| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
LETTERS TO THE EDITOR |
Robert Koch Institute; Department of Environmental Medicine; Berlin, Germany
We recently became aware of the paper by Saito et al. entitled "Symptom Profile of Multiple Chemical Sensitivity in Actual Life" (1). The objective of this study was to monitor everyday symptoms and exposure to environmental chemicals in patients suffering from multiple chemical sensitivity (MCS). Patients were asked to carry an active sampler and two passive samplers for 1 week. Whenever symptoms occurred, the patients turned on the pump of the active sampler and answered questions in an electronic questionnaire (symptom prompts). Background exposure was measured by the passive samplers. Control subjects also carried two passive samplers but none carried an active sampler. Candidate substances were identified by comparing background concentrations with those measured during symptomatic periods. Subjects were also prompted at random times and answered the same questions in the electronic questionnaire to rate mood and symptoms (random prompts).
Although we appreciate the authors’ approach for applying the active and passive samplings, we have some problems with this paper and, therefore, feel compelled to offer some critical remarks and ask some questions.
First, the sample was very small and seems to be extremely biased. Of the 18 patients, not less than 14 (or 78%) suffered actually or had suffered in their lifetime from agoraphobia. Agoraphobia, to our understanding, is the fear of entering public places or streets or using public transport. Both disorders, agoraphobia and MCS, are quite rare in Europe. We can assume that the lifetime prevalence of agoraphobia and MCS is 
5% for each disorder. (A much more likely value is 2%.) Thus, comorbidity for both disorders would account for maximal 0.25% (statistical independency assumed). In other words, of 1000 patients with MCS, only 50 patients are likely to suffer simultaneously from agoraphobia. In the study under question, 14 of 18 patients with MCS initially invited to participate in the study suffered simultaneously from agoraphobia, where we would expect only one (18 x 0.05 = 0.9). This is >14 times over the expectation. According to our experience, agoraphobia in the above stated sense is a rare condition among environmental patients. Applying Composite International Diagnostic Interview, we found only 2.6% had agoraphobia without history of panic disorder (F = 40.00). It is a well-known fact that patients with MCS avoid places where they assume exposure to chemicals, but this does not justify a diagnosis of agoraphobia.
Second, the authors stated that they would conservatively define a chemical as likely to be causative of MCS when the concentration measured by active sampling minus its standard deviation (SD) was greater than the concentration measured by passive sampling plus its SD. If both SD values are equal, this would mean a double SD. It is, however, hard to believe, that all the figures given in Table 3 fulfill this condition. For patient A5, e.g., the toluene concentration was 11.3 µg/m3 when symptoms occurred and 9.1 µg/m3 during symptom-free periods. This would essentially mean that 1 SD (active sampling) plus 1 SD (passive sampling) would add up to <2.2 µg/m3, which is an extremely good and, as we are almost tempted to say, unlikely precision. When using the SD of repeated measurements as an assessment for measurement uncertainty, researchers must make sure that the conditions—at the time assessing the measurement uncertainty—comply exactly with those of the subsequent experiment. A measurement uncertainty elaborated under laboratory conditions cannot be considered to be appropriate for a field study.
Third, sample size and degrees of freedom given in Table 5 are at least confusing. According to this table, the number of random prompts is n = 226 and the number of symptom prompts is n = 128. But the number of degrees of freedom is stated to be df = 11. It is obvious that the effective sample size (for pair-wise testing) was only n = 13, which is the number of patients.
Fourth, if we calculate the SD from the standard error of mean (SEM) by multiplying it with the square root of the sample size, we obtain figures which are again hard to believe. We refer again to Table 5. The symptom, "sore throat", was found to differ statistically significantly between periods with symptoms and symptom-free periods. The values were 32.58 ± 6.52 (symptom periods) and 21.01 ± 32.83 (symptom-free periods). It is conspicuous that the SD value of the symptom periods is so much smaller than that of the symptom-free ones. This would mean that the 13 patients would react extremely uniformly, reaching almost the same end points of discomfort regardless of the kind and concentrations of the chemicals to which they were exposed.
Fifth, there is considerable and confusing discrepancy between the figures given in Table 6 and Table 8, respectively. We refer to the mood measures of the random prompts gained from the patients (n = 226). The estimate (of the mean?) for anxious mood in Table 6 is given as 47.72 (SEM = 28.50) but in Table 8, the corresponding values are 54.88 for the estimate and 26.38 for the SEM, respectively. The same is true for the values given for positive mood, negative mood, and depressive mood. How should this be interpreted? Have the authors, in different tables, used different linear mixed models to evaluate the estimates, and, if so, why did they do this?
Sixth, the authors stated in the Methods section, under Statistical Analysis, that exposure levels (despite being continuous data) have been evaluated by Wilcoxon’s rank sum test, whereas the symptom profile data (being categorical data) have been analyzed applying a linear mixed model. Did the authors make sure that their symptom profile data fulfilled the prerequisites of the statistical model (metrics, normality, linearity, homogeneity of variances or covariances)? Categorical data remain still categorical even after linear transformation, and continuous data should be treated as continuous data. Measurements below the detection limit can be replaced by half of the detection limit.
MCS was diagnosed according to the 1999 consensus criteria (MCS consensus conference). In an earlier paper published by some of the authors in 2004 (2), MCS was, however, diagnosed by detection of eye movement aberrations and the reaction of pupils to light. Do the authors consider both methods equivalent?
In the former study, five (of 15) patients were members of the same family, all simultaneously affected by MCS, and two other patients were a couple. Is this a likely condition that MCS occurs accumulated in families and/or kinship? And, if so, is this accumulation genetically determined or is it rather caused by personal interaction and suggestion? We would like to remark that both accumulation of MCS in kinship and association with mental disorders, like agoraphobia, can be judged as strong hints to a psychosomatic background of MCS.
Numerous other studies support this assumption. Incidentally, it can be assumed that persons fulfilling the MCS case criteria exhibit enhanced attentiveness to environmental exposures and that they, likewise, fear health impairments possibly caused by those exposures. This fear can lead to the expression of psycho-vegetative reactions that are subjectively (but erroneously) interpreted by the patients as symptoms or health impairments. In that way, there would be an association between subjectively perceived episodes of exposure, measured concentrations, and symptoms reported in the electronic questionnaire. The results of Saito et al. are possibly constituted by such psychosomatic or psychophysiological associations. Although this cannot be securely excluded, the results of this paper (let alone the above stated methodical criticisms) are not sufficient to confirm a supposed toxicogenic hypothesis of MCS.
NOTES
EDITOR’S NOTE: Please see the erratum in this issue related to this letter to the editor.
DOI:10.1097/PSY.0b013e31804b21cb
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
