Dear Editor,

I would also like to air my serious concerns regarding the whole concept of the Archimedes section now appearing in Archives of Disease in Childhood. Iain Chalmers and his group rallied for many years against the system whereby experts were invited or chose to write review articles on various clinical topics in peer-reviewed journals. Dr Chalmers’ argument was that these articles were always either intentionally or unintentionally biased and thus likely to mislead.

Bias arose from two main sources; firstly from the expert who would be more likely to quote papers supportive of his or her point of view and less likely to address questions raised by papers which did not support this; and secondly by the peer-review publication system which ensured that trials with positive results were more likely to find a place in the literature than were similar trials with negative results.

Having finally convinced many of the logic of their argument, Chalmers et al went on to found the Cochrane Collaboration and to champion the careful and detailed analysis of all that was published, or available but not published, in respect of a specific question. From this emerged the meta-analysis of randomised studies which has since been recognised (by most) to have great merit.

It is, perhaps, instructive to look back at the story behind one of the most influential meta-analyses published to date, namely that by Patricia Crowley in respect of the effectiveness of antenatal steroids in anticipation of preterm birth [1]. Arguably, one of the contributory factors to the delay in the critical evaluation of this therapy in the UK was the publication of a review article on advances in respiratory distress syndrome in the BMJ in 1982 [2]. This article, from a respected UK neonatologist who effectively pooh-poohed steroids (“the evidence suggests that antenatal steroids are of value only in white males, and even for them the benefit is mainly in those of 30-32 weeks’ gestation…”) in favour of surfactant (“A much more appealing treatment…”), is a good example of the deleterious effect of ‘eminence-based’ medicine.

Should anyone doubt my assertion that such leading articles might be knowingly biased you only have to look in the correspondence following this one to find the letter written in response to a query where the author states “He quotes one paper that shows no effect on blood glucose, and I quoted a paper that did have an effect on blood glucose. You ‘pay your money and take your choice’ how you interpret the literature”.[3]

However, it does not end there. Much of the evidence that Professor Crowley managed to assemble for her meta-analysis was either unpublished or else was published only in obscure, non peer-reviewed publications of very limited circulation. It was Professor Crowley’s urging that persuaded Harold Gamsu to publish his study [4] which, till then, was only published in abstract. Indeed, the 1972 paper by Liggins & Howie [5] contained details of only the first 282 mothers randomised in a study that eventually randomised over 1000 babies. The rest was only published in a Ross symposium [6] and a working party document from the RCOG which is now virtually unobtainable.[7] Even the Royal Society of Medicine does not possess a copy.

To do a proper analysis takes considerable time and effort. It cannot be done merely by a computer and an electronic search strategy. Attempting to do so risks the same problems with bias associated with review articles. My concern is that by printing the Archimedes section in a publication like the Archives of Disease in Childhood, the journal is giving spurious credibility and, dare one say a Cochrane-look, to reviews which are distinctly light-weight. To encourage people to look objectively at the literature is one thing but to imply that it is acceptable to look at it in the dilettante manner that lies behind most of these reviews is quite another.

If Archimedes were alive today I would encourage him to sue.

Dear Editor,

In their commentary in response to our study on the diagnostic utility of anti-basal ganglia antibodies (ABGA) in post-streptococcal movement disorders, Singer and colleagues [1] raise certain points and present laboratory data which require a considered response.

The band at ~55 kDa appears to be in saturation and the number of bands detected raises the issue of non-specific or low affinity binding. This is a common and well documented problem when using enhanced chemiluminescence (ECL) detection systems, and it is for this reason that we do not use ECL routinely in our laboratory for diagnostic screening for anti-neuronal antibodies. Non- specific binding, however, can be reduced by optimising the quantity of protein loaded, using specialised transfer membranes, using specially formulated blocking solutions, optimising the washing steps, reducing the concentration of the second or detector antibody, using more specific detector antibodies or adding a chaotropic agent to the washing solution. It is important to know whether or not Singer and colleagues have examined these strategies and to have more specific details of how the assay was performed.

Singer and colleagues did not describe their Western blotting methods in their commentary, nor the method used to remove lipid from their basal ganglia preparation or the colorimetric agent used in the experiments in Figure 2 [1]. In their most recent paper using this technique they only use a 90 minute incubation period at room temperature to detect ABGA [2]. In our hands we find an overnight incubation at 4 deg. C with agitation to be optimal [3]. In their latest paper they do not explain how they prepared their proteins for electrophoretic separation [2].

Finally they use the second or detector antibody, horseradish peroxide-conjugated sheep anti-human IgG (Amersham International) in a dilution of 1:2,500 [2]. This dilution is more in line with that used for ELISA applications - for ECL applications dilutions of 1:10,000 to 1:50,000 are usually recommended to reduce or eliminate non-specific low affinity bands.

In Figure 2 Singer and colleagues demonstrate that when using second or detector antibody alone two distinct bands are detected [1]. This is a normal and anticipated finding. The antigen source is human tissue and would therefore contain human immunoglobulin. The bands therefore almost certainly represent the heavy and light chains of human immunoglobulin against which the second or detector antibody reacts. The molecular weight of the upper band or Ig heavy chain should be ~55 kDa and not 60 kDa as suggested in their commentary.

The lower band or Ig light chain should have a molecular weight of ~22.5 kDa. The molecular weight of the lower band is marked as being greater than 25 kDa in the figure [1]; this could simply be because the migrations of molecular weight standards vary with different buffer and electrophoresis systems.

Singer and colleagues also question the specificity of our ABGA results because we are detecting ABGA in "different clinical disorders" [1]. It is important not to equate clinical phenomenology with a specific disease. Although these patients had different movement disorders this does not imply they had different disease processes underlying their movement disorder. In well-established diseases of the basal ganglia, e.g. Huntington¡¦s disease, the spectrum of movement disorders is variable.

In conclusion, we acknowledge Singer and colleagues' position on the use of ABGAs in the diagnosis of post-streptococcal movement disorders and encourage other investigators to establish and set-up up ABGA assays. To do this, however, it is important to have free access to the methodological details of the various published assays. Since the publication of these papers we have been in formal contact with Singer and colleagues and plan to initiate a project to standardise our assays and to exchange samples to evaluate their diagnostic utility.

References

(1). Singer HS, Hong JJ, Rippel CA, Pardo CA. The need for caution in considering the diagnostic utility of antibasal ganglia antibodies in movement disorders. Arch Dis Child 2004;89:595-7.

(2). Singer HS, Loiselle CR, Lee O, Minzer K, Swedo S, Grus FH. Anti- basal ganglia antibodies in PANDAS. Mov Disord 2004;19:406-15.

(3). Church AJ, Cardoso F, Dale RC, Lees AJ, Thompson EJ, Giovannoni G. Anti-basal ganglia antibodies in acute and persistent Sydenham's chorea. Neurology 2002;59:227-31.

Dear editor,

We read the article by Barnett et al with interest[1]. We believe prediction of future outcome using developmental tests alone is difficult, and can be erroneous[2]. for the following reasons. Firstly, development is a result of complex and dynamic on-going interaction between genetic potential, pre/peri/post natal variables, socio-economic and environmental factors[2,3]. Secondly, developmental tests, like any other diagnostic test, can only provide a snapshot of this evolving process. Thirdly, developmental tests should not substitute but supplement the ‘full assessment’ which includes history, examination (including neurological examination), careful observations and follow-up[2,3]. Over-reliance on developmental tests alone can be deceiving. ‘Full assessment’ would provide better prediction and could have picked up the ‘missing hemiplegias’ as reported[1].

Moreover comparison between Griffiths scales and other test scores are prone to misinterpretation[3,4], especially when the tests were administered at different ages[1]. This is because the scales differ considerably from one another with regard to selection of test items, test construction, standardisation and inter-rater variability, etc. Results would have been valid had authors used Griffiths scales again at school age, instead they used a cluster of tests that were not used before[5]. The study has attempted to correlate early Griffiths scores with other tests at school age and authors agree that although the results are statistically significant they are of little practical significance[1].

Furthermore it is known that the incidence of minor neurological dysfunction increases with age[6]. As the child is challenged by more complex skill acquisition with increasing age, the possibility of recognising neurological dysfunction increases. Hence Griffiths Scales[7], like several other developmental tests (Denver Developmental Screening Test[8], Bayley Scales of Infant Development[9] etc.,) have identified neuromotor dysfunction at school age in children who were ‘apparently normal’ at 2 years of age. This study demonstrates the above phenomenon and does not provide clear evidence regarding predictability of Griffiths scales.

Finally, infants with neonatal encephalopathy are at significantly high-risk of death, serious disability and later neuromotor dysfunction[1]. The authors rightly point out that failure to identify such children would lead to missed opportunities for early intervention. Accordingly, continued surveillance of such ‘high-risk’ children is important. Therefore the rationalising claim of ‘effective use of limited resources’ in follow-up of such children is questionable.

References

(1). Barnett AL, Guzzetta A, Mercuri E et al. Can the Griffiths scales predict neuromotor and perceptual-motor impairment in term infants with neonatal encephalopathy? Arch Dis Child 2004;89(7):637-43.

(2). Illingworth RS. The development of infant and young child Normal and Abnormal, 9th edn. London: Churchill Livingstone, 1987.

(3). Griffiths R.The abilities of babies. High Wycombe, UK: The Test Agency 1986.

(4). Ivens J, Martin N. A common metric for the Griffiths scales. Arch Dis Child 2002;87(2):109-10.

(5). Sackett DL, Straus SE, Richardson WS et al. Evidence-Based Medicine. How to Practise and Teach EBM. London: Churchill Livingstone, 2000.

(6). Goyen TA, Lui K. Longitudinal motor development of ‘apparently normal’ high-risk infants at 18 months, 3 and 5 years. Early Hum Dev 2002;70:103- 15.

(7). Majnemer A, Rosenblatt B, Riley P. Predicting outcome in high-risk newborns with a neonatal neurobehavioural assessment. Am J Occup Ther 1994;48(8):723-32.

(8). Greer S, Bauchner H, Zuckerman B. The Denver Developmental Screening Test: how good is its predictive validity? Dev Med Child Neurol 1989;31(6):774-81.

(9). Crowe TK, Deitz JC, Bennett FC. The relationship between the Bayley Scales of Infant Development and preschool gross motor and cognitive performance. Am J Occup Ther 1987;41(6):374-8.

Dear editor,

We present a case that illustrates the points made by Baker et al, and emphasises the need for care in the diagnosis of mass lesions.

N.B is a 14 year old girl who presented with a 3 week history of diarrhoea, intial vomiting and malaise. She was seen by her general practitioner on the day of admission as the family were concerned about continued loose bowel motions and her ongoing malaise. An ultrasound arranged by the GP showed 3 pelvic mass lesions. Examination showed an abdomen moderately distended by the masses, which were tense and only mildly tender. CT showed a 10.5x10x12cm mass in the pouch of douglas, with 2 smaller lesions in the left iliac fossa region measuring 6x3.5x5cm and 5x3.5x7cm. On ultrasound they appeared of mixed echogenicity, and the diagnosis of malignancy was considered, but CT showed them to be of fluid density, raising the possibility of infection. There was evidence of bilateral ureteric obstruction.

Laparotomy confirmed the diagnosis of appendix abscesses, with ongoing concern regarding the lesion on the Pouch of Douglas which includes Uterus, ovaries and loops of bowel.

We are grateful to Baker et al for raising the profile of this differential, as the family had assumed the worst in this case, as do many others (including doctors) when mass lesions are identified on clinical and radiological examination.