Send letter to journal:
Re: Cognitive development of ICSI children

We read with interest the leading article in your journal by Sutcliffe[1] in which the various issues associated with intracytoplasmic sperm injection (ICSI) were discussed. We thought that in general it provided a good general overview of the subject. However, we do take exception to Dr Sutcliffe's comments about the supposed limitations of our study,[2] which was the first to be published in which the cognitive development of children conceived using ICSI was compared with a control group.

There were four areas in which Dr Sutcliffe claimed our study had fault: lack of power, multiple observers, unstandardised testing systems, and failure to allow for confounders. We would like to briefly deal with each of those issues.

Power
Our study was well powered to detect the differences found. Considering the sample sizes and standard deviation of the mean MDI value for the total cohort, there was 98% power to detect the 6-point difference in mean MDI that we observed. Similarly the study was well powered to detect the differences in percentage of infants with delayed development, 87% for IVF versus ICSI (2 vs 17%) and 91% for naturally conceived controls versus ICSI (1 vs 17%). These figures compare favourably with Sutcliffe's study,[p3] which had 91% power to detect a five-point difference in GQ.

Multiple observers
There were two testers, both of whom have considerable experience in Developmental assessment using the Bayley Scales of Infant Development and also in the Griffiths Mental Development Scales. We documented in our methods section that inter-rater reliability for assessing the Bayley's MDI by these two investigators was tested using Kendall's tau correlation, and found to be very high at 0.953.

Unstandardised testing systems
The Bayley Scales of Infant Development (2nd edition)[4] is an age-appropriate test that has sound psychometric properties enabling the identification of performance differences among research groups. It was extensively re-standardised in 1991-1992, and this has resulted in an up-to-date, valid and reliable test of infant development. The Bayley Scales of Infant Development (2nd edition) was chosen by us because it was the most recently re-standardised infant test available at the commencement of our study.

Dr Sutcliffe used the Griffiths Scales of Mental Development[5] to test his children. This is also a standardised test, but it is designed for younger infants and does have an age-related "ceiling effect", such that the highest possible score on Griffiths GQ for children in the latter half of their second year is decreased compared with the full possible range of GQ values. For example, the highest possible GQ value is 150, but for children in their 23rd month it decreases to 124. Furthermore, a ceiling of maximum performance in each sub-scale is established when a child fails 6 items in a row. Even advanced infants in their 13th and 14th month may not reach this criterion. Dr Sutcliffe does not report whether any infants in either group failed to reach a ceiling on any of the Griffiths sub- scales. Given the mean age and standard deviation of the mean for Sutcliffe's cohorts it is possible that such ceiling effects may have occurred in the older children, and that this could disguise true differences between ICSI and control children.

Failure to allow for confounders
This is untrue. Our Methods section clearly states that where between-group differences were found for perinatal and demographic variables, these variables were included as covariates in the analyses comparing groups for outcome at 1 year. The main demographic difference that we detected was that ICSI children were more likely to have fathers with an unskilled occupation than either the IVF or naturally conceived children. In our results we documented that we performed a subset analysis in which we excluded all children from all three groups whose fathers had an unskilled population. We documented that for those children whose fathers had a managerial, professional or skilled occupation the differences in proportions with a delayed MDI was still significantly different (16% ICSI, 1% IVF and 1% natural conception). This difference was highly statistically significant (P<0.0001). Given the numbers in this subset, the power for detecting this difference was 86%. Sutcliffe also detected a number of socio-demographic differences between groups in his study, but does not indicate that he allowed for these confounders in his analyses.

We do agree with Dr Sutcliffe that at this point in time it is unsafe to draw any conclusions about the long-term wellbeing of children conceived using ICSI and that the health of the child should be paramount in further developments of new assisted reproductive technology techniques. We are presently reassessing our previously studied cohort at five years of age, and have enrolled additional ICSI and naturally conceived children to address the areas of socio-demographic imbalance observed in the original cohort. Hopefully the results of our study and that of the European collaborative group will help to clarify the issue of cognitive development for children conceived using ICSI.

Garth I Leslie MD BS BSc(Med) FRACP
Clinical Associate Professor
University of Sydney Northern Clinical School
Royal North Shore Hospital
St Leonards, NSW, Australia

Frances Gibson PhD
Psychologist, Department of Neonatology
Royal North Shore Hospital
St Leonards, NSW, Australia

References
(1) Sutcliffe AG. Intracytoplasmic sperm injection and other aspects of new reproductive technologies. Arch Dis Child 2000;83:98-100.

(2) Bowen JR, Gibson FL, Leslie GI, Saunders DM. Medical and developmental outcome at 1 year for children conceived by intracytoplasmic sperm injection. Lancet 1998;351:1529-34.

(3) Sutcliffe AG, Taylor B, Li J, Grudzinskas G, Thornton S, Lieberman B. Children born after intracytoplasmic sperm injection: population control study. BMJ 1999;318:704-5.

(4) Bayley N. Bayley scales of infant development, 2nd edition. San Antonio: The Psychological Corporation, 1993.

Send letter to journal:
Re: Re: Cognitive development of ICSI children

Dear Editor,

I read with interest the comments by Leslie and Gibson concerning my reference[1] to their study.[2] They describe their study as the first on ICSI conceived children with controls but make no reference to the study by Bonduelle et al[3] in the same edition of the Lancet, which had contradictory findings. I previously corresponded[4] at the time their study was published concerning their findings and wish to respond to their points.

Power
In their small single-centre study, a difference of 6.6 was found between the mean mental development index (MDI) of the ICSI group and that of a normally conceived control group (95.9 vs 102.5). However, a power calculation based on this difference and the test intrinsic SD of 15 showed that to achieve 95% CI, at least 135 children would be needed in each group. A control group of 80 children would achieve power of only 79.4%.

It is unfortunate that Leslie and Gibson chose to refer only to the interim findings of the UK population study of ICSI conceived children rather than the final findings for comparisons of power. These findings were presented at ESHRE meeting in Tours, France, 1999[5] at which Gibson was present. These findings are presently being considered by a major international journal but suffice it to say that the 208 ICSI conceived toddlers and 221 normally conceived well matched controls were strikingly similar in their neurodevelopmental abilities (5) at mean age 18 months. In this study there was a power of greater than 99% to detect a 5 point difference in the newly standardised Griffiths scales of mental development between study and control groups.

Multiple observers
In their study there were two observers. In the UK population study only one. Unfortunately a common reason for invalid results in such studies is interobserver error, since 100% agreement between observers seldom occurs.

Unstandardised testing systems
Leslie and Gibson defend their use of the Bayley scales but do not acknowledge that these scales had not been standardised in Australia. The reassuring report from Bonduelle et al referred to above[3] also used the Bayley scales but their translated version had been standardised against the indigent population of Dutch speaking children. To put the significance of non-standardisation into a more general context, during the restandardisation of the Griffiths scales of mental development (1996) the scales were compared to scales standardised in the early 70s. It was noted a toddler of average ability when tested in 1996 (mean score 100) would get a mean score of 107.8 on the old scales. Leslie and Gibson refer to testing children close to twenty-four months and possible pitfalls of this in the application of the scales. Allowance was made in the design of the new Griffiths scales for children of above average ability but I acknowledge that 2 children in each group (2 out of 208 study and 2 out of 221 controls) may have been theoretically affected by this.

Failure to allow for confounders
Unfortunately the Sydney study chose to use control children recruited from an earlier study (ie, retrospective controls). This may have made the study results available quicker but is scientifically unwise. Colleagues in Sydney have also questioned the different socioeconomic features of the neighborhoods from whence the control and study children were recruited.

Of the original 108 pregnancies, I note 38 were from embryos replaced after cryopreservation. Minor developmental deficits have already been noted in children born after replacement of cryopreserved embryos before the introduction of ICSI.[6] Although in the UK study there were a few minor confounders for neurodevelopmental outcome this was allowed for in the analysis.

There is an ongoing European 5 nation study of ICSI conceived children at aged 5 years (ICSI-CFO; International Collaborative Study of ICSI - Child and Family Outcomes). Involving 650 ICSI conceived children, 650 normally conceived control children and a 650 IVF comparison group this should give further information about longer term outcome.[7]

The paper by Bowen, Lesley, Gibson et al[2] has increased awareness of the necessity to enquire about outcome after ICSI. It also had many good points such as the inclusion of multiple births and the blinding of the two observers involved to outcome. However the anxiety the paper caused to the families who have benefited from this treatment and their associated fertility treatment teams worldwide was probably not warranted in view of my comments above.

Alastair G. Sutcliffe MD, MRCP, MRCPCH
Lecturer in Child Health
Centre for Community Child Health Studies
Department of Paediatrics
Royal Free Campus
Royal Free and University College Medical School
University College London, Rowland Hill Street
Hampstead, London NW3 2PF, UK

References
(1) Sutcliffe A G. Intracytoplasmic sperm injection and other aspects of new reproductive technologies. Arch Dis Child 2000;83:98-100.

(2) Bowen JR, Gibson FL, Leslie GI, Saunders DM. Medical and developmental outcome at 1 year for children conceived by intracytoplasmic sperm injection. Lancet 1998;351:1529-31.

(3) Bonduelle M, Joris H, Hofmans K, Liebaers I, Van Steirteghem. Mental development of 201 ICSI children at 2 years of age Lancet 1998;351:1553.

(4) Sutcliffe A G, Taylor B, Li J, Grudzinskas G, Thornton S, Lieberman B, Children conceived by intracytoplasmic sperm injection. Lancet 1998;352:578-9.

(5) Sutcliffe A G, Taylor B, Li J, Grudzinskas G, Thornton S, Lieberman B, UK study of children born after Intracytoplasmic sperm injection [abstract 0-018]. Human Reproduction 1999;14:13.

(6) Sutcliffe AG, DeSouza SW, Cadman J, Richards B, McKinlay IA, Lieberman B. Outcome in children from cryopreserved embryos. Arch Dis Child 1995;72:290-3.