Eye surgery for thyroid disease shifts vision for months, study finds

Orbital decompression surgery, a standard operation for patients with moderate-to-severe thyroid eye disease, does more than reduce proptosis. A retrospective longitudinal study of 57 eyes from 46 patients, published in Eye and Vision, reports that the operation triggers a dynamic, biphasic pattern of refractive change: an initial myopic shift that peaks at roughly two to three months after surgery, followed by partial hyperopic recovery as the eye gradually approaches relative stabilization. The researchers, led by investigators at the 3rd Medical Center of Chinese People’s Liberation Army General Hospital, used generalized estimating equation models with restricted cubic splines to track refractive, axial, and ocular positional measurements over continuous postoperative time rather than at fixed visits.

The clinical stakes are not trivial. Even modest refractive shifts can matter for patients with high visual demands or those scheduled for cataract or corneal refractive surgery, and the authors argue that definitive refractive correction, including spectacle prescription and intraocular lens power calculation, may be better deferred until at least about three months postoperatively, particularly after more extensive decompression. Previous literature had been inconsistent: some early studies reported postoperative myopic shifts of roughly 1.00 to 2.50 diopters, and a larger retrospective cohort found significant early myopic changes in spherical diopter and spherical equivalent within two months of surgery, while other studies reported minimal or negligible refractive change. The new time-resolved analysis suggests a possible explanation, since measurements taken at different postoperative intervals would capture different phases of the same evolving trajectory.

The cohort consisted of consecutive patients with moderate-to-severe thyroid eye disease, defined according to the 2021 European Group on Graves’ Orbitopathy guidelines and Bartley’s diagnostic criteria, who underwent decompression between May 2024 and February 2025. Baseline disease activity was generally inactive, with a median Clinical Activity Score of 1, and mean exophthalmos measured 23.0 millimeters. Most eyes underwent balanced medial-lateral wall decompression, with the remainder receiving medial wall or lateral wall decompression, all combined with orbital fat removal. Preoperatively, the eyes showed mild myopia, with a mean spherical equivalent of -1.57 diopters and a median axial length of 23.75 millimeters.

Descriptively, the refractive changes were most pronounced early. In an earlier postoperative window before two months, covering 44 eyes, median changes in spherical diopter and spherical equivalent were -0.50 and -0.56 diopters respectively, and measured axial length increased by a median of 0.09 millimeters. Clinically significant myopic drift, defined as a spherical equivalent change of -0.50 diopters or more relative to baseline, occurred in 63.6 percent of eyes in that window. In the later window at two months or beyond, covering 42 eyes, the median changes attenuated to -0.25 diopters for both spherical measures and 0.02 millimeters for axial length, with clinically significant drift falling to 26.2 percent of eyes, a pattern the authors interpret as partial recovery and a trend toward stabilization.

The longitudinal modeling confirmed these patterns statistically. Postoperative spherical diopter and spherical equivalent both showed significant nonlinear temporal trajectories, with progressive myopic shift early after surgery, a maximum at approximately two to three months, and partial hyperopic recovery thereafter. Cylindrical refraction, by contrast, changed only mildly and transiently, without reaching statistical significance, suggesting that astigmatic changes are comparatively insensitive to the global biomechanical perturbation of decompression. Axial length and exophthalmos followed parallel nonlinear courses: measured axial length peaked at roughly two to three months before partially reversing, while exophthalmos showed maximal posterior globe displacement early after surgery followed by partial anterior rebound.

Surgical strategy mattered substantially. In the overall analytical framework reflecting routine clinical practice, two-wall decompression was associated with a greater myopic shift than one-wall decompression, with an adjusted difference of -0.279 diopters, and follow-up time was independently associated with progressive myopic change of -0.051 diopters per month. Interaction models showed that two-wall decompression carried a significantly steeper rate of myopic progression, with extent-specific slopes of -0.229 diopters per month for two-wall versus a negligible -0.051 diopters per month for one-wall procedures. Two-wall decompression also carried a markedly higher risk of clinically significant myopic drift, with an odds ratio of 18.236 in unadjusted models and 11.070 after adjustment for the achieved reduction in exophthalmos.

A second analytical framework, adjusting for the achieved exophthalmos reduction as an indicator of decompression extent, clarified the sources of these effects. The main difference in myopic shift between strategies persisted almost unchanged, but the independent time-dependent progression largely disappeared, suggesting that the apparent temporal progression was predominantly related to decompression extent. Yet significant interaction effects and differential temporal slopes persisted even after adjustment, indicating strategy-specific time-varying dynamics independent of extent. Exploratory procedure-level comparisons found that balanced medial-lateral wall decompression consistently showed the largest and fastest myopic drift, though the authors caution that uneven subgroup sizes make these findings hypothesis-generating only.

Mechanistically, the strongest statistical correlate of refractive change was the postoperative change in measured axial length, which showed the largest standardized effect among all evaluated parameters and substantially exceeded every other variable. The authors emphasize that optical biometry captures an effective optical path length rather than true anatomical length, so the axial change should not be read as genuine elongation of the eye. Instead, it likely integrates posterior globe displacement, whole-globe translation within the orbit, and rotational realignment of the visual axis. Supporting this interpretation, greater reductions in exophthalmos were monotonically associated with larger increases in optically measured axial length, linking globe repositioning to the optical manifestation of that movement. Secondary contributions came from corneal biomechanical measures such as deformation amplitude and highest concavity deflection, from anterior corneal curvature changes, and from eyelid position parameters including palpebral fissure height and margin reflex distance 1.

Baseline ocular characteristics also shaped individual susceptibility. Larger baseline exophthalmos, greater margin reflex distance 1, and larger palpebral fissure height were each associated with reduced risk of clinically significant myopic drift, as was a more gradual corneal thickness profile reflected in higher Ambrósio’s relational thickness. Flatter posterior corneal surfaces, a less prolate posterior shape, and larger corneal volume within the 10-millimeter zone also conferred lower risk. Conversely, longer maximum whole eye movement time, higher pachymetric progression indices, steeper anterior curvature, greater anterior and posterior corneal astigmatism, greater thickness-distribution deviations, and deeper internal anterior chamber were each associated with elevated risk. The authors frame these baseline features as an intrinsic biomechanical phenotype, or set of biomechanical set points, that modulates how an individual eye responds to the mechanical perturbation of surgery.

The study has limitations the authors acknowledge directly. It was retrospective, with follow-up concentrated in the early postoperative period and a median observation time of only 1.92 months over a range of 0.90 to 6.67 months, so long-term refractive dynamics remain unresolved. Refraction was measured without cycloplegia, eyelid mechanical pressure was not directly measured, and procedure-level subgroup analyses were limited by modest and uneven sample sizes. Even so, the integration of continuous-time longitudinal modeling with multimodal corneal, orbital, and eyelid measurements offers a coherent systems-level account of why vision shifts after decompression, and it supports individualized surgical planning, perioperative risk stratification, and careful postoperative refractive monitoring, with correction deferred until the eye has had time to settle.

Subject of Research: Longitudinal refractive changes after orbital decompression surgery in thyroid eye disease

Article Title: Longitudinal refractive changes following orbital decompression in thyroid eye disease: dominant role of axial configuration and globe position with contributions from surgical technique and corneal biomechanics

Article References: Yu, M., Tang, J., Li, Y., Li, Y., Jing, G., Ao, Y., Ma, R., Liu, F., Ge, M., Yang, Q., Zhang, M., Peng, M., Nie, B., Yang, X., Wu, W., & Wang, L. (2026). Longitudinal refractive changes following orbital decompression in thyroid eye disease: dominant role of axial configuration and globe position with contributions from surgical technique and corneal biomechanics. Eye and Vision, 13(1), Article 32. https://doi.org/10.1186/s40662-026-00499-9

Image Credits: AI Generated

DOI: 10.1186/s40662-026-00499-9

Keywords: thyroid eye disease, orbital decompression, refractive change, myopic drift, axial length, corneal biomechanics, exophthalmos, surgical technique, GEE models, ophthalmology, Longitudinal, refractive