Dose–Duration Thresholds of Severe Melange Dependency and the Autonomic-First Course of Withdrawal: Arrakis Medical Corps Records of 519 Adults in Three Exposure Bands, 10150–10235 AG

1. Introduction

Melange extends life, sharpens awareness and produces addiction. Heavy users acquire the eyes of the Ibad, sclera and iris saturated to a uniform blue-within-blue, and withdrawal from high dependence is fatal. The glossary tradition of the Imperial period distinguishes two degrees of the effect: mild addiction at small quantities, and severe addiction above roughly two grams daily for an adult of seventy kilograms, about 29 mg/kg/day. That second figure is a dose with no time dimension attached. It says nothing about how long a dose must be sustained before dependency becomes severe, or how a dependent body fails when supply stops.

Prior work has concentrated on the most extreme and least accessible population. A histopathological survey of 41 Spacing Guild Navigators found that scleral pigmentation and dermal vascular change saturate by mid-career and that no sample showed tissue injury, which argues against a simple toxicity model of melange's effects⁸. Navigators, however, cannot inform a study of cessation. They live permanently in tanks of melange gas; no Guild record describes a Navigator withdrawn from spice, and the Guild would not permit one. The nearest comparable work on duration and dependency in the Imperial pharmacopoeia concerns sapho juice among Mentats, where dependency severity continues to rise with years of use after the cognitive benefit levels off⁹.

We write from about 10240 AG, some decades after the Jihad of Muad'Dib, when Medical Corps files from the late Harkonnen administration, the Atreides interval and the Jihad became available together. Two questions follow. How do mean daily dose and duration of use combine to determine the probability of severe dependency, and where does that probability cross one half? In what order do withdrawal signs appear, and which of them precede death? Our contribution is a fitted dose-duration threshold surface read against the glossary's severe-addiction dose, and a withdrawal sequence stratified by exposure band.

2. Methods

The case material is the Arrakis Medical Corps dependency and withdrawal series¹, whose accessions are case bundles rather than single documents, holding intake assessments, follow-up notes and, for supervised cessation episodes, hourly observation sheets. Most patients were assessed with no interruption of supply; only 67 bundles record a cessation episode. Daily dose was reconstructed from CHOAM ration and procurement ledgers for the Arrakis concession², which record melange issued per household or crew, divided by recorded body mass. Fremen intake was not ledgered; dose there was estimated from sietch ration practice in recorded testimony³, checked against the treating clinician's estimate. A few Imperial officers seconded to the Arrakis concession before 10191 AG appear with their own medical returns in the Salusa Secundus archives⁴, an archival finding of this study given the narrow period they cover.

Table 1 summarises the three exposure bands, defined before analysis. The light band comprised off-world residents of Arrakeen and Carthag — CHOAM factors and household staff — who took melange in food and drink. The moderate band comprised occupational users: Harkonnen and later Atreides administrative staff, House household troops, the seconded Imperial officers and spice-factory crews. The high chronic band comprised Fremen adults from sietch communities and a few veteran deep-desert crew, all with at least fifteen years of continuous use. Because melange extends life, this band is chronologically older without matching decline in function; chronological age was therefore tested as a covariate. Reverend Mothers were excluded; their passage through the spice agony is a one-way transformation of a different kind¹⁰.

Dependency severity was scored from the case files as the mean of three 0–10 subscales: tremor, anxiety during missed doses, and a metabolic subscale from resting heart-rate elevation and thermoregulatory instability, the autonomic markers earlier clinical work found most sensitive to chronic exposure⁶. The Corps mandated supervised supply at scores of 7 or above, and we adopted that cut-point for severe dependency. Scleral saturation was graded 0–3 on Imbrell's clinical scale⁵; the term eyes of the Ibad refers throughout to grade 3. Saturation was examined only in a separate within-band model, reported below, and kept out of the threshold model, since pigmentation lies downstream of exposure and adjusting for it would absorb part of the dose effect.

Severe dependency was modelled by binary logistic regression on mean daily dose D (mg/kg/day), duration T (years) and their product. The threshold surface is the contour on which the fitted probability equals 0.5. Calibration was assessed with the Hosmer–Lemeshow test across deciles of predicted risk, and a proportional-odds ordinal regression on the score in four bands (0–3, 4–6, 7–8, 9–10) served as sensitivity analysis. Cessation fatality was modelled separately, by logistic regression on each patient's fitted log-odds of severe dependency, which depends only on dose and duration and not on the severity score. Each of the 67 cessation episodes belongs to a different patient, none having a second recorded episode, so these observations are independent. Withdrawal-sign timing was taken from the 47 episodes with complete hourly records; signs were classed as autonomic (heart-rate elevation, thermoregulatory instability, hypotension) or cognitive (impaired working memory and recall on bedside tests, and anxiety). Proportions carry Clopper–Pearson 95% intervals and means carry t-based intervals.

3. Results

Severe dependency rose steeply across the bands: 12 of 143 light users (8.4%, 95% CI 4.4–14.2%), 143 of 287 moderate users (49.8%, 43.9–55.8%) and 87 of 89 in the high chronic band (97.8%, 92.1–99.7%). Grade-3 saturation was recorded in 81 of the 89 high-band patients. In a model fitted within the moderate band alone, each additional saturation grade raised the odds of severe dependency (OR 2.4, 95% CI 1.7–3.4), though grade and dose were correlated; this term was not carried into the threshold model.

The fitted model was logit(P) = −3.60 + 0.060D + 0.120T + 0.0030D·T. Table 2 gives standard errors and intervals; all three exposure terms are positive and none of their intervals includes zero: risk increases with dose, with duration, and more steeply with duration at higher doses. McFadden's pseudo-R² was 0.41. Mean predicted probabilities were 9.6%, 49.6% and 96.4% in the three bands against observed values of 8.4%, 49.8% and 97.8%, and the Hosmer–Lemeshow test showed no significant lack of fit (χ²(8) = 9.7, p = 0.29). Adding chronological age changed no coefficient by more than a tenth of its value, and the ordinal sensitivity model gave the same signs throughout.

The threshold duration follows as T* = (3.60 − 0.060D) / (0.120 + 0.0030D), for doses below 60 mg/kg/day. At 10 mg/kg/day the surface lies at 20.0 years; at 20, 13.3 years; at 30, 8.6 years; at 40, 5.0 years; at 50, 2.2 years. At 29 mg/kg/day, the dose above which the glossary places severe addiction, half of users reach severe dependency after about nine years. Above 60 mg/kg/day the model places even a new user beyond the threshold, though few records lie in that region and the estimate there is weak. The moderate band's mean position (30 mg/kg/day, 8.1 years) sits just inside the surface, matching its near-even split.

Almost all of the 67 cessation episodes followed involuntary loss of supply: transfer off-planet, the change of fief in 10191 AG, capture, or interrupted supply to Fremen legions off Arrakis. No death occurred among 8 light-band episodes. Three of 39 moderate-band episodes were fatal (7.7%, 95% CI 1.6–20.9%), as were 7 of 20 high-band episodes (35.0%, 15.4–59.2%). Measured against the surface, all ten deaths occurred among the 38 episodes lying beyond it (26.3%, 13.4–43.1%), against none of the 29 inside it (Fisher's exact p = 0.004). Each unit increase in fitted log-odds of severe dependency raised the odds of death by a factor of 1.9 (95% CI 1.2–3.1, p = 0.008).

In the 47 fully timed episodes, the first autonomic sign appeared at a mean of 4.2 hours after the last dose (SD 2.1; 95% CI 3.6–4.8) and the first cognitive sign at 18.3 hours (SD 6.4; 16.4–20.2). Heart-rate elevation was usually earliest, thermoregulatory instability following at a mean of 6.8 hours (SD 3.0; 5.9–7.7). Autonomic signs came first in 44 of 47 episodes (93.6%, 82.5–98.7%): 4 of 5 in the light band, 26 of 28 in the moderate band and 14 of 14 in the high band. The three cognitive-first episodes were non-fatal.

Autonomic severity scaled with exposure. Peak heart-rate elevation above baseline averaged 28 beats per minute (SD 11) in the light band, 43 (SD 15) in the moderate band and 67 (SD 18) in the high band across all 67 episodes (one-way ANOVA, F(2, 64) = 23.4, p < 0.001); for the 20 episodes without hourly sheets the peak came from intake and follow-up notes, which record the highest rate observed but not its onset. Thermoregulatory instability was recorded in 3 of 8, 31 of 39 and 19 of 20 episodes respectively. All ten fatal courses ended in autonomic collapse, with sustained tachycardia, loss of temperature control and terminal hypotension, and none showed cognitive impairment before loss of consciousness. Median time to death was 64 hours (range 38–118).

4. Discussion

This analysis does not confirm the glossary's severe-addiction dose so much as supply the dimension it lacks. At that dose, half of users reach severe dependency only after about nine years, and at half the dose after well over a decade; the figure is best read as a long-run threshold, the dose at which sustained use eventually produces severe dependency, not a level at which a new user is severely addicted. For warnings aimed at off-world personnel, cumulative exposure is the relevant quantity. A household at the light band's typical dose sits far from the surface; a factory crew member on the moderate ration crosses it within a normal tour, and the positive interaction means higher rations shorten that interval disproportionately.

Withdrawal behaves as an autonomic emergency. Control of heart rate and temperature fails first, cognitive signs follow some fourteen hours later, and death follows from the autonomic failure. Supervised management should monitor pulse and temperature from the first hours, not wait for confusion or anxiety. That every death fell beyond the threshold surface also gives clinicians a triage rule when supply is cut.

The Navigator histology survey found saturation without tissue injury⁸, and our fatal cases show no evidence of cumulative poisoning: no death here was attributed to exposure itself, and every fatal course followed cessation, though a design limited to cessation could not have shown otherwise. Both findings fit a dependency in which melange has become a regulatory requirement of the autonomic system. Navigators stand as the limiting case: their exposure lies far beyond the surface, and the Guild's practice of never interrupting supply⁷ treats cessation as unsurvivable.

5. Limitations

Doses are reconstructions: ledgers record melange issued, not consumed, and Fremen intake rests on testimony and clinical estimate, so misclassification would tend to flatten the fitted surface. The cessation series is small and shaped by survival, since patients who died before reaching Corps care, particularly Fremen legionaries off-world, are absent, and ten fatal cases give wide intervals. Few individuals were observed above 60 mg/kg/day, so the surface is least reliable where risk is highest. The composite severity score depends on what clinicians chose to record, and the cut-point of 7 is administrative in origin. Its anxiety and metabolic subscales draw on the same signs later used to class withdrawal as cognitive or autonomic, so score and classification are not wholly independent; the fatality analysis escapes that overlap, resting on log-odds from dose and duration alone. Finally, the timing analysis rests on 47 supervised episodes, which may understate the speed of withdrawal in unsupervised settings.