratory of one author (L.M.B.) and shipped to Wisconsin in in-dividual glassine envelopes, color-coded by tas-tants, in plastic zippered bags to guard against moisture.

Participants were asked to rank intensities using a scale that ranged from no sensation to strongest imaginable sensation of any kind (with a corresponding scale of 0–100 representing the distance from “no sensation”). To familiar-ize participants with the scale, they were asked

to rate the intensity of the sound of the loudest thunderclap s/he can remember, a purring cat held in her/his lap, and a lawnmower across the street. Participants rank ordering them cor-rectly (thunder > lawnmower > cat) proceeded on with the training, while subjects who failed to order the intensities correctly were rein-structed. If the participants failed to rank order the intensities correctly the second time, they did not proceed with the training. Participants successfully mastering the scale (the rank order was correct), were asked to rate the intensity of the sound of snow falling on a calm night, the

asked to rate the intensity of the brightness of the room, brightness of a dimly lit restaurant, brightest light s/he had ever seen, loudness of a whisper, loudness of a conversation, and loudest sound s/he had ever heard.

Then, participants were given each filter pa-per disk to taste in the same order (salt, sweet, sour, bitter, PROP). For each disk, the partici-pant was asked to report the taste quality (salt, sweet, sour, bitter, no taste, other, or unknown) and then rate the intensity of the taste using the gLMS. Between disks the participant was en-couraged to sip room-temperature bottled wa-ter. After tasting the PROP disk, the participant was offered a quick-dissolving peppermint to mask any residual taste. Subjects who reported being pregnant or sensitive to PROP were not eligible to receive the taste disks.

Food Likes and Dislikes

To measure the intensity of food likes and dis-likes a modified gLMS was used (Fig. 1B), with responses ranging from the strongest imagin-able disliking of any kind (−100) to strongest imaginable liking of any kind (100) with zero representing neutral (neither like nor dislike).6

In this study, participants were asked to rate the following foods: mayonnaise, whole milk, black coffee, dark chocolate, salted pretzels, grape-fruit juice, sweets, strawberries, sausage, and milk chocolate. These foods were selected based on one investigator’s (L.M.B.) experience to represent a spectrum of important food experi-ences corresponding to the selected taste sensa-tions being measured as well as fatty foods. For example, pretzels represent a salty food, black

ton swab to the tip of the tongue to provide contrast between fungiform papillae (which appear pink) and other tongue structures (coated blue). We adapted equipment and methods used for ocular examinations and oc-ular images to create a standardized system for obtaining digital images of tongues. An adjustable table was outfitted with a chin and forehead rest (Modified Soderberg LMP-1 Mo-torized Instrument Table [Soderberg Oph-thalmic Services, Minneapolis, Minnesota] and Shin-Nippon Forehead/Chinrest Assembly [Tokyo, Japan]) and equipped with a digital camera system installed on the table using a col-umn support to provide a fixed distance from the person’s face. The camera system consisted of a Canon EOS Digital Rebel XT Body fit-ted with a Canon EFS 60 mm f2.8 Macro lens, Canon MR-14EX Macro Ring Flash, and Canon 52 mm UV filter (Canon U.S.A., Inc., Lake Success, NY). The camera was connected to a computer using Zoom Browser Ex (Canon U.S.A., Inc.) to capture and store images.

The participant was asked to rest his/her forehead against the support while placing the chin on the chinrest, and the examiner adjusted the table height to ensure participant comfort. The participant was asked to stick out his/her tongue and close his/her eyes. The ring flash focusing lamp was turned on and the exam-iner would adjust the focus as necessary to en-sure a sharp image, with minimal glare. After ensuring the contrast was sufficient, the exam-iner held a plastic slide on the tongue tip to the right of the midline applying a slight pres-sure to compress the tongue, and captured the image. Additional blue food coloring was ap-plied as necessary for optimal contrast and, if

excessive amounts were present, the participant was asked to swallow until the desired contrast was observed. After reviewing the original im-age, additional images could be captured, if necessary.

During the process of developing this cam-era system, image quality was reviewed by two authors (L.M.B. and D.J.S.) to ensure compara-bility with existing imaging methods which use specialized operating microscopes. Examiners were taught to follow the standardized testing protocol, and image quality was judged accept-able on five practice subjects before they were considered certified to implement these proce-dures in the field.

Throughout the study, each examiner was observed carrying out the protocols on study subjects, image quality was reviewed, and data were monitored for deviations and drift.

Grading Fungiform Papilla Density

Digital tongue images were transferred to Madison, Wisconsin, for grading using a spe-cially developed application (CanvasTM X; ACD Systems, Inc., Miami, FL). A standard-ized protocol was developed to allow graders to select the best image available for grading us-ing a preview function and standardized crite-ria. The grader evaluated slide placement (the entire width must be contained in the image to calibrate the size of the measurement area), tip visibility (the entire tip should be visible), mea-surement area (the entire meamea-surement area should be visible, in focus, and free of glare, bubbles, or other artifacts), and staining quality (pink circles visible on a blue background) to select the image for grading.

Once the selected image was loaded, the magnification, scale, and color were adjusted according to a standardized protocol, a stan-dard circle was applied (equivalent to a 6-mm diameter), with the right edge of the circle aligned at the midline of the tongue and the edge of the circle at the tip of the tongue. Fungi-form papillae were identified by color

(pink-red), appearance (mushroom-like or vascular-ized), and size (larger than filiform papillae) following a standardized protocol. The total number of fungiform papillae identified in the standard circle was automatically stored as the count.

Graders were trained in the procedures and certified using a standard set of 10 images pre-viously graded by an experienced researcher (D.J.S.). In order to become certified scores must be within 5 of the standard grader’s scores for both the mean difference and mean abso-lute difference, with at least 60% of the scores within 5, and at least 90% of scores within 10. The grader regraded the standard set of images every 3 months throughout the grading period to monitor drift. Intragrader variability (differ-ences between scores measured on separate oc-casions by one grader) was low, with a mean difference of 0.1; mean absolute difference was 3.1; and 80% of gradings matched within 5, and 100% were within 10. Thus grading of dig-ital tongue images is highly reproducible, using this standardized protocol.

Lessons Learned in Beaver Dam

Challenges in Administration In preliminary analyses of 2733 participants, we determined that 53 (2%) were unable to par-ticipate in the taste protocol because of preg-nancy, PROP sensitivity, refusal, or time limita-tions. In field studies the risk for human subjects must be minimized in order to protect subjects from harm and encourage future participation in longitudinal studies. Although the probable risk to pregnant women and fetuses is low, and true PROP sensitivity is rare, a conservative ap-proach is warranted, so it is important to rec-ognize that complete data are unlikely in any epidemiologic study of taste using similar meth-ods. The total examination time for the study approached 4 h, and some subjects were unable or unwilling to give sufficient time to complete the entire examination.

Figure 2. Distribution of intensity of remembered sensations. A, brightness of the room B, brightness of a dimly lit restaurant; C, brightest light you have seen; D, loudness of a whisper; E, loudness of a con-versation; F, loudest sound you have heard. Filled in boxes represent the interquartile range, horizontal lines within the boxes show the medians, plus sym-bols show the means, and whiskers represent the full range of the data.

An additional 415 (15%) failed the practice task for the taste scale and could not proceed to the tasting of paper disks. Most participants (69%) failing to master the scale rated the inten-sity of the purring cat higher than or equal to the sound of the lawnmower across the street. When this problem was identified, the training protocol was revised to allow for reinstructing the participants; however, this modification did not eliminate the problem, as 14.7% continued to order intensities incorrectly. Selecting other experiences for the training could increase the percent of subjects completing the taste testing. Nonetheless, the scale did capture a range of intensity ratings for the remembered sensations as shown in Figure 2. In the sample of 2265 sub-jects with complete data, the median responses ranged from 5 for the loudness of a whisper to 80 for the brightest light ever seen, although the ranges for each sensation were broad as in-dicated by the whiskers. Although respondents correctly ranked these sensations in order with light or sound groups, analyses of taste inten-sities may need to be adjusted based on the scores assigned for the remembered sensations in order to remove variability due to individual differences in the scale range used.7

than that meant for the Wisconsin study. Al-though the time to detection of this problem was short, 226 or 8% of the subjects in this preliminary data set received a sour disk with the incorrect concentration resulting in addi-tional missing data for this taste. This experi-ence highlights the importance of strict quality assurance procedures to detect changes over time.

The distributions of reported taste qualities are shown in Figure 3. Most people correctly identified salt, sweet, and bitter, with a large proportion miscalling sour as bitter and the ex-pected variation in people identifying PROP correctly. Although the taste quality rating may be of limited analytic utility, continuing to col-lect this information provides important data for quality assurance and provides a check on internal validity, namely, that the concentra-tions of the disks were sufficient to correctly recognize the flavor.

The distributions of taste intensity for the five disks are shown in Figure 4 also as box plots to illustrate the range of perceived inten-sity scores. For each taste there is a broad range of scores. The score reflects the proportional distance from “no sensation” to the “strongest imaginable sensation of any kind” recorded as a number from 0–100. There was significant digit preference, with 56–72% of subjects selecting a distance reflecting an intensity score ending in zero and 81–91% of subjects selecting a dis-tance corresponding to numbers ending in zero or five for any taste intensity score. Recording the distance on a scale with broader intervals may be sufficient as it is not known what dif-ference in magnitude of intensity is important when comparing groups.

Figure 3. Distribution of reported taste disk qualities. Each patterned segment represents the proportion of participants reporting that taste quality.

Participants also used a scale to quantify the intensity of liking or disliking a food. Fig-ure 5 shows the distributions in reported in-tensity, with zero being neutral, negative num-bers representing dislikes, and positive numnum-bers representing likes. Again, there was substantial variability in the reports, suggesting that this scale may be useful for studies of food prefer-ences, dietary intake patterns, and health.

The taste disks were easy to ship to the field site, participants did not object to the testing, although some people complained about the bitterness of the PROP disk, and the test was easy to administer.

Density Measures

Our digital tongue imaging system worked well. Although there was a learning curve for applying blue food coloring without making a mess, the examiners quickly became expert in

applying it neatly. The amount of dye needed varied by participant and the time it remained in the mouth also varied, so monitoring the quality of the contrast during the photogra-phy is important. The ring flash system that was selected reduced problems with glare and washout. The image management and grading systems facilitated standardized density mea-sures by ensuring that the grading standard cir-cle was adjusted for scale differences across im-ages. Features of our grading system that were particularly useful were the ability to click on each fungiform papilla to incrementally add to the total score, store the graded images for later review by the epidemiologist for quality assurance purposes, and grade an image multi-ple times (masked to previous results) for addi-tional quality assurance efforts (regrading of a random sample and comparisons between and within graders). The distribution of fungiform papillae obtained in this sample is displayed in

Figure 4. Distribution of taste disk intensities. Filled-in boxes represent the interquartile range, hori-zontal lines within the boxes show the medians, plus symbols show the means, and whiskers represent the full range of the data.

Figure 6 and shows that a broad range of den-sities was detectable. The grading was highly reproducible, and trained graders were com-parable to an experienced researcher (D.J.S.). This relatively inexpensive imaging and grad-ing system provided high-quality images in the challenging setting of a field study.

Measuring Prevalence and Evaluating Associations

The immediate challenge when using these results to report the prevalence of taste

disor-Figure 5. Distributions of food likes and dislikes. A, mayonnaise; B, whole milk; C, coffee; D, dark chocolate; E, pretzel; F, grapefruit; G, sweets; H, strawberries; I, sausage; J, milk chocolate. Filled-in boxes represent the interquartile range, horizontal lines within the boxes show the medians, plus sym-bols show the means, and whiskers represent the full range of the data.

the scale as important, such as those used in studies of refractive error (myopia and hyper-opia in this example), may be useful for stud-ies of taste. Methods used to establish cutpoints based on young “normals” may help to identify important subgroups. Scores for young healthy subjects without olfactory disorders, who do not report problems with taste and use a broad range for scoring the remembered sensations may be used to establish cutpoints for low and high performance on taste intensity scales, which can then be applied to the study sample to classify participants into groups.

Analytic models can then move beyond sim-ple correlations and linear relationships to ex-plore effects of low and high sensation on other health conditions, as well as to explore factors associated with low and high taste sensation. Researchers should evaluate the impact of ad-justing for the scores for the remembered sensa-tions, as this may reduce variability due to dif-ferences in the magnitude of the psychophysical scaling range used.

Additional research is needed to determine if each taste quality has a similar impact on health and similar determinants, which might suggest that responses to each taste disk could be combined in some fashion to identify peo-ple with taste disorders. Analyses of grouped data can identify key factors associated with poor performance and evaluate the impact of performance on health, food likes and dislikes, and dietary intake patterns.

However, these cross-sectional data cannot be used to distinguish participants who have experienced increased or decreased function from individuals with low or high function since birth or early childhood. Longitudinal data

Figure 6. Distribution of fungiform papillae. Numbers shown above bars are the number of participants in each category.

measuring change in function over time will be important to determine how taste function changes with age and to identify factors associ-ated with the development of taste disorders.

Additional Issues

We have demonstrated that taste function can be measured in epidemiologic studies us-ing these simple measures of taste intensity, food likes and dislikes, and anatomic measures of the density of fungiform papillae. However, there remains a need for studies to determine the test–retest consistency of intensity scores. Be-fore applying these tests in longitudinal studies of change in taste function, it is important to know that the short-term variability is low. Be-cause we used one standard order of presenta-tion, it is not known if presentation order affects intensity ratings.

Summary

We have developed and implemented mea-sures of taste function and fungiform papilla density in an epidemiologic study where

exam-inations occur in a community setting. In pilot work we had considered liquid testing, but due to the difficulty of creating and maintaining stock solutions in a field site that consisted of a suite of offices without a laboratory or clean sink area, it was determined that filter paper disks would be preferable. Transporting these disks to the field site was inexpensive, and sim-plified testing for participants. Although spatial testing with liquids may detect clinically sig-nificant alterations in localized oral sensation, our whole-mouth measures are likely to rep-resent the usual experience of an individual. Our imaging and grading methods for eval-uating lingual anatomy offer an inexpensive way to achieve high-quality images and reliable estimates of fungiform papilla density, which will permit studies of the complex relationships of fungiform papilla density and taste inten-sity. These methods may be useful to study the public health importance of differences in taste function, the magnitude of the population with taste impairments, the risk of developing taste impairments with aging, and the relationships between taste perception, food preferences, di-etary intake, and health.

References

1. Vennemann, M.M., T. Hummel & K. Berger. 2008. The association between smoking and smell and taste impairment in the general population. J. Neurol.255: 1121–1126.

2. Cruickshanks, K.J., T.L. Wiley, T.S. Tweed, et al. 1998. Prevalence of hearing loss in older adults in Beaver

the gLMS vs magnitude matching. Physiol. Behav.82: 109–114.

6. Bartoshuk, L.M., V.B. Duffy, J.E. Hayes, et al. 2006. Psychophysics of sweet and fat perception in obesity: problems, solutions and new perspectives. Philos. Trans. R. Soc. Lond. B Biol. Sci.361: 1137–1148.

7. Bartoshuk, L.M., V.B. Duffy, A.K. Chapo, et al. 2004. From psychophysics to the clinic: missteps and ad-vances. Food Qual. Prefer.15: 617–632.