Real-world evidence study · Relapsed and refractory multiple myeloma · cilta-cel
A trial result that held up outside the trial.
1,300 patients treated with ciltacabtagene autoleucel in routine care, read against CARTITUDE-4 and CARTITUDE-1. A broader, higher-risk population reached the same depth of response, with the same manageable safety profile.
- 1,300
- patients treated with cilta-cel
- 92.7%
- overall response rate
- 71%
- complete response or better
- 84%
- alive at 24 months
- Networks
- Lambda + Omega
- Confirmed cohort
- ~17,500 relapsed / refractory
- Benchmark
- CARTITUDE-4 and CARTITUDE-1
- Report date
- 23 July 2026
How we got there
The question
Does the CARTITUDE result survive contact with routine practice, where the population is broader and no protocol selects the patients?
Everything below is the path to an answer: first what cilta-cel is and who receives it, then what the two pivotal trials each measured, then what 1,300 treated patients show across two independent data networks. The answer arrives near the end, in the same words as the question. Lambda and Omega combined · report date 23 July 2026 · benchmark NCT04181827.
Start here · the therapy
A one-time infusion of the patient's own re-engineered immune cells
Multiple myeloma is a cancer of plasma cells in the bone marrow. It responds to treatment, then returns, and each return is harder to control. Cilta-cel is given once, late in that sequence, to patients who have already exhausted the standard drug classes.
-
BCMA
The address on the cancer cell
A protein carried on the surface of myeloma cells and almost nowhere else. It is what the engineered cells are built to find.
-
CAR-T
Collected, re-engineered, returned once
The patient's T cells are collected, given a receptor that recognizes BCMA, grown in a facility, and returned as a single infusion. Bridging therapy holds the disease during the wait.
-
The trade
Deep remissions, paid for with two early risks
Cytokine release syndrome, a systemic inflammatory reaction, and neurological effects ranging from confusion to movement disorders. Both are why the therapy is delivered only at certified centers.
The pathway in order: cells collected · bridging therapy while the cells are manufactured · lymphodepletion · one infusion · no further myeloma therapy for as long as it holds.
The open question · two pivotal trials
The two trials answer for two very different populations
CARTITUDE-1 treated patients at the end of the line, after a median of six prior regimens. CARTITUDE-4 moved cilta-cel much earlier, to patients with one to three prior lines. Both reported strong results. Neither describes the mix of patients a certified center actually sees.
-
Earlier lines
CARTITUDE-4
Lenalidomide-refractory disease after one to three prior lines. Response rate 84.6%, complete response or better 73.1%. Median progression-free survival not reached, against 11.8 months on standard care.
Phase 3, cilta-cel versus standard of care · n = 208 in the cilta-cel arm · NCT04181827
-
Late lines
CARTITUDE-1
Three or more prior lines and exposure to all three drug classes. Response rate about 98%, stringent complete response about 83%. Two-year progression-free survival about 61%.
Phase 1b/2, single arm, heavily pretreated · n = 97
Together the two trials cover 305 patients, every one of them protocol-selected. Routine practice sits somewhere between them, and nobody had measured it at scale.
The benchmark · side by side
How does one real-world cohort compare against both trials at once?
1,300 treated patients — more than four times CARTITUDE-4 and CARTITUDE-1 combined, with more prior therapy than one and far higher cytogenetic risk than the other.
Read the prior lines and cytogenetics rows first
They are what establish that this population is not an easier one. Median prior lines sits above CARTITUDE-4, and high-risk cytogenetics sits at more than double CARTITUDE-1.
Everything after that is measured against a group the trials did not enroll.
| Trio RWE · cilta-cel | CARTITUDE-4 | CARTITUDE-1 | |
|---|---|---|---|
| Patients | 1,300 | 208 | 97 |
| Median prior lines of therapy | 3 | 2 | 6 |
| High-risk cytogenetics | 53% | ~60% | ~24% |
| Overall response rate | 92.7% | 84.6% | ~98% |
| Complete response or better | 71.0% | 73.1% | ~83% sCR |
| Cytokine release syndrome, any / grade 3+ | 67.5% / 0.8% | 76% / 1% | 95% / 4% |
| Neurotoxicity, any | 29.6% | 21% | 21% |
| Median overall survival | not reached | not reached | not reached |
| Survival at last reported timepoint | 84% at 24 mo | 76.4% at 30 mo | 74% at 24 mo |
Real-world percentages are calculated on the patients evaluable for each variable, and every denominator is stated with its figure below. Trial values are as published.
Trial figures are published results · N Engl J Med 2023 and Lancet 2021
Building the cohort · curation
Who counts as a cilta-cel patient, and what can actually be measured on them?
Every patient starts from a documented infusion, not a claim. From there each variable narrows the population differently — a clinician records a cytokine release grade far more reliably than a formal response assessment.
Entry is an infusion that was written down
The index date is the first recorded cilta-cel administration. What can be said about those patients then depends on what else reached the note, which is why the denominators below are not all 1,300.
Response is the narrowest of them, and the report says so everywhere it is quoted.
-
Documented cilta-cel administration
1,300
-
With clinical notes available
1,193
-
Assessable for cytokine release syndrome
1,148
-
Documented IMWG response assessment
397
Response percentages in this study are calculated on the 397 patients with a documented IMWG assessment, and safety percentages on the patients assessable for that event. The cohort is 4.3 times CARTITUDE-4 and CARTITUDE-1 combined, which is what makes the safety percentages stable.
Structured administrations · note availability per patient
Coverage reported per variable, not averaged
An average would hide the one that matters. Cytokine release grade and cytogenetics are documented for virtually everyone; disease-stage and marrow-residual fields are not.
A reader can see exactly which fields carry the weight of a claim.
-
Neurotoxicity
99.5%
-
Cytokine release syndrome
99%
-
Cytogenetics
99%
-
Response and bridging
93%
-
Minimal residual disease
79%
-
Extramedullary disease
69%
-
ISS stage
63%
Note-extracted · AI extraction with clinical QC
Every value comes from a sentence a clinician wrote
A cytokine release grade lives in a progress note as shorthand, alongside the drug that was given and the date it resolved. Fifteen variables were defined before extraction, then rolled up and spot-reviewed by the clinical team.
Cytokine release syndrome: Max grade 2 (12/18). - s/p Toci #1 + 10 mg of IV Dex (12/18)
- CRS grade 2
- CRS management Tocilizumab plus steroids
- Event date 12/18
- Index date and best response From their own lines
One sentence of clinical shorthand, three values. The index date and the best-response assessment come from their own lines in the record — nothing here is inferred from this sentence beyond what it says.
Verbatim note text, anonymized · AI extraction with clinical QC
Who these patients are · 1,193 people
A high-risk population, one dot per patient
More heavily treated than CARTITUDE-4 and far higher risk than CARTITUDE-1 — which is the context every result further down has to be read against.
carries this feature does not, or not documented
-
52.6%
High-risk cytogenetics
627 patients — more than double the proportion in CARTITUDE-1 and close to CARTITUDE-4. Cytogenetics were documented for 99% of the cohort, which is why this figure carries weight.
-
68.1%
Received bridging therapy
813 patients needed treatment to hold the disease between collection and infusion. Bridging is the part of the CAR-T pathway trials describe least, and the part a center plans around most.
-
47.1%
Lenalidomide-refractory
562 patients — and a floor rather than a rate. Prior therapy given outside the network is invisible, so every exposure figure in this study understates the truth.
-
36.5%
Extramedullary disease
436 patients had myeloma outside the bone marrow, a marker of aggressive disease. Median prior lines was three, and 487 patients had already had three or more.
Shares are of the 1,193 patients with clinical notes available, not of all 1,300 treated. Cilta-cel is delivered at certified centers and many patients are referred in, so prior-therapy and refractory figures are floors rather than rates.
The finding · depth of response
Did the depth of response hold up?
Seven in ten reached a complete response — essentially identical to CARTITUDE-4, in a population with more prior therapy.
Depth is what matters here, not just response
Among the 397 patients with a documented IMWG assessment, 368 responded and 282 reached a complete response or better.
A response rate alone would flatter almost any myeloma therapy. The share reaching a stringent complete response is the number that separates them.
-
Stringent complete response
168
-
Complete response
114
-
Very good partial response
73
-
Partial response
13
-
Stable or progressive disease
29
Minimal residual disease negativity was documented in 487 patients across the cohort, on the 79% of records where an MRD result appeared at all.
Note-extracted best response (IMWG) · AI extraction with clinical QC
Against the trials, on the endpoint that counts
Complete response or better, real world against both trials. The comparison is directional — response here is read from clinician documentation rather than central IMWG adjudication — but the direction is not in doubt.
-
71.0%
Real world, of 397 with a documented assessment
-
73.1%
CARTITUDE-4, complete response or better
-
~83%
CARTITUDE-1, stringent complete response
Read against a median of three prior lines and 53% high-risk cytogenetics, which is a harder population than CARTITUDE-4 enrolled.
Trial figures are published results
Durability · 969 patients, 95 deaths
Two years on, where are these patients?
Most are alive, and most are still off therapy. Median survival was not reached; 84% were alive at 24 months and 69% had started no further myeloma treatment.
The second curve is the one patients feel
Anchored at the infusion date. Overall survival is the conventional endpoint, but time to next therapy is what a patient experiences as remission — no infusions, no clinic cycles, no new regimen.
Read the gap above the trial curves as an artefact, not a result. The reason is in the caption.
- Real world, overall survival · 84% at 24 mo
- Real world, therapy-free · 69% at 24 mo
- CARTITUDE-4 survival · 76.4% at 30 mo
- CARTITUDE-1 survival · 74% at 24 mo
⚠️ The vertical axis starts at 50%, not zero — every curve sits high and a full axis would flatten them into one band, but that also exaggerates the gaps, so read the landmark percentages rather than the spacing. Real-world survival is censored at last recorded network activity, so a patient who moved their care elsewhere appears alive; cilta-cel is delivered at certified centers with many patients referred in, which makes that leakage likelier here than in most studies. Trial curves are drawn from published landmark rates, so their shape between points is indicative and CARTITUDE-4 is interpolated back from its 30-month figure.
Structured death records · Kaplan–Meier, time zero at infusion
The landmarks, stated plainly
Time to next therapy is a proxy for progression-free survival, not an IMWG assessment. Median was not reached for either curve, which reflects limited follow-up as much as an absence of events.
-
92%
95% at 6 months
Alive at 12 months
-
84%
Alive at 24 months
-
69%
82% at 12 months
Therapy-free at 24 months
Structured death records and treatment dates
Safety · the two CAR-T-specific risks
How often does the therapy hurt people, and how badly?
Common, almost always low grade, and managed the same way everywhere. Nine patients in 1,148 reached grade 3 or above for cytokine release.
Two thirds had cytokine release syndrome, and it was mild
This is the part of the picture where a thousand patients tell you something a hundred cannot: a grade 3-or-above rate of 0.8% is only measurable at this scale.
The management reflex is remarkably consistent — tocilizumab first, steroids added when it does not settle.
-
Tocilizumab alone
396
-
Tocilizumab with steroids
286
-
Supportive care or none
145
-
Steroids alone
111
Grade 3 or above in 9 patients, 0.8%. CARTITUDE-4 reported 1%, CARTITUDE-1 reported 4%. Management categories are counted as documented and can appear more than once for a patient across episodes.
Note-extracted · CRS grade and management
Neurotoxicity, and why our rate reads higher
Classic ICANS is only part of it. The notes also carry cranial-nerve palsies, parkinsonism and movement disorders, and peripheral neuropathy — the delayed events distinctive to cilta-cel.
They arrive weeks to months after infusion, long after a trial's acute monitoring window has closed.
-
Classic ICANS
209
-
Cranial-nerve palsy
118
-
Parkinsonism or movement disorder
67
-
Peripheral neuropathy
45
Suspected to be late CAR-T effect with element of Parkinsonism
Grade 3 or above in 16 patients, 1.4%. Both trials reported 21% for any neurotoxicity; our rate sits above them because it counts the 230 delayed events beyond classic ICANS in the same figure. Trials count them too, but report them separately.
Note-extracted · event type and grade
Beyond the endpoints
What does the record add that a trial table cannot?
The reason a clinician actually wrote down — for stopping therapy, for bridging, and for the neurological events that arrive after the trial stopped watching.
Why therapy stopped or changed
Trials report discontinuation as counts in categories the protocol defined in advance. Infection and cytopenias together account for 68 stops — neither is a category a CAR-T efficacy table would surface.
-
Progression
108
-
Toxicity
66
-
Death
53
-
Patient choice
53
-
Infection
37
-
Cytopenias
31
-
Cytokine release or ICANS, and other
61
Counted as documented; a patient can contribute more than one reason across episodes. Bridging appears in 93% of records with the regimen named and its complications recorded — the step where a center's real capacity constraints show up, and the one that is invisible in trial reporting.
Note-extracted · reasons as documented
And the part that stops at the edge of a trial dataset
Alongside the endpoints, the notes carry relapse patterns years after infusion, advance care planning, and goals-of-care conversations.
Received cilta-cel >3 years ago and now having biochemical progression.
- Bridging courses documented 813
- Delayed neurological events beyond ICANS 230
- Reasons therapy stopped or changed 409
- MRD-negative results documented 487
Every extracted value in the study retains its verbatim source text, so any figure on this page can be walked back to the sentence it came from.
Verbatim note text, anonymized
The data was always there. Now we can use it.
The answer
It does. Depth of response reproduces CARTITUDE-4 almost exactly, disease control lasts, and severe toxicity stays as rare as the trials reported.
On effectiveness, 71% complete response against the trial's 73.1% — with a median of three prior lines and 53% high-risk cytogenetics. On safety, grade 3 or above cytokine release at 0.8%, measured across 1,148 patients rather than a few hundred. On survival, the real-world curve reads better than either trial and should not be believed as such; censoring at last network activity flatters it. What is genuinely new is the operational record: bridging, the reflex to a cytokine release grade, the delayed neurological events, and why therapy stopped.
What this cannot say
The limits, stated in the same breath as the findings
-
01
Referral makes prior therapy invisible
Cilta-cel is delivered at certified centers. Patients referred in or out have prior therapy and follow-up only partly visible, so prior-line and refractory figures are floors rather than rates.
-
02
Survival reads optimistic
Censoring at last recorded activity biases it upward. Median survival not reached also reflects limited follow-up rather than an absence of events.
-
03
Grades are clinician-documented
Cytokine release and neurotoxicity grades are not centrally adjudicated, and response is IMWG in name but clinician-assessed in practice, only where an assessment was written down.
-
04
The two networks are summed
Counts are combined without cross-asset de-duplication, and one network has no structured deaths table, so its mortality is note-based.
The complete myeloma study
Cohort funnel, documentation audit, the matched comparison, and the head-to-head against CARTITUDE-4 and CARTITUDE-1 — available without a form.