PATIENTS with multiple chemical sensitivity (MCS) (also known as environmental illness and environmental sensitivity) experience a variety of symptoms in response to low-level exposures to a wide range of substances. (1-3) The variety of symptoms reported, the lack of consistency in physical findings or laboratory test results--together with the variability of substances that reportedly provoke symptoms--make it difficult for investigators to formulate a case definition of MCS. (4) However, 7 case definitions have been proposed, (4-10) all of which incorporate several features (Table 1). All definitions recognize 1 essential feature: the link between symptoms and low-level exposures. Other features include the nature of onset of the condition, (6) involvement of the nervous system (4,8); involvement of multiple body systems (4,6,10) chronicity (4,9,10); whether there has been an increase in the number of exposures, which incite symptoms (8). whether avoidance of exposures relieves symptoms (4-10); whether multiple, unrelated substances provoke symptoms (4,6,8-10); whether patients demonstrate addictive responses to any substances (8); and severity of the condition. (4,8)
There is no "gold standard" for diagnosing MCS; therefore, our purpose was to determine which case definitions had the greatest discriminant validity. We predicted that there would be a higher prevalence of patients with MCS in environmental practices than in general practices. (In this study, an environmental practice was defined as a practice run by an environmental physician who had a special interest in treating patients with MCS.) We also assumed that occupational health practices (referred to as "occupational practices" herein) and allergy practices would have an intermediate prevalence of patients with MCS. We proposed that case definitions with the greatest discriminant validity should best distinguish patients who attended practices that we assumed had the highest and lowest prevalences of MCS.
We used the University of Toronto Health Survey (UTHS) self-administered questionnaire to determine which of the existing case definitions best distinguished patients who attended the 4 specified practice types. The questionnaire assessed the prevalence of 171 symptoms, 85 provocative exposures, and particular MCS features described in 7 published case definitions. (4-10) The reproducibility of the UTHS has already been reported. (11)
In this article, we describe the UTHS questionnaire, its overall response rate, and the degree to which features from existing case definitions discriminate among the practice types. In addition, we identified 4 symptoms associated with the likelihood of MCS. The questionnaire is available from the authors upon request.
Materials and Method
Prior to distribution of the questionnaires, we obtained ethical approval from the university and hospital ethical review boards. Completion of the questionnaire constituted informed consent from the respondent.
Questionnaire. The UTHS is a self-administered questionnaire that, in addition to general health status and demographic characteristics, contains questions about symptoms, exposures, and features of published case definitions of MCS. Definitions include those of Thomson et al., (4) the National Research Council (NRC), (5) Cullen, (6) Ashford and Miller, (7) Randolph, (8) Nethercott et al., (9) and the 1999 Consensus. (10)
We assessed features of the case definitions in the questionnaire, as follows:
* To determine the presence of symptoms linked to low-level exposures, as required by all 7 case definitions, (4-10) we asked UTHS respondents if they had experienced any of 171 symptoms during the 12 mo that preceded answering the survey. We also asked respondents if these symptoms were provoked by exposures "at a level that does not seem to bother most people" or if symptoms were relieved if they avoided the exposures.
* To assess the involvement of multiple body systems, as required in 3 definitions, (4,6,10) we grouped symptoms anatomically in the questionnaire. The groups were as follows: eyes, ears, nose, mouth, throat, lung, heart/circulation, blood/glands, muscles/joints, nervous system, stomach/bowel, bladder/genital, and skin.
* We addressed the requirement for nervous system symptoms, as discussed in 2 definitions, (4,8) by including the nervous system among multiple body systems.
* To determine whether the condition was chronic, as required in 3 definitions, (4,9,10) we adopted Thomson's criterion (4); we defined chronicity as being present if exposure-linked symptoms had occurred during 3+ mo of the 12-mo reporting period.
* To determine whether the condition was acquired in response to a particular event, as required only by Cullen, (6) we included questions for which the response would identify the month and year of symptom onset, as well as any exposures or life events associated with symptom onset.
* To assess whether there had been a spreading of incitants, (8) we included questions that addressed whether the number of exposures associated with symptoms had increased over time.
* To determine whether respondents exhibited addictive responses, (8) we included questions that addressed whether the respondents ever craved any food, drink, or smell and whether anything acted as a "pick-me-up."
* To determine whether unrelated substances provoked symptoms, as required in 3 definitions, (6,9,10) we included questions about which of 85 substances provoked symptoms. However, none of the case definitions specifies whether multiple substances must provoke symptoms in individual patients, or whether any one individual may be sensitive to only one substance, but among a population of MCS patients, many substances may provoke symptoms. In addition, no criteria are provided to identify unrelated substances. Therefore, we could not include "unrelatedness" in this analysis.
* To assess the varying morbidity discussed in 2 definitions, (4,8) we included a scale in the questionnaire that indicated the severity of symptoms experienced in each body system. However, given that these definitions provide no guidance about what level of morbidity is most likely associated with MCS, we did not use this feature in the analysis.
Features of case definitions that required physical examination, (4,8) application of diagnostic tests, (4,6) challenge with exposures under controlled conditions, (7,8) or the eliciting of symptoms following demonstrable exposures (6,10) could not be assessed via questionnaire; therefore, we did not consider them any further in this study.
During the development phase, we assessed the UTHS questionnaire for face and content validity and clarity on the basis of (1) 3 pretest studies and (2) repeated consultation with physicians who were experienced in the diagnosis and management of MCS and with representatives of patient support groups.
Study participants. All English-speaking adult patients (ages 16-74 yr) who attended any one of (1) 2 environmental health practices in the Toronto area, (2) 2 general practices, (3) the Occupational Health Clinic of St. Michael's Hospital, and/or (4) 2 allergy practices during the period between January 1, 1994, and December 31, 1994, were invited by their physicians and the study investigators to complete the UTHS questionnaire during the period between January 1, 1995, and April 30, 1995. Each respondent was given a stamped, self-addressed envelope in which he or she could return the questionnaire to the University of Toronto. To enhance the response rate, nonresponders received a telephone contact, a reminder card, and, when necessary, a second mailing.
Analysis. We used a 2-stage procedure to determine whether a respondent satisfied any of the 7 published case definitions, as reflected by the selected features described previously. We first examined responses to determine whether individual features of each case definition were fulfilled. We then combined the features of the case definitions to determine if individual respondents satisfied any of the case definitions.
Discriminant validity of published case definitions. To determine which of the published case definitions (as we applied them) best distinguished patients in the different clinical practices, we used logistic regression analysis to estimate odds ratios for each case definition (separately for men and women). First, we estimated odds ratios; patients of environmental physicians were compared with those of general practitioners (i.e., groups that we expected to differ the most with respect to MCS prevalence). We then ranked the case definitions by magnitude of their odds ratios. Second, we calculated odds ratios to compare patients from general practices with those from the practices of occupational physicians and allergists. Thus, we were able to determine whether the case definitions were ranked similarly in the 4 practice types.
Symptoms that distinguished between clinical practices. To identify symptoms that best distinguished patients in environmental practices from those attending general practices, we used logistic regression analysis to estimate odds ratios for each symptom, comparing patients from environmental with those from general practices. However, for diagnostic purposes, symptoms were needed that not only had a high odds ratio (indicating that environmental patients differed from general practice patients) but were also reported by a high proportion of environmental practice patients (i.e., had high "sensitivity"). Attention was thus restricted to symptoms reported by 50% or more of respondents attending the environmental practices and that also had the 10 highest odds ratios observed in men and women separately. On the basis of these symptoms, multivariate models were developed separately for men and women to predict whether patients came from environmental or general practices.
Response rate. Overall, 61.7% of the patients responded to the UTHS questionnaire. The numbers of questionnaires distributed to each of the 4 practice types, response rates by gender, and mean age of respondents are shown in Table 2. The response rate was lowest from the general-practice setting (i.e., 54% versus 60% elsewhere). The gender distributions among the responders were similar to the distributions observed within the practices in which the responders were based, with the exception of environmental practices. In these practices, gender distribution data were not in machine-readable form, and gender inferences on the basis of names could be made for only 90% percent of potential respondents. In all practices, except occupational health, almost 70% of the patients were female. Mean ages for men and women were younger in allergy and general practices than in the environmental and occupational practices. Given that differences between practices existed in the age distribution of patients, we confined the analysis to persons aged 30-64 yr--an age range that captured about 80% of participants and yielded similar age distributions for analysis among the practices.
Validation of published case definitions. For each case definition, substantially more persons likely to have MCS were identified among patients in environmental practices than in general practices, and, as predicted, the prevalence of such persons in occupational and allergy practices was intermediate between the other 2 practices (Table 3). The prevalence of MCS likelihood in each practice differed markedly across case definitions, as applied in this study. For example, more than 90% of patients in environmental practices satisfied the features from the NRC's and Ashford and Miller's definitions. This percentage was identical for both because only 1 feature was required (i.e., patients report any symptom linked to exposure). In contrast, only approximately 50% of patients satisfied features selected from Cullen's more stringent case definition, which required (1) symptoms linked to low-level exposures, (2) acquisition in response to an event, (3) multisystem involvement, and (4) provocation by unrelated substances.
The odds ratios that reveal the discriminant validity of each of the case definitions are shown in Table 4. The highest odds ratios were observed for individuals who met the case definitions reported by Nethercott et al. and the 1999 Consensus. Men and women who satisfied features from these case definitions were approximately 20 times more likely to be patients in an environmental practice than in a general practice. Features from the Nethercott case definition and the 1999 Consensus also were distinguished most strongly between patients in general practices and those in the occupational and allergy practices for both males and females. Odds ratios for other case definitions were lower, and they did not rank consistently for males and females or between practices. The Nethercott case definition and the 1999 Consensus, therefore, produced the greatest discriminant validity, as assessed by odds ratios (Table 4), compared with other case definitions. Both definitions correctly identified about 90% of females and 80% of males who attended environmental practices and about 73% and 82% of females and males, respectively, who attended general practices.
Symptoms that distinguished between clinical practices. Among symptoms with a high prevalence in environmental practice patients, the 10 symptoms with the highest odds ratios that distinguished environmental from general practice patients are shown in Table 5. Seven of these 10 symptoms were common to both sexes.
On the basis of these subjective symptoms, we used logistic regression to obtain a multivariate model that satisfied the following criteria: (1) the model should discriminate patients who attended an environmental practice from those who attended a general practice, (2) a small number of symptoms should be involved, and (3) the same symptoms should provide good prediction for both males and females.
Models based on the following 4 symptoms satisfied the aforementioned criteria: (1) having a stronger sense of smell than most people, (2) feeling "spacey," (3) feeling dull or groggy, and (4) having difficulty concentrating. Females who reported having a stronger sense of smell than others, or who reported at least 2 of the remaining symptoms (i.e., feeling spacey, feeling dull, or having difficulty concentrating) were more likely to attend environmental than general practices. These 4 symptoms correctly classified 83.4% of responders who attended environmental physician practices ("sensitivity") and 66.5% of responders who attended general practices ("specificity"). Similarly, males who reported any 1 of (1) stronger sense of smell than others, (2) dull/groggy, and (3) difficulty concentrating, or who had at least 2 symptoms (i.e., spacey plus 1 other) were more likely to attend environmental practices. These symptoms led to the correct classification of 77.3% of males from environmental practices and 71.4% percent of males from general practices. The same 4 symptoms identified an intermediate percentage of male and female respondents in occupational medicine practices (41.8% and 58.55, respectively) and allergy practices (47.2% and 48.4%, respectively).
Clinical features from the Nethercott definition and from the 1999 Consensus produced the greatest discriminant validity for distinguishing environmental from general practice patients, as assessed by odds ratios. They also yielded the expected intermediate prevalences in the occupational and allergy practices. In addition, features from these 2 definitions exhibited the greatest reproducibility in terms of patients' self-reporting of the relevant features. Based on UTHS results, the Nethercott et al. case definition and the 1999 Consensus have advantages over other definitions for identifying patients with the greatest likelihood of having MCS (i.e., environmental practice patients) and for excluding patients with the least likelihood of having MCS (i.e., general practice patients). This conclusion is strengthened by the fact that the UTHS displayed good reproducibility with respect to self-report of symptoms and, in particular, achieved good agreement for satisfying not only the Nethercott criteria with Kappa = 0.68 (95% confidence interval = 0.54, 0.82) but also the 1999 Consensus criteria (Kappa = 0.78; 95% confidence interval = 0.66, 0.90). (11)
Replication is necessary if we are to establish whether our findings are generalizable to other populations with different health insurance plans, different types of practices, and, possibly, different patient characteristics. Furthermore, we cannot ignore the fact that respondents who satisfied the Nethercott et al. definition or the 1999 Consensus, as applied in this study, also reported features mentioned in other case definitions. For example, 98.6% of respondents from environmental practices reported symptoms that affected more than 1 body system, 84.7% reported symptoms that affected the nervous system, 60.6% reported symptoms initiated in response to a specific event, 81.4% reported an increase in the number of exposures that provoke symptoms, and 85.2% reported addictive responses. These features may well be associated with MCS, but our findings suggest that the addition of this information does not improve the discrimination achieved with the Nethercott definition or the 1999 Consensus criteria.
The UTHS could not assess case-definition criteria, which require physical examination, (4,8) application of diagnostic tests, (4,6) challenge with exposures under controlled conditions, (7,8) or eliciting symptoms after demonstrable exposures. (6,10) Including these elements of case definitions, which is not possible in a self-administered questionnaire, might alter the accuracy of classification that we reported for the case definitions. Other features of existing case definitions reveal many uncertainties (Table 1). How low is a "low-level exposure?" How quickly must symptoms appear following exposure or disappear after withdrawal in order that they be linked to exposure? Must the symptoms occur only after exposure? Must they disappear entirely after withdrawal of exposure, or only partially? Does multisystem mean anatomic or physiological systems? Does multisystem involvement mean several systems must be involved in one individual? Is single-system involvement sufficient in a single individual, with a range of systems being observed across individuals? For how long and how recently should symptoms have been occurring before they are considered to be chronic? What types of events may provoke onset of symptoms--only those associated with chemicals? How large an increase in the number of substances provoking symptoms over what period of time would confirm that there was "spreading of incitants?" What is the definition of unrelated substances? If morbidity is variable, is any particular level most likely to indicate the presence of MCS? Despite all these unanswered questions, the case definitions, with 2 exceptions, did yield surprisingly good discrimination. Research is needed for the evaluation of these aspects of the case definitions.
Completion of the UTHS took most respondents between 1 and 3 hr. We, therefore, sought to identify a small group of symptoms that might identify patients with the greatest likelihood of having MCS. Models based on 4 symptoms were able to distinguish strongly between environmental- and general-practice patients for both males and females. The four symptoms were (1) having a stronger sense of smell than other people, (2) feeling spacey, (3) feeling dull or groggy, and (4) having difficulty concentrating. In principle, it should be possible to validate at least some of these subjective symptoms. In 1988, Doty et al. (12) investigated the MCS symptom of "greater than normal olfactory sensitivity." A comparison of 18 MCS patients with 18 controls, based on challenges in a controlled environment with 2 chemicals, revealed equivalent olfactory thresholds in the 2 groups. (12) In the UTHS questionnaire, respondents were asked if they had "a stronger sense of smell than others." The responses to such a question may not reflect the results of challenge testing. Perhaps MCS patients experience a greater intensity of an odor without their olfactory threshold being different from controls. In the current study, we evaluated the discriminant validity of reported symptoms for categorizing participants without attempting to evaluate the biological validity of the symptoms. The underlying biological mechanisms for the reported symptoms should be explored.
All 4 discriminating symptoms involved the nervous system, a result consistent with the opinions of Thomson (4) and Randolph, (6) both of whom indicated that neurologic features are an important aspect of MCS. Although these symptoms could be useful for identifying patients who are more likely to have MCS, other combinations of symptoms may also be associated with the condition. Currently, we are conducting additional analyses to identify clusters of symptoms that occur in patients who attend environmental practices.
With MCS, symptoms are linked to low-level exposures; yet, the 4-symptom combinations we have described were based on reported symptoms and not on reported exposure-linked symptoms. This was the case because, in the absence of challenge testing, some respondents might incorrectly attribute symptoms to exposures or fail to recognize that their symptoms are associated with exposures. Given that both bias and random error might affect any reported linkage of exposures to symptoms, we decided to conduct our analysis on the basis of the presence or absence of symptoms, regardless of whether the symptoms were attributed to exposure. It is likely that even higher discriminant validity could be achieved if prediction were based on symptoms linked accurately to exposure.
To our knowledge, this is the first study in which the investigators undertook a comparison of symptoms in patients from practices in which the prevalences of MCS were likely to differ. Although more than 60% of subjects responded to the questionnaire, we cannot rule out the possibility that selection biases occurred among respondents. However, if confirmed in other studies, our findings suggest that there are clinical patterns--especially those reflected in the case definition by Nethercott et al., the 1999 Consensus, and in the UTHS 4-symptom combinations--that are likely associated with MCS. Our conclusions may be helpful in the further investigation of etiology, diagnosis, and therapy for MCS.
(1.) Ashford NA, Miller CS. Chemical exposures: low levels, high stakes. New York, NY: Van Nostrand Reinhold, 1991.
(2.) Randolph TG. Human ecology and susceptibility to the chemical environment. Springfield, IL: Charles C Thomas, 1962.
(3.) Bell IR. Clinical ecology: a new medical approach to environmental illness. Bolinas, CA: Common Knowledge Press, 1982.
(4.) Thomson GM, Day JH, Evers S, et al. Report of the ad hoc committee on environmental hypersensitivity disorders. Ontario Ministry of Health, 1985; pp 17, 18.
(5.) National Research Council, Board on Environmental Studies and Toxicology, Commission on Life Sciences. Multiple Chemical Sensitivities: Addendum to Biologic Markers in Immunotoxicology. Washington, DC: National Academy Press, 1992; pp 5-7.
(6.) Cullen M. The worker with multiple chemical sensitivities: an overview. In: Cullen M (Ed). Workers with multiple chemical sensitivities. IV. Occupational Medicine: State of the Art Reviews. Philadelphia, PA: Hanley and Belfus, 1987; pp 655-62.
(7.) Ashford NA, Miller CS. Case definitions for multiple chemical sensitivity. In: National Research Council, Board on Environmental Studies and Toxicology, Commission on Life Sciences. Multiple Chemical Sensitivities: Addendum to Biologic Markers in Immunotoxicology. Washington, DC: National Academy Press, 1992; p 43.
(8.) Randolph TG. Ecologic orientation in medicine: comprehensive environmental control in diagnosis and therapy. Ann Allergy 1965; 23:11-22.
(9.) Nethercott JR, Davidoff LL, Curbow B, et al. Multiple chemical sensitivities syndrome: toward a working case definition. Arch Environ Health 1993; 48:19-26.
(10.) 1999 Consensus on Multiple Chemical Sensitivity. Multiple chemical sensitivity: a 1999 consensus. Arch Environ Health 1999; 54:147-49.
(11.) McKeown-Eyssen GE, Sokoloff ER, Jazmaji V, et al. Reproducibility of the University of Toronto Self-Administered Questionnaire used to assess environmental sensitivity. Am J Epidemiol 2000; 151:1216-22.
(12.) Doty RL, Deems DA, Frye RE, et al. Olfactory sensitivity, nasal resistance, and autonomic function in patients with multiple chemical sensitivities. Arch Otolaryngol Head Neck Surg 1988; 114:1422-27.
Submitted for publication May 11, 2000; revised; accepted for publication December 4, 2000.
Requests for reprints should be sent to Cornelia J. Baines, M.D., Department of Public Health Sciences, 12 Queens' Park Cres. West Toronto, Ontario, M5S 1A8, Canada.
COPYRIGHT 2001 Heldref Publications
COPYRIGHT 2002 Gale Group