Literature selection

A total of 4917 studies was identified from the selected databases of PubMed, Scopus and EBSCOHost and 696 studies were eliminated due to duplication of titles. The remaining 4221 articles were screened, and 3980 articles were excluded because of irrelevant titles. Meanwhile, 44 articles involving other haemoglobin disorders such as sickle cell disease (SCD) and myelodysplastic syndrome (MDS), 6 non-English articles, 13 abstract-only articles, and 18 review articles were removed after assessing the abstracts. The eligibility of the remaining 160 full-text articles was evaluated. In this process, 82 publications without adherence/compliance measures, 5 studies focused on intervention or healthcare providers’ services, and 53 articles that only evaluated the determinants of chelation measure without comparing the outcomes among the compliant and non-compliant population were removed. Finally, only 20 articles were eligible for synthesis after the culling process. The summary of the literature selection process in this review is shown in Fig. 1.

Fig. 1

PRISMA 2020 Flow Diagram on the literature selection process for a systematic review of the impact of chelation compliance on health outcomes and health-related quality of life on thalassemia patients

Study characteristics

The characteristics of the studies included in the research are tabulated in Table 1. The published articles were of worldwide origins and were conducted in countries including the United Kingdom, Italy, the United States, Canada, Australia, Singapore, Malaysia, Thailand, Iran, Egypt, Syria, and India. Most of the studies were cross-sectional studies (N = 16, 80%) [17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32], followed by cohort studies (N = 3, 15%) [33,34,35], and only one randomised controlled trial [36]. The majority of the studies were single-centred (N = 13, 65%) and about 30% of the studies were conducted in multiple settings [18, 23, 30, 31, 33, 36], while the remaining study did not mention the setting [17].

A variety of medication measures was used across the studies. Most of the studies (N = 7, 35%) examined the rate of compliance using the frequency of ICT’s administration [17, 18, 20, 22, 24, 25, 29], followed by studies that evaluated patients’ compliance status using vial or pill count (N = 5, 25%) [19, 33,34,35,36]. Several studies (N = 3, 15%) used the Likert Scale [23, 27, 32], while some studies (N = 3, 15%) measured compliance using self-reported questionnaires such as Morisky Medication Adherence Scale (MMAS-8) [28], the Medication Compliance Questionnaire (MCQ) [30], and standardised questionnaire (not mentioned in the study) [31]. Meanwhile, a study [26] reported the compliance with medication possession ratio (MPR), and another study [21] did not mention the method of compliance measurement.

Table 1 Study characteristics of published literature on the impact of compliance on health outcome and health-related quality of life on thalassaemia patientsQuality assessment

The details of the methodological quality assessment of the studies are tabulated and summarised in Table 2. To sum up, more than half of the studies (N = 12, 60%) were considered high-quality studies, while the remaining studies (N = 8, 40%) were considered medium-quality studies.

Newcastle-Ottawa scale for cross-sectional study (N = 16)

Overall, most of the cross-sectional studies (N = 11) clearly defined the representative sample of the thalassaemia population, but there were no descriptions of the population in 5 of the studies [17, 18, 26, 29, 32]. Only 5 studies reported having a sufficient sample size following an appropriate formula of sample size estimation [18,19,20, 25, 30].

One study had an unsatisfactory response rate (less than 80%) [31] while there were no descriptions of the response rate nor the characteristics of the respondents and the non-respondents in 4 studies [20, 22, 28, 32].

All of the studies measured compliance rates using validated measurement tools or were able to describe the measurement tool except in the study by Haghpanah et al. (2013) [21].

Aside from that, 6 studies were not adequately designed or analysed to control confounders with regard to demographic factors such as education level, age, income, and treatment type [17, 19, 25,26,27,28]. For the remaining 10 studies, the comparability among the different outcome groups was shown and the confounding factors were controlled. In these studies, recruited samples were matched in age, gender or other significant factors [18, 22, 23, 25]; data analysed in multiple logistic regression while controlling multiple factors such as income levels, and family history that would affect compliance [20, 21, 29, 30]; or the outcome stratified based on gender [32], sociodemographic and clinical factors [31]. The data in all the 16 cross-sectional studies were assessed from reliable resources such as medical records, laboratory investigation, and validated questionnaires. They also used appropriate statistical tests and described the measurement of the relationship among outcomes of interest.

Newcastle-Ottawa scale for cohort study (N = 3)

All the cohort studies [33,34,35] had representative samples of the thalassaemia population and their non-exposed cohort were drawn from the same population as the exposed cohort. During the selection of the cohort study, only Wolfe et al. (1985) [33] reported that all the patients did not have a cardiac disease (outcome of interest). The comparability of cohorts on the basis of the design or analysis was appropriate and justifiable. Two studies recruited both arms from a similar sociodemographic background, while the other study [34] used logistic regression to explain the relationship between multiple variables and the outcome of interest. All the studies assessed the outcome through independent blind assessment. All studies had an adequate period for assessing the outcome ranging between 6 years to 12 years and complete follow-up of the subjects. Wolfe et al. (1985) [33] prospectively followed up the patients for 6 years without any dropouts, while the other studies retrospectively review patients attended the clinic for 12 years [34] and 10 years [35].

Newcastle-Ottawa scale for randomized controlled trial (N = 1)

The sole randomized controlled trial in this study is considered to have high quality [36]. The participants of the study were regular thalassaemia attendees of the Thalassaemia Centers, Ain Shams University, Egypt and Sultan Qaboos University Hospital Oman who had severe iron overload defined as serum ferritin > 2500ng/mL, liver iron concentration > 7 mg/g and cardiac T2* <20 and > 6 ms without heart dysfunction. The case and control groups were adequately defined and appropriately represented the targeted population. The cases and controls were age and gender matched. The data were then assessed through laboratory test and SF-36 health survey for both groups without dropouts.

Table 2 Methodological assessment of the studies through Newcastle Ottawa scaleCompliance towards ICT

The included studies that reported the rate of ICT compliance are summarised in Table 3. Among these studies, 8 evaluated the compliance rate toward DFO monotherapy [17,18,19,20,21, 26, 33, 34], one compared DFO, DFP and DFX [35], and another compared DFO and DFX [22]. The compliance rates were compared between the groups of DFO & DFP and DFO & DFX in one study [36], while a different study compared DFO, DFP and DFX, and measured overall compliance [27]. Two studies [23, 29] measured but did not report the compliance rate while the remaining studies examined compliance toward ICT as a whole.

Generally, the rate of ICT compliance ranged from 20.93 to 75.3%. Specifically, the ICT compliance rate of the different agents ranged from 48.84 to 85.1% for DFO, 87.2–92.2% for DFP, and 90–100% for DFX. Based on the frequency of ICT administration, the rates of compliance ranged from 27.5 to 85.1%. There were a variety of ways to define compliance in the included studies with compliance defined as at least 4–7 days per week on DFO [17, 22], the number of DFO infusions > 50% of the calculated doses per month [29] or > 80% of the prescribed doses per year [18], percentage of the day in a month administering DFO > 90% [20], using the drugs at least 27 out of 36 months [24], and > 50 mg/kg/day of DFO or > 30 mg/kg/day of DFX [25]. The average compliance rates based on the vial or pill count were reported to range from 20.93 to 100% [19, 33,34,35,36]. In comparison, the average compliance rate was reported to be 54.9% in the study using MPR [26].

Meanwhile, six articles evaluated compliance using self-reported measurements. Among these studies, four reported a range between 75.3 and 91.4% of patients being compliant to ICT [27, 30,31,32] with the study by Theppornpitak et al. (2021) [