FDA-Accepted Efficacy Endpoints in Acute Pain Trials: SPID, Time to Rescue, and Rescue Analgesia Handling

Acute pain trial endpoints at FDA: SPID, time to rescue, rescue use, and how reviewers handle rescue analgesia

In acute pain trials, FDA's primary efficacy measure has consistently been a patient-reported pain-intensity measure summed or averaged over a prespecified interval. That usually means the summed (time-weighted) pain intensity difference (SPID), or the area under the curve (AUC) of pain scores for prolonged-duration local anesthetics. Time to first rescue, the proportion of patients who need rescue, and the amount of rescue or opioid consumed have mostly been secondary or supportive measures. The harder question in review is usually how the analysis treats pain scores recorded after rescue. Across two decades of Drugs@FDA reviews, reviewers have moved away from open-ended carry-forward rules. They now favor a pre-rescue score carried forward for a defined rescue-effect window, backed by sensitivity analyses. The May 2026 local anesthetic draft guidance asks sponsors to prespecify rescue handling as an estimand; the acute pain draft guidance does not specify an imputation method.

Current FDA guidance on acute pain endpoints

SPID as the primary analysis

The May 2026 draft guidance Development of Non-Opioid Analgesics for Acute Pain recommends that patients rate their current pain intensity directly, generally on a numerical rating scale. The primary analysis should compare SPID between treatments over a prespecified period that covers at least the expected duration of drug effect ⁴⁶⁴⁷⁴⁸. The guidance gives SPID24 as an example primary analysis, with SPID6 or SPID12 as secondary analyses, depending on the dosing interval ⁴⁶⁴⁷⁴⁸. FDA discourages primary endpoints built on pain relief (a decrease compared with prior pain) instead of current pain intensity, because recall and other factors can influence pain-relief ratings ⁴⁷. It also generally discourages composite primary scales that combine pain, function, sleep, or other domains, because they are hard to interpret ⁴⁷.

AUC of pain for prolonged-duration local anesthetics

The companion May 2026 draft guidance Development of Local Anesthetic Drug Products With Prolonged Duration of Effect recommends the AUC of pain scores, a time-weighted average across the observation window, as the primary endpoint ⁴⁹. AUC can accommodate varying observation times and unscheduled pain assessments taken before rescue. The analysis should emphasize later time points (for example, 48 to 72 hours) that support the claimed duration. A single-time-point comparison is generally not appropriate ⁴⁹.

Secondary endpoints: onset, time to rescue, and rescue use

The acute pain draft guidance lists these informative secondary measures ⁴⁷:

• Time to onset of pain relief. FDA has accepted the two-stopwatch method (first perceptible effect and meaningful relief), and the meaningful-relief result may support a labeling statement such as median time to meaningful pain relief ⁴⁷.
• Time to rescue medication or to a request for the next study-drug dose, used to characterize onset and duration ⁴⁷⁵⁰.
• Assessment of rescue medication use: type, dose, frequency, duration, and timing, plus pain intensity before rescue and throughout the dosing interval ⁴⁶⁴⁷⁵⁰.
• Physical function and patient global impression of change, which should support the primary pain-intensity endpoint ⁴⁷.

FDA explains how to read rescue timing. "A sooner-than-expected first use of rescue medication may suggest that the investigational drug has a delayed onset of pain relief." A second rescue dose that comes earlier than drug exposure would predict raises "waning efficacy" as a potential issue ⁵⁰.

How FDA reviews have used SPID

SPID-based primary endpoints appear across Drugs@FDA reviews in dental, bunionectomy, abdominoplasty, and laparoscopic surgery models.

Two themes recur across these reviews. First, FDA treats SPID as a statistical summary that needs support from time-specific pain-intensity differences before the result is considered clinically meaningful ¹²³. Second, FDA expects the SPID window to match the dosing regimen. In the Advil Dual Action review, SPID6-8 after the first dose was used to support the every-8-hour interval ³¹¹. In the Anjeso review, FDA required a multiple-dose endpoint ²¹.

Responder endpoints built on SPID: the Olinvyk precedent

For oliceridine (Olinvyk, NDA 210730), the applicant's primary responder endpoint required at least 30% improvement in SPID-48 or SPID-24. A patient was also counted as a non-responder for any rescue use, any early discontinuation, or reaching the dosing limit ¹¹⁴¹¹⁵. FDA repeatedly disagreed with this as the primary endpoint. Its reasons were:

• The endpoint was novel for this drug class, and the clinical meaning of a 30% SPID threshold in acute postoperative pain was unclear ¹¹⁴¹¹⁵.
• Dichotomizing a continuous measure discards information ¹¹⁴¹¹⁵.
• Counting any amount of rescue as non-response would underestimate treatment effects ¹¹⁴.

FDA based its efficacy conclusions on continuous SPID analyzed by ANCOVA instead ¹¹¹. FDA also asked for separate results for each responder component, including the rescue component ¹⁰⁴¹⁰⁸.

Time to rescue and rescue-medication use as endpoints

Mostly secondary, occasionally primary

In most reviews, rescue-based measures supported a pain-intensity primary endpoint:

• Ofirmev: Median time to first rescue was 3 hours with IV acetaminophen versus 0.8 hours with placebo. Morphine use over 24 hours was 33% lower. FDA cited both as supportive secondary findings, without a multiplicity correction ⁶⁵⁶⁷.
• Seglentis: Time to first rescue, the proportion with at least one rescue dose, and the number of rescue doses at intervals from 4 to 48 hours were secondary endpoints ⁶⁸.
• Journavx: During protocol development, FDA directed the sponsor to include median time to first rescue in the first 12 hours, and the proportion using rescue in the 0-24 and 24-48 hour intervals, as secondary endpoints ³¹.

A few older programs used rescue measures as the primary endpoint:

• Celebrex (NDA 020998): Time to rescue medication was one of the primary efficacy measures described as "preferred by the Division" ⁷³.
• Chirocaine (NDA 020997): Programs used time to first analgesic request, and separately the proportion requiring rescue within 2 hours, as primary or confirmatory endpoints ⁷²⁷⁵.

In the oxymorphone IR program (NDA 021611), FDA said time-specific pain measurements, onset, and duration were more important than time to remedication for characterizing single-dose effects ⁷⁰. FDA accepted time to discontinuation, triggered by rescue need, as a reasonable multiple-dose endpoint that reduced confounding from rescue medication ⁷⁰⁷⁸.

Clinical relevance over statistical significance

FDA reviewers have judged rescue data on clinical meaning, not p-values alone:

• Anjeso: Meloxicam reduced rescue doses in the first 24 hours with statistical significance (3.66 versus 4.38, and 3.97 versus 5.06). However, median time to first rescue was about 1 to 2 hours in both arms, and patients still needed rescue roughly four times a day. The clinical review found this raised "serious concerns" about clinical relevance and stated that 30 mg IV once daily was not suitable as an analgesic at the proposed dosing interval ⁶⁰. Anjeso was nevertheless approved in February 2020 for moderate to severe pain, alone or with non-NSAID analgesics. A later review linked the frequent rescue use to delayed onset and end-of-dose failure ⁷⁶.
• Journavx: Rescue use with suzetrigine and placebo was nearly identical in the bunionectomy study (69.5% versus 69.4% at 0-6 hours). In the abdominoplasty study, rescue use was lower with suzetrigine (64.5% versus 82.4% at 0-6 hours). FDA used these patterns when reading the pain-time curves ^164165169.

Opioid consumption and opioid-free endpoints

FDA has been cautious about opioid-sparing claims:

• Zynrelef (NDA 211988): Total opioid consumption through 72 hours and the proportion of patients who were opioid-free were key secondary endpoints. FDA found the consumption reductions statistically significant but their "clinical relevance was not clear" ⁶⁴. The difference was concentrated in the first 24 hours ⁸³⁹⁷. FDA viewed opioid-free status through Day 28 as potentially more meaningful, and that result was not statistically significant ⁸⁷.
• Anjeso: FDA stated that an opioid-sparing claim would need replicated evidence of clinical relevance, such as improvement in prespecified opioid-related safety outcomes or a clinically relevant increase in patients needing no opioids ¹⁰².

The 2026 draft guidance formalizes this position. It says "opioid-sparing" is unlikely to be meaningful as a labeling term. Labeling should instead describe the specific benefit of eliminating or reducing the need for opioid analgesics ¹⁴⁷. In acute pain, a claim of reduced opioid use should show a direct patient benefit, such as a clinically meaningful reduction in opioid-induced adverse reactions ⁵⁰.

How FDA handles rescue analgesia in the analysis

Design and data collection

The acute pain draft guidance requires protocols to prespecify the allowed rescue medications, "including the frequency, amount, and threshold of pain" at which they may be given. It warns that heavier rescue use in placebo arms can shrink the apparent treatment difference ⁴⁶. The rescue agent depends on the setting: NSAIDs, or opioids where opioids are typically required ⁴⁶. "All pain intensity measurements, including at baseline, should be obtained before rescue drug administration" ⁴⁶.

The local anesthetic draft guidance asks sponsors to prespecify the estimand, including the strategy for rescue medication ⁴⁹. Under ICH E9(R1), taking additional medication is an intercurrent event. Two strategies apply:

• A treatment-policy strategy uses pain outcomes regardless of rescue ¹³²¹⁴³.
• A hypothetical strategy estimates outcomes as if rescue had not been available. This relies on strong assumptions ¹²⁴¹²⁵.

The acute pain draft guidance does not specify an imputation method for pain scores after rescue ⁴⁶.

Imputation approaches seen in FDA reviews

Drugs@FDA reviews document a wide range of post-rescue rules:

Where FDA reviewers pushed back

Stopping follow-up at rescue. For Combogesic IV (NDA 215320), the applicant used pain data only up to the first pre-rescue score. The statistical reviewer found this unacceptable because it measured improvement only until first rescue, not over the agreed 48 hours ²⁶¹⁸⁷. Rescue was primarily oxycodone, with IV morphine as secondary rescue. The applicant ran a sensitivity analysis carrying the rescue observation forward for 4 hours (4-hour ROCF), and FDA requested a more conservative 2-hour version ^26183187. Under the 2-hour analysis, adjusted mean SPID48 was 36.68 for the combination versus 24.57 (ibuprofen), 19.32 (acetaminophen), and 17.49 (placebo), all p<0.001 ¹⁸⁶. The review describes SPID48 with each pre-rescue score carried forward for up to 2 hours as the primary analysis ^183187193.

Carrying the pre-rescue score forward too far. In the Olinvyk review, FDA rejected carrying the pre-rescue score from first rescue to the end of follow-up. FDA said this approach ignores both the waning of rescue effect and natural postoperative improvement, and "harshly penalizes" patients who used rescue ¹¹¹. FDA's primary approach carried the pre-rescue score forward for 6 hours, the dosing interval of the rescue medication. FDA described this as a hypothetical estimand and paired it with a treatment-policy sensitivity analysis that applied no post-rescue imputation ¹¹¹¹¹³.

BOCF when rescue is common or repeated. In the Zorvolex review, the reviewer noted that replacing all post-rescue scores with baseline flattened the mean curves after about 10 hours, because 82% to 97% of subjects used rescue ²⁴. In the Tivorbex review, the reviewer found baseline replacement after first rescue unreasonable when subjects rescued more than once, and based the efficacy conclusion on sensitivity analyses instead ⁴³.

Single imputation and discontinuations. For Journavx, FDA rejected LOCF for discontinuations due to lack of efficacy. FDA treated these as treatment failures and recommended baseline-based multiple imputation ³¹¹⁷⁵. Sensitivity analyses reduced the suzetrigine-placebo difference (for example, from 29.3 to between 23.7 and 25.3 in bunionectomy), but significance held ¹⁶³. FDA also recommended that, when rescue windows overlap, each pre-rescue score be carried forward only until its own window ends or the next rescue dose, whichever comes first. Results were essentially unchanged ⁴⁵¹⁷⁵. In the Seglentis review, the FDA statistical reviewer added multiple imputation because single-imputation BOCF underestimated variability when missingness was substantial ⁴¹.

Missing pre-rescue scores. For Posimir (NDA 204803), the protocol did not require a pre-rescue pain assessment. FDA advised a post hoc worst-score sensitivity analysis. The reviewer also ran an independent analysis that assigned a moderate score of 5 to 7 within a fixed 4-hour post-rescue window. Conclusions were consistent across both ²⁹.

Using observed post-rescue data as a sensitivity check

Several reviews treated analyses that ignore rescue as a conservative check. In the Zynrelef review, analyzing observed post-rescue scores would improve apparent outcomes in the comparator arms and so shrink the product's advantage. Persistent significance under that analysis was therefore supportive ²⁸. The Zynrelef review also integrated pain AUC and opioid consumption in a ranked analysis (Silverman method), since low pain with low rescue use is more informative than pain scores alone ²⁸.

Practical implications for sponsors

• Use SPID (or AUC of pain for prolonged-duration local anesthetics) over a window that matches the dosing interval and claimed duration. Support it with time-specific PID separation ^1234649.
• Define the rescue regimen and pain threshold in the protocol, and capture a pain score immediately before every rescue dose ⁴⁶⁴⁷.
• Prespecify an estimand-based rescue strategy. Recent reviews most often accept a pre-rescue score carried forward for a window tied to the rescue agent's duration (2 to 6 hours), with treatment-policy and multiple-imputation sensitivity analyses ^{4549111113186}.
• Avoid stopping efficacy follow-up at rescue, unlimited carry-forward, and single imputation for lack-of-efficacy dropouts. FDA reviewers have challenged each of these ^2631111.
• Treat time to rescue, rescue incidence, and opioid consumption as supportive evidence that FDA will judge on clinical relevance. Opioid-use labeling claims need a demonstrated patient benefit ^506064147.