National Health and Nutrition Examination Survey

The Taste and Smell examination component (variable name prefix CSX) was a new health examination component first conducted in 2012 in the second year of the NHANES 2011-2012 cycle. Since only 1-year (2012) of data collection was performed, it is a restricted access dataset with the NCHS Research Data Center (RDC). NHANES previously collected health examination data on most other special senses, including vision, hearing, balance, and peripheral nerve sensation; however, this was the initial NHANES health examination data collection for the senses of taste and smell.

Study Aims:
• To provide reference data for taste and smell testing for U.S. adults ages 40 years and over;
• To examine variations in the ability to smell and to taste salt and bitter tastants and analyze these variations with NHANES hypertension, nutritional, and obesity data; and
• To help estimate the prevalence of U.S. adults who may not recognize the odor of smoke and natural gas, which are important early warning signals for home safety hazards.

The NHANES Taste and Smell examination component was proposed and sponsored by the National Institute on Deafness and Other Communication Disorders (NIDCD) and co-sponsored by the Centers for Disease Control and Prevention.

A full sample of adult participants of both genders aged 40 years and over. Women who were currently pregnant (had either self-reported pregnancy or positive urine pregnancy test) or currently breastfeeding were excluded from the taste and smell examination. Participants who were unable to provide a correct ordinal ranking of the three light intensity standards on the gLMS scale were excluded from taste testing, but not from smell testing.

While quinine is an approved U.S. Food & Drug Administration beverage flavoring, quinine allergy, especially skin hypersensitivity reactions, are well documented in the medical literature (AHFS, 2011; Barr et al., 1990; Bel et al., 2009; González et al. 2002, Kanny et al. 2003). While such reactions are thought to be uncommon, any person with a prior history of allergy to quinine in medicines or beverages was excluded from quinine taste testing. Specifically, participants who answered yes to the quinine allergy question, stated they did not know, or refused to answer the question were excluded from quinine taste testing, but not from any other part of the Taste and Smell examination.

The examination protocol content and sequence is illustrated in the flow diagram in Appendix A. The detailed examination procedures are described in the online NHANES Taste and Smell Examination Component Manual. The exam was performed in the following sequence:

Pre-Exam Procedures:

The participant first answered a short pre-exam questionnaire. This included the two exclusion questions and a short series of questions to screen for certain conditions, such as current nasal symptoms present at the time of the exam. This could potentially influence the interpretation of the results of the smell test. This data may help to distinguish temporary from chronic impariment of smell abilities. The nasal symptom questions were not exclusion questions. The pre-examination symptom questions were administered in a “code all that apply” format. Please see the Taste and Smell Examination Procedures Manual, page 4-3 Exhibit 4-3 for an image of the data collection format screen.

Training and Rating on the Perceived Taste Intensity Scale:

Following the pre-exam questionnaire, the participant received a brief introduction to the generalized labeled magnitude scale (gLMS), used for rating perceived taste intensities (Bartoshuk et al., 2004).This is a verbally anchored quasi-logarithmic response scale with a total of 100 units presented as a vertical line graph, which has the labels "barely,” “weak,” “moderate,” “strong,” “very strong,” and "strongest of any kind" spaced at 1, 5, 16, 34, 53, and 100 units, respectively (Bartoshuk et al., 2004; Green et al, 1996). The participant was instructed to let the top value (100) represent the strongest sensation of any kind that they had experienced, including his/her experiences with light, sound, pain, temperature, or taste.

Two standardized light intensities were then presented to the participant (LED luminescence panel; Illumination Technologies, Elbridge, NY) to provide an initial opportunity to practice using the scale (a lower intensity light [CSQ450] and a higher intensity light [CSQ460]). In previous studies, pure tone audiometric stimuli have been used for this purpose; however, for NHANES, light stimuli were chosen for ease of presentation and because deficits in gross light intensity preception are less likely to be affected by aging as opposed to hearing deficits.

Following the brief practice session, the participant then rated the brightness of a standard series of three lights well separated in intensity. A moderate intensity light was presented first [CSQ470], then a dim light [CSQ480], and then a high intensity light [CSQ490] in that fixed order. If the participant rated these three lights in the correct relative order on the gLMS scale, he/she proceeded to the taste exam. If he/she was unable to, then he/she was excluded from taste testing but not smell testing.

Tongue Tip Taste Testing:

The participant was then presented with two tastants: 1 mM quinine as a bitter taste and then 1 M NaCl as a salt taste in that fixed presentation order. Each tastant was gently applied to the the edge of the circle at the tip of the tongue with a cotton swab applicator in a standardized manner so as not to evoke any tactile stimulation. They were asked to keep their tongue out while rating the tastant intensity. Then they identified it as salty, bitter, sour, some other taste, or no taste. The mouth was rinsed with water before proceeding to the next tasant. The participant was asked if they had any lingering taste effect after the water rinse, and if so it was repeated until the taste was extinguished. There was a minimum wait period of 30 seconds between administering tastants.

Whole Mouth Taste Testing:

When tongue tip taste testing was completed, the participant then proceeded to the primary whole mouth taste testing. Here each participant was presented with three tastants, administered in one of two randomized presentation orders: 0.32 M NaCl, 1 mM quinine, 1M NaCl; or 1M NaCl, 1 mM quinine, 0.32 M NaCl. For each tastant, the participant was asked to take all of the 10 ml tastant solution into his/her mouth but not to swallow it. He/she were instructed to gently swish the solution in his/her mouth for 3 seconds, then spit it out. The participant then provided a tastant-specific intensity rating and identified it as salty, bitter, sour, some other taste, or no taste. The mouth was then rinsed using the same procedure as tongue tip taste testing.

The 8-Item Odor Identification Test:

This test was a brief 8-item "scratch and sniff," forced-choice smell test manufactured by Sensonics, Inc. The smell test contained four food-related and four nonfood-related odors, the latter including the odors of smoke and natural gas. The eight specific odors in their fixed order of presentation were: chocolate, strawberry, smoke, leather, soap, grape, onion, and natural gas.

In the kits, the test odorants are embedded in micro-capsules positioned on scent strips on a paper test booklet. The composition of the sents used in the NHANES study were proprietary to the smell test kit manufacturer. The stimuli to be smelled were released by scratching the odor test strips with a plastic stylus in a standardized manner. Then the participant smelled the odor and identified it. Above each odorant strip there was a list of four possible responses. For example, one of the items reads: “This odor smells most like: a) soap; b) black pepper; c) leather; or d) peanut.” The participant was required to choose one of the four choices presented. If the participant stated he/she could not smell any odor when presented with an odorant, he/she was required to choose one of the four alternatives as an answer. The “forced choice” format was employed because it is thought that even in persons with moderate impairments in olfaction; some residual ability to smell may remain even if the individual is unaware of it.

Whole Mouth Replicate Salt Taste Test:

Following odor identification testing, a single replicate whole mouth test of a salt solution was performed. The participant was randomized to receive either a .32 M NaCl or a 1 M NaCl salt solution. The replicate taste testing protocol was identical to the initial whole mouth tasting testing.

The NHANES health technicians had initial component training program with expert consultants and survey staff. The training included a didactic overview of the component and demonstrations conducted by the expert examiner with volunteer subjects. Supervised practice exercises followed, again conducted with volunteer subjects. During the 2012 survey, the chief health technician at each of the mobile examination centers monitored staff performance in the field. Health technician performance was also periodically monitored by NCHS staff, data collection contractor Project Officers, and the expert consultants.

The NHANES Taste and Smell Examination Component Manual includes detailed descriptions of the quality assurance and quality control measures that were used in the 2012 data collection. Equipment maintenance and calibration was performed on a routine basis by the health technicians and verified by supervisory staff. Periodic review of examination and calibration data was performed to help verify data collection accuracy.

All data were captured into a computerized database system, and automatically uploaded. The 2012 NHANES taste and smell examination data was also verified against the main data collection data file. Computerized data editing was performed to check for logical inconsistencies in the data and health technician errors, and to cross-check other issues potentially affecting data quality. Back-end edits of the data were performed as needed when errors were detected. Otherwise, no data values were excluded from the data release.

The 2012 Taste and Smell dataset, like other 1-year NHANES datasets is accessible as a restricted access NCHS Research Data Center (RDC) data file. The Taste and Smell examination component was designed as a brief screening examination and was developed during two pilot tests in 2010 and 2011, then fielded in the NHANES 2012 survey cycle midway into the ongoing NHANES 2011-2012 data collection. In the MEC, the average time to complete the Taste and Smell exam was 15 minutes. The tastants and odors for the study were selected by NIDCD and the study consultants. Propylthiouracil taste testing was not included since its status as a prescription medication required special informed consent (Cooper, 2005; Bandyopadhyay et al., 2002; NTP, 2013). However, genetic samples were collected in NHANES 2012, and may become available for future study. The following technical notes are provided for data analysts:

i. Pre-Exam Questions:
After the 2011 Taste and Smell pilot test, a change was planned for one of the Pre-Exam questions relating to nasal symptoms: the variable CSQ260j (had nasal symptoms from a head cold or flu) was scheduled to be replaced by CSQ260n (had a runny nose). This change could not be immediately implemented, so CSQ260j continued in the field until it could be replaced with CSQ260n. Because CSX 2012 is a restricted access dataset, both variables have been retained in the dataset for potential analytic use. Analysts should be aware that neither question was administered to a full set of 2012 participants. Please note that in the NHANES 2011-2012 Spirometry exam dataset, there is a related question available (ENQ100).

ii. gLMS Scale Format:
The gLMS scale format used for tastant ratings was similar to that used in previously (Cruickshanks et. al, 2009); however, it was modified based on piloting experience. To counter label preference bias, the scale graphic was revised. The scale labels were enlarged and were moved farther to the right of the scale’s left border and the horizontal lines projecting from the scale labels were shortened and no longer intersected with the left sided measurement line. Also, the term “imaginable” was dropped from the scale instruction text as the meaning of phrase “the strongest imaginable sensation” was not immediately transparent and unambiguous for many participants.

iii. Exam Exclusions:
Three women who were pregnant or breastfeeding were excluded from the exam. A history of quinine allergy was reported by nine participants (0.7% of participants) and these were excluded only from quinine taste testing. After completing the training on the use of the gLMS scale, 5% of examinees failed to rate the 3 standard light intensities in the correct relative order and were excluded from taste testing. In comparison, in the recently reported results for taste testing in a Wisconsin community epidemiology study where a different gLMS scale training and testing protocol was employed, 15% of participants were excluded from taste testing on the basis that they could not successfully demonstrate correct scale use (Cruickshanks et al., 2009; Fischer et al., 2013).

iii. Participant Tastant Ratings:

When participants were observed to select a gLMS scale label for a taste intensity rating, they were actively coached to consider the entire scale for the tastant rating. However, if they persisted in choosing a scale label value as the intensity rating, then it was accepted and recorded as the result. Analysts should therefore consider scale label choice frequency in data analysis. For primary set of three whole mouth taste tests, 18% of participants chose an exact scale label value for one of the three tests; and another 3% chose two or more exact scale labels in rating the 3 tests. Using a wider label match criterion of ± 2 adjacent scale values, 37% of participants chose one scale label and an additional 34% chose two or more labels in primary whole mouth tastant ratings.

Some participants rated whole mouth tastants at a maximum (100) corresponding to their strongest perceived sensation of any kind. In the 2012 data, 2.5% of participants rated 1 mM quinine at “100.” The percentages for 1 M NaCl and .32 M NaCl were 1.7% and 0.3%, respectively. Health Technicians asked the participants to verify this type of rating when it occurred, and asked them to verify that they were not rating their dislike for or a sense surprise at experiencing a strong taste. However, if the participant persisted in choosing the “100” value, then it was accepted and recorded.

iv. Odor Identification Testing:

The 8-item NHANES odor identification test used in NHANES 2012 is similar in format to other smell test kits currently available and employed scents supplied by the manufacturer. The 8 odors fielded in NHANES 2012 had been previously used in larger smell test panel studies; however, they been not fielded previously as a separate smaller 8-item test. In piloting and in the 2012 data, some test odors appeared more recognizable to participants than others. For example, in the 2012 data, in the subset of 432 participants who correctly identified 7 out of the 8 odors, there were clear odor misidentification differences: 36% misidentified the chocolate odor and 21% misidentified the grape odor; whereas less than 2% of participants misidentified the soap scent or the onion odor.

A wide variety of natural gas odorants are in current commercial use and natural gas suppliers routinely formulate and sell mixtures of odorants to improve natural gas safety warning properties. Methyl mercaptan is a widely used odorant; however, ethyl mercaptan and thiophane are commonly used by LP gas suppliers serving rural U.S. areas (Cain & Turk, 1985). The composition of the natural gas scent used in the NHANES study was proprietary to the smell test kit manufacturer. This fact and the wide variety of natural gas odorants currently in commercial use should therefore be considered by analysts in assessing comparability to other research and in generalizing the results of their studies.

Clinical studies indicate that deficits in the ability to detect smoke or natural gas odors are associated with increased risk of injuries (Bonfils et al., 2008, Santos et al., 2004). The ability of adults in the U.S. to smell natural gas odor was previously described in a large survey (Gilbert and Wysocki, 1987; Wysocki and Gilbert 1988), which used a then commercially available mixture of mercaptans.

v. Exam Component Response Rates:

When the Taste and Smell Examination component was introduced in 2012, it was added to an already full examination schedule for the participants, so on any particular day, eligible participants were assigned to the Taste and Smell exam only if there was remaining time during their MEC examination session. These circumstances significantly reduced overall component response rates: of 2,949 adults ages 40 years screened for participation in NHANES 2012, 1912 were interviewed, 1,818 participated in the NHANES MEC examination and of these, 1,351 had either a complete or a partial Taste and Smell exam. Data analysts may need to consider reweighting the data or other adjustment techniques to account for participant non-response.

vi. Statistical Considerations:

Analysis of NHANES examination data should be conducted using NHANES examination weights. The NHANES Taste and Smell data was collected for a single year (2012) rather than the customary 2-year data collection cycle. The single year MEC examination weight, WTSFM should be used for estimating prevalences and testing statistical hypotheses, and the delete 1 Jackknife method (Wolters, 2007) employing the 1-year MEC examination replicate weights, WTSFM01-WTSF15 should be used for variance estimation. The survey strata and primary sampling unit (PSU) variables should be used to calculate the nominal degrees of freedom of the estimated variance, in order to determine the critical value for the Student’s t statistic used to construct confidence intervals. Please refer to the NHANES Analytic Guidelines and the online NHANES Tutorial for further details on the use of sample weights and other analytic issues.

This RDC dataset contains Taste and Smell exam data for participants ages 40+ years of age, however for proper variance estimation with the complex NHANES survey design, the dataset should be merged with the basic demographic data file variables (including sample weights, strata and PSU variables) for the entire set of NHANES 2012 participants. Then in the statistical analysis program used, the domain of interest is specified in a subpopulation statement as the set of adults ages 40+ years. This is to avoid variance underestimation when sampled PSUs contain no observations in a specific analytic domain. Please see the NHANES Online Tutorial and Lohr, 1999 for further discussion.

vii. Related NHANES Datasets

The NHANES 2011-12 Household Interview questionnaire (CSQ) dataset contains data on self-reported symptoms, risk factors, and medical treatment history for taste and smell disorders. Also the AUQ_G Household Interview Audiometry questionnaire collected 2 additional questions: a history of repeated ear infections (AUQ136) and a history of ear tube placement (AUQ138). Other NHANES 2011-2012 datasets that may be pertinent to the analysis of taste and smell data include the datasets on medical conditions (MCQ, KIQ, DIQ, BPQ); blood sugar (BIOPRO, GHB); blood pressure (BPX); body weight (BMQ, WHQ); diet and nutrition (DTQ, DRXDOC, DSQDOC, DBQ); smoking (SMQ, COTNAL); osteoporosis (OSQ, DXX); oral health (OHXDEN, OHXPER); alcohol (ALQ); and prescription medication use (RXQ_RX).