Commentaar SAR en Hirasingh rond scoliose onderzoek + antwoord
The effect of school screening on surgery for adolescent idiopathic scoliosis: reanalysis is needed
With much interest, we read the paper of Wiegersma and coworkers regarding the effect of school screening on surgery for adolescent idiopathic scoliosis. It is based on a comparison of the number of cases of surgery for idiopathic scoliosis in Dutch regions in which the Youth Health Care (YHC) department screens and not screens for this disorders. The topic is important as scoliosis is disabling and may require extensive surgery. An intervention that effectively prevents extensive surgery at reasonable costs should thus be encouraged. Regarding school screening for idiopathic scoliosis (and subsequent conservative treatment) Wiegersma et al. conclude that this does not reduce population rates for scoliosis surgery and should thus be reconsidered. Unfortunately, this conclusion seems to be based on a flawed interpretation by the authors of their own data. Central in the study of Wiegersma et al. are the data as presented in the first table of their paper, which is reproduced here. In their calculations they first assume that a screening for scoliosis of children aged 12 or 13 years will prevent surgery for scoliosis in the age-group 12-19 years. On the basis of this, they conclude that screening for scoliosis has no preventive effect regarding surgery for scoliosis: the odds ratio (OR) comparing regions in which screening is performed with regions in which this is not done, is ‘1.00’ (whereas only values between 0 and 1 indicate a preventive effect). However, their assumption in constructing this table is that a screening of children aged 12-13 years should has an immediate effect for children aged 12 or 13 years as well. This is very unlikely: it means that after detection of early scoliosis, a conservative treatment should be started immediately and should have immediate effects too. In some instances, the screening should even have an effect before the child could actually have been screened.
Wiegersma and coworkers recognise that this line of thinking is logically impossible and therefore do a second calculation in which they exclude children with surgery sooner than one year after possible screening. This calculation seems to be seriously flawed, however. It is based on 58 cases with surgery in the screening regions, and 92 cases in the non screening regions. We cannot derive these numbers of cases in any way from their tables: to be unbiased, regarding cases and controls from similar agegroups should be excluded. We do not know the exact age on which children in the non screening regions would have been screened if there had been a screening in these regions. However, assuming, like Wiegersma et al. do that the screening occurs in the age-group 12-13 year and that the effects of a conservative treatment can only be expected after at least one year, children aged 12-14 years should be excluded from the analysis. In this case the resulting OR (95% confidence interval) is 0.87 (0.571.32).[2 3] This implies a somewhat preventive effect, though clearly without statistical significance due to the small number of cases involved. If the cut-off is set at a higher age, this preventive effect increases though it remains without statistical significance (table, last columns). On the basis of this reanalysis we conclude that the data of Wiegersma and coworkers indicate that screening for idiopathic scoliosis may have some preventive effect. Their study is however to small to yield an accurate estimate of this effect. Furthermore, the design of their study will lead to an underestimation of this preventive effect. Firstly, children may move from a screening to a nonscreening region (and vice versa) in the period in which the effect of the screening on surgery rates would occur. This will always lead to an underestimation of the actual effects of this screening.
Secondly, in the study period almost all Dutch children aged 12, 13 or 14 years received either a screening for scoliosis or a preventive examination by a YHC. Examination of the trunk is usually also part of the latter too. This implies that they compare the effect of a specific screening for scoliosis with the effect of a combined examination, again leading to an underestimation of the net effect of such a screening. Thirdly, even in regions labeled as screening, some children will not be invited or not participate, again giving an underestimation. We invite Wiegersma and coworkers to reformulate their conclusion in a sense that their study gives some, though inconclusive, evidence for effectiveness of screening on idiopathic scoliosis.
Table Distribution of cases and of referents across Youth Health Care department that screen (‘screening) and do not screen (‘non screening) for idiopathic scoliosis; odds ratios and 95%-confidence intervals at each age‑group refer to the number of cases and referents of that age and older
screening no screening OR* 95%-CI
CASES by age
12 years 6 7 1.00 0.74 1.35
13 years 15 15 0.97 0.71 1.33
14 years 14 27 0.86 0.61 1.23
15 years 17 25 0.87 0.57 1.32
16 years 8 19 0.71 0.40 1.26
17 years 7 9 0.74 0.34 1.62
18 years 2 10 0.30 0.07 1.34
19 years 0 1
total of cases 69 113
total of cases 413152 676840
* Odds ratio regarding the number of surgery cases among children of this age and older (but < 20years).
We were pleasantly surprised to learn that more than one year after it’s publication our article still generates enough interest to give rise to letters to the editor. Furthermore, we are grateful to the authors of the letter for bringing forward the interesting point concerning the age difference at surgery and its possible explanation. It provides new insights into the reasons behind the inadvertently adverse effects of youth health care activities, discussed in other publications,[1,2] although we doubt this was the authors’ intention.
Before explaining this in more detail, we will first address the comments of the authors. In their letter Reijneveld and Hirasing state, that we should have excluded the 12 – 14 year olds in toto, because the effect of screening of scoliosis on surgery could only be expected one year after such a screening. This, of course, is a rather curious line of argument. It must be obvious that those screened at age 12 can be included when surgery took place at age 13 or 14, which of course is the reason, why the numbers for the second calculation cannot be derived from the table in question. This misinterpretation of the authors can only be explained by assuming they have somehow overlooked the fact, that information on surgery was obtained at an individual level. This assumption is further substantiated by their comment that in some instances the screening should even have an effect before the child could actually have been screened, although it could also indicate a rather perfunctory reading of the article – see ‘Population and Methods’ section, last two paragraphs. Also, we would like to point out that the remark about the effect being clearly without statistical significance because of the small number of cases involved, seems rather at odds with prevailing epidemiological insights.
Furthermore, in stating that almost all (sic) Dutch children would have had some sort of screening, they demonstrate their – of course understandable – lack of relevant information. Nevertheless, even without the pertinent data, it must be clear that children can have had no screening or physical examination for several reasons, transition to another frequency of activities by the youth health care department being the most apparent.
As for the remarks about nondifferential misclassification, the inclusion of preventive examinations at age 14, and the difference between offering an activity and actual participation, this was extensively discussed in the article itself and in response to an earlier letter to the editor.[3,5] We do not feel it necessary to burden the reader with a reiteration of the arguments.
To return to our opening remark about the reasons behind the inadvertently adverse effects of youth health care activities: surgery for scoliosis is an extremely severe procedure with clear (postoperative) risks, undertaken only in the gravest of cases, and carried out in one of seven specialised centres in The Netherlands, following a rigorous protocol. In their table, the authors of the letter show clearly, that as the cut-off point for age is set higher, the OR decreases. This, coupled with the fact that the mean age at surgery for screened cases was higher than that of cases that were not screened – a fact also demonstrated in said table – indicates that the level of severity of scoliosis, necessitating a surgical intervention, is reached sooner in the screened population. This is a further, worrying, indication that in regions where youth health care workers are active, parents and teachers are lulled into a false sense of security, thus reducing their alertness, which results in a greater occurrence of the behaviour or disorders, targeted by the preventive activities. So, in this particular instance, a patient delay occurs, resulting in a later, and thus less effective, treatment of the scoliosis. In the case of suicide, it resulted in a significantly increased number of parasuicides in regions where open consultation hours were held, and in the case of health compromising behaviour it resulted in an increase in smoking and use of alcohol.
In the light of this new evidence brought to our attention, we would like to stress even more our appeal for a critical reappraisal of youth health care activities, in this case more specifically the screening for scoliosis. In effect, as the OR for surgery for cases one year after possible surgery is 1.03, after reformulation our conclusion would more likely read: ‘our study gives some, though inconclusive, evidence for an adverse effect of screening on idiopathic scoliosis.
- Wiegersma PA, Hofman A, Zielhuis GA. Prevention of suicide by youth health care. Public Health 1999;113:125-130.
- Wiegersma PA. Long term effects of youth health care for schoolchildren in The Netherlands. Nijmegen 1999, Thesis.
- Wiegersma PA, Hofman A, Zielhuis GA. The effect of school screening on surgery for adolescent idiopathic scoliosis. Eur J Public Health 1998;8:237-240.
- Rothman KJ, Greenland S, eds. Modern epidemiology. Philadelphia: Lippincott-Raven Publishers, 1998.
- Wiegersma PA, Hofman A, Zielhuis GA. The effect of school screening on surgery for adolescent idiopathic scoliosis: response to readers’ comment. Eur J Public Health 1999;9:152.