Rheumatology Advance Access published online on October 15, 2007
Rheumatology, doi:10.1093/rheumatology/kem232
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Primary care treatment of knee pain—a survey in older adults
Primary Care Musculoskeletal Research Centre, Primary Care Sciences, Keele University, Keele, Staffs ST5 5BG, UK
Correspondence to:
M. Porcheret, GP Research Fellow, Primary Care Musculoskeletal Research Centre, Primary Care Sciences, Keele University, Keele, Staffs ST5 5BG, UK. E-mail: m.porcheret{at}keele.ac.uk
 |
Abstract |
|---|
 Top Abstract BackgroundMethodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
Objectives. To describe the treatment of knee pain in older adults in primary care and to compare reported practice with published evidence.
Methods. A semi-structured interview of older adults with knee pain about their use of 26 interventions for knee pain.
Results. 201 adults were interviewed. A median of six interventions had been advised for each participant, with heat and ice (84%) the most frequently advised, followed by paracetamol (71%), compound opioid analgesics (59%) and non-selective non-steroidal anti-inflammatory drugs (59%). Three core treatments for knee pain (written information, exercise and weight loss) were advised to 16%, 46% and 39% of the participants, respectively. Half of the interventions had been initiated through ‘self care’. Most core treatments had not been initiated before second-line interventions had been used, paracetamol being the exception. Referral to surgery was commonly initiated before more conservative options had been tried.
Conclusions. Interventions recommended as core treatment for knee pain in older adults were underused—in particular, exercise, weight loss and the provision of written information. There appeared to be early reliance on pharmacological treatments with underuse of non-pharmacological interventions in early treatment choices. Self care played an important role in the management of this condition. The study provides clear evidence on the need to improve the delivery of core treatments for osteoarthritis and highlights the need to support and encourage self care.
KEY WORDS: Knee pain, Knee osteoarthritis, Treatment, Primary care, Older adults
|
Background |
|---|
|
TopAbstractBackgroundMethodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
Adults aged 50 yrs and over (older adults) experience knee pain of sufficient frequency and severity to reduce their ability to participate in social and domestic life [1, 2]. For example, nearly 50% of older adults report knee pain during the course of a year and of these at least 50% report some restriction of daily activity and participation. It is a chronic condition that tends, though not invariably, to worsen over time [3, 4] and commonly presents in primary care: 33% of adults aged 50 yrs and over with knee pain report consulting their general practitioner (GP) about the problem in the previous year [5]. However, sufferers often have unmet health needs and barriers to treatment exist [5, 6]. In most sufferers, the underlying disease is knee osteoarthritis (OA), diagnosed either clinically or radiologically [3], and the knee is the most common site for the presentation of OA [7].
Unlike many other chronic conditions there is no established systematic approach to the ongoing management of OA in primary care. In the UK, although OA is now a recognized public health problem [8], a new incentive scheme (the Quality and Outcomes Framework [9]) for GPs to provide high-quality care does not cover the management of OA. However, two other recent UK national initiatives have made proposals for the organization of services and standards of care for musculoskeletal conditions [10, 11] and the UK National Institute for Health and Clinical Excellence will soon publish guidance on the management of OA [12].
There is no shortage of other guidance about the treatment of knee OA: our recent systematic search found 77 relevant systematic reviews or clinical guidelines [13]. Only a few studies have, however, investigated how knee pain and knee OA are currently treated in primary care [5, 14, 15] and the picture they provide is incomplete. Understanding how knee pain in older people is treated in primary care would inform new initiatives to improve implementation of current or future guidance.
Based on a published systematic search of clinical guidelines and systematic reviews, we identified 26 interventions recommended for the treatment of knee pain in older adults in primary care (Box 1) [13]. There are no published data on the use of the full range of these recommended interventions in older adults with chronic knee pain, nor whether they tend to be initiated by the GP or the patient. Our first two objectives were, therefore, to describe in a sample of older adults with knee pain: (i) the frequency with which evidence-based interventions for the treatment of knee pain in this age group were advised to be used, and (ii) who had first advised the intervention.
BOX 1. Interventions recommended for the treatment of knee pain in older adults and step assigned to in model of care
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||
In addition, we have previously developed and published a model of care, based on consensus within a group of primary care clinicians, which suggested the order in which these interventions could be used in clinical practice to treat older adults with knee pain (Box 1) [13]. The model consists of four steps, with step 1 representing interventions identified in the literature as core treatments which could be offered to all knee pain sufferers in this age group, and with interventions in subsequent steps to be considered if pain and disability persist despite the use of interventions from lower steps. Although the evidence for long-term benefit of a stepped model of care such as this remains unclear, it does provide a list of interventions based on published recommendations, relevance to primary care [16, 17] and an order generated by practitioners with a special interest in musculoskeletal disorders. Our third objective was to describe how current clinical practice might compare with this sequence of interventions.
|
Methods |
|---|
|
TopAbstractBackgroundMethodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
An interview survey of adults aged 50 yrs and over suffering from knee pain was undertaken. Participants were recruited from respondents to a postal survey of knee pain in adults aged 50 yrs and over registered with three general practices in North Staffordshire (Fig. 1) [5]. Participants were invited to attend a survey interview, each one conducted by the lead author. The interview schedule was piloted at one of the three practices (the lead author's practice) through cognitive interviews with 10 patients who reported knee pain [18]. The findings from these were used to refine the schedule prior to the main study at the other two practices. North Staffordshire Local Research Ethics Committee approved the study. REC reference number: 04/Q2604/27 and participants’ written consent was obtained according to the Declaration of Helsinki.
|
At interview, participants were first invited to give an account of their knee problem. They were then asked when the problem had started and when they had first consulted a GP about it, and about their use of the 26 recommended interventions. For each intervention they were asked whether, for their knee problem, they had ever used the intervention, and if not, whether it had ever been suggested. In this article, we have combined responses to these two questions as an overall measure of use of interventions, referred to in the text as ‘advised’ (those used plus those suggested but not subsequently used). If an intervention had been advised, participants were asked who had first advised it. Participants were then asked to rank the order in which interventions were initially advised—hereafter referred to as the ‘initiation rank’.
At the end of the interview participants were asked to self-complete a modified version of the Knee Pain Screening Tool (KNEST) [5] and the Western Ontario and MacMaster Osteoarthritis Index questionnaire (WOMAC) [19]. The KNEST was used to determine the occurrence and duration of knee pain in the last year, whether it was in one or both knees and whether a GP had been consulted in the previous year for knee pain. Participants were graded as having chronic knee pain if pain was reported for 3 months or more in the previous year. The WOMAC was used to determine the current severity of knee pain, stiffness and physical function. Item scores were summed to produce subscales scores (pain range 0–20, stiffness 0–8, physical function 0–68) as recommended by the developers [19]. Participants were graded as having severe pain or disability if one or more pain/physical function responses were recorded as severe on the WOMAC questionnaire [5].
We chose a sample size of 200 as a realistic number of interviews that could be conducted by the lead researcher in the time available. Based on 95% confidence interval (CI), this sample size allows a margin of error of 5% for a frequency of use of 85% (or 15%) and a 7% margin of error for a frequency of 50%. In order to obtain similar numbers from each practice, we invited all those eligible from practice A and a randomly selected sample from practice B, which had more persons eligible for invitation.
Analysis
Data from the 2003 postal survey were used to compare participants in the study with those who reported knee pain in the 2003 survey but were not interviewed. Data (all categorical) were analysed using the chi-squared test or the chi-squared test for trend, as appropriate. A descriptive analysis was then performed of the participants’ characteristics, the frequency with which interventions were advised, who had first advised an intervention and the order in which they were initiated. The responses to ‘who had first advised an intervention’ were grouped by: (i) self/friend/relative, (ii) health care professional and (iii) other/don't know. Median initiation ranks were only calculated for interventions advised to 10% or more of participants.
In addition, we compared the order in which interventions were used in practice with the order suggested in our model of care. We investigated two aspects, namely that (i) the lower step interventions would be used before those from higher steps and (ii) before progressing beyond step 1, all interventions in that step (core treatments) would have been advised. To test this we: (i) compared the recalled sequence of interventions in practice with that suggested by the model; and (ii) determined the frequency with which the interventions in step 1 had been advised among participants who had received one or more treatments allocated to steps 2, 3 or 4. SPSS 13.0 for Windows was used for analysis.
|
Results |
|---|
|
TopAbstractBackgroundMethodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
556 (295 from practice A and 261 from practice B) people were invited to participate and 201 (95 from practice A and 106 from practice B) completed an interview (36.2% response) between September and December 2004. The characteristics of the study population are shown in Table 1. Compared with 2003 survey responders who reported knee pain but were not interviewed, the study sample had a greater proportion of males and persons reporting chronic knee pain and the use of aids, and were younger, from less-deprived areas, from higher socio-economic groups and less depressed at the time of the survey (Table 2). However, there were no differences in anxiety prevalence, pain severity, consultation frequency with a GP and use of drugs or home remedies.
|
|
The mean WOMAC score for pain was 6.2 (S.D. 4.2), 3.1 (S.D. 1.9) for stiffness and 23.0 (S.D. 14.8) for physical function and 51% of participants were graded as having severe knee pain or disability.
Overall participants had been advised a median of six interventions each (interquartile range 4–9) from the list of 26 recommended interventions. The patterns of use and suggested use and the sum of the two (‘advised use’) for each intervention are shown in Table 3. Diacerein, avocado–soya unsaponifiables and intra-articular hyaluronan had not been advised for any of the participants, and very few had been advised group education (2%), capsaicin (3%), restorative sleep advice (4%) or occupational therapy (4.5%). The most widely advised intervention was heat and ice treatment, with 168 (84%) participants having ever been advised to use it, followed by paracetamol (71%), compound opioid analgesics (59%), non-selective non-steroidal anti-inflammatory drugs (NSAIDs) (59%) and walking aids (54%). There were five interventions advised to between a third and a half of the participants: exercise (46%), topical NSAIDs (42%), glucosamine (40%), physiotherapy referral (40%) and weight loss (39%). Surgical referral had been advised for a quarter of the participants.
|
In response to being asked who had first advised the intervention, participants reported that in 50% of instances, treatment was initiated by themselves, or a friend or relative, and in 47% of instances by a health care professional (Table 4). Ten interventions were in most instances initiated as self care (not on the advice of a health care professional), including paracetamol, exercise, topical NSAIDs, walking aids and glucosamine. Eleven interventions were predominantly initiated by a health care professional and included oral NSAIDs, compound opioid analgesics, referral to physiotherapy and referral for a surgical opinion.
|
Seventeen interventions were advised to more than 10% of participants and, together with the first visit to the GP about the problem, are included in the analysis describing the order in which interventions were initiated (Table 5). GP consultation, along with paracetamol, tended to occur at the start of the participants’ use of interventions and treatments. Six interventions (heat and ice treatment, exercise, topical and oral non-selective NSAIDs, compound opioid analgesics and written information) were, when advised, recalled as being initiated after GP consultation or paracetamol. Glucosamine and chondroitin were often not initiated until later. Surgical referral tended to occur earlier than the initiation of treatment with three non-surgical interventions (intra-articular steroid injection, acupuncture and selective NSAIDs).
|
Comparing the order of use observed in practice with that suggested by our model only one step 1 intervention, paracetamol, tended to be initiated before interventions were advised from higher steps. All the other core treatments from step 1 were initiated at the same time or after interventions from higher steps.
188 participants had progressed beyond step 1, and none of them had been advised to use all the interventions in step 1 before progressing to a higher step. A median of three interventions were advised from step 1 (range of 0–7 interventions) in participants who had also used interventions from a higher step (Table 6).
|
|
Discussion |
|---|
|
TopAbstractBackgroundMethodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
Our study has described, in a sample of older adults with chronic knee pain, the pattern of use of 26 interventions recommended in published guidelines or systematic reviews for the treatment of knee OA. No intervention had been used by all participants and only five had been used by more than half of the sample (heat and ice treatment, paracetamol, compound opioid analgesics, non-selective oral NSAIDs and walking aids). Interventions used by the minority of participants included exercise, weight loss and written information. However, overall this was not an uncared-for group, participants had each been advised to use a median of six interventions from the total of 26 recommended.
Interventions were as often initiated by the participant themselves, or on the advice of a friend or relative, as by a health professional. Of the five interventions used by more than half of the participants, health care professionals tended to advise pharmacological rather than non-pharmacological interventions (in 64% of instances and 21%, respectively).
The first intervention participants had tended to use was paracetamol, in addition to consulting their GP. Three interventions (exercise, written information and weight loss) recommended in clinical guidelines for initial use [16, 17, 20] had tended to be initiated at the same time or after topical and oral NSAIDs and compound opioid analgesics, which are generally recommended for use after first-line treatment has been tried [16, 17, 20].
Strengths and weaknesses
This is the first study to compare the use in practice of all these 26 interventions for knee pain in older adults. There were some limitations to the study. Only 36% of those invited were recruited to the study and this may have introduced bias. When compared with those who reported knee pain in the follow-up postal survey but had not been interviewed, study participants did not differ in their use of health care (GP/drugs/home remedies) at the time of the postal survey. However, they did have a greater use of ‘aids’ (for example, knee supports) than non-interviewees and were from less-deprived areas and higher socio-economic groups. In addition, they had been recruited from a cohort study on knee pain and might have been better informed and taken more of an interest in the condition than other knee pain sufferers. The important conclusion, therefore, is that utilization of interventions for knee pain observed in this study is, if anything, likely to be higher than their use among older adults with knee pain in general. In addition, the two practices with which participants were registered are both teaching practices and members of the Keele GP Research Partnership and would be expected to deliver a standard of care at least as good as the average for the UK. It seems likely that any bias would be in the direction of our study having over-estimated the frequency of use of the interventions.
The interview schedule developed for this study had not been previously tested for validity and repeatability. However, cognitive interviews were undertaken to test the validity of the questions and the study interviews were all conducted by the same person (M.P.) who had undertaken training in interview technique (obtained through the Survey Link Scheme—http://qb.soc.surrey.ac.uk/sls.htm).
All the observations are based on recall, which, even though participants were prompted by printed lists, could have resulted in an over- or under-estimation of the true frequency of use of some of the interventions. General practice records could have been used to validate participants’ recall, but as interventions were initiated outside professional care in half of the instances, it would be unlikely that the use of the intervention would have been coded, or even recorded, in the general practice electronic record and so this was not considered to be a robust option for validation. A comparison was made of the frequency of use of the interventions between participants from the different practices. Of the 26 interventions, there were only significant differences for exercise and restorative sleep advice. This consistency suggests there is not a bias dependent on registered practice, but there may be some limited variation between participants in their use of care which is related to the type of care provided by their particular practice.
For some participants, who had a long history of knee pain, management may have begun before there was published evidence of efficacy of a recommended intervention. Use may also be restricted by an intervention's local availability. This was certainly true for group education (with courses only being available to patients on secondary care pain management programmes) and intra-articular hyaluronan (not offered by local clinicians).
Comparisons with other studies and the model of care
Jordan et al. [14] in a postal survey of older adults with a clinical diagnosis of knee OA found that 46% had ever used a NSAID, 43% paracetamol, 16% glucosamine, 13% physiotherapy and 5% chondroitin. Linsell et al. [15] analysed a general practice database to ascertain the use of interventions in the 3 yrs subsequent to a new consultation for knee pain/OA in older patients. They reported that 72% had a record of a prescription for a non-narcotic analgesic, 52% for non-selective NSAIDs, 37% for a topical NSAID preparation and 6% for a selective NSAID. In addition, 18% had received physiotherapy and 10% a steroid injection. These two studies broadly agree with our findings: paracetamol and non-selective NSAIDs used commonly and other interventions less frequently. Obesity is a risk factor for knee OA [21, 22] and has been described as an epidemic that is out of control [23]. At the time of the 2003 follow-up survey three-quarters of the study sample were overweight or obese (body mass index greater than 25) but only two in five of participants recalled being given advice on weight loss. This supports the findings by Mehrotra et al. who found that less than half of obese adults with arthritis recalled receiving advice on weight loss at their last visit to a physician [24].
We compared the frequency of use of the interventions, and the order in which they had been initiated, against our previously developed and published stepped model of care [13]. In broad terms, step 1 highlights 10 interventions considered by our consensus panel as being appropriate core treatment to be offered to all knee pain sufferers. However, in practice, although two of the step 1 interventions were the most commonly advised (heat and ice treatment and paracetamol), only 16% of participants had been given, or sought out, written information, and fewer than half had been advised to use exercise or lose weight. In addition, three interventions considered in the model of care to be appropriate for use after the core treatments (compound opioid analgesics, non-selective NSAIDs and walking aids), had in practice been advised more frequently than the underused step 1 interventions referred to earlier. Only one step 1 intervention (paracetamol) was commonly initiated before interventions from higher steps. Most core treatments from step 1 had been initiated by the patients themselves (or on the advice of a friend or relative) and this suggests that most simple step 1 evidence-based interventions for knee pain have already been tried by patients prior to any advice to use the intervention from health professionals.
There is no evidence as yet that a stepped care approach is more effective than other strategies for selecting interventions for knee pain. Moreover, some of the 26 evidence-based interventions are either only rarely recommended (sleep advice) or are still subject to debate on their efficacy (glucosamine and chondroitin). However, the use of paracetamol, exercise, weight loss advice and the provision of written information are generally considered to be core treatments to be offered to all [16, 17, 20]. In this context, our study highlights the extent to which these core treatments might be currently under-used compared with the frequency of early use of pharmacological treatments such as NSAIDs and compound opioid analgesics.
Implications for future research
Our study has demonstrated that there is a significant gap between evidence-based practice and reported practice. The interventions reviewed here were all drawn from current published evidence and the gap indicates that research findings have either not been widely disseminated, not been actively implemented, or not gained acceptance in primary care. With respect to the order of use, it may be that the consensus model derived as an ideal comes up against barriers of practicality, availability and low expectations of benefit in actual practice. The extent to which interventions were initiated in self care is an important finding and highlights the role of self care in the management of this chronic condition. The next step will be to find ways to support implementation of evidence-based practice, both in primary care and in self care, and to investigate whether benefits in terms of improved clinical outcomes result.
|
Conclusions |
|---|
|
TopAbstractBackgroundMethodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
Interventions recommended for the treatment of knee pain in older adults were underused. This was especially true for core treatments such as exercise, weight loss and the provision of written information. There was an early reliance on pharmacological treatments with under use of non-pharmacological interventions in early treatment choices. Half of the interventions had been initiated in self care. This study has provided clear evidence that policy makers, practitioners and researchers need to address the central issue of how to get simple, achievable and effective treatments for the many sufferers with OA into practice and how best to use clinical practice to support and encourage self care.
|
Acknowledgements |
|---|
|
TopAbstractBackgroundMethodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
We would like to thank the patients, staff and doctors at the two general practices and also the administration team in the Primary Care Musculoskeletal Research Centre, and the Secretariat and members of the Primary Care Rheumatology Society.
Disclosure Statement: The authors have declared no conflicts of interest.
|
References |
|---|
|
TopAbstractBackgroundMethodsResultsDiscussionConclusionsAcknowledgementsReferences |
|---|
- Peat G, McCarney R, Croft P. Knee pain and osteoarthritis in older adults: a review of community burden and current use of primary health care. Ann Rheum Dis (2001) 60:91–7.
[Abstract/Free Full Text] - Wilkie R, Peat G, Thomas E, Croft P. The contribution of peripheral joint pain to participation restriction in community-dwelling adults aged 50 years and over. Arthritis Rheum (2004) 50(Suppl):665.
- McAlindon TE. The Knee. Bailliere's Clin Rheum (1999) 13:329–44.[CrossRef][Medline]
- Peters TJ, Saunders C, Dieppe P, Donovan J. Factors associated with change in pain and disability over time: a community based prospective observational study of hip and knee osteoarthritis. Br J Gen Pract (2005) 55:205–11.[Web of Science][Medline]
- Jinks C, Jordan K, Ong BN, Croft P. A brief screening tool for knee pain in primary care (KNEST). 2. Results from a survey in the general population aged 50 and over. Rheumatology (2004) 43:55–61.
[Abstract/Free Full Text] - Saunders C, Donovan JL, Dieppe PA. Unmet need for joint replacement: a qualitative investigation of barriers to treatment among individuals with sever pain and disability of the hip and knee. Rheumatology (2004) 43:353–7.
[Abstract/Free Full Text] - Felson DT, Lawrence RC, Dieppe PA, et al. Osteoarthritis: new insights. Part 1: the disease and its risk factors. Ann Intern Med (2000) 133:635–46.
[Abstract/Free Full Text] - World Health Organisation Scientific Group. The Burden of Musculoskeletal Conditions at the Start of the New Millennium. In: Technical Report Series 919. (2003) Geneva: World Health Organization.
- Quality and Outcomes Framework Guidance 2006. (2006) Accessed 15 March 2007. British Medical Association. Available at: http://www.bma.org.uk/ap.nsf/AttachmentsByTitle/PDFQOF2006/$FILE/Quality+and+outcomes+framework+guidance+-+Feb+2006.pdf.
- The Musculoskeletal Services Framework. (2007) Accessed 15 March 2007. Department of Health. Available at: http://www.18weeks.nhs.uk/cms/ArticleFiles/gm5hdlexqmhakbqwzzxfb32t19102005124130/Files/MSF.pdf.
- Standards of Care for people with Osteoarthritis. (2007) Accessed 15 March 2007. Arthritis and Musculoskeletal Alliance. Available at: http://www.arma.uk.net/pdfs/oa06.pdf.
- Osteoarthritis: the care and management of adults with osteoarthritis. (2007) Accessed 15 March 2007. National Institute for Health and Clinical Excellence. Available at: http://guidance.nice.org.uk/page.aspx?o=207031.
- Porcheret M, Jordan K, Croft P. in collaboration with the Primary Care Rheumatology Society. Treatment of knee pain in older adults in primary care: development of an evidence-based model of care. Rheumatology (2007) 46:638–48.
[Abstract/Free Full Text] - Jordan KM, Sawyer S, Coakley P, Smith HE, Cooper C, Arden NK. The use of conventional and complementary treatments for knee osteoarthritis in the community. Rheumatology (2004) 43:381–4.
[Abstract/Free Full Text] - Linsell L, Dawson J, Zondervan K, et al. Prospective study of elderly people comparing treatments following first primary care consultation for a symptomatic hip or knee. Fam Pract (2005) 22:118–25.
[Abstract/Free Full Text] - Jordan KM, Arden NK, Doherty M, et al. EULAR Recommendations 2003: an evidence based approach to the management of knee osteoarthritis: report of a Task Force of the Standing Committee for International Clinical Studies Including Therapeutic Trials (ESCISIT). Ann Rheum Dis (2003) 62:1145–55.
[Abstract/Free Full Text] - Altman RD, Hochberg MC, Moskowitz RW, Schnitzer TJ. Recommendations for the medical management of osteoarthritis of the hip and knee: 2000 update. Arthritis Rheum (2000) 43:1905–15.[CrossRef][Web of Science][Medline]
- Ericsson KA, Simon HA. Protocol analysis: verbal reports as data (1984) Cambridge, MA, USA: The MIT Press.
- Bellamy N. WOMAC Osteoarthritis Index. In: A users guide (1996) London (Ont.): London Health Services Centre: McMaster University.
- Walker-Bone K, Javaid K, Arden N, Cooper C. Regular review: medical management of osteoarthritis. Br Med J (2000) 321:936–40.
[Free Full Text] - Gelber AC, Hochberg MC, Mead LA, Wang NY, Wigley FM, Klag MJ. Body mass index in young men and the risk of subsequent knee and hip osteoarthritis. Am J Med (1999) 107:542–8.[CrossRef][Web of Science][Medline]
- Anderson RE, Crespo CJ, Bartlett SJ, Bathon JM, Fontaine KR. Relationship between body weight gain and significant knee pain in older Americans. Obes Res (2003) 11:1159–62.[Web of Science][Medline]
- Haslam D, Sattar N, Lean M. Obesity - time to wake up. Br Med J (2006) 333:640–2.
[Free Full Text] - Mehrotra C, Naimi T, Serdula M, Bolen J, Pearson K. Arthritis, body mass index, and professional advice to loose weight. Implications for clinical medicine and public health. Am J Prev Med (2000) 27:16–21.[CrossRef]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||