Disturbance of cerebral function in people exposed to drinking water contaminated with aluminium sulphate: retrospective study of the Camelford water incident
Esmonde, David, Fahy and McMillan make a number of comments
about the paper in their separate letters, many of which overlap and are
discussed below.
Time interval between incident and our study
We strongly support Professor David's statement that incidents
like Camelford should trigger a rapid and co-ordinated response from
appropriate health professionals. Unfortunately this did not happen after
Camelford and as stated in our paper it was not until three years later
that we were asked to perform these studies, although at that time we
ourselves anticipated that such a delay would favour negative
investigation results. We were surprised by the results of our studies.
Competing interests of authors
In our paper we openly acknowledged that our study was
commissioned by lawyers acting for the plaintiffs and funded through Legal
Aid. This legal action constrained the nature of our study but we do not
believe it biased our conduct consciously or unconsciously. The legal
action affected, in particular, our own wish to study other relevant
groups of randomly selected subjects such as those affected by anxiety,
asymptomatic exposed and non-exposed (to the water contamination) control
subjects, as well as randomly selected exposed individuals irrespective of
symptoms. These suggestions were made to the lawyers. It was also as a
result of the legal action that there was delay in publishing the work.
As we had to accept the constraints imposed upon us and mindful
of previous reports concerning Camelford complainants, we tried to design
a study in which the tests were as objective as possible and could be
compared with our previous studies of dialysis patients. In this way we
hoped to obtain data of clinical and scientific interest. If we had had a
completely free hand we would have studied the additional subjects
discussed above. We have not been engaged to represent the claimants we
studied or others in subsequent legal actions.
The question of competing interests was discussed at length with
the editorial staff of the BMJ before the paper went to print. The details
regarding the commissioning and funding of the studies are explicitly
discussed in the paper and there were no competing interests according to
the guidelines published by the BMJ.
Bias
In our paper we expressed our own concerns over the possibility
of experimental bias caused by the interests of participants who alleged
they were symptomatic. We emphasised that it was the pattern of the
results obtained in the psychometric testing, combined with the results of
the visual evoked responses and the similarity of these to our previous
findings in dialysis patients which was important and suggested that the
Camelford subjects had suffered from exposure to an acute toxic insult. We
therefore suggested that "aluminium poisoning must be considered a
possibility, although other contaminants may have contributed."
One of the reasons why we had chosen the battery of tests from
the Bexley Maudsley package (following advice from clinical psychology
colleagues) was that the symbol digit coding test is particularly
sensitive to organic brain disease. In addition, the manner in which it
tests psychomotor function is subtle to the subject being tested compared
to the more overt tests of memory such as the verbal recognition and
visual spatial recognition memory tests. Both in these and our previous
studies, subjects commented, after testing, that the visual spatial
ability, visual perceptual ability, verbal recognition memory and visual
spatial recognition memory tests were far more daunting and difficult than
the symbol digit coding test. No mention regarding the sensitivity of the
tests was ever made to the subjects and none had any prior knowledge of
what type of testing was going to be done. The tests were run in a calm
but most rigorous and uniform manner, and before each test the subject was
given time to rest and it was explained that they should perform as fast
and accurately as possible. Individuals who have early organic brain
disease exhibit the pattern of abnormality exactly as described in this
Camelford group - a pattern that has been used repeatedly in other studies
of psychomotor function in a variety of contexts including potentially
psychotropic drugs. If litigation-motivated bias had been operating then
one would expect the subject to under-perform on the tasks most overtly
related to memory rather than in the symbol digit coding test. As this was
not the case there is no evidence that such bias affected these results.
The purpose of choosing unexposed siblings, who denied exposure
to neurotoxins or psychotropic drugs of any type, was to provide genetic
(phenotypic) controls as this is a standard way of establishing
abnormalities which might be acquired rather than inherited. There were
only 15 available such siblings and so this was the number that we had to
accept - it would have been wrong not to proceed with this part of the
study and, indeed, we have not been criticised for this or the statistics
thereon by medical statisticians
We did not discuss the very small number of abnormal tap water
aluminium levels as this clearly was open to bias. The samples were posted
to us by the subjects in containers that we gave them. Because we were not
in a position to sample their tap water ourselves we felt that these data
could not be regarded as 100% reliable, and as there were no elevated
serum aluminium levels, we did not follow these up except to recommend to
the patients that they have their water systems overhauled.
We addressed the issue of possible effects from anxiety not just
through the finding of relatively low levels of anxiety, but more
importantly by finding no relationship between the absolute levels of
anxiety, as measured by the SCL90 score, and the parameters we measured.
Flash-pattern differences
In response to Professor David's comments concerning technique we
merely wish to observe that Dr Dhanesha is a fully qualified clinical
neurophysiologist, having trained under the directorship of Professor G
Harding in the Clinical Neurophysiology Unit at Aston University,
Birmingham where she did her PhD studies on visual evoked subcortical
potentials (Aston University 1986). The wave forms were analysed in
exactly the same way as in our previous publications and in the manner
described in many publications from Professor Harding's group.
There is, as we stated, very little normative data published on
large series of visual evoked potential results, either flash stimulated
or pattern stimulated or the difference in timing between them.
It is possible that in a given subject the flash-pattern
difference may be large because of a quicker pattern latency than
"normal". However, as we have remarkably little information about what is
normal we chose to use the difference between the two types of visual
evoked potential and not the absolute measures. In many studies looking at
the relationship between severity of dementia and visual evoked
potentials, it is noteworthy that it is the flash-pattern difference that
correlates rather than absolute values.
Indeed, the magnitude of the flash-pattern difference was
significantly correlated with the magnitude of the symbol digit coding
abnormality (but not the other Bexley Maudsley tests), as in dialysis
patients where it was also related to cumulative aluminium exposure.
If there was a functional relationship between the two types of
visual evoked responses one might expect the two measurements to
correlate, whereas if they were independent of each other there should be
little or no relationship. In our studies there was a very significant
relationship between these two parameters in normals, which was nearly
absent in the Camelford "victims" and completely absent in our dialysis
patients.
How aluminium exposure might lead to this effect on visual
evoked potential is not an area that we addressed in this or previous
papers, but there is no reason to suppose that aluminium toxicity is
manifested solely by causing neuronal death.
We acknowledged the controversy about aluminium and Alzheimer's
disease, but there is no controversy about the potential for aluminium to
be neurotoxic. There are many published reports of visual evoked potential
abnormalities of the sort that we have discussed in dementias including
Alzheimer's disease, some of which were referred to in our paper.
Professor David misquotes our data in his table - we did not
quote absolute latencies for the 55 adults - the 28 cases he quotes were a
separate quality control study and he refers to SD (standard deviation)
where we used SE (standard error), and it is therefore unclear whether the
other data quoted in his table refers to means with SDs or SEs. In
addition there is no mention of the SDs or SEs in the bottom panel of the
table. Our dialysis studies included 16 patients (see reference 24 of web
version paper), and there are no descriptions of pathologically short
pattern evoked visual evoked reponses in the literature.
The ethics of medical publications
Finally, although we were concerned about the possible effects
that the publication of our paper might have on people who were involved
in the Camelford incident, it would have been wrong to withhold it. Had
our studies not been designed to provide scientifically useful information
(despite the inevitable limitations), and only been used to satisfy the
legal process, then the clinical and scientific communities would have
been deprived of results that we find interesting and worthy of debate,
and which we believe are valid.
It is unlikely that another Camelford will occur, so the
opportunity to study the effects will, we hope, not recur.
Rapid Response:
Reply to Electronic Responses to Camelford Paper
Esmonde, David, Fahy and McMillan make a number of comments
about the paper in their separate letters, many of which overlap and are
discussed below.
Time interval between incident and our study
We strongly support Professor David's statement that incidents
like Camelford should trigger a rapid and co-ordinated response from
appropriate health professionals. Unfortunately this did not happen after
Camelford and as stated in our paper it was not until three years later
that we were asked to perform these studies, although at that time we
ourselves anticipated that such a delay would favour negative
investigation results. We were surprised by the results of our studies.
Competing interests of authors
In our paper we openly acknowledged that our study was
commissioned by lawyers acting for the plaintiffs and funded through Legal
Aid. This legal action constrained the nature of our study but we do not
believe it biased our conduct consciously or unconsciously. The legal
action affected, in particular, our own wish to study other relevant
groups of randomly selected subjects such as those affected by anxiety,
asymptomatic exposed and non-exposed (to the water contamination) control
subjects, as well as randomly selected exposed individuals irrespective of
symptoms. These suggestions were made to the lawyers. It was also as a
result of the legal action that there was delay in publishing the work.
As we had to accept the constraints imposed upon us and mindful
of previous reports concerning Camelford complainants, we tried to design
a study in which the tests were as objective as possible and could be
compared with our previous studies of dialysis patients. In this way we
hoped to obtain data of clinical and scientific interest. If we had had a
completely free hand we would have studied the additional subjects
discussed above. We have not been engaged to represent the claimants we
studied or others in subsequent legal actions.
The question of competing interests was discussed at length with
the editorial staff of the BMJ before the paper went to print. The details
regarding the commissioning and funding of the studies are explicitly
discussed in the paper and there were no competing interests according to
the guidelines published by the BMJ.
Bias
In our paper we expressed our own concerns over the possibility
of experimental bias caused by the interests of participants who alleged
they were symptomatic. We emphasised that it was the pattern of the
results obtained in the psychometric testing, combined with the results of
the visual evoked responses and the similarity of these to our previous
findings in dialysis patients which was important and suggested that the
Camelford subjects had suffered from exposure to an acute toxic insult. We
therefore suggested that "aluminium poisoning must be considered a
possibility, although other contaminants may have contributed."
One of the reasons why we had chosen the battery of tests from
the Bexley Maudsley package (following advice from clinical psychology
colleagues) was that the symbol digit coding test is particularly
sensitive to organic brain disease. In addition, the manner in which it
tests psychomotor function is subtle to the subject being tested compared
to the more overt tests of memory such as the verbal recognition and
visual spatial recognition memory tests. Both in these and our previous
studies, subjects commented, after testing, that the visual spatial
ability, visual perceptual ability, verbal recognition memory and visual
spatial recognition memory tests were far more daunting and difficult than
the symbol digit coding test. No mention regarding the sensitivity of the
tests was ever made to the subjects and none had any prior knowledge of
what type of testing was going to be done. The tests were run in a calm
but most rigorous and uniform manner, and before each test the subject was
given time to rest and it was explained that they should perform as fast
and accurately as possible. Individuals who have early organic brain
disease exhibit the pattern of abnormality exactly as described in this
Camelford group - a pattern that has been used repeatedly in other studies
of psychomotor function in a variety of contexts including potentially
psychotropic drugs. If litigation-motivated bias had been operating then
one would expect the subject to under-perform on the tasks most overtly
related to memory rather than in the symbol digit coding test. As this was
not the case there is no evidence that such bias affected these results.
The purpose of choosing unexposed siblings, who denied exposure
to neurotoxins or psychotropic drugs of any type, was to provide genetic
(phenotypic) controls as this is a standard way of establishing
abnormalities which might be acquired rather than inherited. There were
only 15 available such siblings and so this was the number that we had to
accept - it would have been wrong not to proceed with this part of the
study and, indeed, we have not been criticised for this or the statistics
thereon by medical statisticians
We did not discuss the very small number of abnormal tap water
aluminium levels as this clearly was open to bias. The samples were posted
to us by the subjects in containers that we gave them. Because we were not
in a position to sample their tap water ourselves we felt that these data
could not be regarded as 100% reliable, and as there were no elevated
serum aluminium levels, we did not follow these up except to recommend to
the patients that they have their water systems overhauled.
We addressed the issue of possible effects from anxiety not just
through the finding of relatively low levels of anxiety, but more
importantly by finding no relationship between the absolute levels of
anxiety, as measured by the SCL90 score, and the parameters we measured.
Flash-pattern differences
In response to Professor David's comments concerning technique we
merely wish to observe that Dr Dhanesha is a fully qualified clinical
neurophysiologist, having trained under the directorship of Professor G
Harding in the Clinical Neurophysiology Unit at Aston University,
Birmingham where she did her PhD studies on visual evoked subcortical
potentials (Aston University 1986). The wave forms were analysed in
exactly the same way as in our previous publications and in the manner
described in many publications from Professor Harding's group.
There is, as we stated, very little normative data published on
large series of visual evoked potential results, either flash stimulated
or pattern stimulated or the difference in timing between them.
It is possible that in a given subject the flash-pattern
difference may be large because of a quicker pattern latency than
"normal". However, as we have remarkably little information about what is
normal we chose to use the difference between the two types of visual
evoked potential and not the absolute measures. In many studies looking at
the relationship between severity of dementia and visual evoked
potentials, it is noteworthy that it is the flash-pattern difference that
correlates rather than absolute values.
Indeed, the magnitude of the flash-pattern difference was
significantly correlated with the magnitude of the symbol digit coding
abnormality (but not the other Bexley Maudsley tests), as in dialysis
patients where it was also related to cumulative aluminium exposure.
If there was a functional relationship between the two types of
visual evoked responses one might expect the two measurements to
correlate, whereas if they were independent of each other there should be
little or no relationship. In our studies there was a very significant
relationship between these two parameters in normals, which was nearly
absent in the Camelford "victims" and completely absent in our dialysis
patients.
How aluminium exposure might lead to this effect on visual
evoked potential is not an area that we addressed in this or previous
papers, but there is no reason to suppose that aluminium toxicity is
manifested solely by causing neuronal death.
We acknowledged the controversy about aluminium and Alzheimer's
disease, but there is no controversy about the potential for aluminium to
be neurotoxic. There are many published reports of visual evoked potential
abnormalities of the sort that we have discussed in dementias including
Alzheimer's disease, some of which were referred to in our paper.
Professor David misquotes our data in his table - we did not
quote absolute latencies for the 55 adults - the 28 cases he quotes were a
separate quality control study and he refers to SD (standard deviation)
where we used SE (standard error), and it is therefore unclear whether the
other data quoted in his table refers to means with SDs or SEs. In
addition there is no mention of the SDs or SEs in the bottom panel of the
table. Our dialysis studies included 16 patients (see reference 24 of web
version paper), and there are no descriptions of pathologically short
pattern evoked visual evoked reponses in the literature.
The ethics of medical publications
Finally, although we were concerned about the possible effects
that the publication of our paper might have on people who were involved
in the Camelford incident, it would have been wrong to withhold it. Had
our studies not been designed to provide scientifically useful information
(despite the inevitable limitations), and only been used to satisfy the
legal process, then the clinical and scientific communities would have
been deprived of results that we find interesting and worthy of debate,
and which we believe are valid.
It is unlikely that another Camelford will occur, so the
opportunity to study the effects will, we hope, not recur.
Paul Altmann
Frank Marsh
Usha Dhanesha
Competing interests: No competing interests