Convergence and Accommodation Development Is Preprogrammed in Premature Infants

Abstract

Purpose: This study investigated whether vergence and accommodation development in preterm infants is preprogrammed or is driven by experience.

Methods: Thirty-two healthy infants, born at mean 34 weeks gestation (range, 31.2–36 weeks), were compared with 45 healthy full-term infants (mean 40.0 weeks) over a 6-month period, starting at 4 to 6 weeks postnatally. Simultaneous accommodation and convergence to a detailed target were measured using a Plusoptix PowerRefII infrared photorefractor as a target moved between 0.33 and 2 m. Stimulus/response gains and responses at 0.33 and 2 m were compared by both corrected (gestational) age and chronological (postnatal) age.

Results: When compared by their corrected age, preterm and full-term infants showed few significant differences in vergence and accommodation responses after 6 to 7 weeks of age. However, when compared by chronological age, preterm infants' responses were more variable, with significantly reduced vergence gains, reduced vergence response at 0.33 m, reduced accommodation gain, and increased accommodation at 2 m compared to full-term infants between 8 and 13 weeks after birth.

Conclusions: When matched by corrected age, vergence and accommodation in preterm infants show few differences from full-term infants' responses. Maturation appears preprogrammed and is not advanced by visual experience. Longer periods of immature visual responses might leave preterm infants more at risk of development of oculomotor deficits such as strabismus.

Bifoveal fixation is maintained by the precise coordination of vergence, versions, and accommodation to maintain ocular alignment and image clarity. During postnatal development, sensory fusion, motor fusion, and accommodation become more closely coordinated^1–5 as visual experience acts on a basic genetic structure. It is unclear, however, whether these systems and relationships are initially preprogrammed and dependent on physical maturation or are influenced by visual experience from the outset. Comparing performance between preterm and full-term infants provides an opportunity to explore these developmental processes. Figure 1 illustrates the two alternative possibilities.⁶ If responses are mainly preprogrammed, then both full-term and preterm infants will reach maturity at the same corrected (postconceptual/gestational) age, but the preterm infants will be older when compared by chronological (postnatal) age. If responses are more experience dependent, then the two groups will reach maturity at similar chronological ages, but the preterm infants will have reached this at an earlier stage of physical maturation (younger corrected age). 

Figure 1

 

Illustration of differences in hypothetical development of mature responses (vergence and accommodation in this case) between full-term and preterm infants in preprogrammed and experience-dependent scenarios (based on the illustration in Jandó G, Mikó-Baráth E, Markó K, Hollódy K, Török B, Kovacs I. Early-onset binocularity in preterm infants reveals experience-dependent visual development in humans. Proc Natl Acad Sci U S A. 2012;109:11049–11052, with publisher permission). The maturational hypothesis predicts that full- and preterm infants' responses should develop at the same rate when matched by the corrected age (top left), but preterm infants will be chronologically older when they mature (top right). The experience-dependent hypothesis predicts that preterm infants should develop mature responses before full-term infants when matched by the corrected age (lower left) but at the same chronological age (lower right).

Using this paradigm, previous research suggests that most sensory visual development is mainly preprogrammed and that the earlier visual experience resulting from prematurity does not advance most aspects of visual development (for reviews see Refs. 7, 8). The effect of prematurity on development of convergence and accommodation during early infancy has been described only in studies of very small groups, but these also suggest a maturational time course for convergence⁹ and accommodation.¹⁰ 

Importantly for this paper, however, a recent study by Jandó et al.⁶ found that the development of the binocular response to dynamic random dot correlograms (DRDCs) in preterm infants depended on visual experience, not physical maturation. Dynamic random dot correlograms are binocular stimuli that only elicit a characteristic visual evoked potential (VEP) in mature binocular systems¹¹ and are therefore a marker for cortical binocularity in developing infants.^12,13 The same study, however, found that pattern reversal VEP latency, which is a measure of integrity of the visual pathway, was not advanced by premature birth, demonstrating that despite an immature visual pathway, the visual cortex can accept environmental stimulation from birth. These results provided a rationale for more detailed exploration of whether the development of convergence and accommodation is maturational or experiential, but there is also clinical relevance. 

Children born preterm are known to have a higher prevalence of accommodative^14,15 and nonaccommodative^16–18 strabismus. However, what causes this increased prevalence is unclear.^19,20 We know that full-term neonates can have periods of ocular misalignment,²¹ inaccurate vergence and accommodation,^1,3 and even clinically diagnosed eye muscle palsies²² without any apparent long-term harm; but if misalignment persists or increases into the critical period for binocularity, the risk of strabismus, suppression, and amblyopia is known to be severe. Tychsen²³ has suggested that decorrelated sensory input between the eyes in the critical period for binocular vision is “a sufficient cause for infantile esotropia” (p. 564). 

We hypothesized that a mismatch in developmental timing between the sensory and motor components of binocularity could increase the risk of strabismus. If vergence development relates to the corrected age, it would develop later post delivery in preterm infants, so these infants would have longer with imprecise vergence and frequent misalignments. If experience-dependent sensory binocularity,⁶ which normally emerges only once vergence is more stable, emerges relatively earlier, immature vergence, which is normally of little consequence, would become a sufficient cause of decorrelated sensory input and be an additional risk factor for the development of strabismus. 

This paper describes the development of vergence and accommodation in groups of low-risk preterm and full-term infants in order to test the experience-dependent versus maturational hypotheses. 

Methods

The study adhered to the tenets of the Declaration of Helsinki and was approved and scrutinized by institutional and UK National Health Service Ethics Committees. Informed consent was obtained from the parents of all infants. 

Participants

We defined the corrected age and the chronological age as recommended by the American Academy of Pediatrics Committee on Fetus and Newborn.²⁴ The chronological age was defined as the time elapsed from birth, while the corrected age was the chronological age reduced by the number of weeks born before 40 weeks of gestation. The corrected age was calculated from the expected delivery date calculated from the first day of the last menstrual period. Thirty-six preterm infants born between 31 weeks + 2 days and 36 weeks of gestational age (mean 34.09, SD 1.35 weeks) were recruited from a local maternity hospital. Of these, 32 infants were able to be tested at least once. We chose not to study more premature infants in whom high rates of retinopathy of prematurity, general health complications, and later developmental and perceptual difficulties²⁵ might have confounded the data. Three infants were also defined as “small for dates” (low birth weight for their gestational age), and two weighed less than 1500 g (1465 and 1361 g). None had suffered any perinatal or postnatal neurological complications; all were healthy when tested; and none subsequently developed strabismus and at the time of writing all are at least 2.5 years old (corrected age). 

Reasons for preterm delivery were mainly twin pregnancy (53%) and pre-eclampsia (15%). We were unable to analyze the twin data separately. Of the many twins, we collected data from both twins in only six pairs, and rarely from both twins at the same visit. Only one set of monozygotic twins was tested. 

Preterm infants were compared with 45 typically developing full-term infants (born between gestational age 37 weeks + 2 days and 42 weeks + 1 day: mean 40.0 weeks ± 1.6 days), recruited from our departmental Infant Database. Data from these infants contributed to a previous publication, which reported data for the infants on visits when they showed no or minimal (less than +2.0 D) hyperopia.³ This paper reports some additional data from 44 testing sessions in 19 infants (out of a total of 300 sessions) when these infants showed mild hyperopia (up to +3.0 D at 16 weeks of age). 

All infants were recruited soon after birth. We booked the first test at between 6 weeks corrected age for both groups (because younger infants are rarely testable³), although three younger infants were tested in the full-term group, then every 2 weeks until 20 weeks of age, and finally at 26 weeks of age. Since most aspects of binocular vision develop between 6 and 16 weeks,^{3,4,8,12,26,27} we were not expecting that attempting to collect earlier data would help answer our research question. 

Laboratory Testing

A brief history was taken to confirm normal development, and an orthoptic assessment excluded strabismus. 

All infants were tested with a remote haploscopic photorefractor described previously^3,28 (see Supplementary Material). It incorporates a Plusoptix SO4 photorefractor in PowerRefII mode, which continuously and simultaneously records refraction and eye position at 25 Hz; this allows us to calculate accommodation in diopters (D) and vergence in meter angles (MA). The photorefractor is set in a target presentation apparatus consisting of two concave mirrors and a moving monitor. The target appears to move backward and forward in front of the observer between distances of 0.25 and 2 m (presented in a pseudorandom order of 0.33 m [3 D and 3 MA demand], 2 m [0.5 D and MA], 0.25 m [4 D and MA], 1 m [1 D and MA], 0.5 m [2 D and MA]). Meter angles are a preferable measure of vergence as they are a constant measure of response in relation to demand in populations in which interpupillary distance (IPD) varies between participants, and over the course of development. Thus, for example, our 0.5-m target presented to an infant with an IPD of 45 mm would demand 2 MA, 13.5 prism diopters or 7.68° of convergence, while for an adult with an IPD of 60 mm the same target would still demand 2 MA, but 18 prism diopters or 10.2° of convergence. Meter angles also provide an easy comparison between the appropriateness of vergence and accommodation for target demand at each distance. Data from the 0.25-m target were not analyzed for three reasons. Most commonly and importantly, we find an unacceptable loss of data resulting from small pupils at this distance. There is also a small astigmatic error due to the mirror offsets (of subjectively approximately 0.5 D at 25 cm) but this reduces below 0.25 D and is therefore not problematic at the other distances. Thirdly, the fusional stimulus is slightly different at 25 cm because the far edges of the target screen fall slightly beyond the binocular fusional overlap of the lower mirror that is seen in physiological diplopia. We retain the target in the testing order so that a farther target always precedes a nearer one and vice versa. 

Vergence and accommodation responses were measured while the infant watched a binocular cartoon clown target containing a range of spatial frequencies as it moved backward and forward. Some target details were separated by only one pixel (visual angle of approximately 1 min arc at 0.33 m), but the target also contained large elements, high-contrast edges, bright colors, alternating elements, eyes, and a hairline to be maximally interesting to neonates with poorer visual acuity. The target subtended 3.15° at 2 m and 18.3° at 0.33 m. If possible, each child was tested twice in each session and the data were averaged. The Plusoptix monitor allowed the tester to watch the infant in real time to assess attention and fixation and also to follow recording traces even when the accommodation responses exceeded the operating range of the photorefractor. We report data collected only when the infant was observed to have fixated the target steadily for at least 2 seconds at each fixation distance. The Plusoptix SO4 has a linear operating range of −7.0/+5.0 D (i.e., up to 7 D of accommodation and 5 D of hyperopia). Beyond this, our unpublished calibrations and those of others²⁹ demonstrate that although the photorefractor continues to calculate a figure for refraction, this is an underestimation of the true value. This varies between individuals, so without individual calibration it is not precisely quantifiable. Data from infants who demonstrated hyperopic refractive error over +5.0 D (estimated using maximum hyperopic refraction found during testing [MHR]) were excluded before quantitative analysis. We have reported that MHR correlates closely with cycloplegic refraction in other child and infant groups.³⁰ 

Raw data were processed offline.^3,28 Vergence in MA was calculated from the horizontal eye position of each eye, correcting for individually calculated angle λ and interpupillary distance. Individual refraction calibrations and repeatability calculations were not possible for such young infants, but for group comparison studies such as this, averaged data are acceptable.²⁹ We calculated accommodation in diopters, using the increasingly myopic photorefraction that occurs on accommodation, with a correction for a slight systematic error (the photorefractor underestimates accommodative response by approximately 0.5 D) using a formula derived from group calibration studies²⁸ in young adults. Calculations of response gain in relation to target demand (the slope of the stimulus response functions) used at least three data points (four if possible) at the different fixation distances. Where we report responses to particular targets, we have limited them to the nearest (0.33 m, 3 MA and D) and the farthest (2 m, 0.5 MA and D). 

Statistical Analysis and Data Presentation

We present our results in two ways. Firstly we provide descriptive figures to indicate the spread of responses. Since accommodation responses beyond the linear operating range of the photorefractor are likely to underestimate the degree of refraction to an unknown extent, this full dataset was not analyzed statistically. If we had excluded these data completely, however, we felt we would have misrepresented the spread of infant behavior. 

We then calculated group means and 95% confidence intervals (CI) of all data within range. These data were analyzed using two-way between-groups ANOVA (with age group and preterm/full term as factors) to investigate between-group differences in vergence and accommodation responses and gains at intervals of 2 weeks. A main effect of age indicates that vergence and/or accommodation changes with age, and a main effect of group indicates overall differences between preterm and full-term infants. Most importantly, any age × group interaction would suggest that the two groups differ only at certain ages. If more between-group differences in responses are found when groups are compared by their corrected age, this would indicate that development of vergence and/or accommodation is experience dependent. More group differences when groups are compared by their chronological age would suggest that development is more maturational. 

Post hoc testing used Bonferroni correction for multiple comparisons where appropriate. 

Results

Testability and Repeatability

Numbers testable at each age point for both the corrected age and chronological age are illustrated in Table 1. While most infants provided usable data on most visits, only four preterm and 13 full-term infants provided such data at every visit, so data were treated as cross-sectional. Of the maximum potential number of testing sessions over the study period, 55% of the preterm infants and 18% of the full-term infants either were unable to attend or were not able to be tested at all due to being asleep or fretful on a booked session. Premature infants, particularly the large number of twins, were especially difficult to test regularly. These factors added to the normal difficulties of testing infants. But if an infant attended and was attentive, complete runs of targets at the different fixation distances were always recorded. Repeated measurements within a single visit were more often possible for older infants, whether full term or preterm (e.g., 23% repeatable at 6–7 weeks and 58% at 12–13 weeks of corrected age for the preterm infants). Repeated measurements were averaged where available. Variability in repeated measurements within individuals was similar to that between different infants at each corrected-age point (95% confidence intervals were not significantly different), but younger infants were much more variable overall (95%CI for vergence gain at 6–7 weeks: between individuals = ±0.12; within an individual = ±0.09; at 12–13 weeks: between individuals = ±0.045; within an individual = ±0.04). 

Table 1

 

Numbers Testable at Each Age Point

 

Exclusions and Refraction

Myopia did not exceed −0.5 D for any infant tested. Some of the youngest infants behaved myopically (overaccommodated) for distance fixation. However, their accommodation relaxed at least once during testing to an emmetropic or hyperopic refraction, confirming that they were not genuinely myopic. 

One preterm infant appeared consistently significantly more than 5.0 D hyperopic on multiple visits, and this infant's data were excluded completely from further analysis. Two (6.2%) premature infants and four (8.8%) full-term infants showed >5.0 D hyperopia (beyond the linear operating range of the photorefractor) fleetingly (i.e., for a single data point) at some time, all in the first 12 weeks of life, and the data from that single session were excluded (Table 1). No refraction from these infants ever exceeded a photorefractor calculation of +7.0 D hyperopia. No infant whose session data were excluded showed evidence of manifest refraction > +3.00 D by 16 weeks of age, so all had emmetropized to within normal limits. 

The proportion of infants with hyperopia greater than +2.0 D in each group was similar across time when compared by their corrected age, for example, 39% vs. 33%, respectively, at 10 to 11 weeks and 29% vs. 25% at 14 to 15 weeks. At 24 to 27 weeks of corrected age, the infants' mean refraction estimated by the MHR measured during the testing session was +0.18 D (95%CI −0.25 D / +0.66 D) in the full-term infants and +0.28 D (95%CI −0.43 / +0.99 D) in the preterm infants (t(55) = 1.36, P = 0.178, nonsignificant). 

Full Dataset

Figure 2 illustrates the ranges of vergence and accommodation responses at two time points, 6 and 7 weeks of corrected age (which was on average 12–13 weeks of chronological age for the preterm group) and again at 12 and 13 weeks of corrected age (18–19 weeks of chronological age for the preterm infants). We chose these two time points as 6 and 7 weeks is before mature binocular responses develop in full-term infants, while 12 to 13 weeks is when vergence and accommodation are not significantly different from what is seen in adults,³ and sensory binocularity is typically emerging.⁴ 

Figure 2

 

Recorded responses (y-axis) in relation to demand (x-axis), including out-of-linear-range accommodation estimates (gray shaded areas). Black line: Mean response. Left: Full-term infants at 6 to 7 weeks of age (top) and preterm infants of 12 to 13 weeks chronological age (bottom) but equivalent corrected age. Right: Full-term infants at 12 to 13 weeks of age (top) and preterm infants 18 to 19 weeks of age (bottom).

Figure 2 illustrates the whole dataset including out-of-range accommodation estimates (gray shaded areas). Forty-two individual data points (2.3% of the total tested) exceeded the linear operating range of the phororefractor (>7 D accommodation). Twenty-four infants (evenly distributed between preterm and full term) provided these data points fleetingly for the nearest targets in their first 12 weeks (corrected age if preterm), and for all except one infant in each group these were between approximately 7.0 and 10.0 D. The other two infants contributed six data points between approximately 10.0 and 12.0 D. 

There are two important comparisons in Figure 2. The first is a corrected-age match comparison (full-term [top charts] versus preterm infants [bottom charts]), where performances are similar. Many of the youngest full-term and corrected-age preterm infants (left charts in Fig. 2) showed highly erratic accommodation. What we have previously termed “all or nothing” patterns³ were common, where accommodation response to an approaching target was flat for the more distant targets, but then was either appropriate or excessive (and sometimes out of range) for the nearest target, despite concurrent linear vergence. Eleven (5.5%) of the 198 individual data points collected at 0.33 m in the preterm infants and 19 (6.5%) of the 291 points collected in the full-term infants were greater than 7.0 D. Before 12 weeks of age, overaccommodation for the nearest target exceeded 4.5 D at 0.33 m in 28.5% of full-term infants and in 38.5% of the corrected-age preterm infants. 

The second comparison is between full-term infants with preterm infants matched by chronological age. It was not possible to compare full-term with preterm infants at 6 to 7 weeks since insufficient data were collected from the preterm infants, but the comparison at 12 to 13 weeks is illustrated in the top right and bottom left of the figure. This shows that full-term infants' vergence and accommodation are more linear than in chronologically age-matched preterm infants. 

Analysis of Data in Range

For statistical analysis we compared infants matched by both their corrected age and chronological age, considering response gain as well as responses for near (0.33 m) and distance (2 m). Vergence measurements were all within the linear range of the photorefractor across the range tested, so all infants' vergence gains were calculated using responses at four distances. For accommodation, out-of-range points were excluded and gains were calculated from the responses to the three remaining distances. Gains thus calculated are likely to be a slight underestimate of the true gain. Such exclusions occurred most frequently at 8 to 9 weeks corrected age. Here the median accommodation response for the 0.33-m target of the full data set (using out-of-range points that we know are inaccurate) was 0.34 D more than the mean of the more selected data. If the median from the full dataset had been used to calculate the gain, it would have increased the gain by 0.12. At other ages, differences were less. Four accommodation data points were available for 93% of the target runs for the full-term infants and 90% of those from the preterm infants. 

Results of the ANOVAs comparing response gains and responses at 2 and 0.33 m between groups are shown in Table 2, and post hoc significant differences are indicated in Figures 3 (vergence) and 4 (accommodation). 

Table 2

 

Results of ANOVA of Vergence and Accommodation Gains and Responses at 2 and 0.33 m

 

Figure 3

 

Vergence gain (top), vergence responses to target at 2 m (center), and vergence responses to target at 0.33 m (lower). Left column: Responses matched by corrected age. Right column: Responses matched by chronological age. Statistically significant differences on post hoc testing indicated by asterisks. Error bars indicate 95% confidence intervals. *P < 0.05; **P < 0.01.

Again, we compared groups matched by both corrected and chronological age. When matched by their corrected age, there were the expected significant developmental improvements in all infants. Preterm infants relaxed their accommodation significantly less at 2 m than the full-term infants, but there were no other overall group differences. There were significant age × group interactions in four of the six comparisons, but post hoc testing showed that differences were significant only at 6 to 7 weeks of age (Figs. 3, 4), where the preterm infants underconverged for near and overaccommodated for distance targets. Subsequently, up to 24 to 27 weeks, there were no differences in accommodation and vergence responses between full-term and preterm infants matched by their corrected age. 

Figure 4

 

Accommodation gain (top) calculated from at least three fixation distances, and actual responses at 2 m (center) and 0.33 m (lower). Left column: Responses matched by corrected age. Right column: Responses matched by chronological age. Statistically significant differences on post hoc testing indicated by asterisks. Error bars indicate 95% confidence intervals. *P < 0.05; **P < 0.01.

When infants were matched by chronological age, there were significant preterm/full-term group differences for all comparisons except accommodation at 2 m. Full-term infants showed more appropriate responses than the chronologically age-matched preterm infants (gain closer to 1, responses closer to the target demand). There was also a significant age × group interaction for all comparisons except accommodation at 0.33 m. Post hoc testing showed that the majority of significant differences were found between infants aged between 10 and 16 weeks and were particularly clear at 10 to 11 weeks of age. While the full-term infants' responses appeared to have matured (were similar to responses at the oldest age tested), those of the preterm infants were still immature. 

To test the linearity of vergence and accommodative responses for each group, we calculated correlation coefficients (r²) for individual stimulus response slopes where four data points (at 0.33, 0.5, 1, and 2 m) were available. Infants matched by their corrected age demonstrated similar linearity of response; for example, for vergence at 12 to 13 weeks, mean r² were 0.94 and 0.91, respectively, for full-term and the corrected-age preterm infants. However, when matched by chronological age, 12- to 13-week preterm infants demonstrated less linear vergence (r² = 0.77 for preterm infants and 0.94 for full-term infants) (t = 2.57, P = 0.019), not significantly different from full-term infants at 6 to 7 weeks. Similar analysis for accommodation showed that mean r² for the full-term and the corrected-age preterm infants did not differ significantly (0.74 and 0.77 respectively), but preterm infants of the same chronological age had a lower mean r² of only 0.53 (t(39) = 2.4, P = 0.02), again not significantly different from full-term infants at 6 to 7 weeks. 

Discussion

This study investigated the developmental time course for vergence and accommodation responses in full-term and preterm infants matched by both chronological and corrected age. Our results suggest that vergence and accommodation in preterm infants follow a maturational developmental trajectory and that responses are not accelerated by the additional visual experience of earlier birth. Full-term infants show more adult-like vergence and accommodation responses when compared to chronologically age-matched preterm infants. 

These results contrast with those of Jandó et al.,⁶ who showed an experience-dependent development of sensory binocularity, where the additional visual experience in preterm infants resulted in earlier development. Fifty percent of the preterm infants in the study by Jandó et al.⁶ responded to DRDCs by 1.92 months postnatally (approximately 8 weeks). If sensory binocularity develops earlier in preterm infants but accommodation and vergence responses do not, then early development of sensory binocularity is unlikely to be the cause of maturation of vergence and accommodation. Instead, it is possible that the oculomotor system supports or reinforces the development of sensory binocularity. 

Vergence

Vergence accuracy and a gain close to 1 characterize adult-like responses. More recent research has demonstrated that, in full-term infants, vergence is adult-like by 8 to 9 weeks,^1,3 earlier than suggested by older literature in which such young infants were not assessed³¹ or good vergence responses were less commonly found.⁴ The early large neonatal misalignments found in infants younger than 2 months of age are also reducing dramatically.^2,14,31 Thus good alignment for targets at all fixation distances is typically in place before the onset of stereopsis and sensory binocularity (Wong A, et al. IOVS 2008;49:ARVO E-Abstract 3748).^8,26,32,33 In contrast, our preterm infants still showed immature vergence until approximately 15 weeks of age. 

If sensory and oculomotor visual systems had been found to mature in parallel, then the effects of prematurity on visual development would be insignificant, as the onset of critical periods for vergence control and sensory binocularity would be similarly delayed. However, if any aspect of sensory binocularity (with concurrent susceptibility to suppression and amblyopia) can be advanced by experience while oculomotor control is not, a mismatch of developmental trajectories might result in decorrelated input from each eye to the visual cortex, at a time when cortical binocularity is entering a critical period that has been advanced through early visual experience. 

Additional infant studies have demonstrated that development of stereopsis does not depend on the development of vergence.^4,34 Thorn at al.⁴ suggest that good alignment is not necessary for development of the neural mechanisms underlying binocular vision, but is necessary for maintenance of these mechanisms. Tychsen argues²³ that “binocular decorrelation is a sufficient cause of infantile esotropia when imposed during a critical period of visuomotor development” (p. 564). Immature biases to esodeviation such as asymmetrical monocular OKN²⁷ and better convergence than divergence³⁵ may be retained in premature infants, resulting in an increased risk of infantile esotropia. Our findings therefore suggest a mechanism that might account for increased prevalence of strabismus in preterm infants. 

Accommodation

Immature accommodation is more erratic and less linear than vergence at the same age. In preterm infants, this variability is extended for longer after birth. Lower gain was often the result of overaccommodation in the distance; but excessive accommodation for near was also common, often after almost flat responses to the three farther targets, as has been found in previous studies.^3,36 Accommodation development in preterm infants also related to their corrected age rather than their chronological age, with the same gradual increase in accommodation gains over the first weeks that Banks¹⁰ found for two younger full-term infants using dynamic retinoscopy. Banks' research also suggested a similar preprogrammed course of development. We did not detect, however, the same clear developmental trajectory of accommodation development in full-term infants as reported by Banks¹⁰ because most of our full-term infants were already showing response gains of well over 1.0 (which related to their refraction) by 6 to 7 weeks. 

Our results suggest that not only are vergence inaccuracies occurring when cortical binocularity could be emerging, but also that the linkages between vergence and accommodation will be less consistent during this extended period of mismatched retinal input and imprecise accommodation. Although we have reported that mean full-term infant accommodative convergence to accommodation (AC/A) ratios are not significantly different from those of adults,⁵ the variability of response in preterm infants would result in a weaker linkage between vergence and accommodation responses for a greater developmental period. Thus, increased risk of strabismus in preterm infants might also be driven by lack of reinforcement of AC/A and convergence accommodation to convergence (CA/C) ratio linkages. 

Finally, good accommodation is also implicated in emmetropization.^37,38 Previous studies have shown that binocular input dramatically enhances not only vergence but also accommodation in full-term infants,^1,3 older children, and adults.²⁸ As well as inaccurate vergence (and so interocular decorrelation) being a “sufficient” cause of esotropia, any damage to cortical binocularity might then also damage accommodation, and thus be implicated in the defective emmetropzation that is more common in those born both pre-term³⁹ and with strabismus.⁴⁰ Thus, prematurity may not only cause infantile esotropia, but might also be implicated in strabismus with an accommodative element. 

Study Limitations

While comparisons of these data with those of Jandó et al.⁶ support the arguments above, differences in testing paradigm between the two studies might explain apparent differences between developmental time courses of the groups for other reasons. Jandó et al.⁶ measured cortical activity, which required no behavioral response. Visual evoked potential is easier to test successfully in very young infants, and VEP testing is a less demanding task than our paradigm. Our task involves a longer processing time, requires a motor response to a sensory signal, and is more likely to be susceptible to attentional variation. It is therefore possible that the attentional system in premature infants needs to have reached a sufficient level of maturity for them to perform the tests used here. In this case, the difference in timing between full-term and preterm infants might be the result of differences in maturation of higher-order behavioral mechanisms rather than maturation of vergence and accommodation per se. 

All infants, especially preterm twins, present a significant challenge in testing, so a complete set of longitudinal data was rare, and many testing sessions were abandoned or cancelled for reasons unrelated to the study. However, this is likely to affect only the quantity, not the quality, of the results. Despite small numbers in the youngest infant groups, statistical significance was still reached. 

We could not definitively differentiate attentional and physical immaturity, but either means that preterm infants will have inaccurate vergence and accommodation for longer after birth. Immature responses could be due to immaturity of the control mechanisms, so despite sensory detection of the change of target distance, rapid, coordinated physical responses cannot yet occur. Alternatively, acuity, attention, or interest with regard to detailed targets may be insufficiently developed to drive appropriate responses. Accommodation is certainly active in very early infancy, as evidenced by the difference between cycloplegic (generally hyperopic) and noncycloplegic (generally myopic) refraction of neonates (for review see Ref. 41); and convergence is also clearly possible during frequent large neonatal misalignments²¹ but seems poorly controlled. We also accept that the reduction in variability of responses from the older infants could partly be due to averaging of more infants' data, but even the averaged data became less variable with time. 

A major limitation of the Plusoptix photorefractor is its relatively small operating range. Although out-of-range accommodation responses were still collected, we could not measure them accurately because calculations from the Plusoptix become nonlinear, so a reading of 8 D might be the given from an accommodative response between 7 and 9 D, and this error may vary between individuals. By excluding these points our statistical testing used a slightly smaller dataset (and probably underestimated mean overaccommodation), but the type and proportions of excluded data were similar in each group. We continue to use the Plusoptix photorefractor because it is one of the few instruments able to refract and assess eye position binocularly, naturalistically, simultaneously, and continuously. 

We considered excluding the very nonlinear responses, where a pattern of flat or low gain responses was found to targets at 0.5 m or beyond, with a sudden large overaccommodation response to the 0.33-m target. These responses are different from largely linear adult responses and were sometimes out of the linear range of the photorefractor. By excluding them, however, we would inaccurately describe neonatal responses, of which they are a feature. We accept that when excessive near response is out of linear range it is difficult to quantify using our equipment, but it is of interest for two reasons. Flat accommodation responses for more distant targets, followed by appropriate or excessive accommodation for near, suggest that while vergence seems generally well controlled over the linear range of target distances, accommodation can be driven independently once a level of blur (or disparity) reaches a threshold. These responses also have implications for the development of the AC/A ratio because they suggest that the relationship between accommodation and vergence is different at different target demands, suggesting that in infancy A/C linkages are unstable. 

We could also not perform the individual calibrations for accommodation that would have been ideal for such studies,²⁹ although group comparisons are often used in studies such as this. The Plusoptix photorefractor accuracy compares well with refraction derived from retinoscopy (around ±0.75 D),^28,42 while our measure of vergence change is more precise because we correct for variables such as IPD and angle λ.²⁸ There may therefore have been some individual between-participant differences in accuracy of refraction within the operating range of the photorefractor, but there should be no optical reasons why calculation of refraction of younger or premature infants per se should be less accurate (once data are captured). The fact that more linear vergence was demonstrated simultaneously with erratic accommodation shows that the infants were attending to the target and refraction was on-axis, but frequently well outside ranges that could be attributed to measurement error. 

We had too few significantly hyperopic infants to investigate early hyperopia as a separate issue. We had similar proportions of apparently hyperopic infants in each of our groups when matched by their corrected age, so this is unlikely to have affected our results. 

In conclusion, vergence and accommodation follow a preprogrammed developmental trajectory so that preterm infants appear to have longer visual experience of immature responses. This may extend into the period when experience-dependent cortical binocularity emerges. A mismatch in the time course between the development of oculomotor and sensory binocularity might contribute to the increased risk of strabismus in children born preterm. 

Acknowledgments

We thank Greg Boden, MBBCh, consultant pediatrician, and research nurses Sue Hallett and Morag Zelisko at the Royal Berkshire Hospital, Reading, United Kingdom, for help and advice in planning and carrying out recruitment of the preterm infants. 

Supported by UK Medical Research Council Clinician Scientist Fellowship G0802809 (AH, ST). The authors alone are responsible for the content and writing of the paper. 

Disclosure: A.M. Horwood, None; S.S. Toor, None; P.M. Riddell, None