5.1 Polychlorinated biphenyls, dibenzofurans and mortalities
We did a follow-up study comparing the mortality experience of persons with high exposures to PCBs and PCDFs 30 years previously with that of a neighborhood referent group. We found increased all-cause mortality, and specific increases in mortality from malignant neoplasms of the stomach, and of lymphatic and haematopoietic tissue;
diseases of the circulatory system, diseases of the digestive system, and diseases of the musculoskeletal system and connective tissue. In Taiwan in the 1980s, much of the population resided in the 2 largest cities – Taipei in the north and Kaohsiung in the south. Yucheng occurred in central Taiwan, in relatively rural Taichung and Changhua counties. Thus the national mortality rates, dominated by the experience of the cities, were not reflective of the demographic characteristics of the Yucheng subjects. It has always been desirable to have a local referent for mortality, but, until recently, there were too few deaths to allow stable comparisons.
This is the first time the Yucheng follow-up has showed elevated mortality due to malignant neoplasms. They occur mostly in males, which is also true in the 40 year follow-up from Yusho in Japan(Onozuka et al., 2009). The predominant excesses in Japan are in liver and lung cancer, with a small increase in stomach cancer, whereas we see excesses in stomach and haematopoietic malignancies, and no liver cancer excess. It may be worth noting that, throughout follow-up, there has been a larger excess of non-malignant liver disease mortality, but not liver cancer mortality, among Yucheng
subjects compared with Yusho subjects from Japan. The differences might be due to the differences in the degree and character of the contaminants in the rice oil (Masuda et al.,
30
1985), or to the different prevalence of some other etiologic factor, such as aflatoxin (Montesano, et al., 1997), or hepatitis virus (Centers for Disease Control and
prevention, 2008) between Japan and Taiwan. However, the mechanisms underlying both the difference in cancer sites and the much later appearance of malignancy in Taiwan are still unclear.
Several cohorts with exposure to high levels of dioxin-like chemicals show elevated risks of cancer of the lymphatic and hematopoietic system (Steenland et al., 1999;
Flesch-Janys et al., 1998; Bertazzi et al., 2001; Bertazzi et al., 1989; Consonni et al., 2008; Kimbrough et al., 2003; Gustavsson et al., 1986; Gustavsson et al., 1997). For stomach cancer, a mortality study of capacitor manufacturing workers found elevated risk in male workers (Mallin et al., 2004). Incidence of stomach cancer was increased in the highly exposed residents living around a former PCB production site (Pavuk et al., 2004).
We also found excess mortality from circulatory diseases, specifically acute
myocardial infarction and other forms of heart disease, but not cerebrovascular disease.
Among other forms of heart diseases, death from cardiac dysrhythmias was elevated in Yucheng subjects. A systematic review suggested an association between dioxin exposure and mortality from ischemic heart disease, and possibly all cardiovascular disease (Humblet et al., 2008). People living in areas contaminated with persistent organic pollutants, including PCBs and dioxins/furans, had elevated rates of hospital discharge for coronary heart disease and acute myocardial infarction compared to those living in non-contaminated areas (Sergeev and Carpenter, 2005). In the Ranch Hand cohort, exposed to Agent Orange contaminated with TCDD, increased mortality from circulatory diseases was found at 20 years of follow-up (Ketchum and Michalek, 2005).
31
In Yusho subjects exposed to PCBs/PCDFs in Japan, elevated mortality from heart disease was found among females in the period 25–29 years after the incident, but not for other periods (Onozuka et al., 2009). Another study of Yusho (Kashima et al., 2011) reported an increased SMR for cardiovascular disease shortly after the exposure.
Our findings of increased mortality from circulatory diseases support the association of exposure to dioxin-like chemicals with cardiovascular mortality.
Among Yucheng subjects, the SMR for systemic lupus erythematosus was quite increased, with one among Yucheng males, five among Yucheng females; and zero among the neighborhood referents. This finding was consistent with the previous study using the general population as reference group (Tsai et al., 2007). In the Yusho cohort, no mortality due to systemic lupus erythematosus was reported. However, Shimizu reported higher nitrotyrosine in the serum of Yusho subjects (Shimizu et al., 2008).
Nitrotyrosine is a marker of protein oxidation in seraand is associated with disease severity in patients with lupus(Morgan et al., 2005). Further study is warranted to determine whether nitrotyrosine was elevated in Yucheng subjects, which might have contributed to the development of systemic lupus erythematosus.
Diabetes has higher prevalence among Yucheng females, but not males (Wang et al., 2008).However, diabetes mortality was not higher in females and lower in males.
Possible causes of these observations include the low case-fatality rate for diabetes, making estimates of the SMR unstable, and assigning diabetes-related cardiovascular disease to the cardiovascular disease category.
There are strengths in this investigation. The establishment of a registry soon after the exposure allowed for relatively complete and unbiased ascertainment of exposed
32
subjects for mortality studies. The use of neighborhood referents limited bias due to socioeconomic or geographic factors. There are also limitations in this study. First, in the national mortality registry of 1980-84, only 89% of all decedents had complete national identification numbers and ICD-9 codes. Thus, some deceased subjects could have been misclassified as being alive. This would affect both Yucheng subjects and referents, and should cause underestimation of deaths in both groups. Therefore, it is unlikely to produce high SMRs in the Yucheng group. Second, some subjects were excluded due to wrong or missing national identification numbers. This might have caused underestimation of deaths in both groups. Third, although using neighborhood referents may minimize some bias due to socioeconomic or geographic factors, confounding due to other factors such as medical history, smoking status, alcohol drinking, and physical activity could not be controlled. Fourth, selection bias cannot be ruled out, because the selection processes of Yucheng subjects and their neighborhood referents were different. Since one of the selection criteria for the Yucheng subjects was presence of symptoms related to PCB exposure, it is possible that those included in the registry were at higher risk of having some illnesses. However, the most prevalent symptoms among the Yucheng subjects in the registry were skin (chloracne and hyperkeratosis), nail (discoloration), and eye symptoms (discharge and eyelid swelling).
These were more likely directly related to PCB poisoning, and less likely early signs of other important illnesses but unrelated to PCBs. We cannot exclude the possibility that persons with exposure and symptoms were more likely to be registered than those without. There was, however, much local publicity at the time about the etiology of the illness and the specific sites where the contaminated oil was sold or used. About 10%
of those registered had a history of consuming the oil and/or elevated serum levels of PCBs. We do not believe there were many people with high PCB exposure and no
33
symptoms; if there were, our estimates of the SMRs are too high. Finally, some of the referents may have been exposed to the contaminated oil without their knowledge. We believe this to be unlikely. The provincial health department announced the etiology of the episode through mass media and local health units in October, 1979, and offered analysis of serum for PCBs to anyone who thought they had been exposed (Hsu et al., 1985). In 1992, pooled neighborhood blood samples from 56 women were analyzed for PCBs and PCDFs, which showed background exposure only (Guo et al., 1997). If there were many people with exposure who did not participate in the registry but suffered ill effects from their exposure, then our estimates of the SMRs are too low.
34
5.2 Mortalities among Yucheng and Yusho victims
This is the first study examining overall mortality in the two largest exposed
populations to PCBs and PCDFs in the world. Our results suggested elevated mortality
from all-cause, lung cancer, heart disease, and hepatic disease among pooled subjects.
In addition, we found increased all-cause, all cancer, lung cancer, and hepatic disease
among pooled males, and increased liver cancer among pooled females.
Compared to individual mortality studies (Onozuka et al., 2009; Li et al., 2013), this study’s new finding included elevated mortality from liver cancer in pooled
females. In individual studies, significant excess of liver cancer was shown only among
Yusho males (Onozuka et al., 2009). Combining two cohorts increased statistical power
and female liver cancer was found elevated. The literature about PCBs and liver cancer
has been inconclusive. A recent study carried out among subjects with hepatocellular
carcinoma (HCC) living in a highly polluted area by PCBs did not find substantial
differences in total PCB mean and median values with respect to the general population,
and no variation in total PCB serum levels according to HCC etiology (viral vs
non-viral vs unknown) (Zani et al., 2013a). Furthermore, liver cancer mortality
among cohorts occupationally exposed to PCBs has been inconsistent, as summarized in
a recent systematic review (Zani et al., 2013b). Among several follow-up studies of
35
occupational cohorts exposed to PCBs, only one reported significantly elevated
mortality for liver cancer in female (Mallin et al., 2004), but not the others (DeGuire et
al., 1992; Pesatori et al., 2013; Ruder et al., 2014; Kimbrough et al., 2014).A possible
explanation for those findings could be the healthy worker effect, which rendered a
reduced overall mortality and thus non-significant mortality of individual conditions
with moderate elevation. Besides, Yucheng and Yusho cohorts were both exposed to
high dose of PCDFs, which were chemically and toxicologically similar to 2, 3, 7,
8-tetrachlorodibenzo-p-dioxin (TCDD), in addition to PCBs. PCDFs or the combined
effects of PCBs and PCDFs induced the observed results.
Hepatitis B virus (HBV) is one of the major causes of liver cirrhosis and
hepatocellular carcinoma. If Yucheng or Yusho group lived in the area with high
prevalence rate of HBV, selection of referent groups from areas with low prevalence
rates of HBV would have resulted in bias. In Taiwan, the carrier rate of hepatitis B
surface antigen in the general population was as high as 15-20 percent (Chien et al.,
2006). However, in Yucheng mortality study, referents were selected from the
neighborhood (Li et al., 2013). Therefore, biased results due to discrepancy of HBV
carrier rates in Yucheng and reference groups were unlikely. In Japan, referents were
background population of the country. However, only approximately 1.5 million people
36
were infected with HBV, accounting for only 1-2% in Japan general population
(Yokosuka et al., 2009). Due to the low prevalence rate of HBV, the likelihood of bias
induced by HBV is not large. Japan and Taiwan cannot be considered overlapping as
regards HCC epidemiology, as HBV infection has been the main cause of liver disease
in the former and hepatitis C virus in the latter, in the last decades (Kim et al.,
2013). The discrepancies between the two cohorts may also be explained by the
different age at the beginning of the viral hepatitis infection.
Although TCDD and PCBs were both classified as human carcinogens by
International Agency for Research on Cancer (International Agency for Research on
Cancer, 1997; Lauby-Secretan et al., 2013), human studies showed limited information
on specific sites of cancer. Current study provided evidence for reconsideration of
human carcinogenesis in lung or liver.
Several discrepancies in SMR were found between Yucheng and Yusho. Comparing
to their own control population, Yucheng had elevated heart disease mortality, but Yusho
did not. Causes for such observed difference include several possibilities. Japanese
people had lower standardized mortality rate for heart diseases as compared to
Taiwanese people (Department of Budget, Accounting and Statistics, Taipei City
Government, 2008). Even after 40 years of follow-up, exposure to PCBs and PCDFs did
37
not affect the Yusho victims’ heart mortality significantly. On the other hand, heart
diseases were increased in Yucheng, likely as a results of accelerated atherosclerosis, as
reviewed by Humblet et al (Humblet et al., 2008). The finding of elevated heart
mortality in Yucheng is comparable to people highly exposed to Agent Orange
(Ketchum et al., 2005). It is also possible that misclassification in ICD-9 coding
occurred in Japan caused inability to detect elevated heart diseases in Yusho victims
(Onozuka et al., 2009).
The other discrepancy in mortality was elevated liver cancer in Yusho men, but not
in Yucheng men. This might have to do with the observed increased mortality due to
cirrhosis and chronic liver disease occurring early in Yucheng men (Tsai et al., 2007).
Acute insult might have caused deterioration in susceptible Yucheng men, who deceased
from cirrhosis or chronic liver disease, rendering lower number of susceptible victims
who would have develop liver cancer in later years as observed in Yusho men.
The discrepancies in mortality between Yusho and Yucheng study might be caused
due to several reasons. First, in Yucheng study, the rice oil was contaminated by
Kanechlor 500 and their pyrolytic products PCDFs and PCQs, whereas in Yusho study,
Kanechlor 400 was the major contaminant. However, previous studies showed that
Yusho and Yucheng victims had similar level of dioxin-TEQ (Lan et al., 1981; Lambert
38
et al., 2006; Tanabe et al., 1989; Masuda et al., 1998),with an average concentration of
2300 ppt-TEQ in Yucheng victims, and 2630 ppt-TEQ in Yusho victims. Thus
combining two cohorts is reasonable. Second, the different follow-up period between
two cohorts would be an issue. Since the majority of cancers and chronic diseases
appear in older people, the shorter follow-up period in Yucheng study might cause
underestimations when calculating meta-SMR. However we still found some cancers
increased significantly. Finally, although the median exposure levels were close, Yusho
event was caused by higher levels of PCBs and PCDFs in the contaminated oil, and a
shorter period of exposure. It is possible that shorter exposure to higher levels of toxins
might have caused heterogeneity in the outcomes.
This meta-analysis offers some advantages over previously reported studies. 1)
Yucheng and Yusho subjects belong to longitudinal follow-up cohorts, providing better
inference in causal relationship. 2) To our best knowledge, there have been only two
events of this magnitude and severity in the world. Both Yucheng and Yusho victims
were Asian people, and were exposed to similar high dose of PCBs/PCDFs. Thus,
combining two cohorts is reasonable and allows us to detect the mortalities rates more
accurately and powerfully. 3) Many follow-up studies on mortality among people
exposed to PCBs and/or PCDFs were occupational groups, and potentially subject to
39
healthy worker effect. Both Yucheng and Yusho cohorts were not occupational cohorts,
and less likely affected by this effect.
However, there are limitations in this meta-analysis. 1) A shorter follow-up period
(30 years) in Yucheng cohort, as compared to 40 years in Yusho, might not be sufficient
to allow for chronic diseases to develop, even if there was truly effects. Thus, the risk of
mortality of chronic diseases might be underestimated. 2) Despite pooling of the
follow-up person-years, a total of 99,524 person-years still did not provide sufficient statistical
power to detect diseases with medium to low occurrence. 3) Comparing the
age-standardized mortality rates using the 2000 World Heath Organization standard
population as the standard, Yucheng (9442 x 10-5) had higher rates than Yusho (6987 x
10-5) people (data not shown). However, SMR is a relative measurement. When
calculating SMR, the reference group provided information to compute the expected
death in the exposed group. Therefore different overall mortality rates do not bias the
results when combining two SMR values.
40
5.3 Gestational exposure to polychlorinated biphenyls and hearing loss
This is the first paper describing adverse hearing effects in children with gestational
exposure to PCBs and PCDFs. The main damage to hearing threshold by such exposure
was found at low frequencies. Such damage was related to gestational exposure to
2,3,4,7,8-pnCDF, but not to the marker-PCB congeners.
The mechanism of PCBs-induced auditory deficits has been suggested in prenatally
exposed animals. Thyroid hormone is necessary for normal cochlear development
(Uziel, 1986), and perinatal exposure to Aroclor 1254 is known to markedly reduce
serum thyroid hormones in rats (Goldey et al. 1995; Morse et al. 1996). As a result,
perinatally exposed animals had reduced auditory startle amplitudes especially in lower
testing frequencies (Goldey et al. 1995). Those rats perinatally exposed to higher doses
of Aroclor 1254 (4 and 8 mg/kg per day from gestational day 6 through postnatal day
21) had irreversible hearing damage. Replacement therapy with thyroxine injection
ameliorated the hearing loss caused by gestational Aroclor 1254 exposure (Goldey and Crofton, 1998). The cochlea was specifically indicated as the likely site of action since
other investigators found loss of outer hair cells in rats perinatally exposed to PCBs
(Crofton et al. 2000). As for 2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD) and PCDFs,
the auditory effects in rodents of gestational exposure to these chemicals has not been
41
reported, despite several reports documenting reduced serum thyroxine levels in animals
exposed to TCDD (Bastomsky, 1977; Mc Kinney et al. 1985; Potter et al. 1983; Potter
et al. 1986; Gorski and Rozman, 1987; Roth et al. 1988; Lans et al. 1990). Since the
reduction in thyroxine levels was strong in animals exposed to TCDD, one might
assume that gestational exposure to TCDD might cause similar auditory effects. Despite
an elevated rate of reported goiter in Yucheng women (Guo et al., 1999),
hypothyroidism was not increased in Yucheng as compared to their controls. However,
thyroid function was not measured when the Yucheng mothers were pregnant. The
possibility of sub-clinical hypothyroidism in Yucheng mothers as a cause of their children’s hearing damage cannot be totally ruled out.
Gestational exposure to PCBs has been associated with asymmetric hearing loss in
human studies, but the findings were inconsistent. One study in Faroe Islands found that
gestational exposure to wet-weight PCB concentrations was associated with hearing
thresholds at frequencies 250 and 12000, but only on the left side (Grandjean et al.
2001). A study in 8-year-old children showed that higher maternal serum PCB
concentrations were associated with increased hearing thresholds at 2000 Hz in the left
ear, and 4000 Hz in the right ear (Longnecker et al. 2004). In this study, we found
increased hearing threshold more prominently in the right ear in PCB/PCDF-exposed
42
children, as well as clearer association of maternal 2,3,4,7,8-pnCDF with right ear
threshold than with left ear threshold. The mechanism of this kind of asymmetric
hearing loss has not been identified.
Most toxic effects of 2,3,7,8–substituted PCDFs and non-ortho–substituted PCBs
are mediated through the aryl hydrocarbon receptor (AhR), a cytosolic receptor protein
present in most vertebrate tissues (Van den Berg et al., 2006). The most potent ligand
for AhR is 2,3,7,8-TCDD. In environmental and biological media, mixtures of these
compounds are frequently seen. To summarize the overall toxicity in mixtures, toxic
equivalency (TEQ) of these compounds has been applied, which is operationally
defined by the sum of the products of the concentration of each compound multiplied by
its toxic equivalent factor (TEF) value, the latter being the relative effect potency of
each compound as compared to that of 2,3,7,8-TCDD. In the blood of the mothers in
this study, the TEQ were mostly contributed by 2,3,4,7,8-pnCDF, followed by
1,2,3,4,7,8-hxCDF. The TEQ of PCB congeners were rather low. Therefore, it is not
surprising to find strongest relationship between hearing deficits and 2,3,4,7,8-pnCDF,
but not other compounds. This finding implies that the hearing effects of these
but not other compounds. This finding implies that the hearing effects of these