The translational asymmetry of corpus callosotomy: Two literatures that answered different questions

Abstract

Corpus callosotomy is an established palliative operation for drug‐resistant epilepsy dominated by drop attacks, and it is performed in three extents: anterior, usually of the anterior two‐thirds; total, in one or two stages; and selective posterior, which preserves the genu and rostral body. Meta‐analysis finds no difference between them in overall seizure reduction, a difference in drop attack control favoring the posterior approach, and a graded difference in disconnection syndrome, but there has never been a prospective head‐to‐head comparison, and the optimal extent of section remains undefined. A search of PubMed, Embase, and the Web of Science Core Collection identified 62 experimental studies in which the corpus callosum or another forebrain commissure was sectioned with an epilepsy‐related outcome, spanning 1948–2026 across six species and every major seizure model. This literature is neither absent nor recent. What it has almost never done is vary the extent of section: the few studies that varied it took the laterality of kindling transfer or of interictal synchrony as the endpoint, never seizure control and never cognitive cost. No study compares the three extents against a common outcome; none models the selective posterior operation as it is defined clinically; none randomizes between extents; and none combines graded disconnection with time‐locked bilateral electrophysiology, diffusion‐based verification of the fibers divided, and a behavioral read‐out in the same animals. The effect of section is also bidirectional: focal seizures may intensify through loss of transcallosal inhibition, and in the pentylenetetrazol model callosal section paradoxically facilitates convulsions, which limits chemoconvulsant preparations. The asymmetry between the clinical and experimental literatures is therefore one of method rather than of volume or attention: the two answered different questions, and the question the clinical field now asks requires a manipulation the experimental literature has never performed.

Plain Language Summary

Surgeons can treat severe epilepsies by cutting the bundle of fibers that joins the two halves of the brain, but they do not agree on how much of it to cut. We searched three databases and found 62 animal studies of this operation going back to 1948. Almost none of them compared different amounts of cutting, and none looked at the newest version of the operation. Answering the surgeons' question will require a new kind of animal experiment rather than more reading of the old ones.

Key points.

• Callosotomy is performed in three extents — anterior, posterior, and total — but no prospective study has compared them head‐to‐head.

• Our search found 62 experimental studies of callosal or forebrain commissural section with epilepsy‐related outcomes, spanning 1948 to 2026.

• None varied the extent of section against a seizure outcome, and none modeled the selective posterior operation as defined clinically.

• Section of the callosum is bidirectional: it can intensify focal seizures through loss of transcallosal inhibition.

• The gap is methodological, not bibliographic: what is missing is randomized, anatomically verified comparison of extents.

1. AN ESTABLISHED PALLIATIVE OPERATION, THREE WAYS OF PERFORMING IT

Corpus callosotomy is a long‐standing palliative procedure for drug‐resistant epilepsy in which seizures propagate or synchronize rapidly between the hemispheres, and it is particularly effective against the atonic and tonic seizures that produce sudden falls. The rationale is anatomical: the corpus callosum is the principal commissural pathway linking the two hemispheres, and interrupting it attenuates the bilateral engagement that underlies drop seizures. Approximately, one‐third of people with epilepsy develop drug resistance, and a substantial subset — often with Lennox–Gastaut syndrome, West syndrome or other diffuse epilepsies — is not a candidate for curative resection because no single resectable focus can be defined. ¹ For these patients, disconnection and neuromodulation are the principal options, ² , ³ and drop attacks, which cause recurrent injury and severely impair quality of life, are the usual target. ⁴

What is not settled is how much of the callosum to divide. Three procedures are in current use, and they were introduced in this order. Anterior callosotomy, usually of the anterior two‐thirds, has the longest history and the largest accumulated experience; series from the 1980s onward established both its efficacy against drop attacks and its limitations. ⁵ , ⁶ , ⁷ , ⁸ Total callosotomy, performed in one or two stages, offers greater seizure control at the cost of a higher risk of disconnection effects; the largest recent series report drop attack cessation in the majority of patients, with functional recovery within days to weeks. ⁹ , ¹⁰ Selective posterior callosotomy, which preserves the genu and rostral body, was introduced most recently on the hypothesis that dividing the posterior callosum — through which motor and premotor fibers cross — would control drop attacks while sparing prefrontal interconnectivity. ¹¹ , ¹² A fourth variant, extended anterior section sparing only the splenium, has also been described. ¹³

The comparative evidence is more balanced and more uncertain than any single series suggests. A meta‐analysis of 66 studies and 1349 patients found no significant difference between anterior, posterior, and total section in the proportion achieving complete or greater‐than‐90% seizure reduction, but a significant difference in drop attack control favoring the posterior approach (92% vs. 58% for total and 52% for anterior), and a significant difference in disconnection syndrome, which was absent after posterior section but occurred in 9% after anterior section and in 20% after total section. ¹⁴ Heterogeneity was high, and the comparisons are between non‐randomized subgroups. Set against this, in adolescents and adults with severe intellectual disability, one‐stage total callosotomy produced drop attack cessation in 78.6% of patients compared with 10% after anterior section, with no significant difference in the time to recovery of daily living activities. ¹⁵ There has never been a prospective head‐to‐head comparison of total with posterior‐half callosotomy. Table 1 sets out the contemporary clinical evidence on which these comparisons rest.

TABLE 1.

Contemporary clinical evidence on the extent of callosal section.

Study	Design	Extent compared	Seizure outcome	Cognitive/functional outcome	Comment
Blihar 2025 14	Meta‐analysis, 66 studies, 1349 patients	ACC vs. PCC vs. TCC	No difference in seizure score 0 or 1 (p = 0.924); drop attacks PCC 92%, TCC 58%, ACC 52% (p = 0.03)	Disconnection syndrome: PCC 0%, ACC 9%, TCC 20% (p = 0.0045)	High heterogeneity; nonrandomized subgroups
Kagawa 2021 15	Retrospective, 24 patients, IQ <35	ACC (n = 10) vs. one‐stage TCC (n = 14)	Drop attack cessation: TCC 11/14 (78.6%) vs. ACC 1/10 (10%) (p = 0.0009)	No significant difference in ADL recovery period	Favors TCC in severely impaired patients
Kagawa 2025 16	Retrospective, 40 patients, IQ ≤50	One‐ or two‐stage TCC	Favorable outcome 72.5% at 1 year, 57.5% at 2 years; multiple independent spike foci predict failure (OR 6.48)	ΔIQ higher in the favorable outcome group (0 vs. −6; p = 0.0042)	Decline attributed in part to surgery itself
Baba 2018 9	Retrospective, 56 patients, West syndrome, no MRI lesion	Callosotomy, extent not stratified	Seizure‐free 32.1%; spasms eliminated 42.9%; tonic seizures poor	Progressive decline in developmental quotient prevented in responders	Transcallosal bilateralization critical for spasms
Ukishiro 2022 10	Retrospective, 41 patients	One‐stage complete callosotomy	Seizure‐free 5/41; >80% reduction 11; poor 20	Speech at 2.2 days, oral intake 1.6 days, ambulation 5.8 days	Tolerability, not superiority
Purves 1988 5	24 medically refractory patients	Anterior two‐thirds section	Significant reduction in 75%; best results with a frontal focus	No permanent neurological deficit	The anterior literature the review must engage
Spencer 1988 I and II [7,19]	Series, 22 patients	Partial vs. total section	Generalized seizures abolished in 35% after partial vs. 77% after total	Neurological and neuropsychological outcome reported separately	Model for dual‐outcome reporting
Yang 2014 13	29 children with drop attacks	Extended anterior section sparing the splenium	Cessation or >90% reduction in 79.3%	Transient disconnection syndrome in most, resolving within 3 weeks; attention improved	The mirror image of selective posterior section
Novak Filho 2026 17	6 patients, prospective imaging	DTI‐ and neuronavigation‐guided PCC	Complete fiber section in 4/4 → 100% drop attack reduction; partial (82.1%, 89.3%) → 75.0%, 85.7%	Cognition improved in two, unchanged in four; no motor deficit	Extent–outcome relationship, measured in humans
Taylor and Forsyth 2016 18	5 healthy adults, in silico	Simulated graded anteroposterior section	Selective vulnerability of resting‐state networks by extent	Default mode most vulnerable, then frontoparietal, limbic, somatomotor	Extent–cost relationship, simulated

Abbreviations: ACC, anterior corpus callosotomy; ADL, activities of daily living; PCC, posterior corpus callosotomy; TCC, total corpus callosotomy.

The cost of disconnection is also better characterized than the framing of a simple trade‐off allows. The clinical syndrome was described in the human split‐brain series of the 1940s, which already distinguished the effects of partial from complete section, ¹⁹ , ²⁰ and it has been quantified in modern cohorts alongside seizure outcome. ²¹ , ²² , ²³ In contemporary series, acute disconnection effects are common but predominantly transient: after extended anterior section sparing the splenium, most of 29 children had some degree of acute disconnection syndrome, which resolved within 3 weeks, and postoperative attention improved. ¹³ More importantly, seizure control and cognition do not move only in opposite directions. In patients undergoing total callosotomy, the change in IQ at 1 year was significantly better in those with favorable seizure outcomes than in those without, although the authors attribute part of the decline in the unfavorable group to the surgical intervention itself ¹⁶ ; in West syndrome without MRI lesions, progressive decline in developmental quotient was prevented in patients with worthwhile seizure outcomes. ⁹ Cognitive outcome therefore reflects both the cost of disconnection and the benefit of seizure control, and the two must be modeled together.

Technical developments have broadened how the callosum can be divided — endoscopic‐assisted approaches, combined commissurotomy, radiosurgery, and minimally invasive laser interstitial thermal therapy ²⁴ , ²⁵ —and image guidance has begun to make the extent of section a controlled variable rather than an anatomical estimate. ¹⁷ Callosotomy nonetheless remains the epilepsy operation with the lowest overall seizure‐freedom rate in individual‐patient‐data meta‐analysis (43.4%), and its principal benefit remains palliation of falls rather than seizure freedom. ²⁶ Adaptive‐functioning studies underscore that outcomes extend beyond seizure counts to development and quality of life. ²⁷

This review asks what is required to decide, on evidence rather than preference, how much of the corpus callosum should be divided. We take the three procedures together rather than advocating one, and we treat the selective posterior variant as the case in which the question is posed most sharply, because it makes an explicit anatomical claim about which fibers matter. We argue that the clinical literature has taken this question as far as observational and imaging methods allow, that the remaining questions require experimental manipulation, and that the experimental literature—which is substantial, extends from 1948 to 2026, and is largely unread by the clinical field—was built to answer a different question. Figure 1 sets the two literatures side by side. Throughout, seizure types are described in accordance with the current International League Against Epilepsy classification, ²⁸ and generalized seizures are understood in the network sense introduced in the 2010 revision. ²⁹

FIGURE 1.

Corpus callosotomy: Parallel clinical and experimental timelines, 1940–2026. Clinical milestones (upper axis) and experimental studies of callosal or forebrain commissural section with an epilepsy‐related outcome (lower axis, n = 62), identified by structured search of PubMed, Embase, and the Web of Science with forward and backward citation tracking. Column height is the number of studies published in each 5‐year interval, counted from 1940. Forty‐two of the 62 studies fall between 1975 and 2002, the period dominated by kindling preparations. No eligible experimental study was identified between 2003 and 2019; six were published between 2020 and 2026.

2. WHAT THE CLINICAL LITERATURE CAN AND CANNOT ESTABLISH

The clinical logic of callosotomy rests on the corpus callosum as the substrate for rapid interhemispheric engagement. Diffusion‐tensor imaging in operated patients has shown that motor and premotor fibers cross the posterior half of the callosum, providing an anatomical basis for the selective posterior approach and for the observation that extending the section into the anterior body confers little additional benefit for drop attacks. ¹² Intraoperative recording from the callosum itself has shown that callosal neurons participate asymmetrically in bilaterally synchronous discharges, firing more in the more epileptogenic hemisphere even when the scalp EEG appears symmetric, ³⁰ , ³¹ and direct stimulation of the callosum at 1 cm intervals during commissurotomy has been used to map its cortical projections in patients. ³²

Three lines of contemporary work have taken the extent‐of‐section question considerably further than a narrative of clinical empiricism would suggest, and any claim about what remains unanswered must begin by acknowledging them. First, meta‐analytic comparison across the three procedures now provides pooled estimates of both benefit and disconnection cost by extent. ¹⁴ Second, tractography‐guided neuronavigation has been used to define a fiber‐based stopping point for posterior section, and in a small prospective series the completeness of motor fiber section correlated with drop attack reduction: complete section in four patients produced 100% reduction, while partial section of 82.1% and 89.3% of fibers produced 75.0% and 85.7%. ¹⁷ Third, high‐resolution structural connectivity has been used to simulate callosal lesions of increasing anteroposterior extent and to model their effect on resting‐state network integrity, predicting that the default mode network is most vulnerable and that the somatomotor and frontoparietal networks account for the unwanted effects, and that between‐individual variability in callosal connectivity may explain the poor predictability of outcome. ¹⁸ The extent–outcome relationship has therefore been measured in patients and modeled in silico.

What these methods cannot do defines the remaining gap. A clinical cohort cannot randomize the extent of section, cannot obtain the counterfactual within an individual, and cannot vary the extent while holding the underlying epilepsy constant. Tractography establishes which fibers were divided but not, independently, what their division did to propagation: the correlation between completeness and outcome in six patients is exploratory, and the fibers identified as targets were selected on the same anatomical hypothesis being tested. Simulation predicts network cost but has no seizure outcome at all. And none of these methods provides histological ground truth against which the imaging surrogates are validated—a requirement that is not academic, since postoperative imaging establishes the completeness of a callosal section reliably but not that of the anterior commissure, ³³ since partial sections require anatomical verification before their behavioral consequences can be interpreted, ³⁴ since coaptation of the sectioned edges can be misleading on MR, ³⁵ and since commissural reconnection has been proposed as a cause of late failure. ³⁶

Two further considerations argue against treating the callosum as a simple conduit whose division is purely beneficial. The first is that the callosum carries inhibitory as well as facilitatory traffic: focal seizure types have been reported to intensify after callosal section in patients, an effect attributed to the loss of transcallosal inhibition, ³⁷ and the contrasting consequences of bisection—amelioration of generalized seizures on the one hand, intensification of fragmented postoperative seizures on the other—have been recognized as an unresolved problem for four decades. ³⁸ The second is that the callosum is a principal but not exclusive route. Brainstem, thalamic and hippocampal commissural contributions survive callosal section, ³⁹ , ⁴⁰ and the corpus callosum has been proposed to exert a modulatory influence on rapid bilateral synchronization rather than to transfer seizures from one hemisphere to the other ⁴¹ —a reading consistent with the network definition of generalized seizures ²⁹ and with recent evidence dissociating the thalamic from the callosal contribution to generalization in the rodent. ⁴²

The question that remains is therefore narrower and sharper. It is not whether the callosum mediates interhemispheric propagation and synchronization; that has been established in animals and in patients. It is not whether the extent of section relates to outcome; that has been estimated clinically and modeled computationally. It is whether the relationship between the anteroposterior extent of disconnection, the interruption of propagation, and the cost to interhemispheric function can be measured causally in the same subjects, under randomization, with the fibers actually divided, verified anatomically.

3. THE FUNCTIONAL ANATOMY OF THE CORPUS CALLOSUM, AND ITS TRANSLATIONAL LIMITS

Because the optimal extent of section is the central unresolved question, the regional organization of the callosum requires explicit statement. The callosum is not homogeneous: fibers crossing it are ordered along its anteroposterior axis according to the cortical regions they interconnect, and the commonly used partition into five vertical segments corresponds approximately to prefrontal (rostrum and genu), premotor and supplementary motor (rostral body), primary motor (anterior midbody), primary sensory (posterior midbody), and parietal, temporal and occipital fibers (isthmus and splenium) ⁴³ , ⁴⁴ (Figure 2A). Fiber dissection has refined this scheme for the motor segments in particular, localizing the commissural connections of the leg motor area, ⁴⁵ and microsurgical anatomical studies have related these segments to the operative approach. ⁴⁶

FIGURE 2.

Regional organization of the corpus callosum and the translational problem. (A) Human callosum in midsagittal view, partitioned into five segments by the cortical regions they interconnect, with the three extents of section in current clinical use shown below. (B) Rodent callosum, in which the rostrocaudal ordering of callosal fibers is conserved but the anterior cortical territory served by segment I has no homologue, with the three anatomical considerations that limit direct translation. (C) Functional consequence of dividing each segment.

The functional consequence of dividing each segment follows from what it connects (Figure 2C). Division of the prefrontal fibers of the genu and rostrum is held to account for the executive, attentional, and behavioral components of the disconnection syndrome, and these are the fibers the selective posterior operation aims to spare. ¹¹ , ¹² Division of the premotor and supplementary motor fibers impairs bimanual coordination and motor planning, and callosotomy patients show characteristic deficits in the simultaneous initiation of two‐handed movements attributable to anterior and posterior segments differentially. ⁴⁷ Division of the primary motor segment interrupts the interhemispheric motor transfer, whose interruption is held to control drop attacks. Division of the sensory and posterior segments affects somatosensory, visual, and tactile transfer, including the tactile naming and interhemispheric memory phenomena that defined the classical split‐brain syndrome. Stated in this form, the rationale for selective posterior section is a specific anatomical claim: that segments III and IV carry the fibers responsible for the bilateral engagement producing falls, and that segment I can be spared without loss of antiepileptic effect.

That claim is testable, and it is where the experimental question begins. But it also exposes a problem for any rodent model, and the problem is more serious than a difference of scale (Figure 2B). Three points deserve statement, and the first is not the one usually assumed. The rostrocaudal ordering of callosal fibers is conserved: tract tracing and in vivo tractography in the mouse show frontal areas crossing at the genu and parieto‐occipital areas at the splenium, as in the human. ⁴⁸ , ⁴⁹ The difficulty is not that the order is scrambled but that the cortical sheet whose order is preserved is not the same sheet. The rodent has no granular prefrontal cortex homologous to the human dorsolateral prefrontal areas, ⁵⁰ , ⁵¹ so the anterior territory served by segment I in the human has no rodent counterpart, and equal fractions of callosal length in the two species do not divide homologous fibers. Second, and following from this, the prefrontal‐sparing rationale — the entire clinical motivation for the selective operation—has no straightforward anatomical counterpart in the rodent: what a rat model can test is the seizure consequence of dividing posterior versus all callosal fibers, not the preservation of a prefrontal network the animal does not possess. Third, the hippocampal commissure lies immediately beneath the splenium in both, but is proportionally larger in the rodent and is reduced in primates relative to other eutherian mammals, ⁵² , ⁵³ , ⁵⁴ so that the confound it introduces differs in magnitude between the model and the operation being modeled.

The implication is not that rodent models are unusable, but that a rodent section must be defined by the cortical connections it divides rather than by a fraction of callosal length, and that the division must be verified anatomically. Anterograde tract tracing has been used for exactly this purpose in a rodent model of hemispherotomy, identifying cortico‐cortical pathways crossing the midline in controls and their absence after disconnection. ⁵⁵ Diffusion imaging with tractography has been applied to the rodent callosum and has identified the segment interconnecting the cortical regions most involved in seizures in genetic absence strains. ⁵⁶ Both are available; neither has been applied to a comparison of extents of callosal section.

4. AN EXPERIMENTAL LITERATURE THAT ANSWERED A DIFFERENT QUESTION

In 2005, addressing the Asian Oceanian Epilepsy Congress, Wada made the following observation about the two largest meetings in the field held 4 years earlier ³⁸ :

More than 28,000 neuroscientists and 3,000 epileptologists gathered at their respective 2001 meetings of the Society for Neuroscience and the American Epilepsy Society. Yet only six articles, one directly and five indirectly, discussed the corpus callosum. Is not this in itself a remarkable finding?

He went on to note that considerable uncertainty persisted about the rationale for callosal bisection, which produces contrasting effects: amelioration of generalized seizures, at times to the point of seizure freedom, and intensification of postoperatively fragmented seizures, at times leading to status epilepticus. Two decades later, that uncertainty is unresolved, and the reason is not that the experimental literature is absent.

The search reported here was conducted as a structured and reproducible literature search in support of a Critical Review, and not as a formal systematic review as defined by PRISMA: the strategy, the yield of every set and the eligibility classification of every retrieved record are reported in full so that the search can be repeated, but screening and classification were performed by a single reviewer. The limitations that follow from this are set out at the end of this section. We searched PubMed, Embase, and the Web of Science Core Collection on September 7, 2026, without language or date restriction, combining an intervention set with an epilepsy set and a species restriction (the validated animal filter of de Vries et al. in Embase, the corresponding MeSH restriction in PubMed, and a free‐text species set in the Web of Science) in each database, and supplemented the searches with forward citation tracking of 12 anchor studies and with reference list screening. The complete strategy, with the yield of every set, is given in the Supplementary Material. The searches returned 48 records in PubMed, 76 in Embase and 103 in the Web of Science, 227 in total, reducing to 144 unique records after removal of 83 duplicates; only 23 were retrieved by all three databases, while 57 were unique to the Web of Science, 22 to Embase and 8 to PubMed. Forward citation tracking of 12 anchor studies was performed as a completeness check; most of the records it returned were already in the corpus and the remainder were merged into it before screening. We note a limitation of that check: the PubMed “Cited in” index derives from references deposited in PubMed Central and is substantially incomplete for literature published before 2000, so the citation counts it returns for the older anchors are floors rather than true counts. Screening identified 62 experimental studies in which the corpus callosum or another forebrain commissure was sectioned with an epilepsy‐related outcome (Figure 3).

FIGURE 3.

Flow of the literature search, September 7, 2026, in PRISMA style. Records identified in PubMed, Embase, and the Web of Science Core Collection; duplicates removed by DOI, then PMID, then a normalized title key; unique records screened on title and abstract and assigned to one of four categories, defined in Table S1; the eligible corpus is the 62 studies in which a callosal or forebrain commissural section was performed and a seizure‐related outcome reported. The distribution of the 62 eligible studies over time is shown below the flow; the overlap between databases is reported in the Supplementary Material. This is a structured search reported transparently and not a systematic review as defined by PRISMA; screening was performed by a single reviewer, with artificial intelligence assistance in record handling, and the classification of every record is given in Table S1.

This literature begins in 1948, with callosal section in the monkey, ⁵⁷ reported in full in 1950, ⁵⁸ and extends to 2026, though not evenly: no eligible study was identified between 2003 and 2019, a gap examined in Section 5. It spans the cat, rat, mouse, dog, rhesus monkey, and baboon; the alumina cream, penicillin, kainate, pentylenetetrazol, lithium–pilocarpine, and photosensitive models; genetic absence strains and congenitally acallosal animals; kindling of the amygdala, hippocampus, and motor cortex; and in vitro slice preparations (Table 2). It is neither small nor confined to the distant past (Figure 1).

TABLE 2.

Experimental studies of callosal or forebrain commissural section with epilepsy‐related outcomes, 1948–2026, identified by structured search of PubMed, Embase, and the Web of Science Core Collection with forward citation tracking (September 7, 2026).

Study	Species	Model	Extent of section	Principal finding	Bearing on the extent‐of‐section question
Kopeloff and Kennard 1948 57 ; 1950 58	Monkey	Experimental epilepsy	Callosal section	First experimental callosotomy in an epilepsy preparation	Origin of the literature; same group, not an independent replication
Teasdall and Stavraky 1950 59	Cat	Pentylenetetrazol convulsions	Callosal section	Clonic phase augmented, repeated convulsions increased, convulsive threshold lowered by 20–50%	Section worsens the seizure in this model
Séguin 1961 60	Rat	Pentylenetetrazol	Callosal section vs. sham	Threshold lowered, latency shortened, convulsions prolonged; significant against sham under repeated challenge (p < 0.01)	The bidirectional effect, demonstrated experimentally
Wada and Sato 1975 39	Cat	Amygdala kindling	Forebrain bisection	Generalization proceeded despite bisection; midbrain reticular route	Callosum not the exclusive route
McIntyre 1975 61	Rat	Amygdala kindling	Graded commissurotomy	Anterior callosum determines laterality of forelimb clonus	Extent varied; endpoint kindling transfer, not seizure control
Majkowski 1976 62	Cat	Alumina cream	Partial or complete	Seizures developed in ~60%, below intact controls (n = 20, up to 836 days)	Extent varied; not analyzed as a dose–response
Kusske and Rush 1978 63	Rat	Cortical afterdischarge	Callosal section	Propagation of afterdischarge to contralateral cortex and thalamus interrupted	Section vs. no section; extent not varied
Musgrave and Gloor 1980 64	Cat	Generalized penicillin epilepsy	Complete callosal + anterior commissure; massa intermedia alone; partial callosal; cortical slab	Four surgical groups compared for bilateral synchrony	The closest this literature comes to a graded‐extent design
Wada 1982 65	Cat	Amygdala kindling	Forebrain bisection, degree varied	Interference effect maximal with complete bisection	Extent varied; endpoint kindling transfer
Wada and Mizoguchi 1984 66	Baboon (photosensitive)	Limbic kindling	Forebrain bisection	Kindling in the bisected photosensitive baboon	Generalized primate preparation
McIntyre 1985, 1986, 1987 67 , 68 , 69	Rat	Hippocampal kindling	Callosum vs. hippocampal commissure	Dissociates the two commissures; commissurotomy raised the probability of status epilepticus while reducing its severity	Addresses the principal anatomical confound
Fukuda 1989 70 ; 1988 71	Baboon	Photosensitive	Forebrain bisection	Discharge confined to one hemisphere; tonic phase remained bilateral	Generalized model; extent fixed
Vergnes 1989 72	Rat (absence strain)	Spontaneous spike–wave	Callosal transection	Spike–wave desynchronized; residual synchrony persists	Bilateral‐synchronization model; extent fixed
Fernandes de Lima 1990 40	Rat	Hippocampal afterdischarges	Selective commissurotomy	Synchrony depended on the ventral hippocampal commissure	Noncallosal commissural route
Okamoto 1991 73	Cat, baboon, rhesus	Amygdala kindling	Callosal bisection	Callosal role rises up the phylogenetic scale	Bears on rodent‐to‐human validity
Hirsch 1992 74	Rat	Lithium–pilocarpine	Callosal section	57% never developed status; mortality abolished	Hippocampal commissure lesioned in 64% — the confound, measured
Wada group 1993–1999 75 , 76 , 77 , 78	Rat, cat	Amygdala kindling	Brainstem bisection	Dissociates callosal from brainstem contribution	The systematic counterpoint to a callosal account
Kudo 1997 79	Cat	Motor cortical kindling	Callosal bisection	Generalization delayed; propagation persisted subcortically	The focal cortical design, executed
Papo and Quattrini 1997 80	—	Review of the experimental literature	—	Concluded the experimental studies lacked homogeneity and could not guide surgical decisions	The last synthesis before the present one
Tanaka 2001 81	Rat, cat	Kainic acid	Callosotomy	Contralateral propagation blocked; ipsilateral and subcortical activity preserved	Focal model; extent not varied
Ono 2002 82	Rat	Transcallosal volleys	Callosotomy	Transcallosal facilitation of cortical reactivity abolished	Mechanism of the suppressive effect
Medina 2002 83	Acallosal mouse	Pentylenetetrazol	Congenital absence of the callosum	Acallosal mice more susceptible to PTZ seizures	Congenital counterpart of the surgical finding
Wada 2005 38	—	Kindling and clinical synthesis	Anterior callosal bisection	Bidirectional effects; transcallosal secondary antiepileptogenesis	Reframes the extent question
Concha 2006 84	Human (post‐callosotomy)	—	Callosotomy	DTI separates axonal from myelin degradation over time	The imaging read‐out, applied
Chahboune 2009 56	Rat (WAG/Rij, GAERS)	Genetic absence	None (imaging)	Reduced fractional anisotropy in the anterior callosum	Rodent diffusion imaging is feasible
Mishra 2013 85	Rat (WAG/Rij)	Genetic absence	None (imaging)	Increased resting interhemispheric cortical connectivity	Functional read‐out in a bilateral‐synchronization model
Scicchitano 2015 86	Rat (WAG/Rij)	Genetic absence	Unilateral and bilateral cortical resection	Effects on spike–wave discharges	Surgical disconnection used to test a network hypothesis
Matias 2019 55	Rat	Disconnection surgery	Hemispherotomy	Tract tracing and SPECT verification; motor outcome	The verification standard this review proposes
Cotter 2020 87	Mouse (slice)	GABA‐A blockade	Callosal severing in vitro	Callosal circuits direct interhemispheric propagation of epileptiform events	In vitro confirmation of the route
Hasegawa 2021, 2022; Asada 2021; Hasegawa 2025 88 , 89 , 90 , 91	Cat, dog	Natural drug‐resistant epilepsy	Callosotomy; rostral callosotomy	EEG desynchronization; imaging of the extent; 12‐month follow‐up	The translational step between rodent and human
Brodovskaya 2022 42	Mouse	Focal seizure spread	Anterior callosotomy vs. thalamic inhibition	Distinct roles in generalization; challenges the thalamocentric model	The modern mechanistic reference
Tanaka 2025 92	Cat, rat	Kainic acid, sensorimotor cortex	Callosotomy	Contralateral activity abolished; ipsilateral hypermetabolism persisted on 2‐deoxyglucose autoradiography	Metabolic read‐out; extent not varied
Cavalcanti 2026 93	Rat	PTZ kindling	Posterior vs. total	No detectable difference; pilot underpowered, model unsuited to propagation endpoints	The only posterior‐versus‐total comparison

Note: Representative studies, selected to show the range of species, models, and endpoints; the complete screening record for all 144 retrieved records is given in Table S1. No study compares anterior, posterior, and total section against a common seizure outcome; none addresses the selective posterior variant; none randomizes between extents; and none combines graded disconnection with time‐locked bilateral electrophysiology, diffusion‐based verification, and a behavioral read‐out in the same animals.

From the mid‐1970s, the kindling preparation dominated, and a sustained program developed around it. Wada and Sato showed that amygdala kindling generalized rapidly in forebrain‐bisected cats and attributed this to the midbrain reticular formation rather than to transhemispheric relay, ³⁹ a conclusion the same group pursued through a series of brainstem‐bisection studies over the following two decades. ⁷⁵ , ⁷⁶ , ⁷⁷ , ⁷⁸ McIntyre, across studies between 1975 and 1987, dissociated the anterior callosum, which determines the laterality of forelimb clonus, from the hippocampal commissure, which governs the laterality of hippocampal afterdischarge, and found that commissurotomy paradoxically raised the probability of status epilepticus while reducing its severity. ⁶¹ , ⁶⁷ , ⁶⁸ , ⁶⁹ Majkowski followed 20 split‐brain cats with chronic alumina foci for up to 836 days. ⁶² Kudo carried out motor cortical kindling in 12 cats, seven commissurally intact and five callosally bisected, and found that bisection delayed generalization while propagation persisted subcortically. ⁷⁹

Generalized models were not neglected. Vergnes transected the callosum in a Wistar strain inbred for spontaneous absence seizures, the strain from which Genetic Absence Epilepsy Rats from Strasbourg (GAERS) derive, and showed that bilateral spike–wave discharges became asymmetrical, unilateral, or alternating, while occasional synchrony persisted. ⁷² In the photosensitive baboon, forebrain bisection confined light‐induced discharge to one hemisphere while the tonic phase remained bilateral ⁷⁰ , ⁷¹ ; the same preparation was used for limbic kindling. ⁶⁶ In feline generalized penicillin epilepsy, four surgical groups—complete callosal and anterior commissural section, division of the massa intermedia alone, partial callosal section, and a cortical slab—were compared for their effect on bilateral synchrony. ⁶⁴

Nor did the field fall silent after 2002. Diffusion‐tensor imaging has been applied to the time course of axonal and myelin degradation after callosotomy ⁸⁴ and to the callosum of genetic absence strains. ⁵⁶ A standardized rodent model of hemispherotomy has been described with anterograde tract tracing to verify the disconnection, perfusion imaging, and a motor outcome. ⁵⁵ Callosal severing in a slice preparation showed that callosal circuits direct the propagation of epileptiform events between hemispheres. ⁸⁷ The respective contributions of the thalamus and the callosum to seizure generalization have been dissociated in the rodent by mapping the spread of focal seizures, challenging the canonical thalamocentric model. ⁴² Callosotomy has been performed in cats and dogs with naturally occurring drug‐resistant epilepsy, with electroencephalographic confirmation of desynchronization, postoperative imaging of the extent of section, and 12‐month functional follow‐up. ⁸⁸ , ⁸⁹ , ⁹⁰ And a full experimental epilepsy surgery program, including callosotomy with metabolic mapping by 2‐deoxyglucose autoradiography, was reviewed in this journal in 2025. ⁹²

Set against this, the design gap is precise. A small number of studies varied the extent of the section, ⁶¹ , ⁶² , ⁶⁴ , ⁶⁵ and others compared section against no section for the propagation of afterdischarge. ⁶³ In every case, the endpoint was the laterality of kindling transfer, of afterdischarge, or of interictal synchrony. Not one compares anterior, posterior, and total section against a common seizure outcome. Not one models the selective posterior operation as it is defined clinically, by the prefrontal fibers it spares and the anatomical rationale for sparing them; that operation did not exist when most of this work was performed. Not one randomizes animals between extents of disconnection. Not one combines a graded surgical manipulation with time‐locked bilateral electrophysiology, diffusion‐based verification of the fibers actually divided, and a behavioral or cognitive read‐out in the same animals. The single study that treats the anteroposterior extent as its primary independent variable is our own recent pilot, which used a generalized chemoconvulsant model and was underpowered for the purpose ⁹³ ; it compared a posterior with a total section defined by callosal length, not the prefrontal‐sparing operation as it is defined clinically.

The asymmetry, then, is not one of volume, of period, or of attention. Two literatures ran in parallel, at times in the same hands: Papo and Quattrini published on clinical callosotomy in 1987 and 1989 and reviewed the experimental literature in 1997, ⁸⁰ and Wada moved between the operating room and the kindling laboratory for 30 years. They answered different questions. The experimental program was built to establish whether the callosum relays and synchronizes epileptiform activity between the hemispheres, and it answered that: the callosum is a principal, though not exclusive, route. The clinical question that followed—how much of the callosum must be divided, and at what cost—requires a manipulation that literature has never performed.

Four limitations of this search should be stated, because the argument rests on negative evidence. Screening and classification were performed by a single reviewer (VBC), with artificial intelligence assistance in record handling and classification and with every decision verified by that reviewer, and without duplicate screening by a second reviewer or a formal agreement statistic; this is a structured search reported transparently, not a systematic review as defined by PRISMA, and the classification of every retrieved record is given in Table S1 so that any individual decision can be checked. The forward citation tracking used the PubMed “Cited in” index, which derives from references deposited in PubMed Central and is substantially incomplete for literature published before 2000, so the counts it returns for the older anchors are floors rather than true values. Nine eligible records have full texts in Japanese, Polish, Italian, or Russian and were classified from the indexed English title and abstract; none of them carries a load‐bearing claim here. And gray literature, dissertations, and conference proceedings were not searched, which matters more than usual in a field whose early work appeared partly in meeting abstracts—one such report, recovered from the reference list of a full text rather than from any database, is cited below.

5. WHY THE QUESTION WAS NOT ASKED

If the experimental literature is substantial and the clinical question is well defined, the absence of a controlled comparison of the extent of section requires explanation. Five factors account for it:

The first is chronological rather than technical. Most of the experimental program was built between 1948 and 2002, and the selective posterior operation was introduced after it. ¹¹ , ¹² The question of how much callosum to divide could not be posed experimentally in the form it now takes, because the clinical variant that poses it did not exist. The chronology is exact enough to be worth stating: the search identifies no eligible experimental study between 2003 and 2019, and the silence ends in 2020, the year the selective posterior operation entered practice. The experimental literature did not lose interest in the callosum; it had answered the question available to it, and the next question had not yet been asked in the operating room. What the experimental field asked instead—whether the callosum relays and synchronizes epileptiform activity, and through which commissures—was the question available to it. This is the ordinary sequence in which surgical questions become experimentally tractable, and it explains why the two literatures diverged without either being neglected.

The second is a mismatch between the most convenient seizure models and the mechanism callosotomy targets. The chemoconvulsant models most widely used in rodents—pentylenetetrazol above all—produce bilaterally synchronous discharges whose generalization depends substantially on brainstem‐reticular and thalamocortical circuitry rather than on callosal relay. In such models, the callosum is not the principal route of spread, and callosotomy is therefore a poor instrument for interrogating it. The historical record, however, shows something more troubling than insensitivity. Callosal section performed 1–5 months before pentylenetetrazol challenge predisposed Sprague–Dawley rats to convulsions: the convulsant threshold was lowered, the latent period shortened, the convulsions prolonged, and clonic and repeated convulsions more frequent. ⁶⁰ The statistical weight of that study rests on its sham‐controlled arm under repeated weekly challenge, in which the callosotomized animals convulsed more at every injection (p < 0.01); the single‐dose median convulsant dose differed in the same direction but did not reach significance. Nor was the finding new even then: section of the callosum in the cat had been reported in 1950 to augment the clonic phase of pentylenetetrazol convulsions, to increase the proportion of repeated convulsions, and to lower the convulsive threshold by 20–50%. ⁵⁹ In a companion study, callosotomized rats were more sensitive to pentylenetetrazol than sham‐operated animals, with a higher incidence of clonic, tonic, and repeated convulsions, ⁹⁴ and a further study from the same laboratory found a lowered convulsive threshold and prolonged convulsions after section, there under decompression hypoxia rather than pentylenetetrazol ⁹⁵ ; and seizure susceptibility to pentylenetetrazol has been examined in an acallosal mouse strain. ⁸³ Removing the callosum in this model does not fail to change the seizure—it can make it worse. That result is coherent with the transcallosal inhibition hypothesis advanced for the intensification of focal seizures after section in patients, ³⁷ , ³⁸ and it means the pentylenetetrazol model does not merely lack sensitivity to the callosal contribution but may measure it with the wrong sign. The authors of the 1961 study drew the corresponding inference about the clinical operation itself, proposing that the reduction in seizure severity van Wagenen and Herren had reported after section might reflect a selective depression of partially denervated cortical neurons rather than the interruption of propagation alone ⁶⁰ ; whatever the merits of that particular mechanism, the observation that the direction of the callosal effect was contested from the beginning is not a modern qualification of this literature but a feature of it. Our own pilot used this model and found no detectable difference between posterior and total section ⁹³ ; that report states at the outset that generalization in pentylenetetrazol models depends substantially on subcortical circuitry and is therefore not strictly contingent on callosal propagation, and that chronic focal models would be better suited to isolate the callosal contribution.

The third barrier is technical. Stereotaxic section of the rodent callosum is demanding: the target is a thin midline sheet overlying the lateral ventricles and hippocampus, immediately deep to the superior sagittal sinus, and its dorsoventral depth varies along its length. Producing a reproducible, anatomically defined section—let alone distinct posterior and total variants in the same laboratory—requires atlas‐guided coordinates, control of blade depth, and measures to protect the sinus. ⁹⁶ , ⁹⁷ Without a standardized, shareable protocol, each laboratory must reinvent the procedure, and cross‐study comparison becomes impossible. That this obstacle is surmountable is shown by a recently described rodent model of hemispherotomy, in which a reproducible disconnection was defined by atlas coordinates and verified by anterograde tract tracing and perfusion imaging. ⁵⁵

The fourth is verification, and it is both a methodological and an anatomical problem. Demonstrating that the intended section was achieved requires histological confirmation of myelinated fiber interruption, ideally with myelin‐specific staining; fixation and processing artifacts, and the fragility of long‐fixed white matter can compromise the very stains needed to document the lesion, so that completeness is easily asserted and hard to prove. ⁹⁷ Contemporaneous criticism of this literature made precisely this point, noting that several reports omitted the surgical technique and provided no histological evidence of the extent of section. ⁹⁸ The anatomical difficulty is separate and less often acknowledged: the hippocampal commissure lies immediately beneath the splenium, is a delicate membranous structure rather than a discrete bundle, varies between individuals and is reduced in primates relative to other mammals. ⁵² , ⁵⁴ Neither its preservation nor its division can be assumed from the surgical plan, which is the confound that limits interpretation of the lithium–pilocarpine study ⁷⁴ and which the selective commissurotomy experiments identified directly. ⁴⁰ In patients, even postoperative imaging has proved an imperfect guide to completeness, ³³ , ³⁴ and commissural reconnection has been proposed as a cause of late failure. ³⁶

A fifth, less technical factor compounds these. Callosotomy is a palliative operation for a minority of surgical candidates, and experimental surgical models are laborious and low in throughput. In an environment that rewards higher throughput preparations and mechanistic questions of broader reach, a demanding model of a niche operation attracts little sustained investment—the point made forcefully in the observation with which this section opened. ³⁸ Breaking this cycle requires demonstrating that the model can answer a question of clear clinical value that cannot be answered any other way.

6. WHAT A MODERN EXPERIMENTAL PROGRAM WOULD REQUIRE

The gap identified above is narrow and specific: no study has related the anteroposterior extent of callosal disconnection to the interruption of propagation and to the cost of disconnection, in the same animals, under randomization, with the fibers actually divided verified anatomically. Each element of that sentence corresponds to a design requirement. We set them out not as a novel program but as the assembly of components that already exist separately in this literature.

6.1. Seizure models

The model must be one in which callosal fibers demonstrably participate in bilateral engagement, so that disconnection has an interpretable effect. Two classes are required rather than one. Focal cortical models—intracortical kainate, cortical electrical kindling—permit propagation to be measured as a directional quantity, and callosal section has been shown in exactly these preparations to abolish contralateral cortical activity while leaving ipsilateral and subcortical discharge intact. ⁷⁹ , ⁸¹ , ⁹² But callosotomy is indicated clinically for generalized seizure types—tonic, atonic, epileptic spasms—and the callosum has been proposed to exert a modulatory influence on rapid bilateral synchronization rather than to transfer a seizure from one hemisphere to the other, ⁴¹ a reading consistent with the network definition of generalized seizures. ²⁹ A program restricted to focal models would optimize mechanistic interpretability at the cost of face validity. Models of bilateral synchronization are available and have been used with callosal section: the genetic absence strains, in which callosal transection desynchronizes spontaneous spike–wave discharge ⁷² and in which callosotomized animals have since been used as the preparation of choice for studying unilateral discharge generation, ⁹⁹ and the photosensitive baboon. ⁷⁰ Diffusion imaging has already been applied to the callosum of these strains, ⁵⁶ and resting‐state functional imaging has shown increased interhemispheric cortical connectivity in the same model, ⁸⁵ and surgical disconnection has been used in them to test network hypotheses. ⁸⁶ A two‐tier design—a focal model to quantify the blockade of propagation, a bilateral‐synchronization model to test clinical relevance—addresses both requirements.

6.2. Standardized surgical intervention

Progress requires a reproducible, atlas‐referenced protocol defining distinct anterior, posterior, and total sections by explicit coordinates, controlling blade depth along the dorsoventral profile of the callosum, and protecting the superior sagittal sinus with a defined lateral offset. ⁹⁶ Our group has described one such rodent protocol and applied it in a comparative behavioral pilot, both published in 2026 ⁹³ , ⁹⁷ ; these companion reports establish surgical feasibility and a histological verification workflow, not propagation, and the model they used is unsuited to propagation endpoints for the reasons set out above. They are offered as a starting point and are discussed critically rather than as a solution. As Section 3 argues, a rodent section must in any case be defined by the cortical connections it divides rather than by a fraction of callosal length. The independent demonstration that a disconnection procedure can be standardized by atlas coordinates and verified by tract tracing in the rat provides a template for the level of anatomical control required. ⁵⁵

6.3. Read‐outs, and what is already available

Three modalities have been applied to this problem and should be treated as established rather than proposed. Time‐locked bilateral electrophysiology quantifies interhemispheric propagation before and after graded section, and callosal compound action potentials recorded intraoperatively in patients have shown that callosal neurons participate asymmetrically in bilateral synchrony. ³⁰ , ³¹ Diffusion imaging and tractography relate the extent of disconnection to fiber integrity and have been applied both to the time course of degradation after callosotomy ⁸⁴ and to the rodent callosum ⁵⁶ ; this is the modality that allows preclinical and clinical measures to be expressed in a common currency. ¹² , ¹⁷ Metabolic mapping by 2‐deoxyglucose autoradiography offers a whole‐brain image of where activity persists after disconnection and has been applied to experimental callosotomy, showing that contralateral cortical excitation falls while ipsilateral cortical and subcortical hypermetabolism persists. ⁹² Rigorous myelin‐based histological verification must accompany these, and the integrity of the hippocampal commissure must be documented separately rather than inferred.

6.4. Cognitive and behavioral outcomes as a co‐primary endpoint

Structural preservation of anterior prefrontal fibers does not demonstrate preservation of function, and the clinical literature no longer supports a simple trade‐off in which seizure control is bought with cognition. ⁹ , ¹⁶ If extents of section are to be compared experimentally, behavioral and cognitive measures must be collected in the same animals and analyzed alongside the electrophysiological measures. Precedents exist within this literature: state‐dependent learning and avoidance paradigms were applied to commissurotomized animals in the 1980s, ¹⁰⁰ and interhemispheric transfer tasks have a long history in the split‐brain preparation. The absence of such measures from recent experimental surgery programs has been acknowledged by their authors. ⁹²

6.5. A translational step between rodent and human

The surgical feasibility and morbidity of callosal division were first established in normal dogs. ¹⁰¹ Callosotomy is now performed in dogs and cats with naturally occurring drug‐resistant epilepsy, with preoperative electroencephalography, postoperative imaging of the extent of section, and functional follow‐up over months. ⁸⁸ , ⁸⁹ , ⁹⁰ , ⁹¹ These animals have gyrencephalic brains, spontaneous epilepsy, and owners who report quality‐of‐life outcomes. They cannot be randomized between extents of section, but they provide the intermediate validity step that a rodent‐to‐human roadmap otherwise lacks, and this literature is almost entirely absent from the neurosurgical discussion of the operation.

6.6. Design and welfare

Because these models are technically demanding and inherently limited in throughput, studies should be powered on pilot effect sizes, incorporate unoperated and sham controls, and predefine humane endpoints. The comparative design that translates the clinical question most directly is randomization of animals between anterior, posterior, and total section and sham, with the extent verified after the fact rather than assumed from the surgical plan. Adherence to the ARRIVE guidelines and transparent reporting of attrition and verification failures would raise the credibility and reusability of a literature that has suffered from neither being reported. ⁹⁸

A first study follows directly. In a focal, cortically driven model with demonstrable transcallosal spread, animals randomized to anterior section, posterior section, total section, or sham would be compared on time‐locked bilateral electrocorticographic measures of contralateral propagation, on a behavioral seizure metric, on a cognitive or behavioral task sensitive to interhemispheric function, and on the extent of disconnection verified by diffusion imaging and myelin histology, with the hippocampal commissure documented separately. A parallel arm in a bilateral‐synchronization model would test whether the relationship holds for the seizure types for which the operation is actually performed. Such a design would, for the first time, place the three extents of section side by side under controlled conditions and relate each to both the intended benefit and a measured proxy for the off‐target cost.

7. CONCLUSION

Corpus callosotomy is an operation whose clinical evidence base has matured while the experimental question that would allow it to be optimized has remained unasked. The experimental literature is not absent: it is substantial, it extends from 1948 to 2026, and it established what it set out to establish—that the corpus callosum is a principal but not exclusive route for interhemispheric propagation and synchronization of epileptiform activity, and that its section has consequences in both directions. What it never did was vary the extent of section against a clinically meaningful outcome, because the clinical question that makes the extent matter arose after most of that work was done.

A modern experimental program, built on appropriately chosen focal and bilateral‐synchronization models, a standardized protocol defined by the connections divided rather than by callosal length, physiological, imaging and metabolic read‐outs that already exist, cognitive outcomes measured in the same animals, and anatomical verification of what was actually cut, could provide direct experimental evidence on how much of the callosum should be divided and at what cost. Assembling these components into a single randomized design is, in our view, an important next step for the field. It is not the only one: patient selection, long‐term cognitive and developmental outcomes, comparison with neuromodulation, minimally invasive approaches and network biomarkers all remain open, and progress in any of them would be of value. But the extent‐of‐section question is the one that the clinical literature cannot resolve on its own, and it is the one for which the experimental tools are already in place. The proposal is falsifiable, and we state what would refute it: if anterior, posterior and total section did not differ in the interruption of contralateral propagation under randomization, with the extent of each verified anatomically, then the anteroposterior extent would not be the variable the clinical debate has assumed it to be, and the question would have to be reformulated rather than answered.

AUTHOR CONTRIBUTIONS

Vitória Bernal Cavalcanti: Conceptualization, methodology, investigation (design and execution of the literature search, screening, and classification of records), data curation, writing – original draft, and visualization. Paulo Henrique Pires de Aguiar: Supervision, and writing – review and editing. Both authors read and approved the submitted version and agree to be accountable for all aspects of the work.

FUNDING INFORMATION

This research received no specific grant from any funding agency in the public, commercial, or not‐for‐profit sectors.

CONFLICT OF INTEREST STATEMENT

None of the authors has any conflicts of interest to disclose. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

USE OF ARTIFICIAL INTELLIGENCE

During the preparation of this work, the authors used a generative artificial intelligence assistant (Claude, Anthropic) as a research tool, under author direction, to assist with the execution of the pre‐specified database searches, the handling and classification of retrieved records against the authors' eligibility criteria, the verification of references against database and publisher records, the generation of figures from the screening data by scripted plotting, and formatting. All search strategies and eligibility criteria were specified by the authors; every eligibility decision and every reference was checked by the first author, who verified all output. The authors take full responsibility for the content of the publication.

Supporting information

Table S1. Classification of every record retrieved by the search of PubMed, Embase, and Web of Science Core Collection (search date 7 September 2026): 144 unique records after removal of 83 duplicates from 227 records identified. Records are classified into categories A (n = 62, eligible experimental studies), B (n = 20), C (n = 18), and D (n = 44).

Appendix S1.

DATA AVAILABILITY STATEMENT

All data supporting the findings of this review are contained within the article and its Supplementary Material. The search strings as executed, the yield at every stage and the eligibility classification of all 144 retrieved records are provided in the Supplementary Material and in Table S1. No new primary data were generated.