Uncited Press › Star Trek Open the interactive journal →
Trek · Cybernetic & Bioware Augmentation

Removal or Retention After the Collective Lets Go: Nanoprobe Activity, Ninety-Day Survival and Function in 52 Severed Borg Drones, by Duration of Assimilation

Dr. Anjali Varma-Lindqvist1
1 Starfleet Medical, Exobiology Division, San Francisco
Received 11 Aug 2026 · Revised 29 Sep 2026 · Accepted 24 Oct 2026

Abstract

Drones severed from the Collective carry implants that nanoprobes continue to remodel, and surgeons have no estimate of how long assimilation can last before removing every implant becomes more dangerous than leaving some in place. We report a retrospective series of 52 severed drones from the Starfleet Medical record, written in 2379 from Starfleet Medical cases since 2367, Voyager's Delta Quadrant records and cases admitted after the events of 2378. Patients were grouped by duration of assimilation, and we recorded nanoprobe activity, death within 90 days of the first removal procedure and function at day 90 on a 100-point Independence Score. Seven patients died (13.5%; 95% CI 6.7 to 25.3), and mortality rose from 0 of 12 under a year to 4 of 13 at fifteen years or more (trend p = .021). From five years, 5 of 12 patients with a plan of full removal died (41.7%) against 1 of 15 with a staged plan that left some implants in place (6.7%), a difference of 35.0 percentage points (95% CI 2.8 to 61.9; exact p = .060). From five years, survivors of staged plans scored about 10 points lower on function after adjusting for duration group. Surgeons chose the plan, and calendar time and plan cannot be separated. Full removal may cost more lives than it returns to independence at some point beyond five years, most clearly at fifteen years or more, where only 4 patients had it, and the series cannot locate the point more closely.

1. Introduction

Of the 52 severed drones in the series reported here, 7 died within 90 days of their first removal procedure, and 6 of the 7 had been assimilated for five years or longer. Until 2378 a Starfleet surgeon would meet such a patient rarely, and most of the published cases concerned people held by the Collective for days or months, whose implants came out whole. Many of the patients admitted since then, according to the fleet's own logs, had served for decades. The operating team must decide at the first procedure how much of the hardware to take out. Beyond what length of assimilation does an attempt to take out every implant cost more lives than it returns to independence?

Answers fall into two broad classes. In the first, each year of assimilation adds roughly the same increment of operative risk, such that caution should grow gradually and no single duration should change the plan. In the second, risk stays low until the body has handed a circulatory, respiratory or neural task to an implant, and climbs steeply once it has. A surgeon who favours the first will plan a long staged programme for every patient; one who favours the second will operate boldly below the threshold and decline above it.

The two best-documented cases sit at opposite ends of the range. Captain Picard was assimilated for a matter of days across 2366 and 2367, and his implants were removed aboard the Enterprise, and he recovered physically (Starfleet Medical, 2367), although he could still sense the Collective in 2373. Seven of Nine was assimilated in childhood and served as a drone for about eighteen years before she was severed in 2374; Voyager's physician removed most of her implants but left a cortical node, and she requires a regeneration alcove (Starfleet Delta Quadrant Mission Records, 2378).

Neither case settles the matter.

Anatomical work has since described the principal implants, namely the cortical node, the ocular and cardiorespiratory assemblies, and the limb and dermal components (Daystrom Institute Cybernetics Division, 2377). Histological work has shown that assimilation tubules leave scars in which nanoprobes persist (Ibarra-Quennell, 2374), and a follow-up of a single patient recorded probe activity for several weeks after severance (Okonjo-Halloran, 2370). No series has set outcome against the time a person spent assimilated.

We write in 2379. The series draws on Starfleet Medical cases from 2367, on Voyager's Delta Quadrant records and on cases admitted after the events of 2378 that disrupted the Collective, and no source later than the spring of 2379 is used. We describe nanoprobe activity, survival and function by duration of assimilation, and we compare a plan of full removal with a staged plan that leaves some implants in place. We write 'dependence' (in single quotation marks throughout) for an implant that the operating team judged to be carrying a circulatory, respiratory or neural task, since no scan in the series could confirm that judgement before the implant was removed.

2. Methods

The series is the Starfleet Medical record of severed drones admitted for assessment or treatment between 2367 and 28 February 2379. Sixty-one individuals were admitted. We excluded 5 whose duration of assimilation could not be fixed to within a year. We also excluded 4 who died within 72 hours of admission, before any removal procedure was attempted, leaving 52 patients. Records closed on 31 May 2379, and every patient had 90 days of follow-up from the first procedure (the last first procedure took place on 2 March 2379). Twelve patients were treated before Voyager's return: one in 2367, six admitted between 2370 and 2377, and five whose care by the ship's physician is documented in the ship's Delta Quadrant records (Seven of Nine and the four children). The other 40 were admitted from late 2378. The fleet admission logs that we use, our only source for these recoveries, record them as taken by Starfleet vessels from disabled cubes. Thirty-one patients (59.6%) were Human, and the remaining 21 belonged to nine other species.

Duration of assimilation was the time from the documented date of assimilation to the date of severance. A date was documented for 21 patients, from recovered cube logs or from the patient's own account. For the other 31 we estimated it from the maturity of tubule scars and of the tissue around the implants, with an uncertainty that we put at about one year. We grouped patients as under 1 year (12 patients), 1 to under 5 years (13), 5 to under 15 years (14) and 15 years or more (13). The boundaries were set from the spread of durations before outcomes were tabulated.

The operating team recorded its plan at the first procedure. A full-removal plan intended to take out every implant within one admission. A staged plan removed first the implants not in 'dependence' and deferred or declined the rest.

For the 40 patients admitted from late 2378 the choice rested with the operating team under the interim protocol (Starfleet Medical, 2378), and for the 12 earlier patients with the treating physician. Allocation was therefore not random.

Death within 90 days of the first removal procedure was the primary outcome. Nanoprobe activity was assessed by a weekly scan tuned to the probe signature, and a patient was called quiescent after two consecutive weekly scans showed none, and time to quiescence ran from the first removal procedure to the first of those two clear scans. Function at day 90 was scored on the Independence Score, a 0 to 100 rehabilitation measure (Sathyanarayan, 2362). We defined recovery as survival to day 90 with a score of 70 or more, a level that the measure's developers describe as independent living without aid.

Proportions carry score intervals (Teague-Moorcroft, 2374). We tested the rise in mortality and in persisting activity across the four ordered duration groups with a chi-square test for linear trend, scoring the groups 0 to 3. Contrasts between plans used an exact test for two-by-two tables and a score-based interval for the difference in risk. Scores among survivors were compared with an unequal-variance t-test, and because that comparison conditions on survival we treat it as descriptive. We also compared scores within the 5 to under 15 year group and fitted a linear model of score on plan with an indicator for 15 years or more; both are descriptive. None of the contrasts was specified in advance, and we made no correction for multiple comparisons.

3. Results

Seven of the 52 patients died (13.5%; 95% CI 6.7 to 25.3). Mortality rose with duration: 0 of 12 under a year (0%; 95% CI 0.0 to 24.2), 1 of 13 at 1 to under 5 years (7.7%; 1.4 to 33.3), 2 of 14 at 5 to under 15 years (14.3%; 4.0 to 39.9) and 4 of 13 at 15 years or more (30.8%; 12.7 to 57.6) (Table 1). The chi-square test for linear trend gave 5.30 (1 degree of freedom, p = .021). All seven deaths were among patients admitted from late 2378, and none of the five patients treated aboard Voyager died.

Full removal was planned in 35 patients and a staged plan in 17. Six of the 35 full-removal patients died (17.1%) against 1 of 17 staged patients (5.9%). The crude contrast, however, hides the pattern by duration (Table 2). Below 5 years, full removal was planned in 23 patients and 1 died (4.3%; 95% CI 0.8 to 21.0), and both staged patients survived. From 5 years, full removal was planned in 12 patients and 5 died (41.7%; 19.3 to 68.0), against 1 of 15 staged patients (6.7%; 1.2 to 29.8). The difference in risk was 35.0 percentage points (95% CI 2.8 to 61.9; exact test p = .060), and the interval only just excludes zero, whereas the exact test, which is the more conservative, does not.

To set deaths against benefit we counted recoveries. Below 5 years, 19 of the 23 patients with a full-removal plan (82.6%) were alive with a score of 70 or more at day 90. From 5 years the figure was 4 of 12 (33.3%) with full removal and 1 of 15 (6.7%) with a staged plan (exact test p = .139). From 5 years the excess of deaths with full removal (35.0 percentage points) was larger than its excess of recoveries (26.7 points; 95% CI −3.6 to 54.8). At 5 to under 15 years the 8 full-removal patients produced 2 deaths and 3 recoveries, and at 15 years or more the 4 produced 3 deaths and 1 recovery. The 9 staged patients at 15 years or more produced 1 death and no recovery; Seven of Nine belongs to this group and survived. The three deaths after full removal at 15 years or more all occurred in the first nine weeks of the post-2378 admissions, as did one further death after full removal. Of the 12 full-removal patients from 5 years, 10 began treatment in that period, and 4 of them died. After those nine weeks, 12 of the 14 patients from 5 years who began treatment had a staged plan, and the operating notes gave frailty as the reason for a staged plan in 11 of the 17 staged patients.

Function fell with duration among survivors.

The mean Independence Score was 84.2 for the 12 survivors under a year, 72.7 for 12 at 1 to under 5 years, 63.8 for 12 at 5 to under 15 years and 56.9 for 9 at 15 years or more. Among the 21 survivors from 5 years, the 7 who had undergone full removal scored a mean of 68.43 (SD 7.2) and the 14 with a staged plan scored 57.07 (SD 8.8), a difference of 11.4 points (95% CI 3.7 to 19.0; p = .006). Of the 7 full-removal survivors, 6 were at 5 to under 15 years, against 6 of the 14 staged survivors. Had scores depended on duration group alone, this mix would by itself produce a gap of about 3 points. Within the 5 to under 15 year group the difference was 8.0 points (95% CI −2.2 to 18.2). A model that adjusted for the two duration groups gave 10.1 points (95% CI 1.1 to 19.0; p = .029).

Nanoprobes were detectable at day 90 in 2 of 12 survivors under a year (16.7%), 4 of 12 at 1 to under 5 years (33.3%), 6 of 12 at 5 to under 15 years (50.0%) and 7 of 9 at 15 years or more (77.8%), a rise with a chi-square for trend of 8.40 (p = .004; descriptive, since it conditions on survival). Among the 26 survivors who reached quiescence, the median time to it was 35 days under a year (10 patients), 49 days at 1 to under 5 years (8) and 63 days at 5 to under 15 years (6); the two at 15 years or more reached it at 63 and 84 days. Those times count the first of the two clear scans. The second could fall just after day 90. These medians leave out patients active at day 90, and they therefore understate the delay in the longer groups.

4. Discussion

Mortality after implant removal rose with the time a person had spent assimilated, and from about five years a plan of full removal carried a risk of death (41.7%) that the staged plan largely did not (6.7%). Full removal may therefore cost more lives than it returns to independence at some point beyond five years. The pooled excess of deaths (35.0 points) exceeds the excess of recoveries (26.7 points), but by an amount the data cannot distinguish from zero. From 5 years it produced 5 deaths and 4 recoveries, and at 15 years or more 3 deaths and 1 recovery, although only 4 patients were treated that way.

The pattern fits neither class of answer cleanly. A rise of 0%, 7.7%, 14.3% and 30.8% across the four groups looks like steady accumulation. However, the deaths after full removal cluster beyond five years, which is closer to a threshold. The series cannot choose between them, because four ordered groups give too few points to tell a smooth curve from a step, and the trend test does not attempt it. The balance appears to have tipped by fifteen years, where full removal produced 3 deaths and 1 recovery in 4 patients. Between 5 and 15 years it produced 2 deaths and 3 recoveries in 8 patients, against no deaths and 1 recovery in 6 staged patients, which is too close to call, and the data do not narrow the point further.

An alternative explanation is that the difference lies in the period and not in the plan. The three deaths after full removal at 15 years or more all fell in the first nine weeks of the post-2378 admissions, before surgeons had experience of these patients, and all seven deaths were among the 40 patients admitted from late 2378. The selection that we can document runs against the plan effect (teams kept staged plans for many of the patients they judged fragile, a choice that should have made the staged group fare worse). That argument concerns case mix, however, and says nothing about experience. Calendar time and plan are not separated in this series (too small to try), and the period could account for much of the contrast.

A staged plan is not free.

Survivors from 5 years who kept some implants scored about 11 points lower than those who underwent full removal, about 3 points of which the duration mix alone would produce. A patient kept alive by a retained cortical node needs a regeneration alcove (akin to a hull that holds its shape only while its structural integrity field is powered), a burden that a day-90 score may understate. Voyager's medical log records that the retained node of Seven of Nine failed in 2377 and was replaced with one donated by one of the children (Starfleet Delta Quadrant Mission Records, 2378), a case in which a retained implant carried a vital function. Whether the lower function is acceptable is a judgement for the patient. The comparison conditions on survival, and the 7 survivors of full removal are those who had the best chance of doing well.

Persisting nanoprobe activity followed the same gradient as mortality and function. Activity was present at day 90 in 77.8% of survivors at 15 years or more and 16.7% of those under a year. We do not know whether continuing activity contributes to the lower function or marks the same tissue change, and the series cannot say which, though a study that scanned patients before and after each procedure could.

5. Limitations

The series is small, and its intervals are wide; the interval for the difference between plans from 5 years begins at 2.8 percentage points. Allocation was by the surgeon, and the contrasts were not specified in advance or corrected for multiplicity, and the p-values are therefore descriptive. Duration was estimated for 31 of the 52 patients, with an uncertainty near one year, and a patient close to a boundary at 1, 5 or 15 years may sit in the wrong group, and the boundary at one year, no wider than the uncertainty itself, is the least secure. The 40 patients admitted from late 2378 were recovered from disabled cubes according to the logs, and we do not know how long each had been severed before recovery, and their first scans may not show the state at severance. Patients of ten species are pooled, the four children are not analysed separately, and follow-up ends at day 90, which cannot capture later deaths or later gains in function. A prospective register that records the plan, the duration and the calendar date of every first procedure from admission would show whether the plan or the period accounts for the difference.

We hope that grouping patients by the years a device has been carried will help in a second setting, the replacement of VISORs with ocular implants in adults who wore the older device for most of their lives and whose temple connectors now sit in remodelled bone. Their surgeons face a choice of the same form, between taking out every connector and leaving some in place.

Borg assimilationsevered dronesimplant removalnanoprobescortical nodesurgical survival

References

  1. Starfleet Medical (2367). Clinical course and implant removal in Captain Jean-Luc Picard after severance from the Collective. Starfleet Medical Records Archive, record SMR-67-31.
  2. Starfleet Delta Quadrant Mission Records (2378). Medical log of USS Voyager, removal of Borg implants from a severed drone and from four Borg children. Starfleet Delta Quadrant Mission Records, NCC-74656, medical series.
  3. Daystrom Institute Cybernetics Division (2377). Structure and function of Borg cortical, ocular and cardiorespiratory implants, a survey of recovered material. Daystrom Institute Cybernetics Division Reports, DCR-77-09.
  4. Starfleet Medical (2378). Interim protocol for the admission and surgical management of severed Borg drones. Starfleet Medical Records Archive, SM-P-78-03.
  5. Ibarra-Quennell, D. (2374). Assimilation tubule scarring and the persistence of nanoprobes in Human tissue, a histological account. Federation Journal of Xenomedicine, 62(1), 31–58.
  6. Okonjo-Halloran, T. (2370). Residual nanoprobe activity and tissue remodelling after severance from the Collective, a single-patient follow-up. Starfleet Medical Journal, 93(2), 45–63.
  7. Sathyanarayan, R. (2362). The Independence Score, development and validation of a rehabilitation measure after major prosthetic surgery. Journal of Federation Prosthetic Medicine, 10(1), 12–34.
  8. Teague-Moorcroft, L. (2374). Exact tests and score intervals for small two-by-two tables in case series. Proceedings of Applied Speculative Statistics, 22(3), 140–163.
Read this article inside the full journal experience — browse by faculty, search across universes, and filter by discipline.
Open in Uncited Press →