Delay to Repair and Functional Recovery After Wartime Mycotoxin Nerve Damage: A Retrospective Cohort of 142 Patients Repaired in Chiba City, 2033–2039

1. Introduction

Nervous-system injury attributed to a wartime mycotoxin has been one of the more persistent clinical legacies of the War. The registry entries we use describe a toxin of Russian military origin, a label taken from the clinical dossiers and not verified here, which damaged peripheral nerves and, in a smaller number of patients, central pathways. The damage was lasting and, at first presentation, usually functionally severe. Many patients could not manage tasks that depend on fine motor control: sustained grip, small-object manipulation, signature-level handwriting.

Repair became a specialty of the black clinics of Chiba City. These clinics, which operate outside the licensing regime of the municipal hospitals, built their reputation on nerve splicing, and the Chiba City Black Clinic Case Registry holds the dossiers of patients referred to them for neural repair. Clinics report their own methods and results, and the methods differ. Some clinics graft the patient's own nerve tissue; others bridge the damaged segment with synthetic conduits; a third group combines a conduit with cultured donor tissue. Surgical reports describe the graft and conduit approaches (Omodaka, 2038; Okimoto and Adebayo-Kurobane, 2036), but no analysis has compared them across clinics, and the prior literature has said little about when in a patient's course repair was undertaken.

Timing is the variable of greatest practical interest. Patients reached the clinics at very different intervals after exposure. Some were referred within months; others waited years, for reasons of cost, mistrust of unlicensed surgery and the difficulty of linking a progressive neurological decline to a remote exposure (Adeyemi-Ross, 2037). If delay degrades recovery, the referral pathway is itself a clinical problem. If it does not, the pathway's slowness may be a matter of access alone. Earlier surgical series point to a time-sensitivity of regeneration (Shimabukuro, 2037) but were drawn from civilian injuries of other origin, and their relevance to a toxic injury is unestablished.

We therefore ask three questions of the registry. Is the interval from exposure to repair associated with functional recovery at 12 months? Is a recorded proxy for the severity of the acute illness associated with recovery? And do the three repair methods differ in outcome after adjustment for the first two? Outcomes are a continuous recovery score and a binary measure of regained fine motor function. We write in 2041 and report repairs performed up to 2039, with 12-month assessments recorded in the registry until it was closed to analysis at the end of 2040.

2. Methods

Data source and eligibility. The Chiba City Black Clinic Case Registry holds standardised referral and follow-up dossiers contributed by participating clinics under a data-sharing agreement with the Chiba Municipal Medical College (Chiba City Black Clinic Case Registry, 2040). We included every patient whose dossier recorded a repair of peripheral or central neural injury performed between 2033 and 2039, in whom the attribution to a wartime mycotoxin met the registry's own criteria: a documented exposure during the War and a neurological presentation matching the registry's attribution protocol (Chiba City Black Clinic Case Registry, 2038). Patients with a competing neurological diagnosis recorded in the dossier were excluded. The registry's attribution criteria are clinical and were applied by the contributing clinics. We did not review toxicology independently. Identities were removed by the registry before analysis.

Of 171 eligible patients, 142 had a recorded 12-month assessment and form the analytic cohort. The remaining 29 (17%) lacked one, most often because the patient had left the Chiba area. Registry metadata on the excluded group were complete for delay and method. Their median delay was 30 months (IQR 16–42), against 22 months in the analytic cohort (Mann–Whitney p = .460), and their distribution across the three methods did not differ detectably from the analytic cohort (χ² test p = .811). The excluded patients therefore tended to have waited longer, and since longer delay is associated with poorer recovery in our data, their unobserved outcomes were probably worse than the analysed cohort's, so the cohort's mean score is probably an overestimate for the eligible population.

Exposures. Delay was the number of whole months between the exposure date recorded in the dossier and the date of the repair operation. Where a dossier gave an exposure date only to the month, we used that month. The severity proxy was the number of days of acute-phase hospital care documented at first presentation. It is a crude marker, since hospital days depend on bed availability as well as on illness, and we use it only because no other severity measure is recorded for the whole cohort. Repair method was taken from the operative report and classed as autologous nerve graft, synthetic conduit splice or hybrid splice with cultured donor tissue.

Outcomes. The functional recovery score is the registry's 0–100 composite, assigned by the treating clinic at 12 months from standardised motor, sensory and dexterity tasks (Arakawa-Beaumont, 2039). Higher scores indicate better function. The second outcome is the registry's binary entry for regained fine motor function, which records whether the patient could complete a standard small-object sequence within the protocol's time limit (Chiba City Black Clinic Case Registry, 2038). Both outcomes are assessed by the clinic that performed the repair and are not blinded.

Analysis. The score was a continuous outcome and was analysed by ordinary least-squares regression on delay (per 12 months), the severity proxy (per 10 days) and method (autologous graft as reference). Regained fine motor function was analysed by logistic regression with the same predictors. We tested the method term as a block (F-test in the linear model, likelihood-ratio test in the logistic model), tested a delay-by-method interaction and a squared delay term in the linear model, and described the score by tertiles of delay. Confidence intervals are 95% and all tests are two-sided. No corrections were made for multiple comparisons, and the estimates for method should be read as exploratory.

3. Results

The 142 patients were distributed across the three methods as 58 autologous grafts, 52 synthetic conduit splices and 32 hybrid splices (Table 1). The median delay from exposure to repair was 22 months (IQR 16–37; range 5–84). The severity proxy averaged 17.5 hospital days (SD 7.8). The mean 12-month recovery score was 46.1 (SD 14.2), and 32 patients (23%) had regained fine motor function.

In the linear model, delay carried the clearest association. Each additional 12 months from exposure to repair was associated with a 4.96-point lower score (95% CI 3.77 to 6.16 points; p < .001). Each additional ten days of acute-phase hospital care was associated with a 2.75-point lower score (95% CI 0.21 to 5.29; p = .034). The model explained 34% of the variance in score (adjusted R² 0.32; residual SD 11.7 points). Estimates for the conduit splice (−0.58 points against autologous graft, 95% CI −5.01 to 3.85) and the hybrid splice (0.18 points, 95% CI −4.96 to 5.32) were close to zero and imprecise, and the method term as a block was not significant (F(2, 137) = 0.05, p = .950).

Two checks bear on the form of the delay association. A delay-by-method interaction was not detected (F(2, 135) = 2.06, p = .131), so we found no evidence that any method was more or less vulnerable to delay, although the test has little power with groups of this size. A squared delay term was also not significant (estimate −0.30 points per squared year, p = .353). The tertile means, however, are not evenly spaced. Patients repaired within 16 months (n = 47) had a mean score of 51.4, those repaired at 17–31 months (n = 46) had 50.8, and those repaired at 32 months or later (n = 49) had 36.6.

In the logistic model, each additional 12 months of delay was associated with lower odds of regained fine motor function (OR 0.70, 95% CI 0.51–0.97; p = .030). The severity proxy gave an odds ratio of 0.63 per ten days (95% CI 0.37–1.08; p = .095), an interval that includes 1. Compared with autologous graft, the conduit splice had an odds ratio of 0.37 (95% CI 0.13–1.01; p = .053) and the hybrid splice 1.21 (95% CI 0.44–3.31). The likelihood-ratio test for method as a block gave p = .065. In the delay tertiles, fine motor function was regained by 28%, 30% and 10% of patients in order of increasing delay.

4. Discussion

Delay from exposure to repair was the predictor most consistently associated with both outcomes. Longer delay was associated with lower recovery scores and lower odds of regained fine motor function, and the direction held in both primary models. The linear model averages a pattern that the tertile means show to be uneven, so we do not extrapolate its slope to differences across the interquartile range.

The tertile means qualify the linear summary. Mean scores were similar in the two shorter-delay tertiles and lower in the longest, which would fit a loss of recovery concentrated at long delays and not spread evenly across the range. The squared term did not support curvature, and a cohort of this size cannot locate a threshold if one exists. We therefore claim only that longer delay was associated with poorer recovery across the observed range, and that the longest-delay patients did worst.

Severity, as measured by hospital days, was associated with a lower score but not decisively with regained fine motor function. Because hospital days are a coarse and partly administrative measure, the weaker finding in the binary model is unsurprising, and we do not read it as evidence that acute severity is irrelevant to dexterity.

No method showed a detectable advantage. This is an absence of evidence in a cohort where the smallest group held 32 patients, and the confidence intervals for the method contrasts are wide enough to include differences that would matter to a patient. The conduit splice had lower point odds of regained fine motor function than autologous graft, with an interval that came close to excluding equality, and this deserves a larger sample before it is dismissed. The hybrid splice uses non-autologous tissue, and an immunological analysis of bioware grafts found that graft fate was largely predictable from markers read at about the fourteenth week, and if hybrid splices behave likewise, early graft failure would show up in outcomes well before our 12-month endpoint. The registry holds no tolerance markers for the hybrid group, so we cannot say how many of its outcomes reflect early graft failure.

For referral practice, the findings point in one direction. If the association with delay reflects a causal loss of recoverable function, shortening the interval between exposure and repair offers a gain that no choice among the three methods in this registry is shown to match. If it reflects confounding by access or by injury type, it still identifies the patients, those who waited longest, who left the clinics with the least recovery.

5. Limitations

This is a retrospective review of registry dossiers, and every limitation of that design applies. Patients were referred by routes we cannot observe. Those who reached a clinic quickly may differ from those who arrived late in ways the registry does not record, including the extent of nerve damage, and the delay association may partly reflect such differences. We cannot separate the effect of waiting from the characteristics of those who waited.

Attribution to a wartime mycotoxin rests on clinical criteria applied by the contributing clinics. Some patients may have had injuries of a different origin that resembled the registry's pattern, and no independent toxicology was available. Exposure dates, taken from patient history, are uncertain by an unknown margin, and the delay measure inherits that uncertainty. The recovery score is a composite of peripheral tasks and may not register central conduction deficits that persist after peripheral repair (Kanzawa-Lind, 2039).

Both outcomes were assessed by the clinic that performed the repair, without blinding, and clinics may differ in how they score. We could not adjust for clinic, because several clinics contributed fewer than ten patients. The 29 patients without a 12-month assessment are a source of bias even though their registry metadata resembled those of the analytic cohort. Complete follow-up cannot be assumed to be missing at random, and patients with poor outcomes may have been more likely to leave the area.

The logistic model rests on 32 events across four estimated coefficients, about 8 events per predictor, and its estimates are imprecise and may be biased away from zero. The severity proxy is crude. Several comparisons were made without correction. A prospective cohort that records exposure date, baseline neurological examination and tolerance markers at referral would answer most of these questions more directly.