Clinical Benefit and Risk Profile of TV-46000 for Patients with Schizophrenia as Assessed by Number Needed to Treat and Number Needed to Harm

Introduction

Schizophrenia is an often-debilitating, chronic mental health disorder, characterized by symptoms that can be categorized as positive (eg, hallucinations, delusions, disorganized thought, disorganized behavior), negative (eg, loss of motivation, social withdrawal, anhedonia, alogia, affective flattening), and cognitive (eg, impairments of attention, memory, executive functioning, speed of processing, social cognition).^1,2 Schizophrenia is associated with substantial health, social, and economic burden, and the diagnostic criteria include marked interference with at least one area of functioning, such as self-care, interpersonal relationships, education, or work. There is an abundance of evidence linking the symptoms of schizophrenia with reduced quality of life,^3–6 which has, in turn, been shown to affect clinical outcomes.^3–7 In terms of treatment, a meta-analysis found that longer times between first psychosis and first treatment were associated with poorer long-term outcomes, including more severe symptoms, lower likelihood of remission, and reductions in social functioning and global outcomes.⁸

Relapses, defined as exacerbations of symptoms after a period of improvement, are common in schizophrenia and have been shown to be detrimental to patients’ health.^9–12 For this reason, relapse prevention (assessed as time to occurrence) has typically been selected as the primary endpoint in registrational studies of antipsychotics for maintenance treatment in individuals with schizophrenia.^13–16 The exact criteria for relapse vary by study, but generally incorporate a combination of Clinical Global Impression (CGI) and Positive and Negative Syndrome Scale (PANSS) scores, schizophrenia-related hospitalization, suicidality, and violent behavior.^13–19 Due to the costs of relapse to the patient, caregivers, and society, and since relapse avoidance is a major goal of treatment, studies often aim to determine when relapse is imminent to take action, rather than wait until a relapse occurs; in these cases, impending relapse is often identified by similar criteria to relapse itself.^16,20–22 The goal of treatment is to minimize the negative effects of schizophrenia on patients by reducing symptoms and maintaining that reduction, ideally reaching remission and maintaining that stability.^23,24 Similar to relapse, the definition of remission varies across studies, but is generally characterized by sustained reductions in symptoms as measured by PANSS items.^25–27

Antipsychotic medications have demonstrated efficacy in addressing the symptoms of schizophrenia. However, nonadherence to oral formulations, which can result from inadequate efficacy or poor tolerability as well as limited insight, depression, and cognitive impairment, challenges the management of schizophrenia.^28,29 Currently available oral antipsychotics require daily dosing,^30,31 whereas second-generation long-acting injectable antipsychotics (LAIs) can be administered every 2 weeks to every 6 months (depending on the formulation), with newer formulations generally allowing for longer administration intervals.^32,33 Compared with oral antipsychotics, adherence to LAIs is more likely, which improves patient function and reduces the risks of relapse, hospitalization, and mortality.^30,34–38 In real-world analyses, earlier use of LAIs resulted in superior patient outcomes relative to use of oral antipsychotics.^39,40 Although both clinical trial data and treatment guidelines support the use of LAIs in managing schizophrenia, LAIs remain underused.^41–46

TV-46000 is a subcutaneously administered LAI that combines the second-generation antipsychotic risperidone with an innovative, copolymer-based drug delivery technology. TV-46000 is indicated in the United States (US) for the treatment of schizophrenia in adults as well as for maintenance treatment of bipolar I disorder in adults, either as monotherapy or as adjunctive therapy to lithium or valproate.⁴⁷ For the treatment of schizophrenia, TV-46000 can be administered once monthly (q1m) or once every 2 months (q2m), and thus offers another option among the several available risperidone-containing LAI formulations.⁴⁸ TV-46000 does not require oral supplementation or loading doses.⁴⁹ It is administered subcutaneously with a short-length needle, which can be preferred to intramuscular administration by some clinicians and patients.⁵⁰ Options for administration include two potential injection sites (abdomen or back of the upper arm) and doses ranging from 50 mg to 250 mg with a small injection volume.⁴⁹ TV-46000 is supplied in a prefilled syringe that must be kept refrigerated if stored >90 days. Although injection site pain is a common concern among patients, TV-46000 has not demonstrated differences in injection site pain compared with other LAIs.^51,52 TV-46000 has also demonstrated comparable safety and efficacy to other second-generation LAIs overall.⁵²

Regulatory approval of TV-46000 was based, in part, on the results of the randomized, double-blind, placebo-controlled, phase 3 RIsperidone Subcutaneous Extended-release (RISE) trial (NCT03503318).⁵³ In accordance with US Food and Drug Administration recommendations, RISE was designed as a placebo-controlled, randomized-withdrawal, relapse-prevention study that included a placebo group rather than an active control group.⁵³ Similar to other randomized-withdrawal studies, participants first underwent stabilization; in this case, they were first treated with oral risperidone during a 12-week, open-label stabilization stage. Also, similar to other randomized-withdrawal studies, the definition of stability in RISE was based on a combination of PANSS total scores and individual items as well as CGI of Severity (CGI-S) scores, among other criteria (see full definition of stability for RISE provided in Table 1).^14,53–55 After stabilization on oral risperidone, participants in the RISE study were randomized to subcutaneous TV-46000 (q1m or q2m) or placebo for the double-blind, relapse-prevention stage, which was of variable duration. The time to impending relapse (primary endpoint, defined in Table 1) was significantly prolonged with TV-46000 relative to placebo (q1m: hazard ratio [HR], 0.200; 95% CI, 0.109–0.367; P<0.0001; q2m: HR, 0.375; 95% CI, 0.227–0.618; P<0.0001).⁵³ The safety profiles of TV-46000 q1m and q2m were consistent with those of other oral and LAI risperidone formulations, with no new safety signals detected.

Table 1 Criteria for Impending Relapse, Stability, and Remission in the RISE Study53

P-values can indicate the presence of a statistically significant difference between treatment groups, but they do not quantify effect size or clinical impact. Although Cohen’s d is an effect size measure that is commonly reported by researchers, it is not clinically intuitive. Conversely, the number needed to treat (NNT) and the number needed to harm (NNH) are effect size measures that can be easily interpreted clinically, since they are expressed in “patient units” (ie, the number of patients who would need to receive one treatment versus another to observe one additional outcome of interest).⁵⁶ NNT estimates are used to describe beneficial or desired outcomes, while NNH estimates are used for disadvantageous or adverse outcomes. If the calculation does not yield a whole number, the point estimate is rounded up to the next whole number to avoid inflating the effect size. The ratio of NNH to NNT is known as likelihood to be helped or harmed (LHH) and provides an overall indication of benefit versus risk.⁵⁷

NNT and NNH estimates can help healthcare professionals understand the benefits and risks of a treatment and thus inform clinical decision-making.^56–58 Typically, single-digit NNT estimates (<10) for efficacy outcomes and at least double-digit NNH estimates (≥10) for safety/tolerability outcomes are suggestive of a compelling benefit–risk ratio, although clinical context ultimately drives treatment decisions.^56,57 Similarly, although a LHH >1 indicates a greater likelihood of benefit than of harm, the relative clinical significance of the measures chosen determines the practical value of the measure for an individual patient.^57,59 NNT, NNH, and LHH estimates can provide clinicians with clinically meaningful information on whether to use a specific medication for the maintenance treatment of schizophrenia in their adult patients. To this end, data from the phase 3 RISE study were used to calculate NNT, NNH, and LHH estimates for selected efficacy and safety/tolerability outcomes with TV-46000.

Materials and Methods

Patients and Study Design

RISE (NCT03503318) enrolled patients diagnosed with schizophrenia >1 year before screening who had ≥1 relapse in the 24 months prior to screening.⁵³ During the stabilization stage of the study, patients received oral risperidone once daily for 12 weeks, with the dose based on investigator judgement (aged <18 years: 2–4 mg/d; aged ≥18 years: 2–5 mg/d) (Figure 1). During the relapse-prevention stage, patients were randomized to subcutaneous TV-46000 q1m, TV-46000 q2m, or placebo. The primary endpoint was the time to impending relapse, as defined in Table 1, in the double-blind, randomization stage. As previously collected data were examined, separate institutional review board approval was not required for this post hoc analysis.

Figure 1 RISE study design. Abbreviations: EOS, end of study; EOT, end of treatment; PANSS, Positive and Negative Syndrome Scale; q1m, once monthly; q2m, once every 2 months; sc, subcutaneous.

Endpoints

NNT estimates for TV-46000 q1m versus placebo and for TV-46000 q2m versus placebo were calculated for the proportions of patients who did not meet criteria for impending relapse at week 24 and at end of treatment, who maintained stability (defined in Table 1) at end of treatment, and who achieved remission (defined in Table 1) at end of treatment. The all-cause discontinuation rate was evaluated as a measure of acceptability, reflecting both efficacy and safety/tolerability. Other efficacy endpoints included PANSS total, Personal and Social Performance Scale (PSP), CGI of Improvement (CGI-I), and CGI-S scores. The PANSS total score quantifies the severity of individual positive symptoms (7 items), negative symptoms (7 items), and general psychopathology (16 items), and change in the total score is commonly the primary outcome measure in clinical trials of acutely exacerbated schizophrenia.⁶⁰ The PANSS comprises 30 items that are scored on a 7-point scale that represents increasing levels of psychopathology (1=absent, 7=extreme).⁶⁰ The PSP is used to assess personal and social functioning, which is assessed across 4 main categories (socially useful activities, personal and social relationships, self-care, and disturbing and aggressive behaviors).⁶¹ It is a 100-point, single-item rating scale, subdivided into 10 equal intervals, with higher scores representing better functioning.⁶¹ The CGI-I is used to globally evaluate improvements in patients’ symptoms.⁶² This scale rates the improvement relative to symptoms on a 7-point scale, ranging from 1 (very much improved) to 7 (very much worse).⁶² The CGI-S assesses the global severity of a patient’s condition on a 7-point scale, where 1=normal, not at all ill, and 7=among the most extremely ill patients.⁶²

NNH estimates for each TV-46000 regimen versus placebo were calculated for safety endpoints, such as the proportion of patients who discontinued treatment due to an adverse event (AE) and who experienced ≥1 AE. AEs of interest, which included events common to second-generation antipsychotics (eg, akathisia, restlessness, somnolence, sedation), injection site–related AEs, and reproductive system and breast disorder AEs associated with hyperprolactinemia, were also examined. In this post hoc analysis, the specific AEs of clinical interest chosen were selected based on expert feedback and because these AEs are commonly associated with second generation antipsychotics. NNH values were calculated based on the RISE clinical trial safety data for TV-46000. Other safety/tolerability outcomes of interest included Abnormal Involuntary Movement Scale (AIMS) score, ≥7% increase from baseline in weight, and changes in prolactin, fasting glucose, fasting triglyceride, and high-density lipoprotein (HDL) levels. Total cholesterol was not measured in RISE, and low-density lipoprotein (LDL) data were sparse, as LDL levels were only measured in response to above-threshold triglyceride values (≥400 mg/dL) at specific points in the study.

LHH can be calculated for any therapeutic outcome versus any potential harm in order to quantify benefit versus risk. We selected avoidance of relapse and maintenance of stability as measures of overall benefit and compared them with discontinuation due to AEs as a measure of overall acceptability. In addition, we calculated LHH for the aforementioned therapeutic outcomes versus any injection site AE, regardless of whether the injection site AEs led to discontinuation from the clinical trial.

Statistical Methods

Efficacy analyses were conducted on the intent-to-treat population, which comprised all randomized patients. Safety/tolerability analyses were conducted on the safety population, which included all patients who received ≥1 dose of randomized treatment. NNT and NNH estimates were calculated for TV-46000 versus placebo as the reciprocals of the absolute risk differences between treatment groups.⁵⁶ NNT and NNH calculations were performed before percentages were rounded. Fractional NNT and NNH estimates were then rounded up to the next integer to avoid inflating the effect size estimate.

To aid in the interpretation of NNT and NNH estimates, 95% CIs were generated (where possible and appropriate) by taking the reciprocal of the lower and upper bounds of the Wald 95% CIs for the risk differences between TV-46000 and placebo and rounding up to the next whole number. A 95% CI for a NNT or NNH estimate that includes infinity is indicative of a nonsignificant (NS) difference (P>0.05) between treatment groups.⁵⁶ Rather than showing the noncontinuous 95% CI generated when statistical significance was not achieved, we instead denoted “NS.”

LHH is calculated by dividing the NNH estimate for the potential harm by the NNT estimate for the therapeutic outcome of interest. LHH estimates >1 suggest that the benefit will be encountered more often than the harm, whereas LHH estimates <1 suggest that the opposite is true.

Results

Patient Disposition and Characteristics

The intent-to-treat population comprised 543 adult patients (TV-46000 q1m: n=183; TV-46000 q2m: n=179; placebo: n=181). The safety population included 542 patients (TV-46000 q1m: n=183; TV-46000 q2m: n=180; placebo: n=179). Baseline demographic and disease characteristics were generally similar between treatment groups.⁵³

NNT Estimates

At the end of the treatment, 170 (92.9%) patients randomized to TV-46000 q1m, 156 (87.2%) randomized to TV-46000 q2m, and 128 (70.7%) randomized to placebo were without impending relapse, corresponding to NNT estimates versus placebo of 5 for TV-46000 q1m and 7 for TV-46000 q2m (Table 2). Rates of lack of impending relapse at week 24 were similar to those at end of treatment, with NNT estimates of 6 for TV-46000 q1m and 8 for TV-46000 q2m (Table 2). At the end of the treatment, 159 (86.9%), 143 (79.9%), and 110 (60.8%) patients randomized to TV-46000 q1m, TV-46000 q2m, and placebo, respectively, maintained stability, with NNT estimates of 4 for TV-46000 q1m and 6 for TV-46000 q2m (Table 2). Remission rates at the end of the treatment ranged from 16.6–23.5%, and the rate of avoidance of all-cause ranged from 61.2–63.1% (Table 2). There were no statistically significant differences between TV-46000 and placebo for either achieving remission or avoiding all-cause discontinuation.

Table 2 NNT Estimates for TV-46000: Impending Relapse, Stability, Remission, and All-Cause Discontinuation Rate

Proportionally more patients randomized to TV-46000 q1m (154; 84.2%) and q2m (138; 77.1%) avoided ≥20% worsening in PANSS total score than did those randomized to placebo (104; 57.5%) (Table 3). More patients randomized to TV-46000 than to placebo experienced ≥20% improvement in PANSS total score (TV-46000 q1m, 57 [31.1%]; TV-46000 q2m, 63 [35.2%]; placebo, 27 [14.9%]). NNT estimates for avoidance of ≥20% worsening in PANSS total score (4 and 6 for TV-46000 q1m and q2m, respectively) and for ≥20% improvement in PANSS total score (7 and 5, respectively) were in the single digits (and therefore suggestive of the benefit of TV-46000).

Table 3 NNT Estimates for TV-46000: PANSS, PSP, CGI-I, and CGI-S

Between baseline and end of treatment, 29 (15.8%), 23 (12.8%), and 15 (8.3%) patients randomized to TV-46000 q1m, TV-46000 q2m, and placebo, respectively, demonstrated at least one 10-point category improvement in PSP score (Table 3). The NNT estimate was 14 for TV-46000 q1m and not significant for TV-46000 q2m. At the end of the treatment, no statistically significant differences between TV-46000 and placebo were found for the proportions of patients with a PSP score of ≥51 (80.9–87.2%) or ≥71 (23.2–28.5%) (Table 3).

A greater percentage of patients randomized to TV-46000 q1m (28 [15.3%]) and q2m (35 [19.6%]) had symptoms that were considered much or very much improved (a CGI-I score of 1–2) at end of treatment than patients randomized to placebo (13 [7.2%]) (Table 3). The NNT estimate for achieving a CGI-I score of 1–2 was 13 for TV-46000 q1m and 9 for TV-46000 q2m. Approximately 46% of patients randomized to TV-46000 had a CGI-I score of 1–3, meaning that symptoms were at least minimally improved relative to baseline. In comparison, 29.3% of patients randomized to placebo achieved a CGI-I score of 1–3. The NNT estimate for TV-46000 q1m and q2m was 6 and 7, respectively. Proportionally more patients randomized to TV-46000 q1m (166 [90.7%]) and q2m (163 [91.1%]) versus placebo (144 [79.6%]) avoided CGI-I scores of ≥5. On this endpoint, the NNT estimate for both TV-46000 regimens was 9. The majority of patients (93.4–96.7%) in all three treatment groups also avoided CGI-I scores of 6 or 7, meaning that their symptoms were not considered to be much worse or very much worse relative to baseline; the NNT estimates for TV-46000 versus placebo were not statistically significant.

Proportionally, more patients randomized to TV-46000 q1m (171 [93.4%]) and q2m (163 [91.1%]) than to placebo (146 [80.7%]) avoided an increase in disease severity per the CGI-S (Table 3). The NNT estimate was 8 for TV-46000 q1m and 10 for TV-46000 q2m. In our study population, which was prospectively stabilized with oral risperidone prior to randomization, more patients assigned to TV-46000 q1m (42 [23.0%]) than to placebo (26 [14.4%]) experienced a reduction in disease severity, corresponding to a NNT estimate of 12. The difference between TV-46000 q2m (39 [21.8%]) and placebo on this endpoint was not statistically significant. Approximately 70% of patients randomized to TV-46000 compared with 58.6% of patients randomized to placebo had a CGI-S score of 1–3. The NNT estimate for both TV-46000 regimens was 9.

NNH Estimates

For all three treatment groups, the discontinuation rate due to an AE was low (1.7–3.9%), and the NNH estimates for TV-46000 q1m and q2m versus placebo were 190 and 45, respectively, and were not statistically significant (Table 4). In total, 111 (60.7%) patients randomized to TV-46000 q1m, 121 (67.6%) randomized to TV-46000 q2m, and 92 (50.8%) of those randomized to placebo experienced ≥1 AE. For AEs common to second-generation antipsychotics (eg, akathisia, restlessness, somnolence, sedation, extrapyramidal disorder), the NNH estimates for TV-46000 versus placebo were either not statistically significant or exceeded 10, indicating a favorable safety/tolerability profile for TV-46000 (Table 4). The lowest NNH estimate for either TV-46000 q1m or q2m was for extrapyramidal disorder (21 and 26, respectively). A total of 40 (21.9%), 39 (21.8%), and 22 (12.2%) patients randomized to TV-46000 q1m, TV-46000 q2m, and placebo, respectively, reported ≥1 injection site–related AE, and the NNH estimate for both TV-46000 regimens was 11. NNH estimates for specific types of injection site reactions (eg, pruritus, nodule, pain) for TV-46000 versus placebo were either not statistically significant or exceeded 10 (Table 4). There were no notable differences between TV-46000 and placebo on the rates of reproductive system or breast disorder AEs (Table 4).

Table 4 NNH Estimates for TV-46000: AEs, AIMS Score, and Weight Increase

The number of patients who experienced ≥7% increase in weight from baseline was 33 of 155 (21.3%) for TV-46000 q1m, 24 of 143 (16.8%) for TV-46000 q2m, and 16 of 155 (10.3%) for placebo, which corresponded to a NNH estimate of 10 for TV-46000 q1m and 16 for TV-46000 q2m (Table 4).

The first seven items on AIMS relate to movements of facial expression muscles, lips and perioral area, jaw, tongue, upper extremities, lower extremities, and trunk. The proportion of patients with an AIMS score of ≥3 on any of the first seven items or of ≥2 on at least two of the first seven items was 8.7% (n=16), 6.1% (n=11), and 6.1% (n=11) for TV-46000 q1m, TV-46000 q2m, and placebo, respectively (Table 4). The NNH estimate (95% CI) was 38 (not significant) for TV-46000 q1m and not calculable (no difference) for TV-46000 q2m.

Regarding laboratory parameters, statistically significant NNH estimates versus placebo were observed for TV-46000 q1m and q2m on the proportion of patients with a prolactin elevation >1× the upper limit of normal (ULN) at any post-baseline visit (NNH 4 and 8, respectively) and on the proportion of patients with a prolactin increase to >100 ng/mL (males) or >150 ng/mL (females) at any post-baseline visit (NNH 5 and 6, respectively) (Table 5).

Table 5 NNH Estimates for TV-46000: Laboratory Parameters

Not statistically significant were changes in glucose and lipids. NNH estimates for TV-46000 q1m and q2m versus placebo for shifts in fasting glucose (from <126 to ≥126 mg/dL at any time) were 16 and 18, respectively. NNH for fasting glucose ≥126 mg/dL at last observation carried forward were 98 and 30, respectively, NNH for shifts in fasting triglycerides (from <200 mg/dL to ≥200 mg/dL at any time) were 183 and 90, respectively, and NNHs for shifts in HDL (from >40 mg/dL to <40 mg/dL for males or >50 mg/dL to <50 mg/dL for females at any time) were 90 and no difference, respectively (Table 5). Data on changes in laboratory values by gender are summarized in Table 5.

LHH Estimates

The NNH estimate for the proportion of patients who discontinued treatment due to an AE was 197 for TV-46000 q1m, and the NNT estimate for avoiding impending relapse at the end of the treatment was 5, resulting in a LHH estimate of 39.4. This can be interpreted as TV-46000 q1m is 39.4× more likely to result in avoiding impending relapse rather than discontinuation because of an AE. The corresponding NNH and NNT estimates for TV-46000 q2m were 45 and 7, which yielded an LHH estimate of 6.4. Similarly, the LHH for maintaining stability at the end of the study versus discontinuation due to AEs was 49.3 for TV-46000 q1m and 7.5 for TV-46000 q2m. The LHH for avoiding impending relapse versus injection site–related AEs was 2.2 for TV-46000 q1m and 1.6 for TV-46000 q2m, and for maintaining stability versus injection site–related AEs was 2.8 for TV-46000 q1m and 1.8 for TV-46000 q2m.

Discussion

In this post hoc analysis of the phase 3 RISE study, both the q1m and q2m dosing regimens of TV-46000 exhibited a favorable efficacy and safety/tolerability profile as expressed by NNT, NNH, and LHH. NNT estimates for avoiding impending relapse at end of treatment, avoiding impending relapse at week 24, and maintaining stability at the end of the treatment were statistically significant and in the single digits versus placebo for TV-46000 q1m (range: 4–6) and q2m (range: 6–8). The 95% CIs for these endpoints overlapped for the TV-46000 q1m and q2m dosing regimens. These results underscore the efficacy of both dosing regimens in adults with schizophrenia. These results are further supported by the low rate of relapse (1.8%) observed in the 56-week, open-label, phase 3 Safety in Humans of TV-46000 subcutaneous INjection Evaluation (SHINE) study, which randomized patients without relapse in RISE or de novo study participants to either TV-46000 q1m or q2m.⁶³

The ability of TV-46000 to prevent relapse compares favorably with other LAIs. Based on data from placebo-controlled registration studies, NNT estimates (95% CI) for relapse prevention for LAI formulations of paliperidone, aripiprazole, and olanzapine range from 4 (3–5) to 8 (5–26).¹⁹ A recent systematic review and network meta-analysis reported that, relative to placebo, TV-46000 could reduce the risk of relapse at 6 months by 77% (q1m) and 60% (q2m), while relative risk reduction with aripiprazole q1m was 73%, and 66% (q1m) and 70% (q3m) with paliperidone palmitate.⁵² Moreover, all prolonged the time to relapse compared with placebo by as much as 80% with TV-46000 q1m or aripiprazole q1m, 74% with paliperidone palmitate q3m, and 72% with paliperidone palmitate q1m, and 63% with TV-46000 q2m.⁵² In contrast to similar efficacy profiles, differences in AE profiles differ among the molecules available as LAIs, and each formulation has a different array of pragmatic attributes, such as needle gauge and length, injection volume, reconstitution procedures, initiation with oral medication or multiple injections, refrigeration requirements, post-injection observation requirements, drug–drug interactions preventing use or requiring dosing adjustments, adjustments requirements for late or missed doses, and other concerns.⁶⁴

In the present study, the NNT estimate for remission at the end of the treatment was not statistically significant for either TV-46000 regimen; however, this finding was not surprising for several reasons. First, the remission criteria in RISE were stringent, and remission rates among patients with schizophrenia are generally low, even with antipsychotic treatment.⁶⁵ Moreover, patients with schizophrenia who participate in randomized clinical trials, including those assigned to placebo, generally have better outcomes than patients receiving care in the real-world setting, reducing signal detection.⁶⁶

Compared with placebo, significantly greater proportions of patients randomized to TV-46000 q1m or q2m avoided ≥20% worsening in PANSS total score, with NNT estimates of 4–6. Using data from the randomized, placebo-controlled, phase 3 ALK9072-003 trial, the NNT estimate for the avoidance of ≥20% reduction in PANSS total score for LAI aripiprazole was 4 (95% CI: 3–6).⁶⁷ The present analysis revealed that 5–7 patients would need to be treated with TV-46000 to experience ≥20% improvement in PANSS total score, but we found no statistically significant differences relative to placebo for the PSP-based endpoints. That the PANSS-based endpoints yielded significant, single-digit NNT estimates, while the PSP-based endpoints did not may be an artifact of the definition of impending remission, which did not include the consideration of PSP.

Differences between TV-46000 and placebo were not statistically significant for avoidance of a score of 6 (“much worse”) or 7 (“very much worse”) on the CGI-I, although other CGI-I measures examined did find statistically significant advantages for TV-46000 treatment. This is not surprising given the study population where patients with severe illness ordinarily requiring inpatient treatment were excluded. Estimates for all-cause discontinuation are complex to interpret, as patients were followed in a well-controlled, structured trial program with incentives to stay in the study and return for assessment and because reasons for discontinuation varied across patients.

NNH estimates in RISE were generally high (>10), supporting a favorable safety/tolerability profile for TV-46000 at both dosing intervals. The discontinuation rate due to an AE was low (1.7–3.9%) across all treatment arms, and comparable to that of other second-generation LAI formulations,⁵² with NNH estimates of 190 for TV-46000 q1m and 45 for TV-46000 q2m. This result was not unexpected, as only patients who were able to tolerate oral risperidone during the stabilization phase were randomized to receive maintenance treatment. LHH estimates evidenced that TV-46000 q1m and q2m are approximately 40× and 7× more likely, respectively, to help patients avoid impending relapse than to cause an AE that leads to discontinuation.

NNH estimates for TV-46000 versus placebo for movement-related AEs were either not statistically significant or exceeded 10 and were in line with the known safety profile of risperidone.^68–70 Results were similar irrespective of whether AE incidence was evaluated for individual-preferred terms (ie, akathisia, restlessness, extrapyramidal disorder) or for combined terms used in product labeling (ie, restlessness/akathisia, dyskinesia/tardive dyskinesia). The NNH estimate for akathisia in RISE was 63 for TV-46000 q1m and 59 for TV-46000 q2m. Based on data from placebo-controlled studies of the LAIs paliperidone and aripiprazole, the NNH estimates (95% CI) for akathisia were 28 (NS) and 14 (9–33), respectively.^27,67 In line with these results, a recent systematic review and network meta-analysis reported that for treatment-related AEs, TV-46000 q1m conferred a 52% and 38% risk reduction compared with paliperidone palmitate q1m and q3m, respectively.⁵² Injection site–related AEs were nearly two-fold more common with TV-46000 (21.8–21.9%) than with placebo (12.3%), with an NNH estimate of 11 for both dosing regimens. Local injection site reactions were generally infrequent, mild or moderate, and resolved fully; none were considered serious, and few led to treatment discontinuation.⁵³ These results are comparable to other second-generation LAIs which are administered by intramuscular injection.⁵² In the primary analysis of RISE, the most frequently reported treatment-related AE occurring more often in patients administered TV-46000 than placebo was injection site nodules (TV-46000 q1m, 7%; TV-46000 q2m, 7%; placebo, 3%).⁵³ In the present analysis, the NNH estimates for TV-46000 versus placebo for injection site nodules were not statistically significant. The LHH calculations for both lack of impending relapse and maintaining stability compared with any injection site–related AEs were greater than 1 for both TV-46000 q1m and TV-46000 q2m, supporting overall benefits of treatment with TV-46000.

Elevated prolactin is a known adverse effect of risperidone, with increases generally being greater in females than in males.^71–73 We evaluated different thresholds for prolactin increases (eg, 1× ULN, 2× ULN) and found the range of NNH estimates to be narrower for females (6–9 for TV-46000 q1m and 6–11 for TV-46000 q2m) than for males (4–9 and 6–55, respectively). Although NNH estimates for TV-46000 versus placebo tended to be lower for females, most estimates in both genders were not statistically significant, particularly for the proportions of patients with prolactin elevations >3× ULN.

Limitations

NNT and NNH are limited to dichotomous outcomes. The study itself was not powered to detect any differences in the NNT or NNH estimates between any of the treatment arms. LHH calculations are relevant only when the positive and negative outcomes chosen are of clinical importance for the individual patient. The original clinical study design limited the generalizability of our results, as the RISE study enrolled a highly selected population of patients, who achieved stability before randomization, into a placebo-controlled, withdrawal study design, and excluded patients aged >65 years or those with severe illness at screening (PANSS total score ≥100). These criteria were similar to previous maintenance studies, and focused the study on the long-term maintenance setting with patients in whom the efficacy of oral risperidone was previously established.

Conclusion

The robust, single-digit NNT estimates versus placebo for avoiding impending relapse and the maintenance of symptomatic stability support the efficacy of TV-46000 in adults with schizophrenia, while the high NNH estimates versus placebo for safety and tolerability endpoints of clinical interest suggest that TV-46000 is well tolerated. LHH estimates based on avoiding impending relapse and maintaining stability compared with discontinuations due to AEs and any injection site–related AEs were greater than 1 for both TV-46000 q1m and TV-46000 q2m, further supporting overall benefits of TV-46000. Together, these data provide clinically intuitive results on the benefit–risk profile of TV-46000 in the maintenance treatment of people with schizophrenia.