In 1990, Heinz Prechtl recognized that the quality of spontaneous movements provides essential information about the young brain's integrity.¹ His discovery was based on the most frequently occurring movements at an early age, that is, general movements.¹ Prechtl's discovery resulted in the development of the General Movements Assessment (GMA), a tool to assist with predicting neurodevelopmental disorders.^{2, 3} Over the years, the predictive validity of the GMA for neurodisability, as in cerebral palsy (CP), in at-risk infants has been well established.^3-5

The best predictive power of the GMA is during the last phase of general movements, that is, in the period from 2 to 5 months corrected age, when general movements are gradually replaced by goal-directed movements.³ Typical general movements at any age are characterized by movement complexity (the independent exploration of all degrees of freedom in all participating body joints) and variation (the continuous exploration of complexity over time).³ In the last phase, typical general movements are also characterized by fidgety movements, that is, small ‘dancing’ movements that occur irregularly all over the body and form the basic ‘melody’ of general movements.⁶ Atypical general movements are characterized by a marked reduction of movement complexity and variation (in short: general movement complexity) and the absence of age-specific fidgety movements.^{2, 3} Initially, both general movement characteristics were thought to be equally and similarly affected by significant brain lesions. This implied that a marked reduction in general movement complexity would be accompanied by absent fidgety movements and vice versa.⁷ However, this is not always the case. Infants may present with a marked reduction in general movement complexity but nevertheless show fidgety movements;^{8, 9} infants may show absent fidgety movements but not a marked reduction in general movement complexity.⁹ Neurophysiological evidence supports the notion of two separate general movement signs. It suggests two different neuropathological substrates: a marked reduction in general movement complexity reflects impaired subcortical–cortical connectivity, whereas absent fidgety movements are associated with impaired maturation of the cortical networks.³ When infants at 2 to 5 months corrected age present with both general movement signs, the risk for CP is highest.⁸ In infants born preterm, the presence of both signs showed a sensitivity of 98% and a specificity of 91% in predicting CP.⁴ The limited literature available in infants born at term and near-term with neonatal encephalopathy suggests that the predictive values for CP in these infants are more varied and somewhat lower.¹⁰

Clinical application of the GMA is hampered by two factors: (1) it is based on video recordings, which precludes direct feedback to the caregivers during the infant's visit to the clinic; (2) it takes substantial experience to become a proficient GMA assessor.¹¹ Therefore, we wondered whether a standardized infant neurological examination, which is generally easier to apply in clinical practice than the GMA, would result in similar predictive values for an atypical neurodevelopmental outcome as the GMA. A few prospective studies compared the predictive power of the GMA and the Hammersmith Infant Neurological Examination (HINE) for CP in 3-month-old at-risk infants. Prediction using the GMA outperformed prediction using the HINE.^{12, 13} Another study¹⁴ reported that the GMA adequately predicted motor and cognitive delay in at-risk children without CP, but that the HINE only predicted severe motor delay adequately. An explanation for the lesser predictive power of the HINE compared to the GMA could be that the HINE pays less attention to the quality of spontaneous movements (1 of 26 items).¹⁵

The aim of the present study was to compare the predictive value of the GMA for an atypical neurodevelopmental outcome, including CP, at 24 months corrected age (and older) with that of the neurological scale of the Standardized Infant NeuroDevelopmental Assessment (SINDA) in 3-month-old at-risk infants. The SINDA neurological scale is a standardized and validated instrument for infants from 6 weeks to 12 months corrected age. In this age range, it satisfactorily predicts CP and intellectual disability.^{16, 17} A major difference between the HINE and SINDA neurological scales is that a quarter of items in the SINDA address the quality of spontaneous movements, whereas in the HINE only one item addresses movement quality.^{15, 16} In the evaluation of the GMA's predictive properties, specific attention was paid to the contribution of each of the two general movement signs: general movement complexity and fidgety movements.

METHOD

RESULTS

DISCUSSION

In the present study of 3-month-old at-risk infants, the SINDA neurological scale and GMA predicted neurodevelopmental outcome well, including CP, at the age of 24 months (and older). Both definitely abnormal general movement complexity and absent fidgety movements had good predictive properties, with each having its own disadvantage: definitely abnormal general movement complexity had a relatively low positive predictive value and absent fidgety movements had relatively low sensitivities for CP and an atypical outcome. The best predictive values of the GMA were reached by the combination of definitely abnormal general movement complexity and absent or sporadic fidgety movements.

Both the GMA and the SINDA neurological scale had high predictive values for CP and an atypical outcome. Seven of the eight predictive values showed no statistically significant differences. Only the specificity of the SINDA neurological scale to predict CP was lower than that of the GMA, but still greater than 0.900, that is, in the clinically very high range. The finding that the ‘at-risk’ score of the SINDA neurological scale had predictive properties in the same range as those of the GMA in 3-month-old at-risk infants may be clinically relevant. As mentioned earlier, the GMA has two drawbacks (Appendix S3). First, the GMA is based on an off-line assessment of a video recording that is usually made during the consultation. This means that the GMA information is not available during the consultation but only at a later point in time. The lack of direct feedback from the GMA during the consultation may be overcome by asking caregivers to make a general movement video recording before the visit to the clinic, allowing for discussion of the findings during the consultation.^{28, 29} However, this procedure works less well in families of lower socioeconomic status.²⁸ When the SINDA neurological scale is used, the outcome is always immediately available during the consultation. In addition, SINDA does not only provide information on the quality of spontaneous movements, but also insight into the infant's behaviour in four other neurological domains. Second, becoming a proficient GMA assessor requires substantial training.¹¹ It is quicker to learn how to apply the SINDA neurological scale (Appendix S3). In other words, it is relatively easy to learn. The latter is also true for the items on the quality of spontaneous movements, even though these items tap into similar concepts as general movement complexity, that is, movement variation (Appendix S1). Yet, the way in which movement variation is assessed differs between the two methods: in the SINDA neurological scale, items on movement quality are only rated as atypical when the lack of variation or asymmetry in a specific body part is clearly evident, whereas evaluation of general movement complexity variation in all body parts over time needs to be integrated.

It may be argued that a neurological assessment including elicited responses is more difficult to tolerate for young infants than an assessment that is restricted to movement observation. However, the responses elicited by SINDA take little time; when performed in a playful manner, as described in the manual,²² they infrequently result in disturbance of a behavioural state. The latter holds especially true for infants who have reached SINDA's testing age range (≥6 weeks corrected age), when the increased excitability present in the few weeks before and after term age has disappeared.³

Our results confirm the notion that general movement complexity and fidgety movements assess different aspects of the brain's integrity.³ In our at-risk group, absent fidgety movements always coincided with definitely abnormal complexity, a finding that differs from that observed in the general Dutch paediatric population.⁹ Conceivably, the difference is brought about by the difference in proportion of infants born preterm: over 90% in the present study and less than 10% in the general population. Brain lesions resulting in CP that occur most frequently in infants born preterm differ from those in children born at term.³⁰ We also observed that definitely abnormal general movement complexity was not always accompanied by absent fidgety movements, an observation reported previously in high- and low-risk infants.^{8, 9}

Both the predictive properties of definitely abnormal general movement complexity and absent fidgety movements were good, with each having a specific limitation. Definitely abnormal general movement complexity was associated with a relatively high proportion of false positives. However, the term ‘false positive’ refers to the prediction of an atypical outcome, including CP. It does not mean that the assessment's conclusion was false. The presence of definitely abnormal general movement complexity implies that the infant's brain at the moment of the assessment showed dysfunction; it is an indication for early intervention.³ Absent fidgety movements was associated with relatively low sensitivity for CP and an atypical outcome and a relatively high proportion of false negatives for an atypical outcome. It means that the GMA based on the assessment of fidgety movements alone missed some children later diagnosed with an atypical outcome, including CP. The combination of definitely abnormal general movement complexity and absent or sporadic fidgety movements resulted in the highest predictive values for the GMA. Therefore, we recommend that the GMA at a fidgety general movement age is not restricted to the assessment of fidgety movements but also includes the evaluation of general movement complexity.

Abnormal fidgety movements were not often associated with an atypical outcome in terms of CP and general delay at 24 months or later. This does not imply that abnormal fidgety movements does not have any clinical relevance. It has been hypothesized that in infants with high familial risk of autism spectrum disorder, abnormal or sporadic fidgety movements may be an early marker of increased risk for the disorder.³¹ Future studies need to test this hypothesis.

The strength of the study was its evaluation of predictive properties of infant assessments during clinical routines in a non-academic setting for at-risk infants. This means that the outcomes may be generalized to similar clinical settings. On the other hand, the setting of the study does not allow for generalization to the general paediatric population. The clinical setting also had limitations. The paediatricians doing the neurological follow-up were aware of the infant's GMA and SINDA scores. Second, a significant proportion of infants who were doing relatively well were not included in the study because they had no follow-up at the SPZ outpatient clinic. This may have affected the sensitivity and negative predictive values of both GMA and SINDA slightly as we may have underestimated the number of false negatives.

In conclusion, in our study of 3-month-old at-risk infants, the GMA and the SINDA neurological scale predicted atypical neurodevelopmental outcome well, including CP at 24 months (and older). As it takes less time to learn to apply the SINDA than to learn the GMA, and bearing in mind that the GMA mostly does not allow for direct feedback during the consultation whereas the SINDA always does, our findings could have implications for clinical routines. Therefore, we recommend that future studies evaluate both the ease of clinical integration and predictive properties of the GMA and SINDA. Our study also showed that the GMA has the best predictive properties regarding general movement complexity and fidgety movements.