The impact of textual elements on the comprehensibility of drug label instructions (DLIs): A systematic review

Introduction Correct interpretation of drug labels instructions (DLIs) is needed for safe use and better adherence to prescribed drugs. DLIs are often too difficult for patients, especially for those with limited health literacy. What is yet unknown, is how specific textual elements in DLIs (e.g., the presentation of numbers, or use of medical jargon) and patients’ health literacy skills are related to the comprehension of DLIs. In order to provide concrete directions for health professionals on how to optimize drug prescriptions, we performed a systematic review to summarize the available research findings on which textual elements facilitate or hinder the correct interpretation of DLIs in relation to patients’ health literacy. Method A systematic search was performed in PubMed, EMBASE, PsychINFO, and Smartcat (until April 2019) to identify studies investigating textual elements that facilitate or hinder the correct interpretation of DLIs in relation to patients’ health literacy. Results A total of 434 studies were identified of which 28 studies met our inclusion criteria. We found that textual elements contributing to the correct interpretation of DLIs were: using explicit time periods in dosage instructions, using plain language, presenting numbers in a numerical format, and providing DLIs in patients’ native language. Multistep instructions per instruction line, using abbreviations and medical jargon seem to hinder the correct interpretation of DLIs. Although health literacy was taken into account in a majority of the studies, none of them assessed the effectiveness of specific textual elements on patients’ comprehensibility of DLIs. Conclusion Based on our findings, we provide an overview of textual elements that contribute to the correct interpretation of DLIs. Optimizing the textual instruction on drug labels may increase the safety and adherence to prescribed drugs, taking into account that a significant proportion of patients has low health literacy.

Introduction apply health information to exert greater control over life events and situations [20]. In sum, this classification shows functional health literacy is part of the broader construct of health literacy. Moreover, the levels of health literacy progressively allow for greater patient autonomy and empowerment [20,21].
Many different instruments have been developed to assess patients' health literacy which vary in adopted method and purpose [22]. For example, the Rapid Estimate of Adult Literacy in Medicine (REALM), Test of Functional Health Literacy in Adults (TOFHLA), and the Newest Vital Sign (NVS) all focus on functional health literacy, but differ in adopted method. The REALM tests the ability to read and pronounce 66 medical terms [23] whereas the TOFHLA consists of a 50-item reading comprehension and 17-item numerical ability test [24]. The NVS, on the other hand, consists of a reading comprehension and numerical ability test in which patients answer 6 questions about the information in a nutritional label for ice cream [25]. Other instruments, such as the Functional Communicative Critical Health Literacy (FCCHL) and the Set of Brief Screening Questions (SBSQ) assess patients' perceived health literacy skills, but differ in purpose and adopted method. For example, the FCCHL focuses on functional, interactive, and critical health literacy skills. It consists of 14 statements using 4-point Likert scales as response options [19]. The SBSQ only focuses on functional health literacy skills and consists of 3 statements using a 5-point Likert scale as response options [23] In sum, most instruments focus on functional health literacy skills, but the variation between the instruments in adopted method and purpose makes it hard to compare the results from health literacy interventions.
Patients with limited health literacy skills often experience difficulties in reading, understanding, and applying DLIs compared to patients with adequate health literacy [12,14,26]. Therefore, it is important to take health literacy into account when evaluating the comprehensibility of DLI's (cf., [27]). Especially as research shows 46 per cent of the Europeans have limited health literacy skills, varying from 29 per cent in the Netherlands to 62 per cent in Bulgaria [28]. Also, 14 per cent of the American population have limited health literacy skills [29]. As the processing of textual instructions poses a challenge for patients with limited health literacy, much research has been conducted on how visual elements, such as icons, pictograms, and graphics, can be used to improve the comprehensibility of DLIs [7,8,30]. However, mixed results have been found regarding their effectiveness. Wolf et al. [13] found simplified textual DLI with icons were more likely to be interpreted correctly by patients with marginal and limited health literacy compared to simplified textual DLIs alone. However, other studies report that icons on drug labels were frequently misunderstood by patients with limited health literacy [7,8,30]. As far as we know, in most countries the DLI attached on medication packages does not contain pictograms or graphics. Therefore, it is relevant to know how the textual instructions on drug labels are comprehended.
What is yet unknown, is how specific textual elements in DLIs (e.g., the presentation of numbers, or use of medical jargon) and patients' health literacy skills are related to (improved) comprehension of DLIs. While Samaranayake et al. [31] conducted a narrative review on the impact of patient-related factors, such as age and health literacy, and drug label-related factors, such as the use of icons and the format of the label, on the comprehensibility of DLIs, they did not provide an overview of specific textual elements associated with better comprehension of DLIs. Knowledge of which textual elements facilitates patients' comprehension of DLIs, especially those with limited health literacy, will provide concrete directions for health professionals, such as physicians and pharmacists, on how to optimize their drug prescriptions. In summary, the research question is: how are specific textual elements of DLI's and patients' health literacy related to the correct interpretation of DLIs? To answer this research question a systematic review was conducted of studies investigating the relation between the presence of specific textual elements in DLIs and patients' health literacy skills on the comprehensibility of DLIs.

Search strategy
We conducted a literature search in PubMed, Embase, PsychINFO, and Smartcat to identify studies that examined the relation between textual elements and comprehensibility of DLIs. The search was limited to studies published in English or Dutch, up to April 2019. We searched these databases using various synonyms for 'drug labeling' and 'drug label comprehension'. Reference lists of relevant studies were manually checked, and relevant studies were included. The detailed search strategies are presented in Table 1.

Study selection
Two reviewers (EM and HB) extracted data from the identified records, with one reviewer (EM) extracting data, and the other reviewer (HB) checking the information for accuracy. Duplicate publications listed in multiple databases were removed. Publications were eligible for inclusion if they met the following criteria: (1) studies which dealt with the comprehensibility of DLIs, and (2) studies which addressed textual elements of DLIs. Hence, articles focussing solely on the comprehensibility of DLIs in relation to icons, font, format, the way of printing, or other graphical elements were excluded. Also, articles about the comprehensibility of other health information sources than DLIs were excluded (reported as 'other information source' in Fig 1). Furthermore, only actual studies were included, which implies editorials, letters to the editor, or reviews were excluded (reported as 'no study' in Fig 1). Two reviewers (EM and HB) assessed whether studies had to be excluded based on titles and abstracts. To validate the assessment, twenty per cent of the identified titles and abstracts were independently annotated by the two reviewers (EM and HB) after which Cohen's kappas were calculated. The results showed the inter-rater reliability was good (κ = 0.78) [32].
After excluding studies based on titles and abstracts, both reviewers (EM and HB) read the full texts and assessed whether the study should be excluded, based on the eligibility criteria. Both reviewers had to agree on the eliminated studies. Differences in assessment were resolved by discussion or with assistance from a third reviewer (CVH or SB). Multiple publications reporting different analyses of data collected in the same study were counted as a single study with two separate research questions (see also data extraction).

Data extraction
Data from all included articles were extracted using a data extraction sheet in Excel. The data extraction was reviewed by a second author (HB) for 33 per cent of the studies. As there were no major differences between the reviewers, the remaining studies were only extracted by the first author (EM). Data extraction included: study type and setting, population, methods, primary outcome, textual elements, and conclusions. If the study compared an intervention and a control group, the intervention and the primary outcome was expressed as an effect of the intervention compared to the control group. If present, quantitative data were extracted concerning the impact of optimized DLIs compared to standard DLIs and differences between patients with limited health literacy and adequate health literacy on the primary outcome. Instruments measuring participants' health literacy were extracted to explore health literacy as explaining factor for differences in DLI comprehension. Missing information was scored as 'not reported'.

Data analysis
Data were synthesized by narrative and tabular methods. The included studies differed substantially regarding patient population and measurements methods. The study characteristics (primary outcome, study design and sample, patient characteristics, health literacy skills, and textual elements) of the publications were classified ( Table 3). The studies were grouped by DLI type. Three types were distinguished: dosage instructions, auxiliary labels, and drug labels with combined instructions. The latter category was used when a study investigated a combination of auxiliary labels and dosage instructions, or when the DLI type was not specified. -Use simple terms (e.g.
-Translate instructions to patients' native language.
-The amount of content to be retained (dosage, frequency and/or duration of intake) makes instructions complex.
-Explicit dose frequencies were easier to interpret than implicit, vague dose frequencies ('Take one tablet in the morning and one at 5 p.m' vs. 'Take two tablets by mouth twice daily').
(Continued ) -Newly designed DLIs were written in plain language (e.g., 'Take this medication with a full glass of water').
-Use a numeric instead of an alphanumeric presentation for numbers, except for fractions.
-Avoid the term 'as needed'.
-The duration and maximum dose per day should be specified on the label for shortterm drugs.
-Avoid the word Comprehension of a prescribed drug in a patient's regimen was measured at the baseline, 3 months, and 9 months by patient's ability to correctly report, for each medication: 1) How many pills taken per dose.
2) Times per day a medicine was to be taken, specifying the hour of each dose.
3) The total number of pills taken daily. Patients' adherence was measured at 3 and 9 months via: 1) Self-report of missed or incorrect doses in the prior 4 days using the Patient Medication Adherence Questionnaire.
2) Pill count (for diabetes and hypertensive medicines). Instruments measuring health literacy were also identified. We checked if they only measured functional health literacy, or whether they measured interactive and/or critical health literacy skills as well.
As the studies were heterogeneous with respect to type of intervention, primary outcome and study design, pooling of the results was considered inappropriate [36]. We aimed to give a graphical impression of the potential effects of interventions and the influence of health literacy on improved comprehension. Effects and certainty were expressed as presented in the original studies as Relative Risk (RR) or Odds Ratio (OR) with corresponding confidence intervals. Two figures were plotted with the main quantitative outcomes (RR/OR), confidence intervals, number of included patients, and a description of the intervention for the effects of interventions on improved comprehension (Fig 3) and influence of health literacy on comprehension (Fig 4). The studies expressed their primary outcome differently (e.g., patients with correct understanding or misunderstanding), hence we recalculated, if necessary, the relative risks and odds ratios to the effects on improved comprehension. Similarly, we recalculated all comparisons for health literacy to the effects of higher levels of health literacy compared to lower levels of health literacy.
The questions in the JBI checklists for cross-sectional studies, experimental studies, qualitative studies, and randomized controlled trials are specified in the critical appraisal tools available on the website of the Joanna Briggs Institute [33].

Selection of studies
A total of 434 records were identified in our literature search: PubMed (n = 177), SmartCat (n = 109), PsychINFO (n = 16), and Embase (n = 132). Records were eligible for inclusion if they met the following criteria: (1) studies which dealt with the comprehensibility of DLIs, and (2) actual studies which addressed textual elements of DLIs. A total of 60 studies remained for full-text review. Fig 1 presents a complete overview of the article search and review process. Finally, 30 publications addressing textual elements that affect the comprehension of DLIs were included in our review, of which four articles were merged into two studies in Table 3. The articles of Wolf et al. [17] and Davis et al. [26] were based on one study as well as the studies of Shiyanbola [34,35]. Therefore, 28 studies are mentioned in Table 3. Table 3 presents a complete overview of the included studies. The 28 included studies used different research methods: in-person structured interviews (n = 19), surveys (n = 4), discussion groups (n = 3), and randomized controlled experiments (n = 2). The comprehension of DLIs was measured in participant groups varying from 21 to 845 participants. Dosage instructions were most often investigated (n = 16), followed by auxiliary instructions (n = 6), and a combination of instructions (n = 6). Also, most studies investigated DLIs of prescriptions (n = 22) or DLIs that could occur on both prescribed and over-the-counter drugs (n = 5). One study focused on the comprehension of over-the-counter DLIs (n = 1). The included studies used different methods to measure participants' comprehension of DLIs. In 16 studies, participants' comprehension was assessed by rephrasing the instructions in their own words (e.g., 'What do you think this auxiliary label is saying?), whereas four studies asked participants to demonstrate when and how many tablets they would take on a day. Four studies used multiple methods to measure participants' DLI comprehension (i.e., rephrasing the DLI in their own words as well as demonstrating when and how many tablets they would take). In four other studies patients' comprehension was assessed using a survey (n = 2), a discussion group (n = 1) and an interview in which patients' perspectives on the formulation of DLIs were discussed (n = 1). Finally, in 19 studies participants' health literacy skills were assessed.

Interventions and textual elements in DLIs
The following textual elements were studied: the specificity of dosage instructions (implicit vs. explicit formulation), plain language, wording of dose measurements, the number of messages per instruction line, the presentation of numbers (numerical vs. alphanumerical), DLIs in patients' native language, and the use of abbreviations. Fig 2 provides an overview of the frequency of textual elements in the included studies. Also, Table 4 provides an overview of the textual elements investigated in the included studies, and the quantitative data regarding the comprehensibility of DLIs.
In total, seven of the 28 studies investigated the effect of textual interventions in DLIs on participants' comprehension. Fig 3 gives an impression of these quantitative findings: four out of seven showed that interventions in DLIs significantly improved the comprehension, three studies found no effect.

Specificity of dosage instructions
A total of 17 studies investigated the specificity dosage instructions: using implicit dosage intervals or explicit dosage intervals. Implicit dosage intervals only mention the frequency of intake (e.g., '2 times daily'), whereas explicit dosage intervals mention the moment of intake specified by the hour of intake, dayparts, or mealtime anchors (e.g., '1 tablet in the morning and 1 tablet in the evening').
Misunderstanding was less frequent for explicit dosage instructions [9,11,36]. Correct interpretation of dosage instructions increased when providing dosage intervals specified by four distinct time periods (i.e., morning, noon, evening, and bedtime) [9,37,38,49] or hour of intake [36,37]. Only one study on administering liquid paediatric medication using implicit versus explicit dosage instructions concluded that there was no significant difference in patient comprehension [39]. Two studies [36,40] recommended explicit DLIs, however, without defining what implicit and explicit dosage intervals were.

Textual elements
Mentioned � in: Quantitative data about the effect of (combined) textual elements on the comprehensibility of DLIs

17.
Rates of correct interpretation were lowest for instructions that depicted frequency in hourly intervals or the number of times of day (53% and 61%, respectively) and highest for those that used time periods (89%).

19.
DLIs with daily terms were more frequently associated with correct comprehension than meal terms (rates of incorrect responses were 24.2% and 52.2% respectively).

��
Effects of combined elements were studied (Take-Wait-Stop label, TWS). 14% of the participants using the Take-Wait-Stop label exceeded the maximum dose. 31.8% of the participants using the Take-Wait-Stop label exceeded the maximum dose. Difference between TWS-label and standard label was significant (OR = 2.5, 95% CI 1.05-2.70).

24.
Effects of combined elements were studied (Patient Centered Label, PCL). PCLs specify explicit timing with standard intervals (morning, noon, evening, bedtime). PCL labels were more likely to be correctly interpreted than the standard instructions (ARR:

28.
Effects of Universal Medication Scheme (UMS) were studied. Those in the UMS group displayed better understanding of the prescription regimen than those in the usual care group, but this was not statistically significant. (Mean score 9.28 vs. 8.81, p = 0.135). Subgroup analysis: no additional benefit of UMS for patients with limited health literacy (mean score 8.56 vs 9.06, p = 0.514), but rather in those who said that they found instructions on tablets hard to understand (mean score 10.00 vs 8.43, p = 0.019).
10 �� Labels were at greater risk for being misunderstood if they included multiple instructions, or included unfamiliar terms. Rates of correct interpretation of the eight patient warning labels (PWL) ranged from 0% to 78.7%. Rates of comprehension among patients were the lowest for multiple-step PWLs (8.0%, 0%, and 5.3%). Specific PWLs were not understood by most patients. For example, 'For external use only' proved to be difficult for 90.7% of the participants.

11
Simplified text warning labels (with and without icons) were studied. Simplified text + icon and simplified text only warnings more likely to be properly understood compared to standard warnings (92.1%, 90.6%, and 80.3% respectively; p<0.001).

12
Comprehension of the enhanced text + icon label was significantly higher compared to both standard and enhanced text-only labels (icon vs standard

Textual elements Mentioned � in:
Quantitative data about the effect of (combined) textual elements on the comprehensibility of DLIs

Number of messages per instruction line, n = 5 7 ���
No quantitative data presented.

��
To make the new labels more comprehensible, the researchers used single-action directions, plain-language text, and explicit pictorial descriptions of the warning message. According to participants' interpretations of the newly developed labels, the labels understood best were those with the following DLIs (single-action instructions): 'Avoid excessive sun exposure' (n = 61, 50.8%), 'Do not drink alcohol' (n = 49, 40.8%), and 'Take with plenty of water' (n = 49, 40.8%).

10
Understanding of DLIs for liquid medication was studied.
Rates of comprehension were lowest for the three patient warning labels with multiple precautions or steps instructing proper use of medication (8.0%, 0%, and 5.3%). Errors that appeared to be the result of label language were most prevalent on the instruction 'Take two tablets by mouth twice daily'. The repetitiveness between dosage ('two') and frequency ('twice') often led to the common interpretation 'Take a pill twice a day', whereas dosage would go ignored. This was confirmed in the follow-up demonstration task, 'How many pills would you take in one day' with the common incorrect response of 'two' (72% of incorrect responses).

��
Effects of combined elements were studied (Take-Wait-Stop label, TWS). Of the sample, 23% exceeded the maximum daily dose noted on the bottle and for this error type, there were statistically significant differences by study arm (standard label error rate = 31.8% vs. Take-Wait-Stop label error rate = 14%, p = .05). Those exposed to the standard label were 2.5 times more likely to exceed the recommended maximum daily dose (95% CI: 1.05, 7.70, p = .03).
28 �� Effects of Universal Medication Scheme (UMS) were studied. Those in the UMS group displayed better understanding of the prescription regimen than those in the usual care group, but this was not statistically significant. (Mean score 9.28 vs. 8.81, p = 0.135). Subgroup analysis: no additional benefit of UMS for patients with limited health literacy (mean score 8.56 vs. 9.06, p = 0.514), but rather in those who said that they found instructions on tablets hard to understand (mean score 10.00 vs. 8.43, p = 0.019).

Dose measurements
The formulation of dose measurements should be concrete and simple. Research showed words, such as 'teaspoon', 'tablespoon', 'dropperful', 'ml', and 'a small amount', are often misinterpreted due to participants' unfamiliarity with the terminology or found the terminology confusing [5,11,12,18,45,49]. However, the research did not provide recommendations for alternative formulations for dose measurements other than avoiding the above-mentioned examples. Abbreviations, n = 2 6 ��� No quantitative data presented.

22
Effects of simplified text labels (with and without icons) were studied. Simplified text labels, with and without patient-centred icons were better attended to by patients than standard labels (simplified text: AOR: 1.17, 95% CI: 1.02-1.36.
� Numbers refer to the numbers used in Table 3.
�� Quantitative data indicates understanding of several textual elements in DLIs.
��� No quantitative data on the understanding of the textual element(s) in DLIs.

Number of messages per instruction line
Five studies found multi-step instructions (e.g., 'You should avoid prolonged or excessive exposure to direct and/or artificial sunlight') while taking this medication lead to more misinterpretations [14,26,42,46]. Participants became confused when interpreting the multi-step instructions or did not address all messages of the drug label. Placing each part of the instruction on a separate instruction line (i.e., 'carriage returns') was proposed as an effective way to enhance comprehensibility [26] as well as the use of single step instructions (e.g. 'Take with food') [14,27,44].

Presentation of numbers
Four studies recommend the use of numeric over alphanumeric presentation of numbers in DLIs (e.g., 1 vs. one) [5,17,45,46]. However, fractions (i.e., ½) were better understood when presented alphanumerical (i.e., half) as the former lead to more confusion [45]. Research also showed DLIs containing fewer numbers (e.g., 'Take one tablet by mouth once each day') were better understood compared to instructions containing multiple numbers ('Take one tablet by mouth twice daily for seven days').

DLIs in patients' native language
Three studies recommended providing instructions in patients' native language [5,36,41]. Patients were more likely to demonstrate the correct dosing amount when receiving instructions in their native language instead of standard instructions [5,36,41].

Abbreviations
Only one study compared patients' comprehension of the use of Latin abbreviations in DLIs (e.g., 'Take one tablet TID') versus written-out instructions (e.g., 'Take one tablet three times day') [50]. The Latin abbreviations were least understood compared to the written-out instructions. Also, two other studies recommend to avoid abbreviations (e.g., 'ml') [43,45,50].

Health literacy and drug label understanding
Participants' health literacy skills were measured in 19 studies, using the following measures: the REALM (n = 13), the NVS (n = 2), the revised REALM (n = 1), the Arabic Single Item Literacy Screener (n = 1), the Set of Brief Screening Questions (SBSQ) (n = 1), Short Assessment of Health Literacy for Spanish Adults (SAHLSA (n = 1) and the Test Of Functional Health Literacy in Adults (TOFHLA) (n = 1). One study used two measures (SAHLSA and REALM) depending on participants' native language [51]. All instruments measured functional health literacy.
In most studies the effectiveness of specific textual elements on the comprehensibility of DLIs for participants with limited health literacy skills was not measured separately, and, therefore cannot be described in this review. However, five studies compared DLI comprehension between people with lower and higher levels of health literacy. In these studies, health literacy was classified as low, marginal or adequate. A graphical impression of these comparisons is presented in Fig 4 (also see Table 4). For these five studies, the results showed that participants with adequate health literacy comprehended DLIs better than people with limited health literacy. Two out of five studies also showed participants with adequate health literacy understood DLIs better compared to participants with marginal health literacy. However, the three other studies found no difference between these groups.
Two studies investigated the impact of textual DLI interventions on participants' comprehension and took respondents' health literacy into account. In 2010, Wolf et al. [13] investigated the effects of an enhanced label for auxiliary warnings with simplified text and, in addition, simplified icons. As described in Table 4, and shown in Fig 4, in all patients simplified text improved DLI comprehension. In participants with low health literacy, DLI comprehension was improved by the combination of simplified text and icons [OR 3.22, 95% CI 1.39-7.50]-but not for simplified text alone (see Fig 4). In 2016, Wolf et al. [38] found that the Patient Centred Label (PCL) led to slightly better use of drug regimens. The results were not significant for the entire population. However for the subgroup of participants with low health literacy, the intervention improved drug adherence (calculated by pill count) at nine months after the start of the intervention [OR = 5.08, 95% CI 1. 15-22.37].

Discussion
All studies in this review conclude that standard DLIs are too complex for patients which is attributed to their wording. Explicit time periods in dosage instructions, plain language, numbers in a numerical format, and providing DLIs in patients' native language contributed to improved comprehension. Correct interpretation of DLIs was hindered by multistep instructions per instruction line as well as using abbreviations and medical jargon. The included studies differ in the adopted research methods as well as their respondent population, which suggests this is a robust conclusion. Although health literacy was taken into account in a majority of the studies, none of them assessed the effectiveness of specific textual elements on patients' comprehension of DLIs.
The reviewed research is consistent about specifying dosage instructions and using plain language. Dosage instructions indicating more times daily use without specifying the time are often misunderstood (e.g., '1 tablet 2 times a day'). Using dayparts or mealtime anchors to specify the moment of intake facilitates patients' comprehension (e.g., '1 tablet in the morning; 1 tablet in the evening'). Future research should investigate whether dayparts are to be preferred over mealtime anchors, as meals are often skipped or consumed on different moments [52]. However, using time intervals to specify the moment of intake (e.g., 'every 8 hours') should be avoided, as both Hanchak et al. and Davis et al. showed that about 75% of the participants misunderstood these instructions [9,53]. Regarding the use of plain language, the reviewed research recommends to avoid medical jargon and to substitute complex words for simpler ones. However, only few studies provide directions to simplify DLIs or provide an overview of alternative formulations for difficult words and/or medical jargon [5,8,13,41,45]. Moreover, the evidence on the wording of dose measurements, presentation of numbers, presence of abbreviations, number of messages per instruction line as well as the use of DLIs in patients' native language is limited.
Due to the heterogeneity of the included studies with respect to primary outcome and study design pooling of the results was considered inappropriate. The effect of the individual textual elements on the comprehensibility of DLIs was not always presented quantitively, and relatively few studies compared the effect of optimized DLIs to standard DLIs. However, the included studies that allowed a quantitative comparison showed that textual interventions in DLIs are promising: optimized DLIs seem to have a positive impact on comprehensibility, especially in patients with limited health literacy. This might be explained by the fact that patients with limited health literacy experience more problems with the comprehension of DLIs, so there is more room for improvement in this population.
In 19 studies, participants' health literacy was measured with instruments focusing on functional skills. For example, the REALM measures whether participants can read and pronounce medical terms. However, being able to read and pronounce medical terms does not necessarily imply that patients are able to interpret medical terms in DLIs. Moreover, the reviewed research shows patients are able to rephrase DLIs in their own words, but experience difficulty in demonstrating proper drug use [5,46,47,49,51]. Therefore, it is likely that beside functional skills, interactive and critical skills also play a role when interpreting and applying DLIs. Future research should therefore incorporate multiple health literacy measures in order to cover these aspects of health literacy as well as multiple comprehension measures (i.e., reading, interpreting, and demonstrating).
At the start of a new drug therapy, patients usually receive DLIs combined with spoken information from prescriber and written information in patient leaflets. However, approximately 40 to 80 per cent of the information during patient-physician encounters is forgotten or remembered inaccurately [2][3][4]. Unlike patient information leaflets, DLIs present only the most essential information on drug use and are likely to be the last information source patients read before taking their drugs [8]. Hence, DLIs should serve as an independent text which should be comprehensible for all patients. This systematic review shows specifying dosage instructions and using plain language may facilitate patients' comprehension.

Strengths and limitations
As far as we know, this is the first systematic review investigating the relation between textual elements in DLIs and patients' comprehension. A strength of this review is the focus on DLIs instead of focussing on other types of health information sources. Other strengths are the focus on textual elements in DLIs and (if possible), and the inclusion of patients' health literacy in the data analysis.
A first limitation of this review is a consequence of our focus on textual elements: we did not study other elements, such as icons, that may impact DLI comprehensibility. Additional elements, such as icons or pictograms, could facilitate the comprehension of DLIs. For example, a systematic review of Sletvold et al. [54] showed pharmaceutical pictograms combined with written/oral information are useful for patients that are normally at risk for non-adherence. Also, Katz et al. [55] concluded a combination of pictorial aids and textual information facilitate patients' comprehension of medication instructions compared to pictorial aids only. Therefore, future research should focus on the effectiveness of both textual and visual elements in DLIs on patients' comprehension.
Another limitation is that most studies in this review were not performed in a clinical setting and used hypothetical drug labels. Participants had to interpret instructions of drugs they did not use themselves. Therefore, the occurrence of misinterpreting DLIs might be underestimated. This especially goes for patients following multiple drugs regimens and patients with limited health literacy. Future research should therefore investigate patients' understanding of instructions of their own medications (cf. [56]) to increase the ecological validity of research on the comprehensibility of DLIs.
A final limitation is the heterogeneity of the included studies in this review and the differences in data presented on the comprehension of DLIs. This makes it impossible to quantify the effectiveness of textual interventions in general, and to identify the contribution of specific textual elements. Although the data in this review illustrate that (decreased) comprehensibility of DLIs is a substantial problem, especially in patients with limited health literacy, it would be valuable to have pooled data on how (optimized) DLIs influence patients' comprehension, and other outcomes, such as adherence to treatment regimens and even clinical outcomes. We would welcome such comparisons, however, we believe heterogeneity will also be problematic in future evaluations as interventions directed at improved drug use will always be context specific, with multiple factors contributing to the primary outcome.

Implications for future research
Our results underline that the wording of DLIs impacts patients' comprehension. The textual elements found in this review can be used to improve existing drug labels instructions, such as specifying dosage instructions and using plain language. Although health literacy skills are a known predictor of patients' comprehension of DLIs, little is known about the effectiveness of specific textual interventions for patients with marginal or limited literacy skills. Therefore, future research should focus on the effectiveness of specific textual interventions and should include patients' health literacy in the research design. Moreover, the patients' comprehension of DLIs should be tested in a clinical setting using the DLIs of patients' own medication.

Implications for practice
Based on this review, DLIs could be optimized by specifying dosage instructions and using plain language. In many countries, prescribing and dispensing is supported by healthcare information systems [55] which contains a centrally maintained table for all available dosing instructions on DLI with corresponding codes. When processing a prescription, the health care professional will use these predefined DLI codes in order to automatically print the instructions on the drug label. Depending on the health care context, optimized DLIs can be implemented by adapting the associated codes in these information systems. Also, guidelines for the composition of comprehensible DLIs can be developed to support prescribing and dispensing professionals.

Conclusion
The present systematic review documents which textual elements have been investigated to facilitate patients' comprehension of DLIs: specifying dosage instructions and using plain language are promising ways to increase DLI comprehensibility. Especially, patients with limited health literacy might benefit from optimized DLIs. However, the heterogeneity in study design, the textual interventions, and outcomes measured, prevents us from conclusively asserting that specific textual elements are effective in improving patients' comprehension of DLIs.