- Email: [email protected]
Identification of problem foods using food and symptom diaries
identification of problem foods using food and symptom diaries THEODOREKUEPER,BS,DEAN MARTINELLI,MD, FAAOA,WAYNEKONETZKI,MD, FAAEM,RALPHW. STAMERJOHN,BS, and JEANNE B. MAGILL, RN, ASOAT,Walcs, Oconomowoc, and Waukesha, Wisconsin Food and symptom diaries were used to identify problem foods for each of 164 patients with chronic medical problems such as headache, fatigue, congestion, abdominal pain, and sinus problems. A statistical analysis related the total Ioad of 90 biologic families, as well as caffeine, alcohol, and lactose, to changes in symptom intensity during a 2-week diary. The results helped 75% of the patients when used as a guide for elimination diets. Open challenges confirmed 47% of the identified food components. This study required a database and software to estimate recipe components for an average of 243 foods per patient. The analysis of each patient's diary produces a main report that lists suspect food components for each symptom. The report lists components in decreasing order of statistical confidence and gives lag times be~hNeen food ingestion and symptom change. This report also shows the initial direction of the symptom change as a direct or masking effect. Foods that appear "safe" or unrelated to the symptoms are also listed. A second report lists the patient's food sources for each of the suspected food components. The report shows the percentage contribution of source foods and isuseful for patient education and the design of eliminotion diets. (OTO~RYNGOL HEAD NECK SURG 1995;f 12:415-20.]
F o o d and symptom diaries are an established tool for detection of problem foods? -3 Diaries are often not used because interpretation is complicated by multiple foods with mixed ingredients, multiple symptoms, delayed reactions, and questions on ingredient oversights. 4 Food allergy is a complex problem for in vitro testing. The problem has not necessarily been reliability of the individual tests, but each test does not have a wide enough application to cover every facet of food allergy. 50ther types of adver~»e reactions to foods make the food identification problem additionally complicated. Other methods of identifying problem foods are also limited. Food-frequency From Wisconsin Data Laboratory, Ltd. (Mr. Kueper and Mr. Stamerjohn), Central Otologic, Ltd. (Dr. Martinelli), and A1lergy Diagnostic Clinic (Ms. Magill). Dr. Konetzki is in private practice in Waukesha, Wis. Mr. Kueper and Mr. Stamerjohn have a financial interest in Wisconsin Data Laboratory, Ltd., a company formed to process food and symptom diaries with the system described in this article. Received for publication June 2, 1994; revision received Sept. 6, 1994; accepted Sept. 30, 1994. Reprint requests: Theodore Kueper, Wisconsir~ Data Laboratory, Ltd., 415 Cymric Ct., Wales, WI 53183. Copyright © 1995 by the American Academy of OtolaryngologyHead and Neck Surgery Foundation, Inc. 0194-5998/95/$3.00 + 0 23/1/61t945
questionnaires and patient histories depend on recall and are subjective for both the patient and the physician. Elimination diets of many types are used to uncover food sensitivities. These are slow procedures and do not represent the normal dietary load of problem foods. Our diary interpretation has two assumptions: (1) the changes in chronic symptom levels may be statistically related to the amounts of common food components, and (2) the weights for most components of related biologic families 6 may be combined on a dry protein basis. A literature search found no articles during the last 10 years that use these assumptions in food and symptom diaries. A commercial software package (Well Aware Food Sensitivity User's Guide; Positech, Monterey, Calif.) based on ciuster analysis of food cliaries requires the first assumption, but the model and results are not published. This artMe introduces an improved method for discovering relationships between the patient's normal foods and symptoms. The results facilitate the design of elimination or avoidance diets and thus help to control food-caused symptoms. METHODS Patient Selection
Patients were primarily from two Midwestern practices. Orte practice specialized in ENT and 415
416
Otolaryngology Head and Neck Surgery March 1995
KUEPERet al.
allergy. The second practice specialized in environmental medicine and allergy. Most patients had had previous evaluations and immunotherapy treatment for aeroallergens but had had limited benefits. Most otolaryngologists' patients had symptoms related to the head and neck. These symptoms included headache, facial pressure, postnasal drainage, throat clearing, persistent cough, gastroesophageal reflux, ear plugging, and tinnitus. The initial visit included a history and physical examination to screen for other medical or surgical problems. This occasion was also used to discuss the possibility of an allergic basis for the problem. An allergy evaluation sheet and samples of antihistamines, nasal sprays, or decongestants were sent home as part of the first visit. At the follow-up visit (1 to 2 weeks later), the allergy evaluation sheet and an evaluation of sample medications were discussed with the patient. At a third visit, if medication requirements were still daily, an allergy assessment was suggested, including a screening immunoglobulin E radioallergosorbent or skin testing primarily for regional aeroallergens. After evaluating the radioallergosorbent or skin test screening and follow-up titration testing, the patient was placed on hyposensitization injections for a several-month trial. After this, the food and symptom diary was recommended for cases in which symptoms, especially headache, persisted. Patient selection in the environmental medicine practice was based on similar criteria. After a history and physical examination, most patients were examined and treated first for aeroallergens and then for foods with a combination of skin testing, sublingual drops, and food avoidance. If symptoms persisted, the food and symptom diary was recommended. The patient population was 60% female adults (18 years and older), 20% male adults, 10% female children, and 10% male children. Patients were told that the diary would be analyzed by a new computer program to help identify their problem foods. They were also told that the resultant data, coded for anonymity, may be used in publication of the results. Patient Instructions
The physician guided the patient in selecting any of the patient's symptoms that might be related to adverse reactions to foods. Up to five symptoms could be evaluated during the diary. Thus each patient's symptom pattern was specific to that patient. The symptoms were expected to vary in intensity at least four times during the recording period.
Symptom intensity was rated on a 9-point scale (1 = perfect, 9 = very, very sick). All symptoms were to be rated at each meal or snack occasion, and preferentially upon waking and retiring, and also upon a significant change of any symptom. Table 1 gives a summary of patient instructions. Patients were instructed to maintain their regular dietary routines. Patients were instructed to record to the nearest hour all food, drink, and medication for 2 weeks. A summary set of instructions asks the patient (1) to record food, drink, and medications at the nearest hour; (2) to be specific and show brand names and restaurant names; and (3) to list ingredients in nonstandard and home recipes. Single servings for the patient were assumed unless they noted otherwise. Patients were encouraged to clarify recipes by comment and by identification of brand names and restaurants. Patients were contacted by phone to clarify any obvious omissions. Most clarification regarded ingredient alternatives for home recipes. Data Analysis
Each symptom was treated as a dependent variable as long as it was rated for at least 10 consecutive days and there were at least four episodes of symptom change. Each hour with an entry for either foods or symptoms was considered a single record for computer entry. The data matrix for regression analysis consisted of the symptom ratings as dependent variables and a summation of food component indexes for each time period. Six accumulation times (3, 6, 12, 18, 24, and 48 hours) for symptoms were examined. An accumulation at 6 hours, for example, was the total amount of each food component ingested for the current hour and the previous 5 hours. Food records were converted to ingredient lists with a recipe database developed by us. New food recipes were created as required from cookbooks and commercial sources or by consultation with the patient. The recipe ingredients and additives, called components, were summed for a modified list of 90 biologic families6 plus lactose, caffeine, and alcohol. The indexes for summation were grams of protein for the biologic family, lactose, and alcohol, or milligrams caffeine. Refined foods ostensibly containing no protein were assigned a trace (0.01) value of the appropriate biologic families. Table 2 has examples in the column "Components" of the modified list for 37 biologic families. The components were then screened down to the 23 most frequent to evaluate (1) the data with most variation, (2) the variables most common to the
Otolaryngology Head and Neck Surgery Volume t 12 Number 3
patient's diet, and (3) whether they fit the concept that "recurring symptoms may be due to recurring foods." Time of day was included as an independent variable along with the 23 most frequent food components. The use of time of day as a possible factor helps to identify trends in symptoms tlhat may not be caused by foods. There were frequently five symptoms as dependent variables and 24 independent variables. Each symptom was examined with the six accumulation times. The resultant set of 720 t values on standardized coefficients was sorted into a report format that lists significance at low (above t = 2.7 or about 0.01 level of significance) average (above 3.3) and high (above 4.0) confidence. Reports
The main reports a r e a ranking of suspect components and a listing of the patient's sources for these components. Tables 3 and 4 give a typical example. The first column gives the suspected food component and the frequency of ingestion during the diary. Column two identifies the symptoms affected by the component. Column three shows the statistical confidence and either direct for a direct association between food and symptom, or mask for the appearance of symptom relief. In ranking the suspect components, direct was weighted somewhat higher than mask. A mask could represent either the abatement of a withdrawal reaction or an actual easing of symptoms due to exclusion of offending foods. Column four, labeled lag time, shows the time required from ingestion to the symptom change. This lag time may also represent the accumulation time required to produce a burdensome load of a food. The codes for programs used to conduct the process are beyond the scope of this article. The programs were written with Turbo-C, Paradox, and Quattro-Pro (Borland, Inc., Scotts Valley, Calif.) for IBM-compatible personal computers. Quattro-Pro has the regession analysis available. Physician Use and Interpretation
The usual treatment for the now-suspect food components was some combination of avoidance and challenge. Patients were told that these food components were merely suspect and that statistical relationships did not necessarily indicate cause and effect. Although the system is not designed as a nutrition analysis, dietary patterns occasionally revealed a need for nutrition counseling. The physician had the responsibility of interpret-
KUEPER et al.
4t7
Table I. Summary of patient instructions Choose and name up to 5 symptoms on the form Rate your symptoms using the 9-point scale On arising in the morning At each meal Whenever symptoms change significantly Just before retiring at night Record your food, drink, and medications at the nearest hour Be specific, show brand names, restaurant names List ingredients in nonstandard recipes Keep the food and symptoms diary for 2 weeks
Symptom rating scaie: 1, perfect; 3, couid feel better; 5, feeling very uncomfortable; 7, need to take something for problem; 9, ve,'y,very sick.
ing the result as suggesting food aller~, delayed food sensitivity, pharmacologic effect, and so forth. The lag time (from the accumulation time) reported, the direct or masking nature, and the other common characteristics (for example, high fat content) of the components often give support to theories or mechanisms. RESULTS
The most common rated symptoms were headache, fatigue, stuffy or runny nose, bloating, cough, diarrhea, gas and joint pain. One hundred sixty-four patients generated 125 different names for a wide variety of symptoms. The average 2-week diary contained symptom rating sets recorded 83 times, 90 meal or snack occasions, 118 names of different foods, and 243 total foods. Wheat, the legume family, eggs, milk, lactose, the lily family, the mustard family, and caffeine were the most common suspect components. There were an average of 9.1 components identified as suspect per patient and an average of 3.1 symptoms associated with suspect food components for each patient. Forty-nine patients provided in-home uncontrolled elimination diet testing for individual foods. Table 2 shows 105 (47%) of 222 suspected food components were confirmed by these tests. For example it shows citrus, alcohol, the lily family, and pome components were confirmed with 80% or bettet rates. Chicken, the mustard family, and the composite family components had Iess than 20% confirmation rates. The food components with highest levels of confidence gave the highest rates of confirmation. These open confirmations are not of double-blind, placebo-controlled quality. The rate of unequivocal confirmation will require further study. During the in-home test, the patients were instructed to note symptoms while avoiding a select food for 4 to 7 days and then to reintroduce the food.
418
Otolaryngology Head and Neck $urgery March 1995
KUEPERet al.
Table 2. Results of in-home avoidance and challenge tests for 49 patients Component
Examined
Suspected
Challenged
Confirmed
% Suspected
% Challenged
% Confirmed
Legume family Lily family Caffeine Milk Eggs Corn Banana Carrot family Pork Pome family Wheat Beef Gourd family Lactose Grape Citrus family Potato or pepper Tomato Oat or rye Rice Alcohol Chicken Composite family Mustard family Berry family Fish Cashew family Buckwheat family Stonefruit family Poultry (not chicken) Goosefoot family Heath family Walnut Pineapple Laurel family Mint family Other* TOTAL
47 45 39 46 44 42 25 41 34 26 46 44 24 41 23 31 28 23 27 23 6 34 27 41 19 18 3 2 23 10 4 6 1 2 3 5 62 967
23 22 20 22 24 17 14 19 17 13 24 23 9 20 10 13 12 11 21 13 3 22 15 19 8 7 3 1 5 4 4 3 1 1 1 1 28 473
16 12 12 11 12 9 11 9 7 5 11 11 6 4 4 3 7 5 8 7 2 10 8 7 4 2 1 1 4 1 0 0 0 0 0 0 12 222
10 10 7 7 6 6 5 4 4 4 3 3 3 3 3 3 3 3 2 2 2 1 1 1 1 1 1 t 0 0 0 0 0 0 0 0 5 105
48,9 48.9 51.3 47.8 54.5 40.5 56,0 46.3 50.0 50,0 52,2 52.3 37.5 48,8 43.5 41,9 42.8 47.8 77,8 56.5 50.0 64,7 55.6 46,3 42.1 38,9 100.0 50.0 21,7 40.0 100.0 50.0 100,0 50.0 33.3 20,0 45.1 --
69,6 54.5 60.0 50,0 50.0 52,9 78.6 47.4 41,2 38,5 45,8 47.8 66.7 20.0 40,0 23.1 58.3 45,4 38.1 53.8 66.7 45.5 53,3 36.8 50,0 28.6 33,3 100,0 80,0 25.0 0.0 0,0 0,0 0.0 0.0 0.0 42.8 --
625 83,3 58.3 63.6 50.0 66.7 45,5 44.4 57.1 80.0 27,3 27,3 50.0 75.0 75.0 100,0 42.8 60.0 25,0 28.6 100.0 10.0 12,5 14.3 25.0 50,0 100.0 100.0 0.0 0,0 0,0 0.0 0,0 0.0 0,0 0.0 41.7 --
*Other includes yeast, fungi, and combination named components from singularity elimination,
Two repetitions were suggested. This test protocol does not specifically test the overall model, because the model is essentially dependent on the "total load" accumulation of components. In at least one situation, the patient did not confirm individual foods but noted that symptoms disappeared after the suspected components were avoided as a group. The false-positive rate of 52% from the open in-home testing agrees well with a calculated estimate of 51% on the basis of the many relationships being analyzed in the regressions. Most of the patients did not report results of individual food component challenges. About 75% reported improvements due to general avoidance or reduction of suspected foods. The improvements reported by 50% of the patients were clouded by concomitant treatments such as immunotherapy or
new medication. A subset of 25% reported clear, independent improvement that they fully credited to the dietary changes. About 25% reported no improvement or no confirmation of any foods. A 1-year follow-up of patients who noted initial improvements with the food and symptom diary reports revealed no apparent problems with longterm usefulness. The patients continue to report symptomatic control as long as they "watch what they eat." They do not always adhere to the detailed findings, but they feel empowered to control their symptoms and return to a controlling diet as needed. The recording process is tedious for most patients. Two weeks was selected for a reasonably short time period, which gives adequate representation of the dietary and symptomatic changes associated with weekends vs. work or school days. Diary initiation
Otolaryngology Head and Neck Surgery Volume I'12 Number 3
KUEPER et ai.
419
Table 3ù Main report of statistical relationship of food components and symptoms (exampie omits patient and doctor identification details] Food component [no. times in diet) Chocolate (13)
Tea (7) Composite* (6) Fish (5) Beef (11) Time of day Wheat (40) Gourd (4)
Symptom
Regression result
Lag time
Headache Stuffy nase Ears Fatigue Stomach Headache Headache Ears Stomach Headache Headache Headache
Average, mask Average, mask Low, mask Low, mask Average, direct Low, mask Average, direct Average, direct Low, direct Low, direct Low, direct Low, mask
12 to 48 hr 6 to 24 hr 18 to 24 hr 6 to 48 hr 6 to 18 hr 12 to 18 hr 18 to 24 hr 0 to I2 hr 3 to 12 hr lncreasing 18 to 48 hr 12 to 18 hr
Direct, food component associated with adverse reactions; mask, food component temporarily eases symptoms through the lag time but may then cause adverse reactions; high, average, and Iow, refer to the degree of statistical confidence. The following food components were also in your diet and had at least four diary appearances (number in parentheses shows number of times food component was in diet): citrus (13), eggs (26), fungi (35), min (33), mustard* (16), chicken (8), tomato (1I), iegume* (18), corn (27), caffeine (29), carrot* (5), lily (16), potato* (10), mint* (6), park (13), and lactose (33). *Refers to biologicN family or subgroup. See "Food Components" explanation. This report is for your doctor. The report summarizes the results from regression analysis with symptoms as the dependent variables and the food components as the independentvariables. Statistical results da not necessarilycorrespond to cause-and-effect retationships.Your health professionatcan use the information along with other data to cOnSLIt with you on any dietary changes.
difficulties arose because some patients, appeared reluctant to share information about their eating habits. Special events, changes in home situations, or illness were also cited as reasons not to start or to stop keeping records. It seemed important to stress that the computer would be analyzing the data and that there would be no judgments made regarding the contents of the diary. Several unacceptable or incomplete diaries occurred during the early phase of o ür research. Instructions were improved to correct incomplete use of rated symptoms. The concept of rating a symptom when it is temporarily minimal needed special attention.
Table 4. Example of sources for first two components of main report Component Percent«ge Frequency Chocolate
Source
70.3 10.1 8.1
7 1 1
6.3
1
2.8 2.3
1 1
Chocolate Poptart Chocolate mint White chocolate mousse frozen yogurt Drumstick ice cream cone/carmel/nuts Brownie Cake/chocolate
57.1 14.3 14.3 14.3
4 1 1 1
Sun tea/sugar Iced tea/sugar Sun tea Tea
Tea
DISCUSSION The results of this limited group of patients appear encouraging. The inherent inaccuracies of home-recorded data and ingredient and amount estimation may be overcome by the block of data, which allows an objective comparison of food accumulations to the symptoms. The identification of specific food items that contain the suspected food components also appears useful. The model gives total weight to each patient's unique diet during the recording period. No comparison is made to any other patient averages or norms, as is the case for most laboratory tests. The results reflect the relationship between diet and usual symptoms to the degree that the foods selected
during the 2-week period reflect the normal diet for that patient. The most basic assumptions concern the sensitivity of the patient. The model assumes that the symptom levels are related to the accumulated components. Each component has its own weight in symptom production. Components may mask symptoms and directly accentuate symptoms. The timing of effects may be delayed both as direct effects and as masked reactions. One particularly troublesome feature of this model is that it does not differentiate effects duo to masked reactions and to actual symptom relief due to dilution or total replacement of problem foods with tmly safe foods.
Head
420
KUEPER e t al.
The system generates many questions that we hope will be clarified in time and by other researchers. We hope revisions will improve the reliability of these models. At this point of development the model does not attempt to weight the onset of symptoms compared with the sustained presence of symptoms. No attempt is made to recognize effects of medications or environmental or emotional changes. The times selected a r e a compromise to accommodate gut dwell times, normal meal spacing, and many of the delayed mechanisms of food intolerance. Data transformations and other types of statistical analysis may be appropriate. At this stage of our investigation, we cannot separate the individual importance of the major steps in the process. The patients selected were screened for other problems, and perhaps food intolerance was found frequently because other diagnoses were eliminated. Taking the effort to monitor their own diet and symptoms may have affected the patients' belief that foods were a problem. The extensive and objective determination of each patient's most frequent food families could have been another key step. We presume that some of the results are due to placebo effects. The food and symptom diary procedure involves a very personalized computer report. The avoidance and challenge tests include both
Otolaryngology and Neck Surgery March t995
dietary and lifestyle changes. Research has begun with the "safe" foods listed in the food and symptom diary report. These foods would be reversed with the suspected foods, thus providing a control (placebo) report. The in-home doubled-blinded testing with two reports, one true and one false, should clarify this issue. REFERENCES 1. Kettelbut BV, Metcalfe DD. Adverse reactions to foods. In: Middleton E, Reed CE, Ellis EF. Allergies, principles and practice. St. Louis: C.V. Mosby, 1988:1488. 2. American Academy of Allergy and Immunology. Adverse reactions to foods, a patient's guide to problem foods, food additives, diagnosis, treatment. Based on the article: Adverse reactions to food. Bethesda, Md.: National Institutes of Health; July 1984. Publication no. 84-2442;p.1. 3. Golbert TH. Food allergy and immunologic diseases of the gastrointestinal tract. In: Patterson R. Allergic diseases diagnosis and management. Third ed. Philadelphia: J.B. Lippincott Co., 1985:464-8. 4. Rowe AH. Food allergy. Springfield, Ill.: Charles C. Thomas, Publisher, 1972:38. 5. King HC. An otolaryngologist's guide to allergy. New York: Thieme Medical Publishers, 1990. 6. Anderson JA, Sogn DD. Adverse reactions to foods, American Academy of Allergy and Immunology Committee on Adverse Reactions to Foods, 1984. Bethesda, Md.: National Institute of Allergy and Infectious Diseases, United States Department of Health and Human Services, Public Health Service, National Institutes of Health;1984:17-25.
F o o d and symptom diaries are an established tool for detection of problem foods? -3 Diaries are often not used because interpretation is complicated by multiple foods with mixed ingredients, multiple symptoms, delayed reactions, and questions on ingredient oversights. 4 Food allergy is a complex problem for in vitro testing. The problem has not necessarily been reliability of the individual tests, but each test does not have a wide enough application to cover every facet of food allergy. 50ther types of adver~»e reactions to foods make the food identification problem additionally complicated. Other methods of identifying problem foods are also limited. Food-frequency From Wisconsin Data Laboratory, Ltd. (Mr. Kueper and Mr. Stamerjohn), Central Otologic, Ltd. (Dr. Martinelli), and A1lergy Diagnostic Clinic (Ms. Magill). Dr. Konetzki is in private practice in Waukesha, Wis. Mr. Kueper and Mr. Stamerjohn have a financial interest in Wisconsin Data Laboratory, Ltd., a company formed to process food and symptom diaries with the system described in this article. Received for publication June 2, 1994; revision received Sept. 6, 1994; accepted Sept. 30, 1994. Reprint requests: Theodore Kueper, Wisconsir~ Data Laboratory, Ltd., 415 Cymric Ct., Wales, WI 53183. Copyright © 1995 by the American Academy of OtolaryngologyHead and Neck Surgery Foundation, Inc. 0194-5998/95/$3.00 + 0 23/1/61t945
questionnaires and patient histories depend on recall and are subjective for both the patient and the physician. Elimination diets of many types are used to uncover food sensitivities. These are slow procedures and do not represent the normal dietary load of problem foods. Our diary interpretation has two assumptions: (1) the changes in chronic symptom levels may be statistically related to the amounts of common food components, and (2) the weights for most components of related biologic families 6 may be combined on a dry protein basis. A literature search found no articles during the last 10 years that use these assumptions in food and symptom diaries. A commercial software package (Well Aware Food Sensitivity User's Guide; Positech, Monterey, Calif.) based on ciuster analysis of food cliaries requires the first assumption, but the model and results are not published. This artMe introduces an improved method for discovering relationships between the patient's normal foods and symptoms. The results facilitate the design of elimination or avoidance diets and thus help to control food-caused symptoms. METHODS Patient Selection
Patients were primarily from two Midwestern practices. Orte practice specialized in ENT and 415
416
Otolaryngology Head and Neck Surgery March 1995
KUEPERet al.
allergy. The second practice specialized in environmental medicine and allergy. Most patients had had previous evaluations and immunotherapy treatment for aeroallergens but had had limited benefits. Most otolaryngologists' patients had symptoms related to the head and neck. These symptoms included headache, facial pressure, postnasal drainage, throat clearing, persistent cough, gastroesophageal reflux, ear plugging, and tinnitus. The initial visit included a history and physical examination to screen for other medical or surgical problems. This occasion was also used to discuss the possibility of an allergic basis for the problem. An allergy evaluation sheet and samples of antihistamines, nasal sprays, or decongestants were sent home as part of the first visit. At the follow-up visit (1 to 2 weeks later), the allergy evaluation sheet and an evaluation of sample medications were discussed with the patient. At a third visit, if medication requirements were still daily, an allergy assessment was suggested, including a screening immunoglobulin E radioallergosorbent or skin testing primarily for regional aeroallergens. After evaluating the radioallergosorbent or skin test screening and follow-up titration testing, the patient was placed on hyposensitization injections for a several-month trial. After this, the food and symptom diary was recommended for cases in which symptoms, especially headache, persisted. Patient selection in the environmental medicine practice was based on similar criteria. After a history and physical examination, most patients were examined and treated first for aeroallergens and then for foods with a combination of skin testing, sublingual drops, and food avoidance. If symptoms persisted, the food and symptom diary was recommended. The patient population was 60% female adults (18 years and older), 20% male adults, 10% female children, and 10% male children. Patients were told that the diary would be analyzed by a new computer program to help identify their problem foods. They were also told that the resultant data, coded for anonymity, may be used in publication of the results. Patient Instructions
The physician guided the patient in selecting any of the patient's symptoms that might be related to adverse reactions to foods. Up to five symptoms could be evaluated during the diary. Thus each patient's symptom pattern was specific to that patient. The symptoms were expected to vary in intensity at least four times during the recording period.
Symptom intensity was rated on a 9-point scale (1 = perfect, 9 = very, very sick). All symptoms were to be rated at each meal or snack occasion, and preferentially upon waking and retiring, and also upon a significant change of any symptom. Table 1 gives a summary of patient instructions. Patients were instructed to maintain their regular dietary routines. Patients were instructed to record to the nearest hour all food, drink, and medication for 2 weeks. A summary set of instructions asks the patient (1) to record food, drink, and medications at the nearest hour; (2) to be specific and show brand names and restaurant names; and (3) to list ingredients in nonstandard and home recipes. Single servings for the patient were assumed unless they noted otherwise. Patients were encouraged to clarify recipes by comment and by identification of brand names and restaurants. Patients were contacted by phone to clarify any obvious omissions. Most clarification regarded ingredient alternatives for home recipes. Data Analysis
Each symptom was treated as a dependent variable as long as it was rated for at least 10 consecutive days and there were at least four episodes of symptom change. Each hour with an entry for either foods or symptoms was considered a single record for computer entry. The data matrix for regression analysis consisted of the symptom ratings as dependent variables and a summation of food component indexes for each time period. Six accumulation times (3, 6, 12, 18, 24, and 48 hours) for symptoms were examined. An accumulation at 6 hours, for example, was the total amount of each food component ingested for the current hour and the previous 5 hours. Food records were converted to ingredient lists with a recipe database developed by us. New food recipes were created as required from cookbooks and commercial sources or by consultation with the patient. The recipe ingredients and additives, called components, were summed for a modified list of 90 biologic families6 plus lactose, caffeine, and alcohol. The indexes for summation were grams of protein for the biologic family, lactose, and alcohol, or milligrams caffeine. Refined foods ostensibly containing no protein were assigned a trace (0.01) value of the appropriate biologic families. Table 2 has examples in the column "Components" of the modified list for 37 biologic families. The components were then screened down to the 23 most frequent to evaluate (1) the data with most variation, (2) the variables most common to the
Otolaryngology Head and Neck Surgery Volume t 12 Number 3
patient's diet, and (3) whether they fit the concept that "recurring symptoms may be due to recurring foods." Time of day was included as an independent variable along with the 23 most frequent food components. The use of time of day as a possible factor helps to identify trends in symptoms tlhat may not be caused by foods. There were frequently five symptoms as dependent variables and 24 independent variables. Each symptom was examined with the six accumulation times. The resultant set of 720 t values on standardized coefficients was sorted into a report format that lists significance at low (above t = 2.7 or about 0.01 level of significance) average (above 3.3) and high (above 4.0) confidence. Reports
The main reports a r e a ranking of suspect components and a listing of the patient's sources for these components. Tables 3 and 4 give a typical example. The first column gives the suspected food component and the frequency of ingestion during the diary. Column two identifies the symptoms affected by the component. Column three shows the statistical confidence and either direct for a direct association between food and symptom, or mask for the appearance of symptom relief. In ranking the suspect components, direct was weighted somewhat higher than mask. A mask could represent either the abatement of a withdrawal reaction or an actual easing of symptoms due to exclusion of offending foods. Column four, labeled lag time, shows the time required from ingestion to the symptom change. This lag time may also represent the accumulation time required to produce a burdensome load of a food. The codes for programs used to conduct the process are beyond the scope of this article. The programs were written with Turbo-C, Paradox, and Quattro-Pro (Borland, Inc., Scotts Valley, Calif.) for IBM-compatible personal computers. Quattro-Pro has the regession analysis available. Physician Use and Interpretation
The usual treatment for the now-suspect food components was some combination of avoidance and challenge. Patients were told that these food components were merely suspect and that statistical relationships did not necessarily indicate cause and effect. Although the system is not designed as a nutrition analysis, dietary patterns occasionally revealed a need for nutrition counseling. The physician had the responsibility of interpret-
KUEPER et al.
4t7
Table I. Summary of patient instructions Choose and name up to 5 symptoms on the form Rate your symptoms using the 9-point scale On arising in the morning At each meal Whenever symptoms change significantly Just before retiring at night Record your food, drink, and medications at the nearest hour Be specific, show brand names, restaurant names List ingredients in nonstandard recipes Keep the food and symptoms diary for 2 weeks
Symptom rating scaie: 1, perfect; 3, couid feel better; 5, feeling very uncomfortable; 7, need to take something for problem; 9, ve,'y,very sick.
ing the result as suggesting food aller~, delayed food sensitivity, pharmacologic effect, and so forth. The lag time (from the accumulation time) reported, the direct or masking nature, and the other common characteristics (for example, high fat content) of the components often give support to theories or mechanisms. RESULTS
The most common rated symptoms were headache, fatigue, stuffy or runny nose, bloating, cough, diarrhea, gas and joint pain. One hundred sixty-four patients generated 125 different names for a wide variety of symptoms. The average 2-week diary contained symptom rating sets recorded 83 times, 90 meal or snack occasions, 118 names of different foods, and 243 total foods. Wheat, the legume family, eggs, milk, lactose, the lily family, the mustard family, and caffeine were the most common suspect components. There were an average of 9.1 components identified as suspect per patient and an average of 3.1 symptoms associated with suspect food components for each patient. Forty-nine patients provided in-home uncontrolled elimination diet testing for individual foods. Table 2 shows 105 (47%) of 222 suspected food components were confirmed by these tests. For example it shows citrus, alcohol, the lily family, and pome components were confirmed with 80% or bettet rates. Chicken, the mustard family, and the composite family components had Iess than 20% confirmation rates. The food components with highest levels of confidence gave the highest rates of confirmation. These open confirmations are not of double-blind, placebo-controlled quality. The rate of unequivocal confirmation will require further study. During the in-home test, the patients were instructed to note symptoms while avoiding a select food for 4 to 7 days and then to reintroduce the food.
418
Otolaryngology Head and Neck $urgery March 1995
KUEPERet al.
Table 2. Results of in-home avoidance and challenge tests for 49 patients Component
Examined
Suspected
Challenged
Confirmed
% Suspected
% Challenged
% Confirmed
Legume family Lily family Caffeine Milk Eggs Corn Banana Carrot family Pork Pome family Wheat Beef Gourd family Lactose Grape Citrus family Potato or pepper Tomato Oat or rye Rice Alcohol Chicken Composite family Mustard family Berry family Fish Cashew family Buckwheat family Stonefruit family Poultry (not chicken) Goosefoot family Heath family Walnut Pineapple Laurel family Mint family Other* TOTAL
47 45 39 46 44 42 25 41 34 26 46 44 24 41 23 31 28 23 27 23 6 34 27 41 19 18 3 2 23 10 4 6 1 2 3 5 62 967
23 22 20 22 24 17 14 19 17 13 24 23 9 20 10 13 12 11 21 13 3 22 15 19 8 7 3 1 5 4 4 3 1 1 1 1 28 473
16 12 12 11 12 9 11 9 7 5 11 11 6 4 4 3 7 5 8 7 2 10 8 7 4 2 1 1 4 1 0 0 0 0 0 0 12 222
10 10 7 7 6 6 5 4 4 4 3 3 3 3 3 3 3 3 2 2 2 1 1 1 1 1 1 t 0 0 0 0 0 0 0 0 5 105
48,9 48.9 51.3 47.8 54.5 40.5 56,0 46.3 50.0 50,0 52,2 52.3 37.5 48,8 43.5 41,9 42.8 47.8 77,8 56.5 50.0 64,7 55.6 46,3 42.1 38,9 100.0 50.0 21,7 40.0 100.0 50.0 100,0 50.0 33.3 20,0 45.1 --
69,6 54.5 60.0 50,0 50.0 52,9 78.6 47.4 41,2 38,5 45,8 47.8 66.7 20.0 40,0 23.1 58.3 45,4 38.1 53.8 66.7 45.5 53,3 36.8 50,0 28.6 33,3 100,0 80,0 25.0 0.0 0,0 0,0 0.0 0.0 0.0 42.8 --
625 83,3 58.3 63.6 50.0 66.7 45,5 44.4 57.1 80.0 27,3 27,3 50.0 75.0 75.0 100,0 42.8 60.0 25,0 28.6 100.0 10.0 12,5 14.3 25.0 50,0 100.0 100.0 0.0 0,0 0,0 0.0 0,0 0.0 0,0 0.0 41.7 --
*Other includes yeast, fungi, and combination named components from singularity elimination,
Two repetitions were suggested. This test protocol does not specifically test the overall model, because the model is essentially dependent on the "total load" accumulation of components. In at least one situation, the patient did not confirm individual foods but noted that symptoms disappeared after the suspected components were avoided as a group. The false-positive rate of 52% from the open in-home testing agrees well with a calculated estimate of 51% on the basis of the many relationships being analyzed in the regressions. Most of the patients did not report results of individual food component challenges. About 75% reported improvements due to general avoidance or reduction of suspected foods. The improvements reported by 50% of the patients were clouded by concomitant treatments such as immunotherapy or
new medication. A subset of 25% reported clear, independent improvement that they fully credited to the dietary changes. About 25% reported no improvement or no confirmation of any foods. A 1-year follow-up of patients who noted initial improvements with the food and symptom diary reports revealed no apparent problems with longterm usefulness. The patients continue to report symptomatic control as long as they "watch what they eat." They do not always adhere to the detailed findings, but they feel empowered to control their symptoms and return to a controlling diet as needed. The recording process is tedious for most patients. Two weeks was selected for a reasonably short time period, which gives adequate representation of the dietary and symptomatic changes associated with weekends vs. work or school days. Diary initiation
Otolaryngology Head and Neck Surgery Volume I'12 Number 3
KUEPER et ai.
419
Table 3ù Main report of statistical relationship of food components and symptoms (exampie omits patient and doctor identification details] Food component [no. times in diet) Chocolate (13)
Tea (7) Composite* (6) Fish (5) Beef (11) Time of day Wheat (40) Gourd (4)
Symptom
Regression result
Lag time
Headache Stuffy nase Ears Fatigue Stomach Headache Headache Ears Stomach Headache Headache Headache
Average, mask Average, mask Low, mask Low, mask Average, direct Low, mask Average, direct Average, direct Low, direct Low, direct Low, direct Low, mask
12 to 48 hr 6 to 24 hr 18 to 24 hr 6 to 48 hr 6 to 18 hr 12 to 18 hr 18 to 24 hr 0 to I2 hr 3 to 12 hr lncreasing 18 to 48 hr 12 to 18 hr
Direct, food component associated with adverse reactions; mask, food component temporarily eases symptoms through the lag time but may then cause adverse reactions; high, average, and Iow, refer to the degree of statistical confidence. The following food components were also in your diet and had at least four diary appearances (number in parentheses shows number of times food component was in diet): citrus (13), eggs (26), fungi (35), min (33), mustard* (16), chicken (8), tomato (1I), iegume* (18), corn (27), caffeine (29), carrot* (5), lily (16), potato* (10), mint* (6), park (13), and lactose (33). *Refers to biologicN family or subgroup. See "Food Components" explanation. This report is for your doctor. The report summarizes the results from regression analysis with symptoms as the dependent variables and the food components as the independentvariables. Statistical results da not necessarilycorrespond to cause-and-effect retationships.Your health professionatcan use the information along with other data to cOnSLIt with you on any dietary changes.
difficulties arose because some patients, appeared reluctant to share information about their eating habits. Special events, changes in home situations, or illness were also cited as reasons not to start or to stop keeping records. It seemed important to stress that the computer would be analyzing the data and that there would be no judgments made regarding the contents of the diary. Several unacceptable or incomplete diaries occurred during the early phase of o ür research. Instructions were improved to correct incomplete use of rated symptoms. The concept of rating a symptom when it is temporarily minimal needed special attention.
Table 4. Example of sources for first two components of main report Component Percent«ge Frequency Chocolate
Source
70.3 10.1 8.1
7 1 1
6.3
1
2.8 2.3
1 1
Chocolate Poptart Chocolate mint White chocolate mousse frozen yogurt Drumstick ice cream cone/carmel/nuts Brownie Cake/chocolate
57.1 14.3 14.3 14.3
4 1 1 1
Sun tea/sugar Iced tea/sugar Sun tea Tea
Tea
DISCUSSION The results of this limited group of patients appear encouraging. The inherent inaccuracies of home-recorded data and ingredient and amount estimation may be overcome by the block of data, which allows an objective comparison of food accumulations to the symptoms. The identification of specific food items that contain the suspected food components also appears useful. The model gives total weight to each patient's unique diet during the recording period. No comparison is made to any other patient averages or norms, as is the case for most laboratory tests. The results reflect the relationship between diet and usual symptoms to the degree that the foods selected
during the 2-week period reflect the normal diet for that patient. The most basic assumptions concern the sensitivity of the patient. The model assumes that the symptom levels are related to the accumulated components. Each component has its own weight in symptom production. Components may mask symptoms and directly accentuate symptoms. The timing of effects may be delayed both as direct effects and as masked reactions. One particularly troublesome feature of this model is that it does not differentiate effects duo to masked reactions and to actual symptom relief due to dilution or total replacement of problem foods with tmly safe foods.
Head
420
KUEPER e t al.
The system generates many questions that we hope will be clarified in time and by other researchers. We hope revisions will improve the reliability of these models. At this point of development the model does not attempt to weight the onset of symptoms compared with the sustained presence of symptoms. No attempt is made to recognize effects of medications or environmental or emotional changes. The times selected a r e a compromise to accommodate gut dwell times, normal meal spacing, and many of the delayed mechanisms of food intolerance. Data transformations and other types of statistical analysis may be appropriate. At this stage of our investigation, we cannot separate the individual importance of the major steps in the process. The patients selected were screened for other problems, and perhaps food intolerance was found frequently because other diagnoses were eliminated. Taking the effort to monitor their own diet and symptoms may have affected the patients' belief that foods were a problem. The extensive and objective determination of each patient's most frequent food families could have been another key step. We presume that some of the results are due to placebo effects. The food and symptom diary procedure involves a very personalized computer report. The avoidance and challenge tests include both
Otolaryngology and Neck Surgery March t995
dietary and lifestyle changes. Research has begun with the "safe" foods listed in the food and symptom diary report. These foods would be reversed with the suspected foods, thus providing a control (placebo) report. The in-home doubled-blinded testing with two reports, one true and one false, should clarify this issue. REFERENCES 1. Kettelbut BV, Metcalfe DD. Adverse reactions to foods. In: Middleton E, Reed CE, Ellis EF. Allergies, principles and practice. St. Louis: C.V. Mosby, 1988:1488. 2. American Academy of Allergy and Immunology. Adverse reactions to foods, a patient's guide to problem foods, food additives, diagnosis, treatment. Based on the article: Adverse reactions to food. Bethesda, Md.: National Institutes of Health; July 1984. Publication no. 84-2442;p.1. 3. Golbert TH. Food allergy and immunologic diseases of the gastrointestinal tract. In: Patterson R. Allergic diseases diagnosis and management. Third ed. Philadelphia: J.B. Lippincott Co., 1985:464-8. 4. Rowe AH. Food allergy. Springfield, Ill.: Charles C. Thomas, Publisher, 1972:38. 5. King HC. An otolaryngologist's guide to allergy. New York: Thieme Medical Publishers, 1990. 6. Anderson JA, Sogn DD. Adverse reactions to foods, American Academy of Allergy and Immunology Committee on Adverse Reactions to Foods, 1984. Bethesda, Md.: National Institute of Allergy and Infectious Diseases, United States Department of Health and Human Services, Public Health Service, National Institutes of Health;1984:17-25.