Evidence review

A literature search was conducted which identified 814 references (see search strategy for full details). These references were screened using their titles and abstracts and 9 references were obtained and assessed for relevance.

Four RCTs identified from the search (Pi-Sunyer et al. 2015, le Roux et al. 2017, Davies et al. 2015 and Blackman et al. 2016) were included in this evidence summary. A summary of the included studies is shown in table 2 (see evidence tables for full details).

The European Public Assessment Report (EPAR) for liraglutide (Saxenda) states that the clinical development programme included 4 phase III randomised controlled trials (RCTs) in obese or overweight people with or without diabetes and a phase II dose-finding study. Three of the phase III RCTs have been included in this evidence summary (Pi-Sunyer et al. 2015, Davies et al. 2015 and Blackman et al. 2016). The fourth phase III RCT (in people who had already lost at least 5% of their initial body weight on a low calorie diet) and the dose finding study, which were included in the 9 references obtained and assessed for relevance, were excluded (see excluded studies for details).

Table 2 Summary of included studies

Study

Population

Intervention and comparison

Primary outcome

Pi-Sunyer et al. 2015 RCT

Adults (78% female) with either a BMI of 30 kg/m² or above or 27 kg/m² or above in addition to dyslipidaemia or hypertension (n=3,731)

Liraglutide 3.0 mg daily vs. placebo

Co-primary outcomes:

  • weight change from baseline

  • proportion of participants who lost at least 5% of their baseline bodyweight

  • proportion of participants who lost more than 10% of their baseline bodyweight

A fourth co-primary outcome of the study on time to onset of type 2 diabetes is assessed in le Roux et al. 2017.

le Roux et al. 2017 RCT

Adults with prediabetes at baseline in Pi‑Sunyer et al. 2015 (n=2,254)

Liraglutide 3.0 mg daily vs. placebo

Primary outcome: time to onset of type 2 diabetes

Davies et al. 2015 RCT

Adults (50% female) with type 2 diabetes and a BMI of 27 kg/m² or above (n=846)

Liraglutide 1.8 mg or 3.0 mg daily vs. placebo (the 1.8 mg dose is not licensed for weight management and the results from this arm are not discussed in this evidence summary)

Co-primary outcomes:

  • weight change from baseline

  • proportion of participants who lost at least 5% of their baseline bodyweight

  • proportion of participants who lost more than 10% of their baseline bodyweight

Blackman et al. 2016 RCT

Adults (72% male) with moderate or severe obstructive sleep apnoea and a BMI of 30 kg/m² or above (n=359)

Liraglutide 3.0 mg daily vs. placebo

Change in AHI from baseline

Abbreviations: AHI, apnoea-hypopnea index

The remaining 5 references were excluded. These are listed in excluded studies with reasons for their exclusion.

Clinical effectiveness

The EPAR for liraglutide (Saxenda) states that the clinical development programme included 4 phase III RCTs in obese or overweight people with or without diabetes and a phase II dose-finding study. Three of the phase III RCTs have been included in this evidence summary (Pi-Sunyer et al. 2015, Davies et al. 2015 and Blackman et al. 2016), in addition to a fourth RCT (le Roux et al. 2017), an extension of Pi-Sunyer et al. 2015. All of the included studies compared liraglutide 3.0 mg daily with placebo. Davies et al. 2015 included a liraglutide 1.8 mg daily group but the results from that group are not discussed in this evidence summary because that dose is not licensed for this indication.

Weight loss outcomes

Pi-Sunyer et al. 2015 and Davies et al. 2015 both had weight loss primary outcomes; le Roux et al. 2017 and Blackman et al. 2016 had weight loss secondary outcomes. There was a statistically significant weight loss with liraglutide 3.0 mg daily compared with placebo in all 4 studies (an estimated treatment difference of −5.4% to −4.0% in percentage body weight change from baseline across the 4 studies). The EPAR for liraglutide (Saxenda) states that relevant decreases in certain risk factors associated with obesity have been seen with loss of at least 5 to 10 % of initial body weight. Demonstration of a significant degree of weight loss of at least 10 % of baseline weight which is also statistically greater than that associated with placebo is considered to be a valid primary efficacy criterion in clinical trials evaluating new anti-obesity medicines. The NICE guideline on identifying, assessing and managing obesity recommends that orlistat should only be continued beyond 3 months if people have lost at least 5% of their initial body weight since starting treatment (although less strict goals may be appropriate in people with type 2 diabetes).

In Pi-Sunyer et al. 2015 after 56 weeks' treatment, the mean change in body weight from baseline was −8.0% in the liraglutide group and −2.6% in the placebo group with an estimated treatment difference of −5.4% (95% confidence interval [CI] −5.8% to −5.0%, p<0.001). There was a statistically significant higher percentage of participants who lost 5% or more body weight from baseline with liraglutide compared with placebo (63.2% versus 27.1%; difference 36.1%; odds ratio [OR] 4.8, 95% CI 4.1 to 5.6, p<0.001) and who lost more than 10% body weight from baseline (33.1% compared with 10.6%; difference 22.5%; OR 4.3, 95% CI 3.5 to 5.3, p<0.001).

Davies et al. 2015 comprised a population with type 2 diabetes, who were not taking a GLP-1 receptor agonist or insulin at baseline (around 11% treated with diet and exercise alone and around 56% treated with metformin alone). After 56 weeks' treatment, the estimated mean change in body weight from baseline was −6.0% in the liraglutide group and −2.0% in the placebo group with an estimated treatment difference of −4.0% (95% CI −5.1% to −2.9%, p<0.001). There was a statistically significant higher percentage of participants who lost 5% or more body weight from baseline with liraglutide compared with placebo (54.3% compared with 21.4%; estimated treatment difference 32.9%; 95% CI 24.6% to 41.2%, p<0.001) and who lost more than 10% body weight from baseline (25.2% versus 6.7%; estimated treatment difference 18.5%; 95% CI 12.7% to 24.4%, p<0.001).

After 160 weeks' treatment in a population with prediabetes (le Roux et al. 2017), the mean change in body weight from baseline was −6.1% in the liraglutide group and −1.9% in the placebo group with an estimated treatment difference of −4.3% (95% CI −4.9% to −3.7%, p<0.0001). In Blackman et al. 2016, in a population with sleep apnoea, after 32 weeks' treatment the mean change in body weight from baseline was −5.7% in the liraglutide group and −1.6% in the placebo group with an estimated treatment difference of −4.2% (95% CI −5.2% to −3.1%, p<0.0001).

The EPAR for liraglutide (Saxenda) reported that across the clinical trial programme the treatment effect of liraglutide versus placebo for mean body weight change from baseline was −5.2% (5.3 kg). However, there was a high dropout rate in the studies and some of the studies used a last observation carried forward (LOCF) method for missing data (see evidence strengths and limitations). Using a more conservative method for missing data (baseline observation carried forward), the EPAR estimates a treatment effect of −4.28%. The EPAR reported that the efficacy of liraglutide was lower in men; across the clinical study programme the treatment effect for mean body weight change from baseline for women (n=3,759) was −5.83% and the treatment effect for men (n=1,488) was −3.56%.

The EPAR reported that it is not clear if and how long the benefits of liraglutide will persist after treatment discontinuation and that it is likely that weight would return to baseline measures. In Pi-Sunyer et al. 2015, after 56 weeks, participants in the liraglutide group who did not have prediabetes at screening were randomly assigned to continue receiving liraglutide or to switch to placebo for 12 weeks. Participants who switched from liraglutide to placebo gained a mean of 2.91% bodyweight between weeks 56 and 68 compared with 0.69% for those participants who continued on liraglutide.

Additional outcomes

In le Roux et al. 2017, the primary outcome was time to onset of type 2 diabetes in a population of participants with prediabetes (see evidence tables for definition of prediabetes and definition of how type 2 diabetes was diagnosed). Over the 160−week treatment period, 2% of participants in the liraglutide 3.0 mg daily group developed type 2 diabetes compared with 6% in the placebo group (hazard ratio 0.21, 95% CI 0.13 to 0.34, p<0.0001).

In Blackman et al. 2016, the primary outcome was change in the apnoea-hypopnea index (AHI: apnoea or hypopnea events per hour of sleep) from baseline. From a baseline AHI of 49.0 events per hour in the liraglutide 3.0 mg daily group and 49.3 in the placebo group, there was a mean reduction of −12.2 in the liraglutide group and −6.1 in the placebo group after 32 weeks' treatment (estimated treatment difference −6.1 events per hour, 95% CI −11.0 to −1.2, p=0.015). The clinical significance of this difference is unclear; the EPAR reported that the interpretation of the sleep apnoea related outcomes is difficult because there is no established margin for clinical relevance.

In Pi-Sunyer et al. 2015, there were statistically significant changes from baseline with liraglutide compared with placebo for secondary outcomes including change in waist circumference, BMI, change in systolic blood pressure and change in a fasting total cholesterol (see results tables for details).

An overview of the results for clinical effectiveness can be found in results tables.

Safety and tolerability

There are several special warnings and precautions for use in the summary of product characteristics (SPC) for liraglutide (Saxenda). The SPC includes warnings on pancreatitis, cholelithiasis and cholecystitis, thyroid disease, heart rate, dehydration and hypoglycaemia in people with type 2 diabetes. The SPC recommends that a reduction in the dose of concomitant insulin or sulfonylurea should be considered when liraglutide (Saxenda) is initiated to reduce the risk of hypoglycaemia. The SPC reports a number of patient groups for whom the safety and efficacy of liraglutide for weight management has not been established and it states that use in these patient groups is not recommended. This includes people aged 75 years and older, people with severe renal impairment, people with severe hepatic impairment, people with congestive heart failure class III to IV and people with obesity secondary to endocrine or eating disorders or obesity caused by another medicinal treatment. The SPC also recommends that liraglutide is not recommended for use in people with inflammatory bowel disease and diabetic gastroparesis. See the liraglutide (Saxenda) SPC for further details.

The following are reported as very common (1 in 10 or more) adverse reactions in the liraglutide (Saxenda) SPC: nausea, vomiting, diarrhoea and constipation. Common adverse reactions (from 1 in 100 to 1 in 10) reported in the SPC include hypoglycaemia, insomnia, dizziness, dysgeusia, dry mouth, dyspepsia, gastritis, gastro-oesophageal reflux disease, flatulence, eructation, upper abdomen pain, abdomen distension, cholelithiasis, injection site reactions, asthenia, fatigue, increased lipase and increased amylase.

The EPAR for liraglutide (Saxenda) reported that the general adverse event profile is in-line with that for liraglutide (Victoza) for type 2 diabetes. The safety database that was considered in the EPAR for Saxenda included 5,813 people, with 2,341 people taking the 3.0 mg dose for at least 1 year. The most commonly reported adverse events were gastrointestinal disorders. These included nausea, diarrhoea, constipation and vomiting (reported respectively in 39.3%, 20.9%, 19.4% and 15.7% of participants taking liraglutide 3.0 mg compared with 13.8%, 9.9%, 8.5% and 3.9% of participants taking placebo).

In people without type 2 diabetes, no severe hypoglycaemic events were reported. In people with type 2 diabetes severe hypoglycaemia was reported by 0.7% of people treated with liraglutide 3.0 mg and 1% of people treated with liraglutide 1.8 mg and occurred only in people taking concomitant sulfonylureas. The EPAR stated that injection site reactions were reported more frequently with liraglutide than placebo (13.9% compared with 10.5%). The EPAR further comments that this incidence rate was higher than that reported in completed studies with liraglutide (Victoza) for type 2 diabetes (which was 2.9% for liraglutide (Victoza) compared with 1.5% for comparator).

Overall, the EPAR concluded that the higher liraglutide (Saxenda) dose did not appear to increase adverse event rates compared with the lower liraglutide (Victoza) dose, apart from gastrointestinal disorders which were more frequent. The EPAR stated, however, that there is currently insufficient data to assess if uncommon events (pancreatitis/neoplasms) occur more frequently with liraglutide 3.0 mg daily compared with liraglutide 1.8 mg daily. In addition, the EPAR reported that gallbladder adverse events occurred more frequently with liraglutide 3.0 mg daily compared to 1.8 mg daily. The EPAR stated that the increase in pulse rate associated with liraglutide does not appear to be dose-related.

In Pi-Sunyer et al. 2015, the most common adverse events were gastrointestinal disorders. However, 94% or more were classed as mild or moderate in severity. Gastrointestinal adverse events were the most common reason that participants in the liraglutide group withdrew from the study (6.4% [159/2,481] with liraglutide compared with 0.7% [9/1,242] with placebo). Nausea and vomiting occurred most frequently within the first 4 to 8 weeks after starting treatment with liraglutide. The change in mean resting pulse (beats per min) from baseline was greater in the liraglutide group compared with the placebo group (statistically significant), estimated treatment difference: 2.4 beats per min (95 CI 1.9 to 3.0, p<0.001).

Gallbladder adverse events were more common in the liraglutide group than the placebo group (2.5% [61/2481] compared with 1.0% [12/1242]). There were 11 confirmed cases of pancreatitis that occurred during the study and follow-up period, 10 occurred in 2,481 participants in the liraglutide group (0.4%, 0.4 events per 100 patient-years at risk) compared with 1 case in the 1,242 participants in the placebo group (less than 0.1%, less than 0.1 events per 100 patient years at risk). Nine of the 10 cases in the liraglutide group were graded as mild. Six participants (5 in the liraglutide group) had gallstone related pancreatitis.

The incidence of neoplasms was similar in the liraglutide group and placebo group (1.9 per 100 patient years at risk and 2.4 events per 100 patient years at risk, respectively). There were more cases of malignant and premalignant breast neoplasms in the liraglutide group (10 events in 9 women in the liraglutide group compared with 3 events in 3 women in the placebo group). Most women with events had above average weight loss. There were no cases of medullary thyroid carcinoma or C-cell hyperplasia and liraglutide treatment did not increase serum calcitonin concentrations. Four confirmed thyroid disease events occurred in 3 participants in the liraglutide group (3 cases of thyroid cancer and 1 case of autoimmune thyroiditis, which occurred after treatment had been stopped).

le Roux et al. 2017, was an extension study to Pi-Sunyer et al. 2015 which included participants diagnosed with prediabetes at baseline. Participants included in this extension study had liraglutide 3.0 mg for up to 160 weeks treatment. Again, gastro-intestinal disorders were the most common adverse events reported and the most common cause of withdrawal from the study (8% [118/1,501] in the liraglutide group compared with 2% [11/747] in the placebo group). Gallbladder related adverse events were more common with liraglutide than with placebo (5% [74/1,501] with liraglutide compared with 2% [13/747] with placebo). Pancreatitis and neoplasms were assessed over 172 weeks. There were 10 confirmed cases of pancreatitis in the liraglutide group (0.7%, 10/1,501) and 2 in the placebo group (0.3%, 2/747). Eight of the cases in the liraglutide group occurred during the first year of treatment. As seen in the 56-week period of the study (reported in Pi-Sunyer et al. 2015), more cases of malignant and pre-malignant breast neoplasms were reported in the liraglutide group than the placebo group (10 events in 9 women, 7 of which occurred in the first year of treatment, compared with no events in the placebo group).

Gastrointestinal adverse events were also the most common adverse events reported in Davies et al. 2015 and Blackman et al. 2016. In Davies et al. 2015, gastrointestinal disorders occurred more frequently in the liraglutide 3.0 mg group than the liraglutide 1.8 mg group (65.2% [275/422] compared with 56.2% [118/210]). There were more severe hypoglycaemic events in the liraglutide groups than the placebo group in Davies et al. 2015 (which was conducted in a population with type 2 diabetes). There were 5 events in 3 participants in the liraglutide 3.0 mg group, 3 events in 2 participants in the liraglutide 1.8 mg group and no events in the placebo group. All of the severe hypoglycaemic events occurred in participants who were taking concomitant sulfonylureas.

An overview of the results for safety and tolerability can be found in results tables.

Evidence strengths and limitations

The clinical development programme for liraglutide (Saxenda) for weight management included 4 double-blind RCTs, 3 of which are discussed in this evidence summary (Pi-Sunyer et al. 2015, Davies et al. 2015 and Blackman et al. 2016). A fourth RCT (le Roux et al. 2017), an extension of Pi-Sunyer et al. 2015, is also included. These RCTs included adults who were obese or overweight with a variety of weight-related co-morbidities including dyslipidaemia, hypertension, type 2 diabetes and sleep apnoea. In le Roux et al. 2017, participants had treatment with liraglutide 3.0 mg daily for up-to 160 weeks. In all of the studies, participants received regular structured lifestyle advice on weight loss interventions based on increasing physical activity and diet. It is unclear, if similar weight losses to those seen in the strictly controlled situations of clinical trials would be achievable in practice.

All of the studies discussed in this evidence summary compared liraglutide with placebo. There are no published double-blind RCTs which compare liraglutide (Saxenda) to other medicines for weight management. A phase II dose ranging RCT and extension study (Astrup et al. 2009 and Astrup et al. 2012) did include an orlistat arm. However, numbers in each treatment group were small and the orlistat arm of the study was open-label.

In Pi-Sunyer et al. 2015 and Davies et al. 2015 the primary outcomes on weight loss were assessed after 56 weeks treatment in all patients, which does not reflect the product license for liraglutide (Saxenda). The summary of product characteristics recommends that treatment should be discontinued after 12 weeks on the 3.0 mg daily dose (recommended maintenance dose) if patients have not lost at least 5% of their initial body weight. However, as highlighted in the EPAR for Saxenda, the 'stopping rule' for discontinuing liraglutide after 12 weeks in participants who have not responded to treatment to protect people from long-term use of a non-effective therapy, was only agreed during the marketing authorisation assessment process.

For all 4 studies, people who had taken part in an organised weight loss programme or taken any other medicine for weight management in the previous 3 months were excluded from the studies. In clinical practice, it is likely that adults who are being considered for treatment with liraglutide (Saxenda) for weight management will have previously tried other weight management options.

There was a high drop-out rate in all of the 4 studies discussed in this evidence summary. This ranged from 26% to 47% in the liraglutide 3.0 mg group and from 21% to 55% in the placebo group. In Pi-Sunyer et al. 2015, le Roux et al. 2017 and Blackman et al. 2016, missing values were accounted for using the last observation carried forward (LOCF) method, which carries forward the last post-baseline outcome or measurement for participants who have dropped out of the study. As reported in the EPAR for Saxenda, LOCF is not considered to be a conservative method for dealing with missing data. For example, when considering the weight loss outcomes, it could be expected that after treatment discontinuation baseline weight would be reached again after a few weeks or months. The EPAR considered that baseline observation carried forward would be a more conservative method for dealing with missing data. Davies et al. 2015 used a multiple imputation method for missing data for the 3 co-primary outcomes, which may have reduced bias.

In le Roux et al. 2017, the primary outcome of time to type 2 diabetes was assessed at week 160. Participants and investigators were masked to treatment allocation during the entire trial (160 weeks plus the 12-week off treatment follow-up period). However, the manufacturers were unmasked to treatment allocation at week 56, which may have introduced bias. The manufacturers provided logistical support during the study, collected the data and performed the statistical analysis.

The primary outcome in Blackman et al. 2016, was change in the apnoea-hypopnea index (AHI: apnoea or hypopnea events per hour of sleep) from baseline. The outcome was assessed during an overnight clinic stay which may not be applicable to clinical practice. In addition, as highlighted in the EPAR interpretation of this outcome is difficult as there is no established minimal clinically importance difference for AHI.

In Davies et al. 2015, secondary endpoints were considered exploratory because no controls were applied for multiple comparisons. Comparisons between the liraglutide 1.8 mg and 3.0 mg doses were considered exploratory for the same reason. Pi-Sunyer et al. 2015, le Roux et al. 2017 and Blackman et al. 2016 also had no adjustments for multiple testing for the secondary outcomes. Therefore, any statistical analysis provided for these outcomes should be interpreted with caution.

The SPC for liraglutide (Saxenda) reports a number of patient groups for whom the safety and efficacy of liraglutide for weight management has not been established. This includes people aged 75 years and older, people with severe renal impairment, people with severe hepatic impairment, people with congestive heart failure class III to IV and people with obesity secondary to endocrine or eating disorders, or obesity caused by another medicinal treatment.

An overview of the quality assessment of each included study can be found in evidence tables.