Subjects and examination items
After the occurrence of Kanemi Yusho, health examinations for designated Yusho patients (hereinafter referred to as Yusho examinations) have been conducted since 1986. We used data on the presence or absence of symptoms in physical, blood, dermatological, dental and ophthalmological examinations in Yusho patients who had undergone annual examinations in 1986-1989 and 2001-2004, at least once in each period, and whose PeCDF levels were measured. Subjects included both designated and undesignated patients. The number of subjects who underwent examinations in both periods was 302. To analyze the relationship with recent PeCDF levels, representative values were obtained for each item in the physical, blood, dermatological, dental and ophthalmological examinations for each year in 1986-1989 and 2001-2004, for each patient. We defined abnormality as one or more abnormal values in 4 years. Yusho examinations consisted of a total of 241 items, including 52 items in a questionnaire, 55 in physical and laboratory examinations, 21 in dermatological examinations, 108 in dental examinations, and five in ophthalmological examinations [1, 7].
Methods for data analysis
We found laboratory parameters that were related strongly to recent PeCDF level among the past and recent data. We used association analysis [12] to identify symptoms that were strongly related to PeCDF level. Association analysis was applied in medical domain [8, 13]. The numerical value that indicated the level of association with PeCDF level was compared between the strongly associated past and recent symptoms. The proportion of patients with high PeCDF levels and those with each symptom were calculated. If a high level of PeCDF and the presence of each symptom were not related, the rate of patients with a certain symptom among those with high PeCDF levels was calculated as the product of both proportions, and is binomially distributed. If the sample size was adequate, the binomial distribution could be considered as a normal distribution. The relation of symptom to the PeCDF level was judged based on how the symptom was distributed in the probability distribution. Z-score was used to show the difference from the mean in the normal distribution [8]. Association analysis revealed highly-related combinations of symptoms and PeCDF levels, which were defined as a z score ≥ 1.645 (one-tailed significance level, 5%).
Classification of examination results
Designation of Yusho was based upon the diagnostic criteria in Additional File 1 [1]. The diagnostic criteria largely consist of "onset conditions," "important findings" and "reference symptoms and findings." Among these, severity of "acneform eruptions" and "pigmentation" ("important findings") was determined differently by the doctors who conducted the examinations, whereas measurement of "abnormalities of blood PCB properties and levels," "abnormalities of blood polychlorinated quarterphenyl (PCQ) properties and levels" and "abnormalities of blood PeCDF levels" can be expressed numerically, thereby playing an important role in determining designation.
To conduct association analysis, the numerical data were classified into categorical data such as "within the normal range" and "abnormal values". For general blood examination, standard values that the study group used in examinations were used to define "normal range". Findings from the physical, dermatological, dental and ophthalmological examinations were each classified by the presence or absence of abnormalities, and their strength.
With regard to PCBs, it is known that chromatographic analysis patterns in Yusho patients differ from those in the general population. Patterns are classified into four types: Type A, which is specific to Yusho patients; Type C, which is specific to the general population; Type B, which is intermediate between Types A and C; and Type BC, which is intermediate between Types B and C.
For PCQ levels, it was considered that ≥ 0.1 ppb was abnormally high; ≤ 0.02 ppb (detection limit) was normal; and 0.03-0.09 ppb was borderline. As for PeCDF level, "50 pg/g lipid or above" and < 50 pg/g lipid were classified as high and normal level, respectively.