Introduction At least 68% of persons with aphasia (PWA) experience reading difficulties. Even though strategy-based interventions are a promising treatment approach for text level reading comprehension deficits in PWA, empirical evidence for their efficacy remains rare. The primary objective of this study is the analysis of the efficacy of a strategy-based intervention on text-level reading comprehension and on reading activities in PWA.
Methods and analysis In a repeated measures trial, 24 PWA will first participate in a waiting period and then in a strategy-based intervention (14 face-to-face-sessions, 60 min each). We will apply two combinations of strategies to treat either the microstructure or the macrostructure, respectively. Participants will be randomly allocated to two parallel groups that will receive these combinations in interchanged sequences. Assessments will be implemented before and after each period as well as 3 and 6 months after the intervention. The primary outcome measure is text-level reading comprehension measured with a German version of the Test de Compréhension de Textes (TCT-D) and represented by the score TCT-D Total . A non-blinded and a blinded rater will evaluate the primary outcome measure. Secondary outcome measures will address specific reading functions, reading activities and cognitive functions. The sample size was determined with an a priori power analysis. For statistical analysis, we will use contrast analyses within repeated measures analysis of variance models. We expect significant improvements in primary and secondary outcome measures during the intervention as compared with changes during the waiting period.
Ethics and dissemination This study was approved by the ethics committee of Deutscher Bundesverband für akademische Sprachtherapie und Logopädie (20–10074-KA-MunmErw+Ko). Results and relevant data will be disseminated in peer-reviewed journals, at conferences and on the Open Science Framework.
Trial registration number DRKS00021411 (see Supplementary Table 1).
- adult neurology
- rehabilitation medicine
- clinical trials
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
Strengths and limitations of this study
This study will support clinical decision making in aphasia therapy since it is the first repeated measures group study that evaluates effects of strategy-based interventions on text level reading comprehension.
Based on theoretical models of text comprehension, the effects of two different combinations of strategy-based interventions on microstructure and on macrostructure will be explored.
Outcome measures do not only include reading functions and cognitive functions but also reading activities.
Blinding of speech and language pathologists and patients will not be possible due to the nature of speech and language interventions but we will perform a blind duplication of the evaluation of the primary outcome.
The effect size of the a priori power analysis was estimated based on more general treatment approaches in persons with aphasia due to the novelty of the specific investigated treatment approach.
Background and rationale
Acquired brain injury can impair reading from grapheme or word levels to the comprehension of written texts in the context of cognitive communication disorders or aphasia. At least 68% of persons with aphasia (PWA) experience reading difficulties.1 With regard to more complex reading functions, prevalence may be even higher.2 Text-level reading difficulties constrain activities such as reading medication schedules, text messages, books, job requirement or contracts, which in turn restricts participation in life domains like communication, self-care, family roles, employment as well as domestic and social life.3–5 Despite these severe consequences, reading impairments following stroke are often treated insufficiently or remain untreated.6 Systematic reviews of Purdy et al7 and Watter et al8 indicate that the evidence for the efficacy of text-level reading comprehension treatments in PWA is limited. They suggest further investigation with more rigorous study designs as well as research on the contribution of cognitive and linguistic factors to text comprehension impairments in PWA.
Current cognitive theories suggest that reading is a constructive process based on an interaction between reader and text.9 10 Reading comprehension processes depend on visual processes, the conversion of visual input into a linguistic representation, general knowledge, and a reader’s mental lexicon as well as orthographic and linguistic systems.11 When reading a text, readers form three levels of mental representation: The surface level, the textbase and the situational model. While the phrases ‘funny girl’ and ‘the girl is funny’ differ in their superficial appearance of exact words and phrases (the surface level), they convey the same semantic content. The semantic content of a text is represented in the textbase in propositional form, thus in an idea-level format.9 12 Propositions are ‘linguistic units consisting of a relational term (or predicate) and one or more arguments’,12 for example, (FUNNY, GIRL) or (EAT, GIRL, CAKE), and allow to work with the semantic content of a text regardless of the exact wording. In a text, every proposition is linked to at least one other proposition by a relational term or a shared argument, which establishes local coherence.13 Following the Construction-Integration-Model by Kintsch,12 the textbase consists of microstructure (the whole ‘network of propositions that represent the meaning of a text’,12 that is, all detailed information of a text) and macrostructure (the most relevant propositions, the gist).12 Although the process of generation is unclear, the resulting macrostructure indicates that microstructure is reduced according to specific macrorules: selection (selection or deletion of (ir)relevant propositions), generalisation (replacing subordinate by superordinate propositions) and construction (general proposition instead of a sequence of propositions). Due to spreading activation in the propositional network, several plausible meanings may be constructed in the textbase. The activation values of incorrect elements decrease with richer context, and they will be suppressed in the integration phase.12 At the third level, the information from the textbase, the context and the reader’s pre-existing knowledge are integrated continuously taking into account cognitive schemata like scripts (eg, processes) and frames (eg, situations, objects, institutions or character types).9 Thus, the meaning of a text is constructed and represented in individual coherent situational models in the episodic memory. Situational models can be considered as a form of inference.12 Inferences are cognitive processes or results that provide information not explicitly stated in a text.14 They may involve the automatic or controlled generation of new information, or the retrieval of pre-existing knowledge from long-term memory to bridge gaps.12
Thus, text comprehension depends on a complex interaction of linguistic and cognitive resources. In PWA, on the one hand, cognitive resources such as verbal working memory, episodic memory, monitoring and cognitive flexibility may be impaired.15 16 On the other hand, receptive and productive language skills can be affected to various degrees.16 17 Even though general aphasia tests (which often assess phonology, morphology or syntax) may detect only minor or no difficulties, PWA may encounter substantial problems in text comprehension, including a reduced reading speed.15 18 Moreover, incomplete lexical and syntactic information may affect the construction of the propositional textbase, especially the microstructure.15 19 Furthermore, local and global inferencing skills as well as the construction of macrostructures and of situational models may be impaired.15 16 This may be explained by anomalous macrostructural processes or—since knowledge about the world is usually better preserved than linguistic components in PWA—as a consequence of impaired lexical and syntactic processing.19
To treat reading comprehension in PWA, oral, strategy-based, cognitive and hierarchical reading treatments can be distinguished.7 In oral reading treatments, PWA read texts repeatedly aloud, either independently or together with a speech and language pathologist (SLP), to facilitate whole-word recognition and top-down processing (ie, using prior knowledge to enhance understanding).7 8 This may be particularly helpful to improve reading comprehension in severe aphasia.20 Hierarchical reading treatments21 22 were provided with systematic computerised visual-matching and reading comprehension tasks. Stimuli are available in English language and consisted of symbols, letters, numbers, words, phrases and sentences, whereas text level reading comprehension was neither included in stimuli nor in outcome measures.21 22 Cognitive treatments focus on underlying cognitive functions and were conducted with Attention Process Training II/III (computerised exercises to train specific domains of attention)23–25 or sequenced exercises addressing working memory (sentence grammaticality judgement and semantic categorisation of words across several sentences).26 They may be particularly beneficial for persons with mild aphasia.7 In strategy-based treatments, reading strategies are used to facilitate reading comprehension.8 Reading strategies are specific, target-oriented cognitive techniques applied either automatically or consciously and flexibly. They support linguistic and cognitive processes such as the maintenance of text contents and the construction of the textbase and the situational model.9 27 Watter et al8 distinguish visual strategies (eg, underlining, highlighting), content strategies (eg, identification of key words, summarisation, preview of headings) and (meta)cognitive strategies (eg, rereading, recall, self monitoring).
To date, the evidence base is limited: cognitive and strategy-based interventions have been explored in (multiple) case studies, oral and hierarchical reading treatments in randomised controlled trials (RCTs). A gold standard cannot be concluded.7 8
This study focusses on strategy-based treatments in PWA. With regard to PWA, reading strategies may be a suitable option to take into account both linguistic and cognitive processes. Strong evidence exists already for other populations such as children with reading or learning disorders.eg,28 29 Moreover, strategy-based interventions are cost-effective due to their relatively low dosage.8 In contrast to other approaches, reading strategies are not confined to a predefined software or to specific languages, items or text types. They permit the lecture of authentic texts that are age appropriate, relevant and interesting for individuals, which may raise the motivation to engage in therapy. Despite these potentials of strategy-based interventions and despite PWA’s ability to apply reading strategies,30 the latest systematic review by Purdy et al7 identified only four case studies31–34 that evaluated strategy-based treatments with a total of six PWA. Four of the patients improved their reading comprehension.7 Furthermore, Rogalski and Edmonds35 and Rogalski et al36 found that two out of three patients improved in text-related measurements with Attentive Reading and Constrained Summarisation which contains the strategy of summarisation.
The strategies that were applied in the studies listed above are provided in online supplemental table 2. In these studies, mainly content strategies (identifying/writing down key words and key messages, summarising paragraphs and chapters, creating mind-maps and using advance organisers) as well as a few visual strategies (highlighting key words, using cards to block lines above and below focused lines) and the cognitive strategy (re)reading sections aloud and silently were explored. Furthermore, SLPs implemented strategies that can only be applied with a second person, such as discussions to increase motivation, reading/writing aloud to the patient, shared reading and writing, foreshadowing, modelling of reading and writing, providing metaliteracy comments and providing questions about the text.31–36
Elements that require either a second person or text-specific additional material contradict autonomy and may not be useful in authentic reading situations. The strategies explored in PWA to date have predominantly focused on the macrostructure and on the situational model as well as on the integration of background knowledge and (meta)cognitive abilities. They have not been explored in group studies with PWA yet. Considering that the microstructure is particularly prone to impairment in PWA, these strategies could be supplemented by strategies that target the microstructure, that could be used without a second person or additional text-specific material, and that have been suggested for other populations. Strategy-based programmes constructed for other populations include SQ3R (survey, question, read, recite, review), PQRST (preview, question, read, summarise, test) and reciprocal teaching (prediction, clarification, summarisation, question generation).10 37 Based on their meta-analysis, Mayer and Marks28 developed a strategy-based concept for children that includes the activation of prior knowledge, mental imagery, comprehension monitoring, summarising and asking questions. In students with learning disabilities, Souvignier and Antoniou29 found highest effects for the strategies activation of background knowledge, summarisation, monitoring, asking questions, identification of the topic as well as strategies focussing on the text structure. Schmidt38 suggests the use of referential links as a reading strategy. It is important that strategy combinations are more effective than the use of only one strategy.28 In view of these suggestions, we think that asking questions and using referential links could be implemented in ways that facilitate understanding of the microstructure.
Based on this preliminary evidence and on specific impairments in PWA in linguistic as well as cognitive resources, we will combine specific reading strategies that have been investigated in PWA or that have been suggested for other populations. We will group the strategies in (1) macrostructure and situational model and (2) microstructure and surface structure. To raise motivation and to take advantage of possibly unimpaired schemata, we will embed the strategies in product-oriented goals.39
Objectives and research hypotheses
The primary objective of this study is to determine whether the effects of a systematic strategy-based intervention on text-level reading comprehension and on activities and attitudes related to reading in PWA are superior to spontaneous recovery.
Secondary objectives are (1) to compare the effects of two types of strategy-based interventions on the comprehension of the microstructure and the macrostructure; (2) to investigate the relationship between microstructure, macrostructure and the situational model on the one hand and specific cognitive functions such as working memory, episodic memory, executive functions and selective attention on the other hand.
We expect the superiority of the intervention compared with the waiting period with regard to improvements in text-level reading comprehension and in activities and attitudes related to reading. Furthermore, we anticipate specific effects on the microstructure or the macrostructure depending on the strategies used.
Methods and analysis
The study protocol is based on the Standard Protocol Items: Recommendations for Interventional Trials 2013 statement.40
The trial is designed as a repeated measures study in a superiority framework with a primary endpoint of text-level reading comprehension. The effects of a strategy-based intervention and of a waiting period will be compared in a single group design. The strategy-based intervention will consist of two elements (Intervention Micro and Intervention Macro). To explore strategy-specific effects and to control for order and sequence effects, participants will be allocated to two parallel groups who will participate in these two elements in counterbalanced order after the waiting period (see figure 1).
Participants will be recruited in Germany (regional focus on urban and rural areas of Munich and Erfurt). Assessment and intervention sessions will be implemented in their regular rehabilitation centres, at the University of Erfurt, at the Ludwig Maximilian University of Munich or at participants’ homes.
We will include persons with mild to moderate aphasia. For recruitment, we will advertise the study in newsletters of professional associations, send a flyer to rehabilitation centres, and contact aphasia support groups at meetings, by phone and e-mail. Table 1 shows the eligibility criteria.
A priori power analysis using G*Power V.3.141 revealed a minimum sample size of N=21 to detect an effect size of d=0.58 with a power of .80 in a repeated-measures analysis of variance (ANOVA) (α=0.05; two measurements: eg, T2, T4; correlation among repeated measures=0.60). The effect size of the power analysis was informed by the effect size of previous research on speech and language interventions for PWA.42 We expect a drop-out-rate of 10% (n=3). Thus, we will recruit a total sample size of N=24.
Assignment of interventions
Participants will be allocated to two groups with the ratio 1:1 by stratified permuted block randomisation. With regard to prognostic factors,17 43 44 we will stratify for age, severity of aphasia (age-adjusted score in the subtest Token Test of Aachener Aphasie Test45) and the TCT-D Total score. Randomisation and allocation will be implemented independently by Dr Thomas Lauer, scientific managing director of the ErfurtLab at the University of Erfurt, Germany. SLPs will enter enrolled participants’ identification numbers along with their stratification factors into a table that Dr Thomas Lauer will use to generate blocks and to randomly allocate the participants to one of the two groups. The block sizes will not be disclosed, to ensure concealment.
Intervention procedures and material
The interventions will start in the beginning of 2021. They will be implemented by ST and by qualified SLPs who work in a setting licensed by the German health insurance system. Between T1 and T4, the participants will not receive any other aphasia treatment.
Strategy-based intervention. The intervention period consists of two elements, Intervention Micro and Intervention Macro (see table 2).
To promote transfer, we will only use strategies that participants may apply without a second person, and we will inform participants about the possible use of reading strategies in everyday reading activities. Intervention Micro and Intervention Macro will each consist of seven 60 min sessions, provided twice a week in individual therapy. Additionally, the participants will apply the same strategies as homework for 60 min after each intervention session. Homework will be monitored with a timetable. During the intervention sessions, SLPs write down the answers of the participants.
All external SLPs who deliver the treatment will participate in a preparatory workshop. They will be provided with
A manual that contains instructions, examples, structured guidance for facilitation and worksheets.
Stimulus material: articles from magazines with increasing difficulty (based on readability index, word count, number of pictures) in a predefined randomised order for each participant.
Procedures of allocated interventions may not be modified. Whenever a participant cannot solve a given task, hierarchical cues will be offered:
Authentic situations, for example, ‘If you tell your partner later on what that paragraph was about, which key word would you say?’
Semantic cues, for example, ‘You said ‘A hiker marched towards…’. Now we have to replace ‘his house’. So, we are looking for another building. Which building does the hiker want to go to?’
Phonological cues, for example, ‘I am thinking about a building that starts with c.’
Waiting period. The participants will not receive any form of aphasia treatment during 4 weeks prior to the intervention.
The outcome measures (see table 3) were chosen with regard to the levels functions and activities of The International Classification of Functioning, Disability and Health.5 Assessments will take place at six points (T1, T2, T3, T4, T5, T6) with T6 being optional (see figure 1). Instructions will be standardised without explicit encouragement to use reading strategies postintervention.
The primary outcome is text level reading comprehension defined as the score TCT-D Total measured with TCT-D. TCT-D allows to analyse specific text-level reading functions that may be impaired in PWA. TCT-D Total is the sum of microstructure and macrostructure scores across all subtests of TCT-D.
As secondary outcomes, we will collect scores of microstructure, macrostructure, the situational model and reading time with TCT-D. Reading activities as well as attitudes towards reading will be measured with an adapted German version of the Comprehensive Assessment of Reading in Aphasia (CARA) Reading Questionnaire2 which was developed specifically for the needs of PWA. At T2 and T4, we will measure inferences and macrostructure using the MAKRO-Screening46 and verbal episodic memory using Wechsler Memory Scale IV (WMS IV) 47 (logical memory 1). Cognitive functions that have been shown to be relevant for reading in PWA will be measured with subtests of the German versions of the WMS-Revised48 (digit span forward and backward), Farbe-Wort-Interferenz-Test49 and a subtest on attention of Aphasie-Check-Liste.50 We will analyse change in individual scores across the whole group and across the subgroups between specific assessment points. We will explore individual effects on functions, activities and participation as well as the potential generalisation of strategy use to everyday reading materials with a semistructured interview at T4.
Participants will be blinded with regard to the specific hypotheses of the study. Due to the nature of behavioural speech and language therapy, neither participants nor SLPs can be blinded with regard to the intervention they receive or provide. Analysis of the main outcome measure TCT-D will be duplicated by blinded assessors for T1, T2 and T4 based on audio recordings and the level of agreement will be reported in the final trial report.51 52
Data collection and participant retention
For all participants, the same data collection methods will be used. Data collection will be performed by ST and by participating SLPs based on manuals and protocol booklets to ensure standardised procedures. We will collect the data described in table 3. Points of assessment are provided in table 4. For additional information on the assessment tools, see online supplemental table 3.
Participants may withdraw from the study at any time. Participants that drop out after enrolment will be asked to participate in the TCT-D and in the CARA questionnaire for a final assessment. They will be reported in the final trial protocol with reasons for drop out. Temporal deviations of up to 3 weeks, for example, in the case of illness of participants or SLPs, will not result in drop out and planned intervention or assessment sessions will be continued. At least 75% of the assigned homework must be completed until the last therapy session of each intervention period for participants to be included in the per protocol analysis. We will undertake all reasonable efforts for participant retention during the study period:
Treatment locations close to participants’ homes.
Option to perform interventions at participants’ homes.
A worksheet for structuring homework.
SLPs will support the participants in integrating the homework into their daily routines (eg, setting an alarm, asking family members to remind the participants, preparing memos) and call participants prior to follow up appointments, if requested.
A data management plan will be provided on Open Science Framework (OSF)53 on the publication of the final trial report. Subsequent to informed consent, the data described above will be recorded with an audiorecording device, transcribed and entered in SPSS.54 Collected data will be described by metadata and will be pseudonymised by assigning a number to each participant. The list that links personal data to other data will be destroyed for the sake of anonymisation after the data analysis. Data will be stored on a laptop, on an external hard-drive and on a decentral server of the university of Erfurt and will be archived for at least ten years on millenial discs. Backups will be implemented regularly.
Primary and secondary analyses will be conducted using contrast analyses within repeated measures ANOVA models. We will apply a two-step approach to address the primary objective. First, changes in the primary outcome (TCT-D Total) will be analysed over time. Three contrasts (T2–T4; T2–T5; T2–T6) will be used to analyse whether the intervention led to any short-term and/or long-term benefits. Second, we will investigate superiority to spontaneous recovery by comparing the change scores from step one (T2–T4; T2–T5; T2–T6) with the change score from the waiting period (T1–T2). To address the secondary objectives, we will conduct the same analysis with the outcomes of the German version of CARA Reading Questionnaire to investigate superiority of the intervention with regard to reading activities. Furthermore, we will compare contrasts (T2– T3; T2–T4) between groups to explore strategy-specific effects on macrostructure and microstructure as well as on reading activities. Potential moderator effects will be explored using severity of aphasia, gender, age and education as moderators. We will use a 0.05 significance level for all models. We expect dropouts to be minimal. In case of missing data points, we will use listwise deletion and conduct analyses on just the metrics that have both pretreatment and post-treatment data points available.
ST or the Deutscher Bundesverband für akademische Sprachtherapie und Logopädie (dbs) ethics committee may withdraw particular participants if any new information indicates that continuing participation is not in their best interests. Adverse events or unintended effects reported by participants will be described in the final trial report. Adherence to the therapy manual will be monitored by ST through a standardised therapy documentation form and weekly supervision for participating SLPs. There will be neither independent auditing procedures, nor an official data monitoring committee or an interim analysis because the study is pure behavioural, non-invasive speech and language therapy and therefore, we expect no risks or harms.
Patient and public involvement
Patients and/or the public were not involved in the design, conduct, assessment of the burden, reporting or dissemination plans. Support groups with PWA will be involved in the recruitment of participants.
Ethics and dissemination
The study has been approved by the ethics committee of dbs (reference number: 20-10074-KA-MunmErw+Ko). Participants may be informed about their individual results after the final assessment on request. They will be advised on how to proceed in case of unexpected findings. There are no publication restrictions. Results will be disseminated as soon as possible regardless of the magnitude or the direction of effect in journals that focus on speech and language therapy or related subjects. Anonymised participant-level datasets and statistical codes will be provided on OSF repository after journal publication according to findable, accessible, interoperable and reusable principles with a digital object identifier and a CC-BY Creative Commons Licence. Authorship in the final trial report will be granted according to standards presented in McNutt et al55. Professional writers and editorial services will not contribute to the study.
Important protocol modifications will be registered on Deutsches Register Klinischer Studien (see online supplemental table 1) and communicated to the ethics committee.
Consent to participate
SLPs will obtain informed consent from participants or their representatives with forms approved by the ethics committee (see online supplemental file 1 for English language example). Signed forms will be collected by ST.
Consent for publication
Not applicable. No details relating to individual persons have been included.
Personal data will be collected on paper and on electronic data storage devices. Sensitive data will be stored in password-protected digital files or in a locked cabinet. Pseudonymised data may solely be shared according to specific regulations approved by the dbs ethics committee.
To date, evidence for text-level reading comprehension treatments in PWA remains rare and precludes sound clinical reasoning. In this study, we will investigate the effects of a strategy-based intervention on text level reading comprehension in a repeated measures study. The results will contribute to a better understanding of the role of cognitive functions and of reading strategies as well as to decision-making in text level reading comprehension treatments in PWA. Effects on specific reading functions and activities may allow SLPs to combine strategies according to individual reading profiles.
Due to the lack of a German-language tool to measure text level reading comprehension in PWA, we will use the translated version of the TCT13 for which no German normative data exists yet. We will address this issue by collecting data from a parallelised healthy reference group. Furthermore, we will use additional tools that have been validated and normalised with German samples in the assessment sessions T2 and T4 before and after the strategy-based intervention (see table 4).
Three parallel versions of the TCT-D exist, whereas at least five assessment sessions will take place. To estimate the possible impact of retest effects, each participant of the healthy parallelised reference group will be tested twice with the same version of the TCT-D in the same interval as participants of the intervention groups.
If we find positive effects of the strategy-based intervention, future research should compare strategy-based interventions with other treatment options in RCTs.
We would like to thank Dr. Thomas Lauer, scientific managing director of ErfurtLab, who provided his expertise on randomisation processes and who will perform the randomisation and allocation procedures. Thanks to Prof. Dr. Andreas Mayer, Ludwig-Maximilians-University Munich, and to Stiftung Pfennigparade, Therapiepraxis, for providing access to facilities for speech and language therapy. Furthermore, we express our gratitude to Janet Webster and her colleagues who supported the process of translation and adaptation of the Reading Questionnaire of CARA , to Niamh Petri who supported its back translation, to Antonia Wolf for feedback from an SLP’s perspective on its German version as well as to Corinna Thomas, Nora Bauer and Rebecca Neueder for supporting testruns with its pilot version. Thanks also to Luisa Scherf and Jula Carlsen for their valuable comments on the therapy manual and to Rahel Wacker, who prepared two beautiful drawings for the therapy manual. Thanks to Jan Stratil for helpful discussions about the study design, to Maria Ewald and Anna Thomas for performing pretests with TCT-D, and to Sofie Bale for discussions about wording and orthography in British English. Thank you to the peer reviewers for their helpful and constructive comments.
Contributors S-MT conceived the study and led the protocol development. S-MT, PS, SC and FD developed the study design. FD helped with implementation. SC supported the translation and adaptation process of TCT (13). S-MT is responsible for data acquisition and data management. PS provided statistical expertise in clinical trial design. All authors contributed to refinement of the study protocol and approved the final manuscript. No other groups or individuals will be involved in overseeing the trial.
Funding This work was supported by a Christoph-Martin-Wieland scholarship of the University of Erfurt granted to S-MT.
Disclaimer This source of funding did not and will not influence any aspects in study design, conduct, data analysis, interpretation, manuscript writing, and dissemination of results.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.