Studies evaluating pediatric patients (ages 17 years and younger) with GHD and other chronic conditions (MS, JIA, or IBD) requiring daily or weekly injectable treatments were eligible for inclusion in this review. Studies that included young adults (e.g., ages 18–25 years) were eligible if they also included patients younger than 18 years of age, meaning that some patients in the transition period could be included. Studies reporting on self- or caregiver-administered rhGH or an injectable drug, via an SC route, for ongoing daily or weekly treatment of chronic conditions (MS, JIA, or IBD) in pediatric populations were eligible for inclusion. Studies with and without comparators were eligible.

Publications were eligible for inclusion if they reported one or more of the following outcomes: (i) clearly defined standardized measures of adherence/non-adherence or compliance/non-compliance AND reported prevalence of adherence/non-adherence or compliance/non-compliance to injectable drugs or (ii) identifiable and measured (a) barriers to adherence OR (b) characteristics of patients, families/caregivers, providers, or institutions associated with adherence/non-adherence to treatment OR (c) properties of the treatment associated with adherence/non-adherence. Observational studies (prospective or retrospective; including cohort studies, cross-sectional studies, or surveys), or randomized controlled trials (RCTs) or non-RCTs (if reporting medication adherence or compliance) were eligible for inclusion. Publications were required to have been published in peer-reviewed journals between 2015 and 2020.

Searches were conducted in the following databases: Embase^® via Ovid and MEDLINE^® via Ovid. A detailed search strategy for each database is presented in Supplementary Methods 1 and Supplementary Tables S1, S2 . From the records identified in the database searches, one researcher (CR) excluded records that were irrelevant, which were then checked by a second researcher (KS). One researcher (KS) screened the titles and abstracts (Level 1) of the identified publications against the eligibility criteria. A second researcher (CR) screened 20% of the records to ensure agreement of screening decisions. The records identified in Level 1 were then subjected to full-text screening (Level 2) using the same process (initial screen by KS followed by validation of 20% of the records by CR). Screening discrepancies were discussed and resolved.

In addition to the Embase and MEDLINE searches, a targeted search of the Chinese literature was also conducted to identify additional publications on adherence to rhGH therapy in pediatric populations. An author fluent in Chinese (DL) conducted a literature search for publications in Chinese in the Wan Fang database (http://www.wanfangdata.com.cn/index.html).

One researcher (KS) extracted data from the included studies into a pre-specified data-extraction table (DET) in Microsoft Excel. A second researcher (CR) quality-checked 100% of the extracted data against the original publications for accuracy. For consistency, where available, data were extracted for 12 months or for the latest timepoint available. A list of data points extracted is provided in Supplementary Methods 2 . For studies reporting mean adherence, the population-weighted mean across studies was calculated. The population weights were the sample size divided by the total number of patients with mean adherence over 12 months in the included studies. The weights were multiplied by the adherence in each study, then summed to obtain the weighted mean.

As part of the data-extraction process, one researcher (KS) quality-assessed the RCTs for bias using the Cochrane Risk of Bias tool 2 (RoB2). Observational studies, including cohort studies, surveys, and cross-sectional studies, were quality assessed using the Newcastle–Ottawa Scale (NOS) for observational studies. A second researcher (CR) checked all of the quality assessment ratings; any differences were resolved through discussion. Risk-of-bias assessment was not conducted for studies identified in the targeted search of the Chinese literature.

A total of 1058 references were identified from the database searches; following removal of duplicates, 946 references underwent title/abstract screening, and 893 were excluded ( Figure 1 ). Full-text screening was then performed on 55 references (53 from literature databases and two identified from hand search), and a total of 23 publications met the inclusion/exclusion criteria and were subjected to data extraction and quality assessment.

The details of each study are provided in Supplementary Table S3 . Of the 23 studies that met the inclusion criteria, only one study (11) was in an indication (MS) outside of pediatric growth disorders. Study sample sizes ranged from 30 (12) to 13,553 (13) patients, with a median sample size of 95 patients. More than half of the sample population was male in all studies (median 41.5% female) except for the MS study, which was 70% female (11). The mean age of patients among the 17 rhGH studies that reported age ranged from 6 years (for patients receiving daily rhGH) (14) to 12.3 years (15). A total of 6 (26%) studies were from Italy (11, 16–20), five (22%) were from multiple countries (13, 14, 21–23), and three (13%) were from Spain (4, 12, 24). The remaining nine studies were each from a different country (15, 25–32).

Two of the studies were clinical trials. REAL 3 (14) was a Phase 2 randomized, active-controlled, open-label study of short-acting (daily) vs. long-acting (weekly, three different dosages) rhGH. SYNERGY (27) was a randomized, open-label study of growth outcomes following 12 months of daily rhGH treatment compared with 6 months’ delay followed by 6 months of daily rhGH; adherence for the 12-month treatment arm is reported in this review. Other than the REAL 3 trial (14), which reported adherence to short-acting vs. long-acting rhGH, the remaining 21 rhGH adherence studies were of daily GH.

The 23 included studies were subject to critical appraisal/quality assessment using the quality assessment tool appropriate for each study type. The reviewers (KS and CR) initially agreed on all ratings except for 4 of the observational cohort studies, which they then discussed until agreement was reached. Both of the RCTs (14, 27) were identified as having a moderate risk of bias using the RoB2 tool ( Supplementary Table S4 ). Both RCTs collected self-reported adherence, which could have been subject to bias. Further, the 2018 study by Chung et al. (27) was an open-label study, and the 2020 study by Sävendahl et al. (14) was partially blinded, which could have introduced bias. Study quality for the 18 observational cohort studies was assessed using the NOS ( Supplementary Table S5 ). Four studies were rated as good quality, and 14 as fair quality; none of the observational cohort studies had a comparator intervention or control group. Of the 18 studies, 12 did not control for differences among patients in the cohort, which may have influenced results. The quality of the three survey studies (15, 16, 21) was assessed using a modified NOS ( Supplementary Table S6 ). One study was of fair quality, and two were of low quality; none of the survey studies reported participation rates or information on non-respondents.

Broadly, adherence in these studies was assessed by several different methods (depending on the study): number of administrations/prescribed doses, as monitored by a medical device [n = 12/23 (52.2%)], self-reported number of missed doses for a given time period [n = 7/23 (30.4%)], or quantity of pharmacy fills or product supplied/quantity prescribed (n = 4/23 [17.4%]). Details of the definition of adherence used in each study are provided in Supplementary Table S7 . All of the rhGH studies that monitored adherence using a medical device used the easypod™ (Merck) electronic drug-delivery device. The studies also differed in terms of how they reported adherence; some studies reported mean or median adherence rate, whereas others reported the percentage of patients within different pre-specified categories (which varied across studies). Some studies reported adherence using more than one of these measures.

Of the 22 rhGH studies, 11 reported the 12-month mean adherence rate to daily rhGH ( Figure 2 ). One study compared daily rhGH against weekly rhGH (three different doses) and found that the mean 12-month adherence rate was 91.8% for daily treatment vs. 97.5%, 98.6%, and 96.3% for the weekly doses (14); the remaining rhGH publications reported adherence to daily rhGH. The population-weighted mean adherence rate among the 11 studies was 79.3%. Mean adherence rates across the 11 studies ranged from 73.3% (29) to 95.3% (24).

Of the 11 studies, Farfel et al. was different from the others because it measured adherence to rhGH based on the number of months with a pharmacy fill (and not days’ supply) (29). Eight of the 11 studies measured adherence using a medical device (all easypod™); the mean adherence rate across these eight studies ranged from 80.8% (in year 0–1, n = 95 patients) (22) to 95.3% (n = 232 patients) (24). In the single study on MS, the mean 12-month adherence rate (measured using a medical device; Supplementary Table S7 ) to interferon beta-1a among 40 pediatric patients was 67.5% (11).

A total of 8/22 rhGH studies reported median adherence to daily rhGH ( Figure 3 ), and of these, most (6/8) studies used a medical device (easypod) to monitor adherence. Median adherence to daily rhGH was high, ranging from 91% in a single-center retrospective observational study from Germany (31) to 99.2% in an open-label, multicenter RCT (14) ( Figure 3 ).

Many of the studies referred to adherence categories, using terms such as “adherent” or “non-adherent” or “poor,” “fair,” “good,” or “excellent” adherence; details are described in Supplementary Table S7 . Of the studies reporting adherence as a category, the largest number of studies defined patients as adherent (having “good adherence”) to daily treatment if they missed <1 dose per week or were administered >85% of prescribed doses (12, 13, 20, 24, 26, 30, 31). Six studies reported the percentage of patients with adherence >85% or with <1 missed dose per week ( Figure 4 ) (4, 12, 13, 16, 30, 31). In a 2015 study of 103 patients that defined adherence as possession of >85% of prescribed doses (pharmacy fills), Lass and Reinehr reported that 51% of patients were considered adherent (31). In the largest study population included in this review, Koledova et al. reported in 2020 that 77% of 8163 patients with 12 months of data had injected >85% of the prescribed dose of rhGH, as measured by the injection device (13).

A clinical trial by Chung et al. (27) defined non-adherence as receipt of <75% of expected injections based on self-reported diary entries. Based on this definition, adherence (receipt of ≥75% of expected injections) in patients who received treatment for 12 months was 93.27%. Using the same definition, adherence in the control group (patients untreated for the first 6 months and then treated with rhGH for 6 months) was 95.69% (27). Adherence would have been lower if the more commonly used definition, receipt of >85% of expected injections, was used to define adherence.

Two studies used ≥80% to define good adherence (15, 32). The study by Michaelidou et al. (32) was a single-center study from the United Kingdom that measured adherence based on the proportion of days covered, defined as the quantity of jet-injector device-heads delivered, multiplied by the length of time each head should last (1 week), divided by the number of days rhGH treatment was prescribed during the treatment period ( Supplementary Table S7 ) (32). According to this definition, 57.5% of patients had sufficient device-heads to be considered adherent.

The medical device easypod was the most commonly used method of measuring adherence. Of the 22 included rhGH studies, 11 used easypod to measure adherence to somatropin; these studies (identified by **) are described and adherence reported in Supplementary Table S7 ; most of these studies were sponsored by Merck. ECOS was a Phase 4, open-label study spanning 5 years and 24 countries that measured adherence to rhGH (Saizen, Merck) administered by the easypod. Mean adherence in the primary ECOS (23), which included patients with GHD, small for gestational age, and Turner syndrome from 24 countries, was 84.1%, with a median adherence of 93.7% (23).

In addition to the primary ECOS (23), five different ECOS sub-analyses (all prospective observational studies) met the SLR inclusion criteria. One of these was a 24-country sub-analysis restricted to children with idiopathic isolated GHD (22), wherein the mean (standard deviation [SD]) adherence in the first year was 80.8% (31.1%) with a median of 95.1%. In the second year, the mean (SD) adherence was 81.5% (23.0%), and the median was 92.9%. The remaining four ECOS were specific to certain countries (Mexico, Greece, Spain, and Italy), with a mean adherence ranging from 85.7% (Mexico) (25) to 95.3% (95% confidence interval: 93.3–97.2) (Spain) (24). Median adherence was above 92% for all four studies and ranged from 92.3% (Italy) (18) to 99.1% (Spain) (24).

In addition to the six ECOS, there were five other studies that used easypod to collect adherence data; three from Italy (17, 19, 20), one from Spain (12), and one large multi-country study of 13,553 patients described above (13). All were retrospective studies except for one by Loche et al. (19), which was a prospective observational study. Of the three studies that reported mean adherence rate, values ranged from 70% ± 13% (time period not reported) (17) to 96% over 12 months (20).

The targeted search of the Chinese literature identified two publications, only one of which met the SLR inclusion criteria. A 2020 randomized controlled trial by Li and Liu compared long-acting polyethylene glycol rhGH with short-acting daily rhGH. The investigators reported the number of missed doses out of the total number of prescribed doses for each treatment arm over 26 weeks (33). Adherence was very high in both treatment groups (>99%); however, the difference between adherence in the combined long-acting rhGH (4/1066 missed doses, or 99.62% adherence) and short-acting rhGH (70/7280 missed doses, or 99.04% adherence) was statistically significant (33).

Several barriers to adherence were identified in the included publications. Older age/adolescence was identified as a barrier in five studies (4, 12, 20, 31, 32). One study by Maggio et al. that looked at different age groups found a lower adherence in patients aged ≤9 years or ≥14 years compared with those aged 10–13 years (20). A longer treatment duration was also associated with a lower adherence, as described in six studies (4, 12, 16, 20, 29, 31). A number of barriers were identified in single studies, and these included: self-administration (vs. caregiver-administered) of rhGH (31); lower education level of the patient’s mother (4); prescribed seven doses of rhGH compared with six doses administered per week (20); and lack of family education/awareness and engagement (16). Device design could potentially also be a barrier as patients who considered the injection device to be ‘not convenient at all’ were less likely to be adherent when compared with patients who considered the device ‘very convenient’ (16). None of these studies reported statistically significant differences in adherence by gender or growth outcomes.

The studies identified in this review recommended multiple methods of improving adherence ( Table 2 ). Broadly, most of the recommendations were aimed at interactions between patients and the health care team, in addition to the use of reminder tools to improve adherence. Recommendations were also proposed for the injection device and the drug product ( Table 2 ). Recommendations for the injection device focused on improving device design in order to reduce pain and needle anxiety and to increase convenience. It was suggested that adherence might also be improved if the drug product did not require cold storage and was able to be injected less frequently than once per day.

One of the main findings from this review was the substantial heterogeneity in the way adherence was defined, measured, and reported in recent publications on rhGH. This heterogeneity makes it difficult to compare studies and to uncover trends and patterns across the literature. Therefore, we propose the following recommendations to enable more effective characterization/comparison of adherence data across future studies and publications. First, we propose standardizing the way adherence is defined, measured, and reported. In addition, adoption of standardized study designs, including adherence measures, definitions of “adherent” vs. “non-adherent,” and outcomes, is also recommended. Quantitative data describing the impact of missed doses on growth would also be valuable for determining the threshold of adherence to be reported. Furthermore, at a minimum, studies should report the presence or absence of any association between adherence and study duration, patient age, and rhGH indication. Lastly, in studies where the primary purpose is to study adherence, adherence should be measured using more than one method — agreement between different methods in the same study should be assessed. We believe that these measures should help address the substantial heterogeneity among the studies identified in the published literature.