TOPAZ will test whether litifilimab can deliver reproducible efficacy in SLE. We put 40.5% on both Ph3 trials meeting the primary endpoint, and about two-thirds of that comes from one judgment: how much of LILAC’s effect survives. What a positive result would be worth depends on the size and consistency of benefit, supportive outcomes, safety and treatment burden, in a lupus market where Roche’s obinutuzumab now awaits an FDA decision.
Biogen has been here before in SLE. Dapirolizumab pegol, its UCB-partnered anti-CD40L antibody, missed its Ph2b dose-response primary endpoint in 2018, yet the partners advanced it, and the Ph3 PHOENYCS GO later met its BICLA primary (Lancet, Jun 2026). Litifilimab now faces its own test of how much encouraging early evidence survives Ph3. Its systemic signal, +26.4 pts on SRI-4 at Wk 24 over placebo in LILAC, came from a secondary endpoint in a Ph2 enriched for arthritis and rash. TOPAZ-1 (n=548) and TOPAZ-2 (n=562) must show that it replicates in broader enrollment over a full year of treatment (SRI-4 at Wk 52). Dapirolizumab’s success does not validate litifilimab’s different mechanism; the parallel is Biogen’s development judgment, not the biology.
Biogen expects registrational data by the end of 2026. We put the probability of both trials hitting at ~40% (Exhibit 1). About two-thirds of that comes from the retained-effect state, so the forecast depends mostly on how much of the Ph2 effect survives.
Our assumption. Our central retained effect is 18.6 pts, an analyst-selected assumption, not an estimated effect. With more favorable retention and dilution assumptions, the retained effect reaches 23 pts and P(both) rises from 40.5% to 45.8%. If alpha is split between the two doses, it falls to 34.5%. These two sensitivities show where the judgment sits; the full range and the construction are in Sections 3 and 4.
A double would still leave the product question open. Breadth of benefit across outcomes, safety and treatment burden will shape differentiation in a market that obinutuzumab, with an FDA decision expected by Dec 2026, may soon join. This preview assesses the clinical outcome distribution and illustrates the SLE opportunity; it does not value Biogen or make a stock recommendation.
1. What TOPAZ means for Biogen
A reproducible TOPAZ result would give Biogen a late-stage SLE asset with Ph3 support and extend its immunology portfolio. It would not change Biogen's revenue mix for years.
Biogen's MS franchise is shrinking: MS product revenue was $1.921bn in H1 2026, down 6.8% Y/Y (Biogen Q2 8-K), or ~$0.14bn lower than a year earlier. Litifilimab cannot offset that near term: even our mid-adoption case reaches ~$0.9bn of sales (US, EU5 and Japan) only at a mature peak around 2035 to 2037 (Section 2).
Litifilimab is not Biogen's only SLE program. Biogen co-develops dapirolizumab pegol, a CD40L antagonist, with UCB. Its first Ph3, PHOENYCS GO, met its BICLA primary endpoint at Wk 48 (Lancet, Jun 2026), and a confirmatory Ph3, PHOENYCS FLY, is recruiting, with primary completion estimated for May 2028 (UCB/Biogen). A TOPAZ miss would not end Biogen's presence in SLE, and a win would leave it managing two programs and any overlap between them. We do not model the collaboration economics.
2. What a valuable result looks like, and what the SLE opportunity could be
Statistical success needs ~10 pts of SRI-4 delta in each trial. A commercially valuable result needs more, because it must stand against belimumab and anifrolumab today and against newer agents such as obinutuzumab, now under FDA review.
The bar and the benchmark
Each trial needs an observed delta of ~10 pts over a ~41% placebo to clear two-sided 0.05, assuming 1:1:1 allocation (182 and 187 per arm, from registry actual enrollment) and the high dose against placebo; a Bonferroni split for two doses raises the bar to ~11.5 pts. That is an estimate under our assumptions, not a universal hurdle (power in Appendix A1). The redacted TOPAZ-2 protocol plans about 540 randomized, or about 180 per arm if the groups are equal, against 187 in our model (the registry shows 562 actually enrolled, about 4% over plan); the bar moves by 0.2 pt at most. Across the eight past trials in our model class (seven at Wk 52 SRI-4, ADDRESS II, a Ph2b, at Wk 24) the placebo-adjusted delta averages 10.8 pts (12.9 pts without TULIP-1); the set omits failures such as LOTUS (−12.1 pts). Belimumab ranged from +7.1 to +14.0 pts across five trials, and anifrolumab was +18.2 pts in TULIP-2 and −4.2 pts in TULIP-1. A litifilimab double at ~10 to 14 pts would sit within the belimumab range and support a potential filing, subject to the prespecified testing hierarchy and the full evidence package. Cross-trial deltas cannot show equivalence, and such a result would not by itself differentiate litifilimab on efficacy.
The contemporary benchmark is higher. In ALLEGORY (NEJM 2026), Roche's obinutuzumab produced SRI-4 at Wk 52 in 76.7% against 53.5% on placebo (+23.1 pts, 95% CI 12.5 to 33.6), with BICLA +21.9 pts and sustained steroid reduction to ≤7.5 mg/day +26.1 pts. The FDA accepted the filing in Apr 2026, with a decision expected by Dec 2026.
Benchmark, not comparator. ALLEGORY is not a like-for-like comparator. It required serologic activity (autoantibodies plus low complement) and SLEDAI-2K ≥8, enrolled a more active population (mean SLEDAI-2K ~13; musculoskeletal and mucocutaneous BILAG A or B in ~90% to 95%; about half Hispanic or Latino), scored rescue therapy and early D/C as non-response, and gave IV infusions after methylprednisolone premedication. Under a treatment-policy analysis the delta was +16.8 pts. ALLEGORY shows what a modern Ph3 can deliver and where a TOPAZ result will be judged, not how litifilimab compares, and we keep it outside the statistical reference class (Exhibit 3).
What would differentiate
Four features would shape litifilimab's differentiation beyond a positive primary endpoint.
- Same-dose success. A clean double means the high dose beats placebo on SRI-4 in both trials; a low-dose win in one trial is a split-dose result, which we treat as lower quality. The registered statistical plan (EU Clinical Trials Register) uses a graphical multiple-testing procedure that controls the study-wise two-sided α at 0.05 across the primary endpoint, the key secondaries and both dose comparisons, in 12 hypotheses. The α allocation between the two doses, the ordering of the key secondaries and the weights are left to the SAP, which is not public. The redacted TOPAZ-2 protocol (Protocol 230LE304, Version 6.2; NCT04961567) states the primary comparison as SRI-4 at Wk 52 for each litifilimab dose against placebo, at a two-sided α of 0.05 for the study. That fits a study-level α but does not show how α is allocated between the doses, so the high dose at the full 0.05 and the split-alpha case both remain modeling conventions. We assume TOPAZ-1 follows the same plan.
- Corroboration from a second composite. BICLA requires BILAG improvement in every affected organ, so it can disagree with SRI-4, which rests on SLEDAI-2K. It is complementary rather than independent, since it scores the same patients at the same visits. LILAC's BICLA confidence interval included zero (Section 3).
- Key secondaries. BICLA at Wk 52 is a key secondary per Biogen. Joint-50 at Wk 52 and a sustained steroid taper (≥10 mg to ≤7.5 mg by Wk 40, held to Wk 52 with no disease worsening) are multiplicity-adjusted in the design abstract, and annualized flare rate is also registered. TOPAZ-1 allows up to 20 mg/day of steroid at entry. The SAP is not public, including how D/C, rescue medication and missing data are scored (the registered plan sets baseline as the last measurement before the Day 1 dose and uses the 52-week double-blind period for efficacy). The registry's SRI-4 and BICLA definitions in both trials include “no violation to protocol-specified medication rules”, so these rules enter response classification, but their consequences are not disclosed; ALLEGORY's estimand gap above shows such rules matter.
- Safety and dosing burden. TOPAZ doses subcutaneously every 4 weeks with a Wk 2 loading dose through Wk 48 (design abstract); the mg doses are undisclosed. Safety is summarized over the 52-week period and a 24-week safety follow-up. Biogen reports the AMETHYST Ph2 52-week safety as consistent with prior studies, with no new signals (release of 2 Oct 2026). In LILAC Part A (pooled litifilimab n=76, placebo n=56, 24 weeks), adverse events occurred in 59% and 68%, serious events in 5% and 11%, and discontinuations (D/C) in 3% and 5%. Herpes zoster occurred in 3% and 4%, with one case of herpes keratitis on litifilimab, and there were no deaths on litifilimab against three on placebo. That exposure is too limited to assess the risk of rare or opportunistic infections over 52 weeks of TOPAZ dosing. For comparison, ALLEGORY showed serious infections in 8.6% against 4.6% on placebo and infusion reactions in 11.9% against 3.3%, with D/C for adverse events at 4.0% in both arms. Litifilimab does not deplete B cells, but that is a mechanism, not evidence of a safety advantage. Anifrolumab, which also acts on the type I interferon pathway, carries a herpes zoster imbalance in its label, and LILAC cannot show whether litifilimab shares it. TOPAZ will give the first 52-week placebo-controlled safety data for litifilimab in SLE.
Population
TOPAZ-1 enrolls moderately to severely active SLE: modified SLEDAI-2K ≥6, modified clinical SLEDAI-2K ≥4, BILAG A ≥1 or B ≥2, and steroid ≤20 mg/day. We expect a label to follow that population. LILAC also required active arthritis and an active skin lesion, which TOPAZ-1 does not (Section 3).
SLE opportunity (illustrative)
Our illustrative US/EU5/Japan funnel gives ~$0.3bn, ~$0.9bn and ~$1.5bn of peak SLE sales in low-, mid- and high-adoption cases, conditional on approval and launch (CLE excluded). The cases are defined by the market position each requires, not by a TOPAZ outcome (Exhibit 2). The mid case puts litifilimab on ~37,000 patients in the peak year and produces ~$0.4bn of mature annual operating profit after an assumed 5% royalty.
Prevalence
We use 575,000 prevalent SLE patients as a planning estimate (US 250,000, EU5 250,000, Japan 75,000), not a reported market count. The US figure sits above the CDC estimate of ~204,000, which rests on a 2018 estimate. The Japan figure follows a 2025 national claims study of ~74,000 treated patients. The EU5 figure reconciles country studies with different definitions and sums to ~250,000. We test 200,000 to 300,000 for the US and EU5 and 60,000 to 90,000 for Japan; country rates and sources are in Appendix A9.
We assume 40% of prevalent patients are adults with clinically suitable moderate-to-severe SLE despite standard therapy, or 230,000 patients. A literature review found ~43% with moderate-to-severe disease, with wide heterogeneity. The 40% is a composite allowance for disease activity, adult treatment and clinical suitability. It is not a verified TOPAZ-label population, and we test 30% to 50%.
Price
Actual competitor net prices are not public. We assume annual manufacturer net revenue per full treatment-year of $45,000 in the US and $15,000 in each of EU5 and Japan (2026 dollars). The US anchor is Benlysta's SC list price of $1,337.01 per dose, or ~$69,525 for 52 weekly doses, so $45,000 implies ~65% realization against that benchmark. This is an assumption about litifilimab, not an estimate of GSK's net realization.
Funnel and peak sales
Scope of the sales figures. US, EU5 and Japan only, not global; other markets and CLE are excluded. The mid case of ~$0.9bn splits into the US ~$0.62bn (~70%), EU5 ~$0.21bn and Japan ~$0.06bn. All figures are in 2026 dollars, conditional on approval and launch, at a mature peak around 2035 to 2037.
| Row | Low adoption | Mid adoption | High adoption | Basis |
|---|---|---|---|---|
| Prevalent SLE, US / EU5 / Japan | 250k / 250k / 75k (575k) | Planning estimate from reported studies | ||
| Clinically suitable population (40%) | 230k | Clinaptis assumption | ||
| Advanced-therapy uptake in suitable population at peak | 50% | 65% | 75% | Clinaptis assumption |
| Patients on an advanced therapy in the peak year | 115k | 150k | 173k | Calculated |
| Litifilimab share of that population | 10% | 25% | 35% | Clinaptis assumption |
| Litifilimab patients treated in the peak year | 11.5k | 37.4k | 60.4k | Calculated |
| Average annual treatment exposure | 80% | 85% | 90% | Clinaptis assumption |
| Full treatment-year equivalents | 9.2k | 31.8k | 54.3k | Calculated |
| Annual net price per treatment-year, US / EU5 / Japan | $45k / $15k / $15k | Clinaptis assumption | ||
| Peak SLE sales | $258m | $891m | $1,524m | Calculated |
| Mature operating margin, after 5% royalty | 35% | 45% | 50% | Clinaptis assumption |
| Illustrative operating profit at peak | $90m | $401m | $762m | Calculated |
Sales = prevalence × suitable fraction × advanced-therapy uptake × litifilimab share × annual exposure × net price, summed by geography. Advanced therapy means a biologic or targeted therapy added to standard therapy (antimalarials, steroids and conventional immunosuppressants); share is of patients on an advanced therapy in the peak year, counted as unique patients treated at any point in that year. Illustrative; not a forecast.
Average annual treatment exposure replaces a persistence input: at 85%, each patient treated during the year contributes ~10.2 months of revenue-generating treatment. It is not an observed persistence rate, and we would drop it if we modeled average patients on therapy throughout the year, to avoid counting discontinuation twice.
Why these adoption assumptions, and where they are vulnerable
The mid case needs advanced-therapy use in suitable patients to reach 65%. In the observational SPOCS moderate-to-severe cohort, 21.2% of patients were on a biologic (Lupus Sci Med 2025). That is cohort evidence and not today's market share, but it shows that 65% is roughly a tripling.
We read the share assumptions as market positions. At 10%, litifilimab is a useful additional option, well below an equal split among three advanced therapies. At 25%, it holds an equal split among four, in a market whose advanced-therapy use has roughly tripled; that requires credible efficacy, steroid and flare support, acceptable safety and workable access against belimumab, anifrolumab and newer entrants. A 35% share exceeds an equal split among three therapies and requires sustained differentiation against those competitors. Uptake and share also pull against each other: the more effective therapies reach the market, the more plausible 65% uptake becomes and the harder a 25% share is to hold.
The FDA approved weekly SC self-administration of Saphnelo in Apr 2026 (FDA record 24 Apr; AstraZeneca announced 27 Apr), after a complete response letter in Feb 2026, so any convenience edge from monthly dosing is monthly against weekly, subject to litifilimab's eventual device and label. Obinutuzumab is under FDA review on the ALLEGORY data, and dapirolizumab, Biogen's own partnered program, is in a confirmatory Ph3. We keep 25% for the mid case as an illustration. It is the assumption most exposed to how many effective therapies reach the market by the mid-2030s.
In the mid case, five more points of litifilimab share add ~$178m of annual sales, and a 20% cut to all net prices lowers sales from $891m to $713m. Population, suitability and price ranges are in Appendix A9.
Illustrative operating contribution
The operating-profit figures are illustrative mature product economics. We apply mature operating margins of 35%, 45% and 50% after an assumed royalty of 5% of sales to Royalty Pharma, inside the disclosed mid-single-digit description. The margins and the royalty rate are our assumptions. The agreement also entitles Royalty Pharma to approval-based fixed milestone payments of up to $250m (Biogen 10-K), separate from the development funding it provides. The disclosure ties them to approval in the US or certain major markets without an allocation by indication, so we do not charge them to SLE. They belong in launch-period cash flows, and we exclude them from the peak-profit figures.
Existing lupus sales show the category can support large products: Benlysta had 2025 worldwide sales of £1.773bn (US £1.464bn, including lupus nephritis) and Saphnelo had $686m. They are scale checks only, since geography, renal indications, prices and treated populations differ. Realized earnings would depend on launch timing, uptake, commercial expenditure and continuing development costs.
Clinical states and commercial cases
The clinical states and the commercial cases are separate judgments, and we do not pair them. SRI-4 magnitude does not set adoption: a large double with a safety or access problem could sit in the low case, and a modest double with clean safety, steroid sparing and convenient dosing could reach the mid case. Observed deltas overlap across states. Conditioning on both trials clearing the ~10 pt bar selects results that tend to exceed the underlying true effect; a ~15 pt double could therefore arise from any of the three states. The central retained state supplies ~65% of modeled doubles, which is not a probability of any commercial case (Section 4).
| Clinical state | Decayed (7.7 pts) | Baricitinib-like (11.5 pts) | Retained, central (18.6 pts) | Retained, alternative (23 pts) |
|---|---|---|---|---|
| Weight in our model (20% no effect) | 15% | 30% | 35% | 35%, replacing 18.6 |
| P(both hit) if the drug has this effect | ~19% | ~38% | ~75% | ~90% |
| Share of modeled doubles | ~7% | ~28% | ~65% | ~69% of the 45.8% alternative |
Clinaptis TOPAZ lean Monte Carlo v4 (modeled), slope 0, scatter 3 and 6 pts averaged, 21 Oct 2026. In the 23 pt alternative the two decayed states supply ~6% and ~25% of doubles. Not observed data.
3. What LILAC supports, and where translation could fail
LILAC shows biological activity in arthritis- and skin-active SLE. It does not establish the size of a Wk 52 SRI-4 effect in TOPAZ. Every TOPAZ scenario is centered below LILAC’s +26.4 pts.
What LILAC reported
In LILAC Part A, SRI-4 at Wk 24 was 56.3% (36/64) on litifilimab 450 mg against 28.6% (16/56) on placebo, an adjusted difference of +26.4 pts (95% CI 9.5 to 43.2). SRI-4 was a listed secondary endpoint, and the paper states no plan to adjust secondaries for multiple comparisons. The primary endpoint was changed by protocol amendment, from CLASI-A to active joint count, before Part A or Part B data were unblinded; the authors attribute the change to slow recruitment and the aim of assessing arthritis. It was not an outcome-driven rescue. A separate selection risk remains: the TOPAZ design abstract says SRI-4 was chosen as the primary endpoint based on its performance in Part A, so the +26.4 pts that motivated the endpoint is the estimate most exposed to winner's curse.
SRI-4 requires a SLEDAI-2K fall of at least 4 points with no worsening on BILAG or PGA, so the response rests on SLEDAI-2K by construction. The check is whether other measures agree. BICLA, which requires BILAG improvement, was 44.3% against 29.6% (+16.3 pts, 95% CI −0.5 to 33.1; 27/61 vs 16/54), and the interval includes zero. Stricter SRI-5 and SRI-6 thresholds gave +21.0 and +21.7 pts, which supports the SRI-4 result. A post hoc sustained-SRI-4 analysis gave a hazard ratio of 2.45.
Placebo and enrichment
LILAC's placebo response was 28.6%, against a 40.7% pool across seven Ph3 placebo arms (Exhibit 3). A 56.3% active response over a 40.7% placebo is +15.6 pts. That is arithmetic, not a forecast.
LILAC also enrolled for active arthritis (≥4 tender and ≥4 swollen of 28 joints) and an active skin lesion; baseline arthritis was 97% and 93% and rash 98% and 100% in the two arms. TOPAZ-1 requires neither, and the TOPAZ organ mix is not public. Arthritis scores 4 SLEDAI points and rash 2 under standard weights, so an enriched population reaches the ≥4 point SRI-4 threshold more easily. Diluting +26.4 for a TOPAZ population that is 50% to 65% arthritis-active, with non-arthritis patients getting none to half of the effect, gives ~13 to ~22 pts; both ranges are our assumptions. The nearest proxies use broader definitions than LILAC's: musculoskeletal BILAG A or B was 66% to 71% in BLISS-76 and 92% to 95% in ALLEGORY. We do not stack this dilution on the baricitinib-like state, because baricitinib's ratio already contains enrichment loss. The anifrolumab-like state is not diluted either, although its Ph2 (MUSE) was not enriched; it supplies ~7% of modeled doubles.
What past programs did
Ph2 SRI effects mostly shrank in Ph3. Across five programs the Ph3 delta kept a median 0.38 of the Ph2 delta, ranging from −0.42 to 0.99 (Exhibit 4). Baricitinib is the closest structural analog: an arthritis- and rash-enriched Ph2 with SRI-4 as a secondary endpoint, followed by a Ph3 that dropped the enrichment. It kept 0.38 (+16.0 pts in Ph2; +6.2 pts mean in Ph3, with BRAVE-I at +10.8 and BRAVE-II at +1.5). Applied to LILAC through the log-odds ratios of 0.37 (baricitinib) and 0.25 (anifrolumab), the class gives 11.5 and 7.7 pts (10.0 and 7.1 on the risk-difference scale). With five programs of differing endpoints, time points and populations, and a leave-one-out median of 0.33 to 0.60 on these ratios (0.31 to 0.58 on the log-odds ratios), we use it as context, not as a calibration.
Our scenarios, and the one that carries the forecast
We assign 20% to no effect, 15% to 7.7 pts, 30% to 11.5 pts and 35% to a retained effect of 18.6 pts (Exhibit 5), a mixture mean of 11.1 pts with no effect counted as zero, or 13.9 pts if the drug works. Shrinking LILAC toward past trials gives 13.4 to 18.1 pts, and the 13.9 pt working-drug mean sits inside that band. That is a consistency check on the mean, not a test of the mixture's shape. The mixture is bimodal, and the retained state is the bullish judgment. It rests on dapirolizumab (0.99) and telitacicept (0.82), neither of which had an enriched Ph2, and neither is a clean precedent: dapirolizumab's Ph2 missed its dose-response primary endpoint, and its retained +15.0 pts is a top-dose pairwise BICLA comparison; telitacicept's figures are China-only company data. LILAC's SRI-4 was an uncorrected secondary in an enriched population, the pattern in which baricitinib decayed.
How the retained state is built. We carry LILAC's odds ratio of 3.5 to a 40.7% placebo, as for the other two states, which turns +26.4 into ~30 pts before any decay. The retained precedents keep 0.79 to 1.01 of that on the log-odds scale (24 to 30 pts), and the enrichment dilution then applies in full, because neither precedent's ratio contains enrichment loss. Across those ranges the retained effect runs from ~12 to ~25 pts. We use the midpoint of each range (retention 0.90, arthritis share 57.5%, non-arthritis effect 25%), which gives 18.6 pts. That is an analyst-selected assumption, not an estimated effect: the arthritis share and the non-arthritis effect are our own inputs, and litifilimab's skin activity in CLE suggests non-arthritis patients could respond more than the midpoint allows. 23 pts is reachable within the same ranges under more favorable inputs, for example retention of 1.01 with a 57.5% arthritis share and half the effect in non-arthritis patients (23.8 pts), or a 65% share and a one-third effect (~23 pts). We keep 23 pts visible throughout as the alternative: P(both) is 40.5% at 18.6 pts and 45.8% at 23 pts, with the weights, no-effect chance and testing rule unchanged (Exhibit 6A).
An 18.6 pt delta on a 40.7% placebo implies an active-arm response of ~59%, and 23 pts ~64%. The highest active-arm response among the seven Wk 52 trials in our class is 61.4% (BLISS-SC), and the largest Ph3 delta is +18.2 pts (TULIP-2). ALLEGORY reached 76.7% and +23.1 pts outside the class, in a more active, serologically selected population under a composite estimand; a delta of that size is attainable in a modern SLE Ph3, but that says nothing about whether a BDCA2 antibody reaches it. We keep the retained state at 35% because two of five programs retained 0.8 or more, and LILAC's own confidence interval (9.5 to 43.2) contains both levels. Cutting the retained weight to 25% takes P(both) from ~40% to 35.8%.
Dose adds a second uncertainty. LILAC’s 50 mg and 150 mg arms enrolled six patients each, leaving 450 mg as the main basis for assessing SLE efficacy; the TOPAZ doses are undisclosed. We model the low dose at 0.75 of the high-dose effect and report the high dose as the headline (Section 4).
4. The probability framework and its largest assumptions
How the model works
A hit means the high dose beats placebo on SRI-4 at Wk 52, two-sided 0.05. That gives the high dose the full alpha, as a hierarchy that tests it first would. The registered plan controls α at 0.05 across both doses, but the allocation between them sits in the unpublished SAP, so the headline is our assumption and the split-alpha case brackets it. The registered primary analysis adjusts for baseline and randomization strata (logistic regression, with the rate difference from a standardized estimator and a Wald test); our simulation compares two proportions and does not model that adjustment. Both trials hitting is the clean double defined in Section 2. We simulate 5,000 trial pairs per scenario. Each pair first draws whether the drug works; if it does not, its true effect is centered on zero with SD 3 pts, so a no-effect run can still carry a small true effect. Each pair then draws a shared true effect, adds trial-to-trial scatter (SD of 3 or 6 pts, averaged), draws a trial-specific placebo response around 40.7%, and adds binomial noise at 182 and 187 patients per arm. The 1:1:1 allocation is our assumption; the registry does not state it, although the redacted TOPAZ-2 protocol gives one per-group size for both doses and placebo (about 540 randomized, against 562 actually enrolled per the registry), which is consistent with equal allocation. The two trials also differ in geography. China and Japan account for 58 of 177 TOPAZ-2 sites (32.8%) and none of TOPAZ-1's 164; US sites are 53 of 164 in TOPAZ-1 (32.3%) and 44 of 177 in TOPAZ-2 (24.9%). These are site shares, not patient shares. They support keeping trial-level uncertainty but do not validate our scatter size. Parameters are in Appendix A6.
Where the ~40% comes from
If the drug works (80% in our weighting), both trials hit in ~50% of runs: ~19% at 7.7 pts, ~38% at 11.5 and ~75% at 18.6. The retained state carries 44% of the working-drug weight and the highest hit rate, so one judgment, a 35% weight on a retained 18.6 pt effect, supplies ~26 of the ~40 pts of P(both), about 65%; the 7.7 and 11.5 pt states supply the other ~35%. At 23 pts the retained state would hit in ~90% of its runs and supply ~69% of a 45.8% blend.
The issue with the mixture is its shape, not its mean. We assign 56% of the working-drug weight to scenarios centered at or below 11.5 pts and 44% to one centered at 18.6 pts, with no intermediate scenario center. We have not run a prior concentrated near 15 pts, so we cannot say how the outcome split would change under one.
Sensitivity
The ~40% is a central case, and its largest inputs are our judgments. The ~36% to ~50% range is the span of the Exhibit 6A variants, all of which give the high dose the full 0.05 (35.8% to 49.7%); it excludes the earlier-weights calibration row (32.5%). It does not cover a 23 pt retained effect combined with a 45% retained weight, which reaches 52.0% (Appendix A5), or the split-alpha convention, which lowers those variants to 29.8% to 45.9% (26.8% for the calibration row). None of these figures is a confidence interval. Exhibit 6A lists the variants and Exhibit 6B shows the heatmap behind the blend.
Other model choices
Scatter. The TULIP-1 to TULIP-2 gap (22.4 pts) is ~3 standard errors of the difference, so it is more than sampling noise. A random-effects estimate across seven trials gives ~5 pts and falls to 0 without TULIP-1. This is approximate and rests on one trial pair, so we bracket at 3 and 6 pts. P(both) is 42.3% at 3 pts and 38.6% at 6.
Multiplicity. If the two doses split alpha at 0.025 each (the headline gives the high dose the full 0.05), the high dose must clear ~11.5 pts in both trials and P(both) falls to 34.5%, ~6 pts below the headline. Allowing either dose to win both trials under that split gives 36.7%. Counting a win at either dose in each trial gives 38.1%, including ~1.4 pts from wins at different doses. Section 2 treats those split-dose results as lower quality, so this broader success definition understates the cost of multiplicity. The registered plan fixes the study-wise α at 0.05 across both doses but not the split between them.
Placebo. At a fixed absolute delta, placebo level barely moves the ~10 pt bar; it matters mainly if a higher placebo also compresses the drug effect. The reference pool may understate a modern placebo (BRAVE-I and BRAVE-II ran 46%, LOTUS 56.0% and ALLEGORY 53.5%), and the slope is unobserved (r=0.07, 95% CI −0.67 to 0.74 across 8 trials). At an unshifted 40.7% placebo, slopes from −0.5 to +0.25 move the central case by ≤1 pt. With the placebo center shifted up 5 pts to 45.7%, the central case is unchanged at slope 0 (40.7%) but falls to 35.7% at slope −0.25 and 30.8% at slope −0.5 (Appendix A2).
Reported, modeled and unknown. Reported: the LILAC results, the TOPAZ registry design (actual enrollment of 548 and 562, SRI-4 at Wk 52), the redacted TOPAZ-2 protocol's three-arm design (two litifilimab doses and placebo, about 540 randomized, SRI-4 at Wk 52 tested for each dose against placebo at a two-sided α of 0.05 for the study), the registered analysis methods (a graphical multiple-testing procedure across 12 hypotheses and both doses, covariate-adjusted logistic regression, mixed models for continuous endpoints and Cox models for time to event), the seven placebo arms and the eight past deltas. Calculated: the ~10 pt bar. Modeled: every probability in this note, none of which is observed TOPAZ data. Unknown: the SAP details (α allocation between the two doses, key-secondary ordering and weights, estimand, D/C, rescue and missing-data rules), any serology entry requirement, the exact allocation ratio and per-arm size (redacted; equal groups implied), the mg doses, the protocol's assumed placebo rate, treatment difference and power (redacted), the organ and regional mix, and the low-dose effect, which we assume at 0.75 of the high-dose effect.
5. What each outcome would mean, and what to listen for on 28 Oct
The count of positive trials decides less than the quality of each result. Exhibit 7 maps the three outcomes; the determinants follow.
Both hit (~40%)
The five checks in Exhibit 7 decide the value, with the size of the delta read against belimumab's +7 to +14 pts and ALLEGORY's +23.1 pts, neither a head-to-head comparison. About one in three doubles in our model comes from true effects of 11.5 pts or less (~14 of the ~40 pts), so a double at ~10 to 12 pts without BICLA or steroid support would add little differentiation, and we would value it below its headline. A double near 20 pts or more, supported by BICLA and steroid outcomes, would shift weight toward the retained state and strengthen the case for higher commercial adoption, which would still depend on safety, access, dosing and competition.
Exactly one hits (~27%)
A split result fits a smaller true effect. Among runs where the drug works, exactly one trial hits in ~40% of runs at 7.7 and 11.5 pts, ~22% at 18.6 pts and ~9% at 23 pts, so a split moves weight toward the decayed states. Whether one positive trial plus supportive data supports a filing is a regulatory call we cannot make.
Neither hits (~32%)
About 18 of the ~32 pts come from the 20% no-effect state. The retained rows of Exhibit 1 show ~21% and ~19% neither only because they include that state; a drug that genuinely works at 18.6 pts misses both trials in ~4% of its runs, and at 23 pts in under 1%. The other ~14 pts represent runs in which a working drug, mostly in the decayed states, misses both trials. The two cases read differently (Exhibit 7). A consistent near-zero result would end the SLE case on this primary endpoint; a miss at a modest effect would raise the question of a further trial, which would push any commercial ramp well beyond the scenarios in Section 2.
SLE and CLE read-through
Biogen's AMETHYST Ph2 52-week data (2 Oct 2026) show continued improvement in CLE on litifilimab from Wk 24 to Wk 52: clear or almost clear skin (CLA-IGA 0-1) in 27.2% vs 19.0% at Wk 24, and CLASI-70 in 28.8% vs 21.7%. Participants who switched from placebo improved from as early as 4 weeks, with no new safety signals. After Wk 24 every participant is on litifilimab, so these comparisons are within-group with no placebo arm and say nothing about systemic efficacy. An SLE miss would weaken the mechanism case for CLE without erasing it: CLE is a different disease with a different endpoint, and the FDA granted litifilimab Breakthrough Therapy designation in CLE on 28 Jan 2026. Biogen expects AMETHYST Ph3 results in H1 2027.
What to listen for on the Q3 call (28 Oct)
Biogen reports Q3 2026 results on Wed 28 October, before the market opens, with a webcast at 8:30 a.m. ET.
- Timing language. Biogen previously indicated H2 2026, specified Q4 2026 in its 6 February 2026 earnings release, and reiterated an expectation of data by year-end on 29 July. These statements are consistent; any subsequent change to the window matters. The registry primary-completion estimates (30 Sep 2026 for TOPAZ-1, 27 Sep 2026 for TOPAZ-2, last updated Feb and Jan 2026) are not company guidance.
- Trial conduct. Comments on enrollment, placebo behavior or regional mix.
- Broader business context. Updated MS trends and management's priorities for investment in growth products.
- Filing intent. What Biogen says it needs from the readout to file.
- Competitive context. How Biogen positions litifilimab against newer SLE therapies and alongside dapirolizumab. Obinutuzumab's FDA decision is expected by Dec 2026, around the TOPAZ data window.
Conclusion: what TOPAZ would settle, and what would change our view
TOPAZ settles less than its headline suggests. Two positive trials would establish reproducible systemic efficacy and support a potential filing, but not differentiation, approval, a label or a market position; the commercial cases in Section 2 depend on adoption, access and competition that the readout informs but cannot settle. For Biogen, a win adds a late-stage SLE program alongside dapirolizumab, with mature economics a decade away; a miss removes one SLE option, not Biogen's presence in the disease. We put ~40% on both trials meeting the primary endpoint, with substantial sensitivity to the size and weight of the retained effect and to the testing rule: 34.5% if alpha is split between the doses, ~46% with a 23 pt retained effect.
The principal clinical uncertainties are the size of the retained effect, the α allocation between the two doses, and placebo behavior. We would lower the odds of a double if we saw:
- TOPAZ placebo above ~49% with visible effect compression;
- a disclosed testing rule that splits alpha between the two doses;
- a TOPAZ organ mix much less arthritis- and rash-heavy than LILAC's.
We would read a double as lower quality, without changing the headline odds (which count the high dose only), if we saw:
- wins at different doses in the two trials;
- low-dose data that contradict our 0.75 dose ratio.
The most important commercial assumptions are the share of suitable patients on an advanced therapy at peak (50% to 75%), litifilimab's share of them (10% to 35%; the mid case uses 65% and 25%), net price per treatment-year ($45,000 in the US; $15,000 in EU5 and Japan), and the number of effective therapies on the market by the mid-2030s.
Limits. 1:1:1 allocation is assumed, although the redacted TOPAZ-2 protocol implies equal groups of about 180 (we assume TOPAZ-1 mirrors it); the protocol's placebo-rate, treatment-difference and power assumptions are redacted, so our ~10 pt bar cannot be checked against the sponsor's own sizing; the α allocation between doses and the ordering of key secondaries sit in the unpublished SAP; the two trials' site footprints differ, so a regional difference in placebo response or effect is possible but unmeasured; the past-trial class is survivor-heavy, and the placebo pool excludes recent and failed trials with higher placebo responses; the decay class has five programs; the retained tail rests on two imperfect precedents and an analyst-selected central level; and the commercial sizing is illustrative, resting on our assumptions for uptake, share and net price.
Appendix
A1. Single-trial power
Closed-form normal approximation, no simulation. TOPAZ-1 (182 per arm) gives power of 21.5%, 49.3%, 77.4% and 93.7% at true deltas of 6, 10, 14 and 18 pts. TOPAZ-2 (187 per arm) gives 22.0%, 50.3%, 78.5% and 94.3%. High dose against placebo only, with placebo at 40.7%. A Bonferroni rule for two doses raises the observed delta needed from ~10 to ~11.5 pts.
A2. Placebo coupling
If placebo and drug effect are unlinked (slope 0), P(both | drug works) stays at ~49% to ~51% across placebo levels. If a higher placebo also shrinks the effect, it falls to ~39% (slope −0.25) or ~27% (slope −0.5) at ≥49%. Built on the v4 simulation at the central weights (retained 18.6 pts), scatter 3 and 6 pts averaged, placebo center shifted +5 pts to 45.7% so the high bins fill, 5,000 trial pairs per cell. Bins with fewer than 30 draws are blank. Across all runs, the same shift leaves P(both) at 40.7% at slope 0 and cuts it to 35.7% (slope −0.25) and 30.8% (slope −0.5).
A3. Reference-class inputs
| Placebo SRI-4, Wk 52 | % | Past delta (placebo-adjusted) | pts |
|---|---|---|---|
| BLISS-52 (n=287) | 43.6 | TULIP-2 | +18.2 |
| BLISS-76 (n=275) | 33.5 | ADDRESS II (Ph2b, Wk 24) | +14.8 |
| BLISS-SC (n=279) | 48.4 | BLISS-52 | +14.0 |
| BLISS-NEA (n=217) | 40.1 | BLISS-NEA | +13.7 |
| EMBRACE (n=149) | 41.6 | BLISS-SC | +13.0 |
| TULIP-1 (n=184) | 40.2 | BLISS-76 | +9.7 |
| TULIP-2 (n=182) | 37.4 | EMBRACE | +7.1 |
| Pooled (random effects) | 40.7 | TULIP-1 | −4.2 |
| Mean (12.9 without TULIP-1) | 10.8 |
LOTUS (ustekinumab; 43.9% vs 56.0% placebo; −12.1 pts) was stopped for futility and is not in the class. EMBRACE placebo n=149 checked against Ginzler 2022.
A4. Ph2-to-Ph3 decay class
| Program | Ph2 delta, pts | Ph3 delta, pts | Ph3 / Ph2 | Log-OR ratio | Source tier |
|---|---|---|---|---|---|
| Baricitinib | +16.0 | +6.2 (BRAVE-I +10.8, BRAVE-II +1.5) | 0.38 | 0.37 | 1/2 |
| Anifrolumab | +26.0 | +7.0 (TULIP-1 −4.2, TULIP-2 +18.2) | 0.27 | 0.25 | 1/2 |
| Ustekinumab | +29.0 | −12.1 | −0.42 | −0.41 | 1/2 |
| Dapirolizumab pegol (BICLA) | +15.0 | +14.9 | 0.99 | 1.01 | 1/2 |
| Telitacicept (China only, company data) | +41.9 | +34.4 | 0.82 | 0.79 | 2/1 |
Median ratio 0.38, mean 0.41, leave-one-out median 0.33 to 0.60 on these ratios (0.31 to 0.58 on the log-odds ratios). Endpoints, time points and populations differ across programs. Anifrolumab uses the MUSE day-365 anchor (day 169 gives 0.42). The baricitinib Ph2 SRI-4 figures (64% vs 48% at Wk 24, 67/104 vs 50/105, a secondary endpoint) are from Wallace et al., Lancet 2018, Table 2.
A5. Weight sets: full results
| Set | P(both) | Exactly one | Neither | At least one | High dose at 0.025 (both) | Same dose at 0.025 (both) |
|---|---|---|---|---|---|---|
| Central case (retained 18.6 pts) | 40.5% | 27.2% | 32.3% | 67.7% | 34.5% | 36.7% |
| Retained weight 25% | 35.8% | 29.0% | 35.2% | 64.8% | 29.8% | 31.9% |
| Retained weight 45% | 45.2% | 25.3% | 29.5% | 70.5% | 39.1% | 41.5% |
| No-effect chance 25% | 38.2% | 25.8% | 36.1% | 63.9% | 32.5% | 34.7% |
| No-effect chance 15% | 43.1% | 28.3% | 28.6% | 71.4% | 36.8% | 39.1% |
| Retained 18 pts | 39.7% | 27.7% | 32.6% | 67.4% | 33.4% | 35.6% |
| Retained 20 pts | 42.6% | 25.5% | 31.9% | 68.1% | 36.8% | 38.9% |
| Retained 23 pts (alternative) | 45.8% | 23.0% | 31.2% | 68.8% | 40.8% | 42.7% |
| Retained 23 pts, weight 25% | 39.6% | 26.0% | 34.4% | 65.6% | 34.3% | 36.2% |
| Retained 23 pts, weight 45% | 52.0% | 19.9% | 28.1% | 71.9% | 47.3% | 49.1% |
| Retained 23 pts, no-effect chance 25% | 43.1% | 21.9% | 34.9% | 65.1% | 38.5% | 40.3% |
| Retained 23 pts, no-effect chance 15% | 48.7% | 23.9% | 27.4% | 72.6% | 43.5% | 45.5% |
| Retained undiluted (26.4 pts) | 47.7% | 21.2% | 31.1% | 68.9% | 43.5% | 45.0% |
| Five-program class | 49.7% | 19.3% | 31.0% | 69.0% | 45.9% | 47.3% |
| Earlier weights | 32.5% | 28.3% | 39.2% | 60.8% | 26.8% | 28.8% |
High dose vs placebo at two-sided 0.05 unless stated. “Same dose at 0.025” counts a double in which either the high dose or the low dose wins both trials at 0.025 each; a win at different doses in the two trials adds a further ~1 to 1.6 pts and is excluded. Within the central case, the no-effect state produces a positive trial in ~9% of its runs and a double in ~0.4%, because its true effect is centered on zero with SD 3 pts. Slope 0, scatter 3 and 6 pts averaged. Source: Clinaptis TOPAZ lean Monte Carlo v4 (modeled), 21 Oct 2026.
A6. Model parameters and reproducibility
Trial-pair Monte Carlo, 5,000 pairs per cell, seed 20261016, run 21 Oct 2026; 23 weight sets. Every cell starts from a fresh generator with the same seed, so adding or reordering cells cannot change any cell, and sets that only re-weight the same cells differ by an exact function of identical draws. Cells with different effect sizes share draws only until the first active-arm binomial draw, so comparisons across effect levels carry Monte Carlo error of ~±0.7 pts per cell. The model reproduced every cell of the earlier 18 Oct 2026 run exactly and every blend within rounding. Parameters: standard deviation of the true effect around each scenario 4 pts; no-effect state centered on zero with standard deviation 3 pts; low dose at 0.75 of the high-dose effect; 1:1:1 arms (182 and 187 per arm); placebo center 40.7%; scatter 3 and 6 pts; placebo slopes −0.5, −0.25, 0 and +0.25, anchored at a 40.7% placebo; two-sided 0.05, with 0.025 per dose for the split-alpha events. Headline convention: slope 0, scatter averaged. These parameters apply to every exhibit unless stated; figures labeled modeled are model outputs, not observed TOPAZ data. Weights are our judgment. The model does not describe the actual TOPAZ analysis plan. The retained effect follows: RD at a 40.7% placebo of LILAC's odds ratio 3.5 raised to the retention ratio, times [s + (1 − s) × r], with retention 0.79 to 1.01, arthritis share s of 0.50 to 0.65 and relative non-arthritis effect r of 0 to 0.5. The central 18.6 pts uses the midpoint of each (0.90, 0.575, 0.25); 23 pts is reached with more favorable inputs within the same ranges, for example retention 1.01, s 0.575 and r 0.5 (23.8 pts).
A7. Inputs and provenance
| Input | Status |
|---|---|
| LILAC Part A: SRI-4, BICLA, SRI-5/6, baseline arthritis and rash | Reported (NEJM and supplement) |
| LILAC sustained-SRI-4 hazard ratio 2.45 | Reported; post hoc analysis |
| TOPAZ actual enrollment, endpoints, entry criteria, site footprints | Reported (ClinicalTrials.gov) |
| TOPAZ-2 protocol sample-size paragraph: three arms, about 540 randomized, SRI-4 at Wk 52 for each dose against placebo, two-sided α of 0.05 for the study | Reported (redacted TOPAZ-2 Protocol 230LE304, Version 6.2); TOPAZ-1 assumed parallel; placebo rate, treatment difference, power, per-arm size and dose strengths redacted |
| TOPAZ statistical analysis methods: baseline definition, graphical multiplicity procedure (12 hypotheses), analysis models | Reported (EU Clinical Trials Register, TOPAZ-2); α allocation and secondary ordering in the unpublished SAP |
| TOPAZ dosing (every 4 weeks, Wk 2 load) and multiplicity-adjusted secondaries | Reported (design abstract, Lupus Sci Med 2023); mg doses not disclosed |
| Royalty Pharma funding (up to $250M over six quarters) and mid-single-digit royalty | Reported (Royalty Pharma release, Feb 2025; Q1 2025 deck) |
| Royalty Pharma approval-based milestones (fixed, up to $250M) | Reported (Biogen FY2025 10-K, Royalty Pharma Funding Arrangement); the amount is not in Royalty Pharma’s release |
| LILAC odds ratio 3.5 (95% CI 1.6 to 7.8) for SRI-4 | Reported (NEJM supplement, Table S7) |
| ALLEGORY design, efficacy and safety | Reported (NEJM 2026; Roche release, 6 Mar 2026) |
| Obinutuzumab FDA filing acceptance and expected decision | Company statement (Genentech release, 20 Apr 2026) |
| Dapirolizumab PHOENYCS GO result and PHOENYCS FLY status; Ph2b dose-response miss | Company statements (UCB/Biogen) |
| Placebo pool 40.7%, class mean delta 10.8 pts | Calculated from reported trials |
| ~10 pt bar; power figures | Calculated under stated assumptions |
| Decay ratios; shrinkage to 13.4 / 14.9 / 18.1 pts; scatter of 5 pts | Calculated or modeled; approximate |
| Scenario weights; 0.75 dose ratio; 1:1:1 allocation; testing convention (high dose at full 0.05); TOPAZ arthritis share (50% to 65%) and non-arthritis effect (0 to 50%) | Clinaptis assumptions |
| All probabilities (40.5% / 27.2% / 32.3% central) | Modeled; not observed data |
| Retained effect of 18.6 pts (central) and 23 pts (alternative) | Clinaptis assumptions, not estimates: 18.6 pts is the midpoint of our retention and dilution ranges; 23 pts is reached with more favorable inputs within the same ranges |
| AMETHYST Ph2 52-week data; Breakthrough designation (28 Jan 2026); Q3 date, webcast time and timing | Company statements (Biogen releases) |
| SLE prevalence (CDC; German, Italian and Japanese studies); competitor sales; MS H1 2026 revenue; Benlysta list price | Reported (sources in A9) |
| Planning population 575k; 40% suitable; uptake, share, exposure, net prices, margins, 5% royalty, 2035 to 2037 peak | Clinaptis assumptions |
| Patient counts, peak sales, operating profit | Calculated from the assumptions above |
| EU5 prevalence (250k) | Country-level reconciliation: ~250,000 (range ~200,000 to ~320,000); evidence in A9 |
A8. Sources
Litifilimab for systemic lupus erythematosus, NEJM (doi 10.1056/NEJMoa2118025) and supplement. ClinicalTrials.gov: NCT04895241 (TOPAZ-1), NCT04961567 (TOPAZ-2). Protocol 230LE304, Version 6.2, Phase 3 Study to Evaluate Litifilimab (BIIB059) in Active SLE (TOPAZ-2, NCT04961567; redacted version; sample-size paragraph). EU Clinical Trials Register, TOPAZ-2 (EU CT 2023-505696-74-00; EudraCT 2020-005776-35), statistical analysis section (graphical multiple testing procedure after Bretz 2011 and Xi and Bretz 2019; standardized estimator after Steingrimsson 2017). TOPAZ design abstract, Lupus Sci Med 2023;10(Suppl 1):A157.1. Biogen Q2 2026 results 8-K (29 Jul 2026: registrational SLE data "expected by end of 2026"; H1 2026 MS revenue $1,920.8m vs $2,060.2m) and Biogen Q4 2025 results release (6 Feb 2026: both Phase 3 SLE studies "expected to read out in Q4 2026"). Benlysta SC list price: GSK pricing page (undated; "data on file as of 1 July 2026"; $1,337.01 per dose), accessed 21 Oct 2026. Biogen release, 2 Oct 2026, "Biogen's Litifilimab Demonstrates Rapid and Durable Efficacy in New 52-Week Phase 2 Data from Ongoing Phase 2/3 AMETHYST Study" (investors.biogen.com). Biogen release of 16 Sep 2026 (Q3 results 28 Oct 2026, before the financial markets open; webcast 8:30 a.m. ET). Biogen release, 28 Jan 2026 (Breakthrough Therapy designation for litifilimab in CLE). TOPAZ registry status and completion estimates as of 21 Oct 2026. Royalty Pharma Q1 2025 earnings deck. Past-trial data: ClinicalTrials.gov results, FDA reviews and published trial reports, including Wallace et al., Lancet 2018;392:222-231 (baricitinib Ph2, Table 2). Furie et al., Efficacy and Safety of Obinutuzumab in Active SLE (ALLEGORY), NEJM 2026;395:243-54; Roche release, 6 Mar 2026; Genentech release on sBLA acceptance, 20 Apr 2026 (FDA decision expected by Dec 2026). UCB/Biogen release on the Lancet publication of PHOENYCS GO (Jun 2026); UCB/Biogen dapirolizumab Ph2b topline release (2018). PHOENYCS FLY: ClinicalTrials.gov NCT06617325 (recruiting; estimated primary completion 31 May 2028; record posted 15 Oct 2026). BLISS-76 baseline organ involvement (PMC5007058, Table 1). Royalty Pharma release, 12 Feb 2025 (R&D funding collaboration with Biogen). Biogen Form 10-K for FY2025, Royalty Pharma Funding Arrangement (approval-based milestones of up to $250M). SPOCS: Lupus Sci Med 2025;12:e001336. Clinaptis TOPAZ lean Monte Carlo v4 (modeled), 21 Oct 2026, with the 18 Oct 2026 run reproduced. Clinaptis Ph2-to-Ph3 decay reference class, 18 Oct 2026 (calculated). Not observed TOPAZ data.
A9. Commercial inputs: evidence, ranges and sources
| Geography | Evidence found | Planning input | Range tested |
|---|---|---|---|
| US | CDC: ~204,000 people with SLE (2018-based estimate) | 250,000 | 200,000 to 300,000 |
| EU5 | Germany claims: 59.9 per 100,000, ~49,800 patients (2019). France national claims: 54,804 patients, 81.6 per 100,000 (2020). UK CPRD: 107.1 per 100,000 (2020). Italy (Veneto): 70.6 per 100,000, standardized (2020). Spain: ~58 per 100,000 (primary-care database, 2022); EPISER2016 survey 210 per 100,000 (95% CI 110 to 400). Country sum ~250,000; low-rate set ~200,000; EPISER ~320,000. Definitions and periods differ. | 250,000 | 200,000 to 300,000 |
| Japan | 2025 national claims (NDB) study: prevalence 60 per 100,000, or ~74,000 patients at Japan's population | 75,000 | 60,000 to 90,000 |
| Net price per treatment-year | Benlysta SC list $1,337.01 per dose (~$69,525 for 52 doses). NICE 2021 appraisal: 200 mg pen £222.75 (~£11,583 for 52 doses) before a confidential discount. | US $45k; EU5 $15k; Japan $15k | US $35k to $50k; EU5 and Japan $10k to $20k |
Calculated sensitivities, mid case ($891m base): population ranges above move sales to $713m to $1,069m; a suitable fraction of 30% to 50% gives $668m to $1,114m; the price ranges above give $663m to $1,050m; five more points of share add $178m; a 20% cut to all net prices gives $713m.
Cautions. Do not count all lupus nephritis patients as an incremental litifilimab opportunity without renal evidence. A history of renal involvement should not exclude a patient whose current extrarenal disease could be treated. Prices are manufacturer net revenue per full treatment-year in 2026 dollars; a further 20% reduction for payer and competitive pressure is a separate stress. Other economics: Royalty Pharma provides up to $250M over six quarters in exchange for mid-single-digit royalties on worldwide sales plus regulatory milestones (release, Feb 2025). Biogen’s FY2025 10-K puts the approval-based fixed milestone payments to Royalty Pharma at up to $250M; they are excluded from peak profit.
Commercial sources: CDC; German study; Italian study; Japanese study; severity review; GSK pricing; NICE appraisal; SPOCS; Saphnelo approval; GSK FY2025; AstraZeneca FY2025; Royalty terms; Biogen disclosure; Biogen 10-K; Biogen Q2 8-K. Checked at source on 18 Oct 2026: the CDC, German, French, UK, Veneto and Japanese prevalence figures, the NICE pen price, 2025 Benlysta sales (worldwide and US), the Saphnelo SC US approval. Checked at source on 21 Oct 2026: Biogen's H1 2026 MS revenue ($1,920.8m, down 6.8% from $2,060.2m; Biogen Q2 8-K), the Q3 date and webcast time, the litifilimab CLE designation date, the Gazyva sBLA acceptance and decision timing, the Saphnelo SC approval date and prior CRL, the TOPAZ and PHOENYCS FLY registry records, the Benlysta SC list price (GSK, as of 1 Jul 2026), SPOCS biologic use (21.2%), 2025 Saphnelo revenue ($686m worldwide; AstraZeneca does not split it by region in its FY2025 results announcement) and the Royalty Pharma milestone amount (Biogen FY2025 10-K; Royalty Pharma’s own release does not state it).