Immunologic, Psychological, and Neuropsychological Factors in Multiple Chemical Sensitivity: A Controlled Study

Methods

Study Sample

Patients with chemical sensitivity were recruited from the practice of a community allergist with special interest in treating chemical sensitivity. All patients with a recorded diagnosis of chemical sensitivity seen between January 1989 and June 1990 were identified from computerized billing records. We then reviewed standardized history questionnaires included in clinical records for the following eligibility criteria: duration of illness 3 months or more; symptoms reported in at least three organ systems, including the central nervous system; and reported sensitivity to 4 or more substances from a list of 14 common exposures including fresh paint, newspapers, perfume, hair spray, and solvent fumes. All potential participants were sent a letter requesting their participation. Of the 76 patients with chemical sensitivity identified from records, 32 did not respond to letters and 3 were judged to be ineligible (2 reported complete resolution of chemical sensitivity and 1 described respiratory symptoms only). The remaining 41 patients (56% of those potentially eligible) were enrolled.

Controls were selected from two university-based clinics for patients with musculoskeletal injuries: an occupational musculoskeletal clinic and a back injury specialty clinic. We reviewed clinic charts to identify all patients seen between January 1989 and June 1990 whose ages were within the range of the chemical sensitivity cases and who resided within 25 miles of the study clinics. Patients with any systemic illness (such as rheumatoid arthritis) likely to affect the results of immunologic testing were excluded, and potential controls were frequency matched to cases by sex and 5-year age strata. This procedure identified 85 potential controls (34 from the occupational clinic and 51 from the back clinic) who were sent a letter requesting participation in the comparison group for a study of chemical-related health problems. Forty-nine did not respond (17 from the occupational clinic and 32 from the back clinic) and 2 patients (both from the back clinic) were judged ineligible because of reported symptoms of chemical sensitivity. The remaining 34 patients (41% of those potentially eligible) were enrolled. The primary diagnosis was low-back pain for all 17 control patients from the back clinic and for 4 from the musculoskeletal clinic. Of the remaining control patients, 8 had repetitive-motion syndromes and 5 had other chronic symptoms following work-related musculoskeletal injuries.

Laboratory Methods

All immunologic studies were performed by a commercial laboratory with special interest in the evaluation of chemical sensitivity. Laboratory personnel assisted in design of the immunologic battery but remained blinded to case/control status during testing. Total leukocyte count, red cell count, hemoglobin, and hematocrit concentration were measured on a Coulter T140 Cytometer (Coulter Electronics, Hialeah, Florida).

For cell-surface marker studies, mononuclear cells were isolated using a Ficoll-induced density gradient followed by 50% dilution in saline and triple washing in Hanks balanced salt solution to remove serum protein. Phenotypic marking of lymphocytes was performed using monoclonal antibodies against CD5 (total T cells), CD4 (T-helper cells), CD8 (T-suppressor cells), CD19 (total B cells), and CD25 (interleukin-2 receptor-positive cells) (all from Becton Dickinson; Los Angeles, California), then counterstained with mouse antihuman IgG coupled to fluorescein. The monoclonal antibody against CD26 (TA1-positive cells) (Coulter) was directly coupled to rhodamine. The percentage of labeled cells was determined using fluorescence microscopic examination.

Autoantibodies against parietal cells, mitochondria, smooth muscle, brush border, and nuclear components (Medica; Carlsbad, California) were screened at a dilution of 1:20 in phosphate-buffered saline, pH 7.0, by incubation with species substrate in a room temperature humidifier. Counterstaining was with Evans blue, and fluorescence was classified as present or absent. Each assay run included positive and negative controls.

For assay of interleukin-1 generation, peripheral blood monocytes were isolated from sodium heparinized blood by differential centrifugation, washed, and isolated by adherence to tissue culture plates. Monocytes were incubated overnight at 37 C in the presence of lipopolysaccharide (1500 g/mL) and phorbol (1 mg/mL) in 10% fetal calf serum. The supernatant was collected and frozen for later assay. Generated interleukin-1 was measured with an enzyme-linked immunoassay technique using microtiter plates coated with a monoclonal capturing antibody and polyclonal second antibody (Cistron; Pine Brook, New Jersey) linked to peroxidase. Each run included standard and control specimens.

Psychological and Neuropsychological Evaluation

Patients in both groups completed a questionnaire assessing demographic data, work and chemical exposure history, as well as type, frequency, and duration of chemical sensitivity symptoms.

Psychological evaluation included the Hopkins Symptom Checklist-90 (SCL-90), a self-report measure of psychiatric symptoms [23], and the Diagnostic Interview Schedule [24], a highly structured psychiatric diagnostic interview based on the criteria of the American Psychiatric Association's DSM-IIIR [25]. Diagnoses assessed by the structured interview included panic disorder, generalized anxiety, depression, and somatization (a tendency to seek care for physical symptoms that have no apparent medical explanation). The standard Diagnostic Interview Schedule was supplemented with questions to determine the age of onset for individual somatization symptoms. Because the structured interview includes specific questions about medical history, the interviewer could not be blinded to case/control status.

Neuropsychological evaluation included the Logical Memory and Visual Reproduction subtests of the Wechsler Memory Scale-Revised (WMS-R), evaluating immediate and 30-minute recall for contextual verbal and diagrammatic visual information; Trails A and B, timed paper-and-pencil tests of visuomotor speed and mental flexibility; the Rey Auditory-Verbal Learning Test, examining verbal learning and memory for disassociated verbal material; the Wechsler Adult Intelligence Scale-Revised (WAIS-R) Digit Span, evaluating auditory attention and concentration; and the WAIS-R Digit-Symbol, a timed paper-and-pencil test of concentration and visuomotor speed.

Data Analysis

We used unpaired t-tests to compare group means for normally distributed variables and Wilcoxon tests to compare variables with skewed distribution. Group proportions were compared using contingency tables and chi-square tests. Linear regression and analysis of covariance were used to examine adjusted casecontrol differences.

Initial analysis of psychological data included all structured interview symptoms in the assessment of psychiatric diagnoses. This approach may inflate the prevalence of psychiatric disorders among patients with chemical sensitivity because some psychological symptoms might reflect neuropsychiatric effects of chemical sensitivity. Similarly, our requirement that patients with chemical sensitivity report symptoms in three or more organ systems might artificially increase the prevalence of somatization disorder by structured interview. Consequently, further analyses of structured interview data attempted to exclude symptoms temporally associated with chemical sensitivity. The reported date of onset of chemical sensitivity for each case was used to identify structured interview symptoms and diagnoses occurring before chemical sensitivity (pre-existing psychiatric symptoms). Each control was then paired with a case of the same sex and age (within 5 years) using the 34 cases for which ages matched most closely to those of controls. Pre-existing psychiatric symptoms for each control were identified using the date of onset for the paired case. This procedure allowed computation of the prevalence of psychiatric diagnoses and somatization symptoms that began before the onset of chemical sensitivity (for cases) or before an analogous censoring date (for controls).

Results

Among those completing the evaluation, demographic and background characteristics were similar for cases and controls. Both groups were predominantly female (85% of cases, 82% of controls), with the mean age slightly greater for cases (46.4 9.5 [SD] years) than for controls (42.7 10.4 years). Educational attainment was similar; 14.6 years (2.7 years) for cases and 15.1 years (2.6 years) for controls. Controls were more likely to report any current alcohol use (24% for cases versus 56% for controls), but the proportion of participants who reported consuming more than three drinks per week was similar in the two groups (17% for cases versus 21% for controls). The two groups did not differ in frequency of current tobacco use (17% for cases versus 25% for controls) or current caffeine use (61% for cases versus 78% for controls).

Those included in the study sample did not differ from nonparticipants in age or sex. Patients with chemical sensitivity who did not respond had a mean age of 44.3 years (12.1 years), and 78% were female. Comparison patients who did not respond had a mean age of 42 years (10.9 years), and 76% were female.

Comparison of immunologic studies revealed few differences between groups (Table 1). Prevalence rates for five of six autoantibodies were strikingly similar in cases and controls. The only statistically significant difference was in the frequency of antibodies to brush border, with the direction of the difference (controls greater than cases) being the opposite of predicted [5]. Mean total lymphocyte counts as well as mean percentages of T cells, B cells, CD8+, and interleukin-2+ cells were similar in the two groups. The mean percentage of CD4+ cells was somewhat higher among cases, but the absolute counts were nearly identical. The mean percentage of TA1+ cells appeared to differ between cases and controls, but the direction of the difference (greater in controls) was opposite that predicted by earlier reports [5, 7, 8]. Cases had significantly lower levels of interleukin-1 generation by cultured monocytes. Because cases were evaluated earlier in the study period, it was considered whether temporal factors, such as changes in laboratory methods or materials, might have contributed to the observed difference. Cases were evaluated between day 1 and day 146 (mean, 68; 49 days), whereas controls were evaluated between day 24 and day 365 (mean, 227; 92 days). Values for interleukin-1 generation tended to increase with time among both case and control groups, suggesting that such a temporal effect might exist. When analysis of covariance was used to examine the joint effects of time and case/control status, interleukin-1 generation was strongly associated with evaluation date (P = 0.003) but not with the presence of multiple chemical sensitivity (P > 0.2).

Current psychological distress was higher among cases than controls, but this difference did not appear to pre-date the onset of chemical sensitivity (Table 2). Scores on the SCL-90 questionnaire revealed more frequent symptoms of anxiety and depression and a greater tendency to report physical symptoms among patients with chemical sensitivity. Structured interviews revealed higher prevalence of current (within the last month) diagnosis for each disorder examined. Patients with multiple chemical sensitivity had many more physical symptoms judged to be medically unexplained by the somatization section of the interview. Psychiatric morbidity among patients with chemical sensitivity, however, was not universal. Twenty-three of 41 cases had no current psychiatric diagnosis, and 10 of those had SCL-90 anxiety and depression scores less than one standard deviation above the normal population mean (that is, no statistically significant current distress). When analysis was confined to symptoms that predated the onset of chemical sensitivity (or occurred before an analogous date for controls), the two groups showed similar prevalence of preexisting anxiety or depressive disorder. Cases, however, reported a markedly higher number of somatization symptoms preceding the onset of chemical sensitivity symptoms. When all symptoms beginning after the onset of chemical sensitivity were eliminated, 9 of 34 cases (and no controls) satisfied DSM-IIIR criteria for somatization disorder.

Mean neuropsychological performance scores for cases and controls fell within an average range (1 SD) for age, with few differences noted between the two groups (Table 3). Patients with chemical sensitivity showed slightly poorer immediate verbal recall, but the portion of initially memorized information remembered after a 30-minute delay (83%) was essentially identical for cases and controls. Cases also showed slightly poorer performance on one of the five initial trials (data not shown) and on the postdistraction trial of the Rey Auditory Verbal Learning test. For all other measures of attention, visual memory, visuomotor speed, and mental flexibility, case and control performances were indistinguishable. When analysis of covariance was used to adjust neuropsychological performance scores for level of psychological distress (SCL-90 anxiety and depression scores), casecontrol differences in cognitive function were no longer apparent (data not shown).

The technique described above for evaluating temporal trends in measured values of interleukin-1 generation was repeated for psychiatric and neuropsychological measures. Measures of psychiatric morbidity tended to decrease over time in the entire sample but showed no temporal trends within the case or control group, suggesting a casecontrol difference rather than a temporal change in methods. Neuropsychological measures showed no relationship to date of evaluation.

Discussion

The design of our study addresses some of the methodologic shortcomings of previous work. Patients with chemical sensitivity were systematically selected from a defined population and compared with an appropriate medically ill control group. Structured psychiatric assessment and use of self-report measures reduced the potential for interviewer bias in ascertainment of psychiatric disorder. Controlled and blinded laboratory evaluation avoided the potential biases of previous immunologic reports.

Low response rates among cases and controls pose the greatest threat to the validity of these results. Participation was not related to age or sex, but we lack the data to determine whether nonrespondents differed clinically from those in the study sample. The results presented here may misrepresent the population of eligible patients if willingness to participate was related to psychological distress, cognitive impairment, or immunologic abnormality. This sample of patients with chemical sensitivity, however, is probably more representative of the entire population than previous samples selected from referral clinics [20], psychiatric clinics [18, 19], compensation evaluations [21], and advertisements [22]. In the comparison sample, results of neuropsychological tests were all within established normal ranges. Controls did report higher levels of psychological distress than did most community or patient samples, but any resulting bias would act to reduce observed casecontrol differences.

The results of psychological evaluations are consistent with previous reports of excess psychological morbidity among patients with multiple chemical sensitivity [18-22]. Patients with chemical sensitivity showed higher levels of current psychological morbidity on every measure examined as well as a greater tendency to report medically unexplained physical symptoms. These findings are striking given the conservative choice of a control group likely to have increased levels of psychological distress [26]. Controls in this study, in fact, reported greater prevalence of psychiatric disorder than did typical medical clinic samples [27], had higher mean SCL-90 scores [23], and more functional somatic symptoms [28] than did community samples.

These findings do not necessarily imply that anxiety or depression or both are primary causes of chemical sensitivity. The higher levels of psychological morbidity among cases might be either causes or effects of chemical sensitivity. The finding of more current symptoms of anxiety and depression among cases is consistent with either view. Psychologic and cognitive symptoms are prominent in most case descriptions of chemically sensitive patients and were included in the selection criteria for this study. Most patients with multiple chemical sensitivity readily acknowledge symptoms of anxiety and depression but attribute these symptoms to the neurotoxic effects of chemical exposure. Finally, more than one half of the cases did not fulfill criteria for any current psychiatric diagnosis, and almost 25% showed no evidence of clinically significant psychological distress.

Some psychological findings, however, are consistent with a pre-existing pattern of illness behavior among patients with chemical sensitivity. Cases reported more frequent episodes of seeking care for medically unexplained physical symptoms before the onset of chemical sensitivity symptoms. These results do not necessarily demonstrate a psychological cause for chemical sensitivity, but they suggest that psychological factors or health beliefs may predispose some individuals to respond to precipitating events by developing a generalized chemical sensitivity. The findings regarding pre-existing symptoms or illness, however, must be interpreted cautiously. These data depend on retrospective accounts of physical and psychological symptoms occurring years earlier. Variable recall of remote health events could bias results in either direction [29], and current psychological distress may increase the likelihood that past symptoms or health events will be recalled [30]. Consequently, the higher reported prevalence of preexisting somatization symptoms among cases may simply reflect higher levels of current anxiety or depression. Alternatively, the belief that a symptom was caused by chemical sensitivity might have made case patients less likely to recall an occurrence of that symptom before the onset of multiple chemical sensitivity. Such an effect would lead to underestimation of preexisting symptoms among case patients. Only a prospective study starting before the onset of chemical sensitivity symptoms could overcome these potential biases inherent in a retrospective design, but such a study would require follow-up of an impossibly large population at risk. Review of medical records might yield some data on pre-existing symptoms, but medical records rarely contain sufficient information formal rating of psychiatric diagnoses.

The immunologic evaluation revealed no meaningful differences between patients with chemical sensitivity and those with musculoskeletal injury in the prevalence of autoantibody reactions or in lymphocyte surface marker studies. This controlled and blinded comparison did not support any of the immunologic findings reported in previous case series using similar or identical methods [3-9]. Although we cannot exclude the possibility of an immunologic abnormality not detected by this evaluation, the battery of immunologic tests in this study included all those recommended for evaluation of chemical sensitivity by a laboratory prominent in this area. These findings strongly militate against the immunologic mechanisms of chemical sensitivity that have been proposed. Autoantibody and lymphocyte surface marker studies appear to have no value in the routine diagnostic evaluation of patients with chemical sensitivity.

The significance of the observed casecontrol difference in interleukin-1 generation by cultured monocytes remains unclear. The difference was statistically significant in this small sample even after adjustment for multiple comparisons. Examination of time trends, however, suggests that changes in laboratory methods or materials during the course of the study probably caused or contributed to the intergroup differences. In addition, no published data on interleukin-1 generation in patients with chemical sensitivity are available for comparison. With no a priori hypothesis of higher or lower interleukin-1 generation rates in chemically sensitive patients, the possibly artifactual differences observed here can only suggest hypotheses for further study.

We chose the brief neuropsychological battery to evaluate the impaired memory and concentration often described in chemically sensitive patients. Of the nine subtests, case patients and controls differed only on measures of immediate contextual and discrete verbal memory. Cases still performed within the average range, and the two groups did not differ with respect to percentage of information retained during a period of 30 minutes. Performance by the cases on the Trails B test, regarded as highly sensitive to diffuse organic brain dysfunction, was within the average range and indistinguishable from controls. Case-control differences appear relatively minor considering the frequency of anecdotally reported cognitive impairment among patients with multiple chemical sensitivity. Resolution of these differences after adjustment for SCL-90 anxiety and depression scores suggests that these small differences in cognitive performance may reflect psychological distress.

Our findings have substantial implications for the clinical care of patients with multiple chemical sensitivity. In this sample, psychological distress was the only factor that clearly distinguished patients with chemical sensitivity from those with musculoskeletal injuries. The anxiety and depressive symptoms identified among patients with chemical sensitivity cause suffering and disability. Although the pathophysiologic mechanism of chemical sensitivity remains controversial, the prominent role of psychological symptoms in the syndrome has been affirmed repeatedly. For lack of any more specific treatment, relief of psychological symptoms should be a central component of care for chemically sensitive patients.

The findings of this study do not directly address the utility of the strict avoidance regimens advised by some practitioners, but they do militate strongly against immunologic injury as a justification for avoidance. Strict avoidance of low-level chemical exposure may only reinforce social withdrawal and disability without actually preventing any biological injury. Although no investigation can prove the absence of any biological injury from continued low-level chemical exposure, no studies have found plausible evidence of any such effect. In the absence of such evidence, the speculative benefits of severe avoidance do not justify the accompanying disability and isolation. Short-term avoidance of specific exposures may have some role in the initial treatment of some patients with chemical sensitivity, but effective care for chemically sensitive patients should emphasize relief of symptoms and progressive return to active life.