The NRC Wants to Change How You Release Therapy Patients. Here Is What Is Actually in the Draft.

On August 11, 2026, the NRC published Draft Regulatory Guide DG-8070, proposed Revision 2 to Regulatory Guide 8.39, "Release of Patients Administered Radioactive Material" (91 FR 51749, Docket NRC-2026-3730). The comment period closes at 11:59 p.m. Eastern on September 10, 2026.

Revision 1 has been the operating document since April 2020, and most of it traces back to the 1997 patient release rule and NUREG-1492. If you run a Lu-177 program, the release worksheet sitting in your radiation safety binder right now almost certainly rests on a table that never listed Lu-177 at all. This draft changes that, and it changes a good deal more besides.

I read both documents side by side. Below is what moved, why the numbers shifted the way they did, and where the draft still has rough edges worth commenting on.

First, the part that is easy to miss

DG-8070 is guidance written to match a rule that is not final yet.

The draft's Discussion section states that 10 CFR 35.75 "was revised" to incorporate the concept of a caregiver. The Federal Register notice announcing the draft says something more precise: DG-8070 aligns with proposed revisions to 35.75 contained in the proposed rule "Reforming and Modernizing the NRC's Radiation Protection Framework" (91 FR 43456, July 15, 2026), which responds to section 5(b) of Executive Order 14300.

Nothing in your release program changes until that rule is final. The tables in this draft are not usable today. This is also the second attempt at a Revision 2; the NRC put out DG-8061 and did not finalize it after commenters raised concerns about technical complexity and asked for more worked examples.

Two limits where there used to be one

Revision 1 applies a single limit: 5 mSv to any individual other than the patient.

The draft splits that population in two. Bystanders remain at 5 mSv. Caregivers , defined in the proposed 10 CFR 35.2 as an adult providing the patient support or comfort for non-commercial gain, may receive up to 50 mSv. Where caregiver dose is expected to exceed 5 mSv, both the patient and the caregiver must give written consent and must be instructed on radiation risks and on how to manage that exposure.

Evaluating the caregiver does not replace the bystander evaluation. The draft is explicit that a licensee must still demonstrate compliance for bystanders separately. Practically, identifying a spouse as a consenting caregiver removes the co-sleeping problem from the bystander calculation, which is where most therapy release math falls apart.

For patients who cannot be released, 10 CFR 20.1301(c) carries its own pair of numbers: 5 mSv for a visiting bystander, 20 mSv for a visiting caregiver.

The change that will actually cost you time

The limits in the proposed rule are stated per administration regimen, not per release.

An administration regimen is the full prescribed course of a radiopharmaceutical. Four infusions of Lu-177 dotatate are one regimen. Six infusions of Lu-177 vipivotide tetraxetan are one regimen. The draft directs licensees to consider the total activity administered across the entire regimen, or the combined dose rate contribution from all administrations.

In the worked examples, this means dividing the tabulated threshold by the number of planned administrations before comparing anything. Example B releases a Lutathera patient by taking the caregiver dose rate threshold of 0.22 mSv/h, dividing by four, and comparing the measured 0.020 mSv/h against 0.055 mSv/h. The instruction threshold gets divided too, which is how that patient ends up released with instructions rather than without them.

This is the item I would build into a worksheet first. It means release documentation now has to track cycle number, and it means a patient who was comfortably releasable on cycle one is being measured against a threshold one-sixth the size.

Why the numbers moved, and why some went down

The tables were not simply relaxed. Some thresholds rose by a factor of four. Others fell by a factor of three. Two separate changes are responsible, and they push in opposite directions.

Occupancy. Revision 1 used an occupancy factor of 0.25 at 1 m for radionuclides with a physical half-life longer than one day, and 1.0 for half-lives of one day or less, on the reasoning that short-lived activity delivers its dose during the window when contact is heaviest. Revision 2 drops that special case and applies 0.25 uniformly to the bystander tables. Every short-lived radionuclide therefore gains a factor of four.

Dose rate constants. Revision 1 used exposure rate constants in R/mCi-h at 1 cm, drawn from the 1970 Radiological Health Handbook, NUREG/CR-4444, and calculations in NUREG-1492, with 1 R taken as 10 mSv. Revision 2 replaces these with dose rate constants in mSv/GBq-h at 1 m, computed from ICRP Publication 107 decay data and including photon and electron emissions with bremsstrahlung. Where the older constant understated photon output, the new threshold drops.

You can see both effects cleanly in the table.

Rev. 1 values from RG 8.39 Rev. 1 Table 1, column 1. Rev. 2 values from DG-8070 Tables 1 and 3, column 1. The Se-75 caregiver entry is discussed under errata below.

Tc-99m and I-123 are pure occupancy effects. Both rise by almost exactly four.

Tl-201 is the clearest example of the other mechanism. Converting Revision 1's exposure rate constant of 0.447 R/mCi-h at 1 cm into the new units gives roughly 0.0121 mSv/GBq-h at 1 m. The draft's Appendix A lists 0.0405, about 3.4 times higher. Revision 1's 16 GBq divided by 3.4 is 4.8 GBq, which is the number in the draft. Se-75 behaves the same way, with the new constant about 2.8 times the old one.

I-131 barely moves, because its old and new constants agree to within a few percent and its half-life was already over a day. That is worth knowing if you have been assuming this revision loosens everything.

Lu-177 is finally in the tables

Revision 1's Table 1 lists 24 entries. It contains no Lu-177, no Y-90, no Ra-223, no F-18, no Ga-68, and no Zr-89. Every program running radioligand therapy has been deriving its own thresholds or releasing on measured dose rate and keeping a record each time.

The draft's tables run to roughly 50 entries and include Lu-177, Ra-223, At-211, Bi-213, Cu-67, Ho-166, I-124, Sc-47, Sm-153, Y-90, Zr-89, F-18, Ga-68, and additional implant forms.

Alpha emitters are handled by exclusion. Ac-225, Th-227, Pb-212, and Ra-224 are omitted from every threshold table, with a note that activity and dose rate limits are not applicable because of the minimal external exposure from the microcurie dosages typically administered. Ra-223 is included, since it has a photon component worth tabulating. If you are standing up an alpha therapy program, the external release tables will not be your constraint, but you still owe a written basis.

Patient-specific calculations get a real framework

Revision 1 gave you effective half-life, three fixed occupancy values (0.75, 0.25, 0.125), a two-component biokinetic model specific to I-131, and tissue shielding.

Revision 2 replaces that with four modifying factors applied to the basic dose equation: biokinetics, occupancy, geometry, and attenuation. Appendix B provides methods for each, including a table of geometry factors at defined separation distances and a second table keyed to standard US bed sizes. A queen bed is treated as 0.7 m of separation. A twin is 0.3 m.

Appendix C handles exposure that changes over time, so you can weight a two-week instructional period differently from the behavior that follows it. Appendix D provides a graphical template for deriving a biokinetic factor from measured patient retention. Appendix E works through eight examples, including screening criteria for Lu-177 dotatate and Lu-177 vipivotide tetraxetan by name. The draft also adds a hold time equation, which lets you tell a patient before administration roughly how long they will need to stay.

One caution the draft states plainly and which is easy to violate: the four factors are conservative only as a set. If you use real patient information to relax occupancy or biokinetics, you can no longer assume unity for geometry and attenuation without verifying it. Setting the geometry factor to 1 while claiming a 0.25 occupancy for a patient who co-sleeps in a twin bed will underestimate dose.

Breastfeeding guidance changes materially

Revision 1's Table 3 combines instruction thresholds, record thresholds, and example interruption durations, in MBq, and resolves several agents to "complete cessation for this infant or child."

Revision 2 splits this into Table 7 (thresholds, in GBq) and Table 8 (interruption times), and adds equations for effective half-life and interruption time so you can calculate rather than look up. Several of the cessation recommendations become finite numbers.

• Lu-177 octreotate: Revision 1 recommends complete cessation. The draft gives 350 hours for 5 mSv and 540 hours for 1 mSv at 7.8 GBq.
• Ra-223 dichloride: Revision 1 groups it with all alpha emitters under complete cessation. The draft gives 1,400 and 1,700 hours at 0.10 mCi.
• Zr-89: 28 days becomes 140 and 270 hours.
• I-131 sodium iodide: complete cessation becomes 1,000 to 1,900 hours depending on indication and activity.
• F-18 FDG: 4 hours becomes a calculated 3 hours at 0.74 GBq.
• Ga-68 octreotate: 4 hours becomes no interruption required at 0.19 GBq.

Recordkeeping shrinks considerably

Revision 1 requires a three-year, patient-level record whenever release is based on retained activity, measured dose rate, an occupancy factor below 0.25, effective half-life, or tissue shielding. Because most therapy patients are released on a measured dose rate, that has meant a record on essentially every therapy release.

The draft's recordkeeping section requires two things. The written release procedure, and every superseded version of it, retained for the duration of the license. And the authorized user's justification for caregiver exposure when visiting a patient who cannot be released, retained three years.

Routine patient-level release calculation records are not clearly required anywhere in that section. If this holds in the final version, it is a meaningful reduction in administrative burden for a community program. It is also the provision I would read most carefully before assuming it, because the draft contradicts itself on it.

A new section on sources separated from the patient

The draft adds a dedicated section for implants and seeds that become detached. The framing is that once a source leaves the patient, ordinary 10 CFR Part 20 public dose limits apply rather than the patient release limits.

It covers verifying attachment before release, labeling and recovery instructions, returning temporary implants, reporting lost sources under 10 CFR 20.2201, and determining under 10 CFR 35.41 whether source movement constitutes a medical event.

Rough edges worth putting in a comment

The draft is provisional and reads like it. The title page still says "Month 2026." Beyond that, a few errors are substantive enough that anyone building from these tables should know about them.

Numerical. The Se-75 caregiver threshold is listed as 0.80 GBq (2.2 mCi) in Tables 3 and 5. Those two values differ by a factor of ten; 2.2 mCi is 0.081 GBq. The Xe-133 caregiver threshold in Table 3 is 12 GBq (320 mCi), identical to the Tl-201 row directly above it, and inconsistent with both the 2.5x relationship that holds for every other radionuclide and with Table 6's own Xe-133 entries. The Ru-106 instruction threshold in Table 1 reads 14 GBq (3,900 mCi), where 3,900 mCi is 144 GBq.

Worked examples. Example F.2, the Lu-177 dotatate screening case, contains the sentence: the baseline threshold from table 5 column 1 is 4.1 GBq, therefore the licensee uses 1.9 GBq (4.7 GBq divided by 4 equals 1.2 GBq). Three different numbers appear where one should. The calculation that follows uses 1.2 GBq. The same example later concludes that "275 mCi is greater than the administered activity of 41 GBq," comparing a per-cycle threshold against a tabulated total.

Internal consistency. Table 7 labels its 5 mSv breastfeeding column as the activity "requiring a record," while the recordkeeping section imposes no such requirement. This is the single inconsistency I would most want resolved, because it sits directly on the recordkeeping relief described above. Section 4 points readers to "section 5.1" for recordkeeping guidance; recordkeeping is section 7. The Overview cites 10 CFR 37.5(a) for the written release procedure, which should be 35.75(a). Section 3 repeatedly cross-references Tables 1 and 2 where Tables 3 and 4 are intended, and states in one place that the demonstration is that bystander dose will not exceed 50 mSv. Appendix B contains two sections numbered B.4. The breastfeeding example in Appendix E is captioned Figure C-1 and cites Table 3 columns for values that live in Table 7.

None of these are fatal. All of them are the kind of thing that gets fixed when someone who works with the tables daily says so in the docket.

What I would do between now and September 10

Read Appendix E first. The eight worked examples tell you more about how the NRC expects this to function than the body text does, and two of them are Lu-177 therapy cases with named drugs and real activities.

Then take one recent multi-cycle therapy patient and run them through the draft twice, once as a bystander-only release and once with a consenting caregiver. That exercise surfaces the per-regimen division problem faster than reading about it, and it tells you whether your current worksheet can absorb the change or needs rebuilding.

Post your comments. We would love to hear from you.

---