Introduction

Osteoporosis is the most common bone disease and represents a major public health problem. [1, 2] It affects about 20% of women and 7% of men aged over 50 years, with half of them suffering an osteoporotic fracture (particularly hip, wrist and vertebral fractures) [3–7]. In 2000, nine million incident osteoporotic fractures worldwide were estimated [8].

Osteoporotic fractures result in physical and psychosocial consequences for the patient and place a major economic burden on healthcare systems [2, 3, 5, 8–12]. As a consequence of demographic changes and the resulting ageing population, the prevalence and incidence of osteoporosis and its related fractures will increase dramatically in the future [3, 10, 13].

A major problem regarding osteoporosis management is the poor adherence of patients to medical therapy and other treatment recommendations which can result from a lack of patient information [14–17]. Patient education programmes aim to improve osteoporosis knowledge among patients, their adherence, health beliefs, motivation and behaviour [16, 18, 19]. There is moderate to strong evidence that patient education is effective in other chronic diseases [20–23]. Also, several systematic reviews indicate that educational interventions may improve osteoporosis knowledge, adherence and health-directed behaviour (including calcium and vitamin D intake and physical activity) [24–31]. However, these reviews included studies with diverse patient groups, patient education was not always assessed as a single intervention and some of the reviews included both randomised controlled trials (RCTs) and observational studies.

The objective of this systematic review is to assess the effects of patient education on osteoporosis prevention and treatment results for people diagnosed with or at high risk of osteoporosis. We hypothesise that patient education will improve osteoporosis prevention and treatment results, i.e. increase the initiation of and improve the adherence to medical therapy (e.g. bisphosphonates), vitamin D and calcium intake as well as increase physical activity and thereby reduce the fracture rate and improve quality of life (QoL). In contrast to previous reviews, this review focusses on patient education onlyFootnote 1 assesses multiple endpoints relevant for osteoporosis management and, additionally, aims to provide a more detailed evaluation of the methodology and internal validity of included trials.

Methods

Two reviewers independently conducted the literature search, the risk of bias assessment and data extraction. Discrepancies were solved by discussion until consensus was reached.

Literature search

A literature search was carried out in various databases including PubMed, CINAHL, Embase, Cochrane Library (Cochrane Database of Systematic Reviews, Cochrane Central Register of Controlled Trials) and Education Resources Information Center (ERIC) to identify relevant studies. Search terms included osteoporosis, osteoporotic, education, educational and educate as well as terms describing the study design such as random, controlled and clinical. The primary search strategy for PubMed, using a combination of MeSH and free text terms, was as follows: (((((osteoporosis[MeSH Terms]) OR osteoporosis[Title/Abstract]) OR osteoporotic[Title/Abstract])) AND (((education[MeSH Terms]) OR education[Title/Abstract]) OR educat*[Title/Abstract])) AND ((((random*[Title/Abstract]) OR “random allocation”[MeSH Terms]) OR “randomized controlled trial”[Publication Type]) OR “controlled clinical trial”[Publication Type]). To achieve a high sensitivity, alternative search strategies have been applied, which consisted of less specific and alternative search terms such as patient information, self-management, fragility fracture, low bone mass/density and bone loss. Subsequently, Embase, the Cochrane Library, ERIC and reference lists of included studies were searched for further relevant literature.

Inclusion and exclusion criteria

Studies had to fulfil the following inclusion criteria to be classified as relevant:

  • Patients

    • Studies with a mixed or Caucasian population of (i) postmenopausal women, (ii) men and women aged 50 or above or (iii) men and women with confirmed osteoporosis or a history of fragility fractures.

  • Intervention

    • Educational interventions addressing osteoporosis mainly targeted at patients (i.e. interventions primarily targeted at healthcare professionals as well as studies with non-educational measures as the main element of the intervention were excluded)

    • Educational interventions including a personal part, i.e. a face-to-face delivery to patients, either individual- or group-based (interventions that solely consist of written or audio-visual material, patient and/or physician reminders and notifications, case manager or self-referral interventions were not considered)

  • Comparison

    • No intervention or usual care (i.e. written materials such as information sheets or brochures about osteoporosis)

  • Outcomes

    • At least one of the following outcomes regarding osteoporosis diagnosis and management: initiation of and adherence to pharmacological therapy, physical activity, calcium and vitamin D intake, changes in smoking behaviour, fractures, QoL and osteoporosis knowledge

  • Study design

    • Randomised controlled trials (including individual and cluster-randomised trials)

    • Studies written in English, French, Italian, Spanish and German with no restrictions on publication date

Methodological assessment

To assess the internal validity and potential limitations of the included trials, the Cochrane Collaboration’s tool for assessing the risk of bias has been applied [32]. Where possible, study protocols and information from trial registriesFootnote 2 were obtained, especially to assess the risk of reporting bias. Otherwise the judgement was based on a comparison of pre-specified outcomes in the methods section of the publication with those reported in the results section. The risk of bias assessment was done in accordance with the recommendations of the Cochrane handbook. [32]

The domain “blinding of outcome assessment” was assessed separately for two outcome groups. The first group included subjective outcomes such as QoL as well as behavioural outcomes such as exercise behaviour, calcium and vitamin D intake, medication adherence, etc. The second group covered objective outcomes such as bone mineral density (BMD) test, initiation/prescription of drug therapy, fractures and osteoporosis knowledge.

Data extraction and presentation of study results

The following information and data were extracted from each included study: study design, country, publication year, recruitment process, inclusion criteria, sample size, patients’ characteristics (e.g. age, gender), drop-out rate, length of follow-up, characteristics of the intervention (incl. delivery mode, scope and length, educator, educational material, non-educational components, follow-up interventions), comparison and outcome data. The presented outcome data are based on intergroup differences between the intervention and the control group (CG). Intragroup changes were not considered. The outcome data presented were based on the results at the end of the follow-up. That is, in trials with multiple time points of measurement, only the findings at the latest time point were consideredFootnote 3 Where outcome data was presented as percentages, the absolute differences and the 95% confidence interval (CI) were calculated if applicable.

All outcome results were summarised in the following outcome categories: osteoporosis management, lifestyle modifications, fractures, knowledge and QoL.

Statistical evaluations and meta-analyses were not feasible in this review due to the wide variety of applied endpoints. Even in cases where similar parameters have been assessed, varying outcome definitions, evaluation tools and time frames as well as different approaches of reporting data made statistical summaries difficult.

Results

Literature search

The literature search was carried out on the 29th of October 2016. Figure 1 shows the literature search and study selection process. The primary search strategy on PubMed identified 310 records, of which 30 were included after screening titles and abstracts. Overall, 35 publications were included for full-text screening after searching all predefined databases. Out of these, 20 were excluded and 15 fulfilled all inclusion criteria after full-text review. A list of excluded studies including the primary reason of exclusion is provided in appendix A. No additional articles were identified by screening the reference lists of included studies. In total, 15 articles (of 13 different trials) were included in this review.

Fig. 1
figure 1

Flow chart of the literature search and study selection process

Full size image

The 15 included articles were published between 2001 and 2013 and the majority of trials originated from the USA (n = 5) [34–38]. The other studies were conducted in Canada [33, 39, 40], Turkey [41, 42], Australia [43], Spain [44], Denmark [45, 46] and Finland [47]. For trials published twice, for this review, the analyses will be based mostly on Nielsen et al. (2010) [46] and Yuksel et al. (2010) [39] as they represent the main reports of the two trials.

Study characteristics

Twelve of the included RCTs used individual randomisation and one (Guilera et al. [44]) applied cluster randomisation of primary care practises.

A variety of recruitment processes was applied in the included trials. Participants were recruited from outpatient practises (n = 4) [38, 41, 44, 46], through public campaigns such as distributing flyers in the community or newspaper articles (n = 3) [34, 36, 43], from electronic databases (n = 2) [37, 47], hospitals (n = 1) [35] and pharmacies (n = 1) [39]. The majority of studies (n = 9) [33, 36, 38, 41, 43, 46, 47] were conducted in a single site, and three trials [39, 42, 44] were carried out in multiple sites.

The inclusion criteria varied across studies. About half of the trials included both men and women (n = 7) [33–36, 39, 43, 46], while the remaining studies included women only. While most studies did not specify whether their participants had osteoporosis or were at increased risk of osteoporosis/osteoporotic fractures, four studies only included patients diagnosed with osteoporosis [41, 42, 44, 46], and one trial solely focused on patients at high risk for osteoporosis and osteoporotic fractures [39].

Study characteristics regarding the sample size, patients’ characteristics, the intervention and comparison as well as the follow-up period are shown in Table 1.

Table 1 Study characteristics of included trials
Full size table

Methodological assessment

Eight trials [33–35, 38, 39, 41–43] used adequate methods for allocation sequence generation, and three [38, 39, 42] out of these also applied adequate methods of allocation concealment. The other five [33–35, 41, 43] did not provide sufficient information on allocation concealment and therefore the risk of bias remains unclear. The most commonly used method for sequence generation was computer-generated random number scheme, and the methods used for allocation concealment were either opaque, sealed envelopes or central randomisation. Three trials [36, 44, 46] did not report sufficient details on their randomisation system, and the risk of bias was unclear regarding both the sequence generation and allocation concealment process. In the remaining two trials [37, 47], the risk of bias was found to be high for these two aspects.

Because blinding of patient education is impossible, there was a high risk of performance bias and detection bias (for subjective and behavioural outcomes)Footnote 4 Among the 11 trials which measured objective outcomes, the risk of detection bias was low in 5 [33, 37, 39, 41, 47], unclear in 4 [34, 42, 43, 46] and high in 2 studies [35, 38]. In these two trials, the risk of bias was considered to be high as the study investigators relied on patients’ self-report (who were not blinded) instead of using more objective measures (e.g. obtaining the relevant data from medical records), which are less prone to bias.

The risk of bias due to incomplete outcome data was low in eight trials [33, 35, 36, 38, 39, 41, 43, 46], while it was considered to be high in the remaining five [34, 37, 42, 44, 47]. Reasons for a high risk of bias include overly high proportions of missing data and a considerable rate of drop-outs in the study groups.

Almost all studies (n = 11) presented all prespecified outcomes (prespecified in the trial register, study protocol or the “Methods” section of the report) and reported non-significant outcomes in the same manner as significant outcomes. In two trials [34, 45, 46], the risk of bias was considered to be high.

As patients, personnel and outcome assessment of subjective and behavioural outcomes could not be blinded in any trial, no study is completely free of bias. Overall, the general risk of bias is considered to be moderate to high in the included trials.

A summary of the risk of bias evaluation is shown in Fig. 2 and detailed risk of bias assessments of each included trial is provided in Appendix B. Blank spaces in the two subgroups of the domain “blinding of outcome assessment” indicate that the trial did not assess any outcomes of the respective outcome group.

Fig. 2
figure 2

Summary of the risk of bias assessment (RevMan [48] was used to create this figure) plus sign: low risk of bias; question mark: unclear risk of bias; minus sign: high risk of bias

Full size image

Study results

A great variety was observed regarding the types of outcomes assessed (Table 2). Eight different primary outcomes were assessed in 11 different trials. Some of these outcomes covered several aspects which were analysed separately in the trials (e.g. the outcome ‘appropriate osteoporosis management’ included initiation of pharmacological therapy as well as calcium and vitamin D intake). Only two primary outcomes were used in more than one trial: medication adherenceFootnote 5 (n = 4) and the composite endpoint of dual x-ray absorptiometry (DXA) scan/initiation of pharmacological osteoporosis therapy (n = 2). Two trials [34, 36] did not explicitly distinguish their outcomes as primary and secondary outcomes. Overall, the most common outcomes (primary and secondary) were calcium intake (assessed in six studies) as well as pharmacological treatment and vitamin D (each assessed in five studies).

Table 2 Outcomes, assessment methods and times of measurement of included trials
Full size table

Collectively, 17 different methodsFootnote 6 and instruments have been applied to analyse outcomes of interest. Most outcome data has been measured through patients’ self-report (questionnaires or interviews). In some studies, outcomes were measured at baseline and at the end of the study period, whereas in others, outcomes were assessed up to five times throughout the study (Table 2).

The outcome ‘composite endpoint of BMD testing or initiation of pharmacological treatment’—which was considered in two trials [35, 39]—revealed statistically significant intergroup differences in both of them. The intergroup differences with respect to calcium and vitamin D intake as well as osteoporosis knowledge were found to be statistically significant in favour of the IG in ≥50% of studies analysing these outcomes. In contrast, differences between the IG and the CG regarding pharmacological treatment, medication adherence, physical activity, fractures and QoL were found to be statistically significant in fewer than 50% of trials examining these endpoints. Overall, no clear association between statistically significant results and the delivery mode (group-based vs. individual education), the length or complexity of the educational programme, the sample size, the study duration, the type of outcome (primary vs. secondary outcome) or trials with CG which received educational material could be observed.

Osteoporosis management

Overall, the majority of trials (n = 9) analysed osteoporosis management outcomes.

Two trials [39, 35] had a composite endpoint, defined as the percentage of patients who received a BMD test (DXA scan) or osteoporosis treatment (i.e. a new prescription). Both trials [35, 39] showed statistically significant results in favour of the IG, i.e. the educational programme doubled the number of patients in whom osteoporosis was addressed (i.e. patients received a DXA scan or bisphosphonate therapy) compared to usual care (Table 3). In the trial of Yuksel et al. [39], the differences were driven by BMD tests, whilst in the trial of Gardner et al. [35], 47% of patients in whom osteoporosis was addressed received both (BMD testing and pharmacological treatment), 33% only had a BMD test and 20% received bisphosphonate therapy only.

Table 3 Differences in pharmacological treatment and medication adherence at the end of the follow-up
Full size table

The percentage of patients who received a DXA scan was analysed as a secondary outcome in Yuksel et al. [39]. A BMD test with DXA was performed in 22% of the IG patients and in 10% of the CG patients (p = 0.01, absolute difference 12%, 95% CI 3 to 21%).

Pharmacological osteoporosis treatment was examined in five trials [33, 37–39, 47]. Overall, the differences between the IG and the CG were statistically significant in favour of the IG in four out of nine outcomes (Table 3). When this parameter was restricted to the three trials [33, 37, 39] that used patient records or pharmacy data to collect outcome data, the findings regarding pharmacological treatment still remain inconsistent.

Medication adherence, compliance and/or persistence were analysed in four trials [38, 42, 44, 46]. The outcome definitions as well as the inclusion criteria with respect to osteoporosis medication varied considerable across studies. The results of these trials are summarised in Table 3 (except those of Tüzün et al. [42] who used a different approachFootnote 7 of outcome measurement without reaching statistical significance. Only one trial [46] reported a statistically significant higher proportion of adherent patients in the IG. When comparing the adherence rates across trials, a substantial variation can be found: the proportion of adherent patients varied between 16 and 92% in the IG and between 22 and 80% in the CG.

Lifestyle modifications

Calcium intake was assessed in six trials [33, 36–39, 47] with almost all showing a significantly higher proportion of patients taking calcium in the IG compared to the CG. Detailed results of this outcome are shown in Table 4 (except those of Plawecki et al. [36] who used a different approach of outcome measurementFootnote 8 without demonstrating statistically significant intergroup differences).

Table 4 Differences in lifestyle modifications at the end of the follow-up
Full size table

Vitamin D intake was evaluated in five trials. [33, 36, 37, 39, 47] Three [33, 37, 47] out of these five studies (60%) reported a significantly higher proportion of patients taking vitamin D in the IG compared to the CG. Detailed results of this outcome are shown in Table 4 (except those of Plawecki et al. [36] who used a different approach of outcome measurementFootnote 9 without demonstrating statistically significant intergroup differences).

Results for the outcomes physical activity and changes in smoking behaviour are shown in Table 4.

Alp et al. [41] also assessed physical activity in both study groups but did not analyse it statistically. Therefore, their results are not shown in Table 4. They report that 74% in the IG participated in regular physical activity (balance and weight-bearing exercises two or three times a week), while no behavioural changes occurred in the CG. However, it is important to note that the participants in the CG were specifically instructed to maintain their sedentary lifestyle.

Fractures

Fractures were assessed in four studies [33, 41, 42, 47]. Pekkarinen et al. [47] assessed hip fracture incidence as a primary outcome based on data from the National Hospital Discharge Register in Finland. During the 10-year follow-up, 12 women (1%) in the IG and 29 women (3%) in the CG sustained a hip fracture. This difference was statistically significant (p = 0.04, absolute difference: 2%, 95% CI 1 to 3%). After adjusting for baseline differences, the risk of hip fractures was reduced by 55% with the educational intervention. When any other fractures were also considered, significantly less women in the IG were admitted to hospital (59 (6%) vs. 95 (8%), p = 0.045). In the remaining trials, any differences were not statistically significant (p > 0.05).

Overall, the trial [47] with a long-term follow-up showed a statistically significant decrease in fractures in elderly women, while trials [33, 41, 42] with a short-term follow-up did not show a clear effect.

Osteoporosis knowledge

Osteoporosis knowledge was assessed in four trials [34, 39, 43, 46]. All four trials used multiple-choice questionnaires with true-false schemes, consisting of 20 to 27 questions. Although Gaines et al. [34] and Yuksel et al. [39] both used the “Facts on Osteoporosis Quiz” (FOQ), the results were presented in a completely different manner: Gaines et al. [34] reported the mean knowledge scores of the two study groups, while Yuksel et al. [39] stated the percentage of patients answering the questionnaire correctly, making direct comparison difficult.

Table 5 summarises the differences in osteoporosis knowledge between the IG and the CG at the end of the follow-up. In two [43, 45, 46] out of four trials (50%), the analyses showed significantly better knowledge scores in the IG at the end of the follow-up compared with the CG.

Table 5 Differences in osteoporosis knowledge at the end of the follow-up
Full size table

Quality of life

QoL was assessed in four trials (Table 6) [39, 41, 42, 44]. One of those assessed QoLs as a primary outcome and reported significant improvements in all domains of the SF-36 in the IG compared to the CG [41]. In the other three trials, there were no statistically significant differences [39, 42, 44].

Table 6 Differences in QoL at the end of the follow-up
Full size table

CG control group, IG intervention group, SF-36 36-Item Short Form Health Survey, QUALEFFO-41 41-item Quality of Life European Foundation for Osteoporosis questionnaire, OPTQoL Osteoporosis-Targeted Quality of Life questionnaire

Yuksel et al. [39] additionally assessed the generic health status (SF-12). The mean mental and physical component scores were similar across groups at the end of the follow-up (p = 0.97 and p = 0.98, respectively).

Discussion

Summary of main results

This systematic review assessed 13 different trials evaluating the effects of patient education on osteoporosis prevention and treatment results. In summary, patient education resulted in improved osteoporosis management (i.e. earlier BMD testing or initiation of pharmacological treatment), calcium and vitamin D intake as well as osteoporosis knowledge in more than 50% of the included studies. In contrast, differences between the IG and the CG regarding pharmacological treatment, medication adherence, physical activity, fractures and QoL were improved in fewer than 50% of trials. No clear association could be found between statistically significant results and characteristics of the intervention such as the delivery mode (group-based vs. individual education), the length or complexity of the educational programme. Because of these inconsistent results and moderate to high risks of bias of the included studies, a conclusive statement about the effects of patient education on osteoporosis prevention and treatment results cannot be drawn, and the results need to be interpreted with caution (Table 7).

Table 7 List of Excluded Studies
Full size table

The inconsistency (heterogeneity) across the included trials with respect to several aspects (e.g. design of the interventions) made it difficult to assess why some education programmes led to improvements in the outcomes of interest, while others did not. In cases of large or unexplained inconsistency, the GRADE group even recommends rating down the quality of the overall evidence which would weaken the potential conclusions deducible from this review [49]. However, the transparent disclosure of methodological drawbacks as summarised below is important for balancing the principal opportunities and risks associated with patient education in osteoporosis prevention (Table 8).

Table 8 Risk of Bias Assessment of Included Studies (The table has been adapted from Higgins et al. [23])
Full size table

Limitations and risks of bias of included trials

The overall risk of bias of the included trials was graded as moderate to high due to methodological limitations especially in the areas of randomisation processes, blinding of participants as well as personnel and outcome assessment.

As blinding of participants and personnel may not be feasible in educational intervention trials, it introduces a high risk of bias. Therefore, it is even more important to ensure objective and blinded outcome assessment where applicable [32, 50–52].

Certainly, outcomes such as QoL cannot be assessed objectively due to their subjective nature. Additionally, outcomes regarding lifestyle changes are difficult to assess objectively as the only objective means would be direct observation, which is not feasible. The risk of bias could be reduced, if patient surveys and interviews are conducted by trained and blinded interviewers, and if validated questionnaires/instruments are used more increasingly [32, 50, 51].

Several other limitations of the included trials need to be considered. First of all, there was considerable variance in how outcomes were defined and measured. For example, in some trials, calcium intake was assessed without differentiating between dietary sources of calcium and calcium supplements [33, 37–39], while in others, researchers focused on supplementary or dietary calcium intake only [47]. Similar issues exist for other outcomes such as vitamin D or medication adherence. A further reason for the variance of outcome measurement was a broad range of assessment methods (17 methods/instruments) used in the included trials, including the use of non-validated questionnaires or questionnaires developed by the study investigators [33, 35–38, 41, 42, 46, 47].

To allow for direct comparison and meta-analysis of study results, it is important for researchers to find consensus regarding the terminology and definition of outcome measures as well as on the most appropriate assessment methods for educational intervention trials. For instance, Madureira et al. [53] provide an overview of osteoporosis-specific QoL questionnaires, which may help investigators to select the most appropriate questionnaire for their trials.

Secondly, some reports lacked important information about the study population or the intervention, and the presentation of outcome data varied across studies. For example, important information about patients’ characteristics (e.g. existing osteoporosis risk factors) or about the intervention (e.g. about the educator, group size, scope and length) were not reported in all included trials. Heterogeneous reporting and limited information about specific aspects in some trials made direct comparisons at least difficult, if not impossible.

Third, some trials had multiple primary endpoints [33, 37, 38, 41] or did not differentiate between primary and secondary outcomes [34, 36]. Multiple analyses of the same data, without adjustment of sample size, are at risk for type I error (false-positive rate) [54]. However, chronic diseases such as osteoporosis and their management are multifaceted and the usage of appropriate statistical methods is important. Appropriate statistical methods include the frequentist/Bayesian approach, the Bonferroni or the Hochberg procedure and others [55–57]. None of the included trials with multiple primary endpoints considered the related issues or described any statistical methods undertaken to address them.

Fourth, the follow-up durations of most trials may be an issue of concern because more than half of the studies had follow-up periods of 6 months or less. As a result, the short follow-up times may limit the ability to detect important long-term effects of patient education. In particular, clinically relevant behavioural changes may require longer periods to be achieved, or in the case of significant behavioural changes observed during short study periods, a long-term evaluation is needed as some benefits may fade as time passes [58].

Fifth, the external validity of the included trials may be limited, which was acknowledged by the majority of authors. The study populations of the included trials may not be representative of osteoporosis populations due to specific settings and recruitment processes applied.

Finally, only one study assessing the incidence of osteoporotic fractures as a primary endpoint could be identified. Although the result of this trial [47] showed a significant reduction of hip fractures, this has to be interpreted with caution due to high risks of selection, performance and attrition bias.

Results of additional conference abstracts

During the literature search, conference abstracts of two RCTs were found of which the full text has not been published yet. Lin et al. [59] conducted a trial to evaluate an osteoporosis education programme with a 1-year follow-up. Participants were postmenopausal osteoporotic women (n = 120), and the educational intervention was delivered either individually or in group sessions. At the end of the follow-up, participants of the IG were more compliant than those in the CG (no intervention) [59]. In the second trial, McLeod et al. [60, 61] assessed the impact of a theory-based osteoporosis education and screening programme on calcium and vitamin D intake in men and women aged 50 years and above (n = 203). At the 6-month follow-up, they observed a statistically significant difference (p = 0.03) in calcium and vitamin D intake between the IG and the CG (usual care).

Comparison with other reviews

The results of this review are in line with the findings of a systematic review conducted by Jensen et al. [25], who investigated the effectiveness of multifaceted osteoporosis group education, including RCTs and observational studies. The authors reported that group education interventions may have a positive effect on lifestyle changes, adherence, knowledge and QoL, but no clear conclusions could be drawn from the included trials. In contrast, Smith et al. [26], who conducted a systematic review about healthcare professional-led education, found that eight out of nine trials showed improved adherence to osteoporosis medication. However, the review was based on different levels of evidence (i.e. RCTs, two quasi-experimental trials, and comparative studies).

Four other systematic reviews [24, 27, 29, 31] that investigated the effects of multiple interventions to improve osteoporosis treatment varied in some way in their inclusion criteria from this review, but their results were also consistent with our findings.

Strength and limitations

To our knowledge, this is the first comprehensive systematic review that focuses on the effects of patient education only, was not restricted to publication date, assessed multiple endpoints based on evidence-based recommendations in clinical guidelines and was based on RCTs only. Noteworthy, the RCTs included in the analysis undertook a critical assessment of their methodological quality.

Besides these strengths, this review also has some limitations. First, the search was restricted to specific languages. Although a publication bias could not be completely ruled out, our literature search identified no relevant articles in any other language, and no study was excluded due to language restrictions only. Secondly, study heterogeneity made direct comparison of interventions difficult and precluded any statistical aggregation and meta-analysis of study results.

Conclusions

To summarise, this review indicates that it is still unclear whether patient education is beneficial and whether it has a significant and clinically important impact on osteoporosis management results. Educational programmes for osteoporosis require further investigation within the context of well-conducted RCTs with longer follow-up periods (at least 4 or 5 years) and larger sample sizes. Furthermore, future trials should aim to recruit higher proportions of men because a substantial number of men are affected by osteoporosis.

The study methodology of future studies needs to be improved to minimise the risk of bias. Future studies need to apply adequate randomisation procedures and improve the description thereof. Additionally, future trials should include standardised, adequate and detailed descriptions of the education programme, the participants and the outcome measures. For instance, Jensen et al. [62] provide a comprehensive description of the educational programme, which may provide a good example of how to describe educational programmes in future studies. Researchers also need to find consensus on the terminology and definition of outcomes as well as on the most appropriate assessment methods and instruments. Several initiatives started developing standardised outcome sets, which will facilitate comparability of future RCTs (COMET Initiative, ICHOM)Footnote 10 Where available, researchers should refer to already existing guidelines and use standardised and validated instruments.