OUP user menu

A Narrative Review of Intra-Articular Corticosteroid Injections for Low Back Pain

Nikolai Bogduk MD, PhD, DSc
DOI: http://dx.doi.org/10.1111/j.1526-4637.2005.00048.x 287-296 First published online: 1 July 2005


Objective. To summarize and to analyze the available literature on the efficacy of intra-articular injections of corticosteroids for low back pain.

Design. Publications, in English, French, and German, were obtained that reported the proportions of patients who obtained complete relief of pain following intra-articular steroids, and that provided any form of follow-up. These publications were analyzed to determine the rationale, indications, and outcomes of the treatment.

Results. The only rationale for intra-articular steroids appears to be the expectation that they should work. The most commonly used indication has been back pain, for which no specific diagnosis has been made. When the results of observational studies are pooled, they paint a picture of impressive immediate responses, but a rapid decay of outcomes by three and six months. Initial responses, however, are dissonant with the literature from controlled studies of the prevalence of lumbar zygapophysial joint pain. Moreover, controlled trials have shown that there is no attributable effect to the injection of steroids.

Conclusion. The apparent efficacy of lumbar intra-articular steroids is no greater than that of a sham injection. There is no justification for the continued use of this intervention. Better outcomes can be achieved with deliberate placebo therapy.

  • Back Pain
  • Zygapophysial Joints
  • Injection
  • Corticosteroids
  • Intra-articular


The lumbar zygapophysial joints are paired synovial joints formed by consecutive lumbar vertebrae and between the fifth lumbar vertebra and the sacrum [1]. Under fluoroscopic guidance, a needle can be passed into the cavity of any of these joints, and used to inject any of a number of agents, such as contrast medium, local anesthetic, or corticosteroids [2,3]. Contrast medium can be injected either to obtain an arthrogram, or simply to verify that the needle has been correctly placed inside the joint, before injecting a subsequent agent (Figure 1). When corticosteroids are administered, the procedure is known as intra-articular injection of steroids. According to one source [4], lumbar intra-articular injection of steroids is one of the more common procedures practised in interventional Pain Medicine, amounting to over 175,000 billings under Medicare in the United States, in 2001.

Figure 1

An oblique view of an arthrogram of an L5–S1 zygapophysial joint (courtesy of Dr. Charles Aprill, New Orleans).

For those familiar with the literature, the popularity of intra-articular injection of steroids for the treatment of back pain is dissonant with the evidence concerning their efficacy. The present review was undertaken to underscore this dissonance, and to examine why a costly procedure continues to be practised.


The literature on intra-articular injection of steroids for low back pain was obtained from the personal library of the author. This library was established before intra-articular steroids were used for back pain, and papers reporting their use were added to the library as they appeared in the literature. As a check, an electronic search of Medline was conducted from 1966 to April 30, 2004, using the headings: facet joint, zygapophysial joint, injections intra-articular, and corticosteroid, with back pain.

Publications were examined for information on the rationale and indications for the procedure, and for data on its outcomes. The examination was not restricted to controlled trials. All studies that provided information or data in the targeted domains were included.

For the assessment of outcomes, papers were included for analysis if they used validated outcome measures and instruments. If validated measures had not been used, papers were included that reported the proportion of patients with complete relief of pain, and that provided any form of follow-up. Papers were not included if they reported outcomes in terms of “partial,” “good,” or “distinct” relief, on the grounds that these descriptors cannot be reliably and validly interpreted, unless accompanied by quantitative data such a visual analog score or numerical pain score. Complete relief of pain was adopted as a cardinal outcome measure also for the reason that intra-articular steroids are portrayed as a treatment specific for the source of pain, in which case complete relief should be an appropriate outcome.


The personal library of the author contained 19 relevant papers [5–24]. The electronic search produced 157 titles. Of these, 135 were not appropriate because they addressed issues other than lumbar intra-articular steroids (Table 1). Of the 22 potentially relevant titles, 16 were already held [6–15,18–23]. Two others were in Norwegian [25,26] and one in Czech [27]. These were not retrieved because of the inability of the author to translate them. Two were in French [28,29] and one in German [30]. These were obtained. The bibliography of the French papers identified two additional French titles [31,32], which were obtained.

View this table:
Table 1

Results of an electronic search of the literature for studies of lumbar intra-articular injection of steroids

Topic of PaperNumber
Not primary study of lumbar intra-articular steroids
  Essay, review, or letter, with no original data  34
  Diagnostic block only  21
  Facet denervation  16
  Clinical features only  14
  Technique only   6
  Agent other than steroid used   1
  Complications only   6
  Basic sciences   3
  Treatment of synovial cyst   7
  Prolotherapy   1
  Not lumbar spine
    Cervical spine  19
    Thoracic spine   1
    S1 spinal nerve   1
    Sacroiliac joint   4
    Transforaminal injection   1
Possible study of lumbar intra-articular steroids
  Already held in library  16
  In French   1
  In French-Canadian   1
  In German   1
  In Czech   1
  In Norwegian   2

The papers that were held or retrieved all described the technique of intra-articular steroids, and most addressed the rationale to various extents. Not all, however, provided useful data on outcome. Some reported outcomes only immediately after treatment [10,28], with no subsequent follow-up. Others did not report the proportion of patients who obtained complete relief of pain, nor did they use validated instruments by which partial relief of pain might be reliably gauged [18,24,29–32].


There is no explicit rationale for lumbar intra-articular injection of steroids. In theory, steroids may be used to suppress inflammation, but there is no evidence that back pain emanating from the lumbar zygapophysial joints is caused by inflammation of these joints. Nor has anyone claimed that inflammatory arthropathy is the particular reason why they use lumbar intra-articular injections of steroids. Nor has “inflammation” been explicitly used prospectively as an indication for intra-articular steroids in the treatment of low back pain.

The only rationale for lumbar intra-articular injection of steroids seems to be the belief that they should work, probably on the basis that intra-articular steroids have a reputation for reliving pain in the shoulder, and in the knee. The frailty of this rationale emerges if the wider literature on intra-articular steroids is consulted. Controlled trials have shown that for shoulder pain, intra-articular injection of steroids has no attributable effect beyond that of injecting local anesthetic alone, or even normal saline [33–35]. Any differences in effect are rapidly extinguished within weeks after treatment. For no condition that does not involve overt inflammation have intra-articular steroids been shown to be curative.

Otherwise, those authors who addressed the rationale for their use of intra-articular steroids all referred to previous reports of their efficacy [6–9,12–14,16–19,28,32]. Eventually all such citations trace back to Mooney and Robertson [5], who did not conduct an outcome study. They studied the pain patterns of the lumbar zygapophysial joints, and only incidentally reported that steroids appeared to work, as a treatment.


The logical indication for lumbar intra-articular injection of steroids would be that the joint to be injected has been shown to be the source of the patient's pain. This has rarely been the case in the literature [22]. In most instances, intra-articular steroids have been administered presumptively without regard to diagnosis. A diagnosis of pain stemming from the lumbar zygapophysial joints was not made. Instead, either the patients treated were assumed to have pain from the zygapophysial joints, or they were treated with injection of steroids without establishing either the cause of back pain or its origin. In effect, back pain of unknown origin was the only indication for treatment.

There are no symptoms or signs by which lumbar zygapophysial joint pain can be diagnosed clinically. No particular site of tenderness, no pattern of aggravation of pain by particular movements distinguishes lumbar zygapophysial joint pain from other causes of back pain [36–39]. Certain features correlate with patients not having zygapophysial joint pain [40], but are not diagnostic of those who do [41].

The radiographic appearance of lumbar osteoarthrosis is equally present in subjects with and without back pain [42,43] and so, cannot be used to diagnose a painful zygapophysial joint. Nor does any morphological appearance on computed tomography correlate with whether a lumbar zygapophysial joint is painful or not [44].

Controlled diagnostic blocks are the only means by which a putative diagnosis of zygapophysial joint pain might be made. Diagnostic blocks must be controlled, in each and every case, because single diagnostic blocks have a high false-positive rate [45]. A positive response to a single, diagnostic block does not necessarily mean that the joint is painful. Patients may have a placebo response to an initial block; they may have a positive response even if a placebo block is administered subcutaneously [38]. In formal studies, false-positive responses and placebo responses have been encountered in as many as 25–41% of patients tested [45–47].

Controlled blocks do not ensure a true-positive response, but they militate against false-positive responses. A patient is more likely to have genuine zygapophysial joint pain if his/her pain is consistently relieved each and every time that joint is blocked, and the duration of relief is concordant with the duration of action of the local anesthetic agent used. Moreover, that relief should be complete. Partial relief of pain is not compelling evidence that the joint is the source of pain.

If a patient obtains only partial relief of pain (e.g., 50% relief), two problems arise. First, the source of their remaining pain remains unaccounted for. Partial relief does not necessarily imply a separate, additional source of pain, for it does not exclude a partial placebo response. For example, the patient may report “some” relief in order to satisfy the expectations of their doctor, but fundamentally the pain persists because the actual source has not been anesthetized. Second, if the relief is only partial after a block, treatment targeting the joint blocked cannot be expected to produce more than partial relief of pain. Only if complete relief of pain is obtained from diagnostic blocks might complete relief be expected following treatment.

In that regard, zygapophysial joint pain is not common. Formal studies, using controlled diagnostic blocks, place its prevalence at 15% (95% confidence interval: 10–20%) among injured workers [37], at 40% (27–53%) in an older population, without a history of trauma [38], and 45% (39–54%) in a heterogeneous population attending a pain clinic [39,46,47]. Each of these studies, however, used only 50% relief as the criterion for a positive block. If complete relief is used as the criterion the prevalence is considerably lower. Studies, using single diagnostic blocks, have found that only 6%[22] or 7%[48] of patients from a general population with back pain obtained complete relief of pain. A study, using controlled blocks, however, did find that 34% of an elderly population obtained at least 90% relief of pain [38].

Indication: Back Pain

When back pain has been the sole indication for treatment with intra-articular steroids, observational studies have reported various results. No study reported outcome measures such as physical function in a quantitative manner. Only one study reported restoration of normal activities of daily living [13], and only one study referred to return to work [9]. Relief of pain was the cardinal outcome measure of all studies, and most often was the only measure. No study, however, provided quantitative data, such as numerical pain scores. Outcomes were reported only in categorical terms such as “excellent” or “good.” Nevertheless, most studies did provide sufficient information to determine what proportion of patients obtained complete relief.

Nine studies reported complete relief of pain after injection in various proportions of patients, ranging from 0% to 68% (Table 2). One reported greater than 70% relief, and another used resumption of normal activities as the outcome. Follow-up varied from six weeks to over six months, with various proportions of patients maintaining relief over this period.

View this table:
Table 2

Results of observational studies of lumbar intra-articular steroids

StudyAgentSelection Criteria*NOutcome
DefinitionImmediate ProportionFollow-Up
Carrera 1980 [6]L + MP“Clinical”20Complete relief65%30%6–12 months
Destouet et al. 1982 [7]B + MPTenderness41Complete relief46%15%6–12 months
Carrera and Williams 1984 [8]L + MP“Clinical”93Complete relief68%37%>6 months
Lippit 1984 [9]L + MP“Clinical”99Complete relief + RTWNA17%>3 months
Lau et al. 1985 [12]B + MPTenderness50>70% relief38%38%<1 month
30%1–6 months
24%>6 months
10%5–18 months
Lynch and Taylor 1986 [14]MP“Clinical”17Complete relief53%82%3months
82%6 months
Lewinnek and Warfield 1986 [13]L + MPTenderness20Normal activities30%10%5–18 months
Taylor et al. 1987 [17]B + MP“Clinical”24Complete relief or marked improvement42%17%6 weeks
8%<6 months
Nash 1990 [16]L + B + MP“Clinical”33Complete relief 0%
Marks et al. 1992 [23]L + MP“Clinical”42Complete relief26%7%1 month
5%3 months
  • * Patients were selected on the basis of undefined clinical criteria, said to be suggestive of zygapophysial joint pain, or on the basis of tenderness overlying a zygapophysial joint.

  • The proportion of patients with initial complete relief was not reported.

  • Figures not reliable, because they may include patients additional to the 17 originally treated with intra-articular injection.

  • L = lignocaine; B = bupivacaine; MP = methylprednisolone; RTW = return to work.

When the observational data are pooled, they paint a picture of some 47% of patients having an initial response, with a decay to 24% at one month or six weeks, stabilizing at 18–22% at three months and six months (Figure 2). In essence, slightly less than half the patients with an initially favorable response maintain that response for three to six months.

Figure 2

Pooled data from observational studies on the effectiveness of lumbar intra-articular injection of steroids. The bars indicate the 95% confidence intervals of the proportion of patients obtaining complete relief of pain in individual studies. The diamonds represent the mean proportion and 95% confidence intervals of the pooled results at the times indicated.

One exception to that pattern is the results of Lynch and Taylor [14]. They found an increasing number of patients with successful outcome over time. This arose because patients with initially partial responses gradually became pain-free and increased the number of successful outcomes. It is not evident from the published paper, however, how many of these patients stemmed from the group originally treated with intra-articular injections, or from the group treated with extra-articular injections. The correct denominator for calculating the proportion of successful outcomes therefore cannot be determined from the data provided.

In the one controlled trial, reported in three forms [19–21], Lilius et al. compared the outcomes of patients treated for back pain by intra-articular injections of steroids, intramuscular injection of steroids, or intra-articular injection of normal saline. Figure 3 illustrates the outcomes for pain scores [20].

Figure 3

Pain scores after lumbar intra-articular (IA) injection or intramuscular (IM) injection of steroids or intra-articular injection of normal saline obtained in the study of Lilius et al. [20].

The graph shows no differences in outcome between the three groups. All groups showed a decrease in pain immediately after the injection, but at follow-up the response decayed rapidly. At no point were there any statistically significant differences between any of the three groups. Although a smaller proportion of patients treated with normal saline maintained prolonged relief than did patients treated with steroids, the differences were not statistically significant (Table 3). Moreover, even among the patients treated with steroids, the proportions obtaining relief were appreciably less than those recorded at similar times in the observational studies.

View this table:
Table 3

Outcomes of two controlled trials of lumbar intra-articular injection of steroids, with follow-up periods measured in months

Back pain as the sole indication
Lilius et al. 1989 [20]B + MPIA28Pain (0–100)4533324146
B + MPIM395227364243
B + MPIA or IM67% with “relief”4719121016
NSIA4223 5 5 4 9
More than 50% relief from prior diagnostic blocks as the indication
Carette et al. 1991 [22]MP + NSIA49Pain634540
MP + NSIA49% with improvement423646
  • B = bupivacaine; MP = methylprednisolone; NS = normal saline; IA = intra-articular; IM = intramuscular.

These results indicate that intra-articular steroids have no attributable effect when used for low back pain without a diagnosis of lumbar zygapophysial joint pain being made. Whatever benefit occurs is indistinguishable from the effect of simply performing a sham injection.

This interpretation is corroborated by a study that was not intended as a controlled trial. Marks et al. [23] compared the diagnostic yield of intra-articular blocks with that of blocks of the medial branches of the dorsal rami that supply the lumbar zygapophysial joints. However, after the initial, diagnostic response, they followed their patients. It transpired that the proportions of patients obtaining either excellent or simply good relief of their pain did not differ for the two groups (Table 4). The therapeutic effect of intra-articular steroids was not demonstrably better than the effect of a diagnostic block.

View this table:
Table 4

Outcomes after intra-articular injection of steroids (IAS) or medial branch blocks (MBB), stratified for two different grades of relief of pain. Based on the data of Marks et al. [23]

Excellent ReliefGood Relief
Immediate118 5 3
2 weeks 431417
1 month 3012 9
3 months 20 7 6
χ2 = 2.86, P = 0.24χ2 = 0.78, P = 0.68

Thus, the evidence from controlled trials refutes the observational literature. When used for back pain, without first making a diagnosis of zygapophysial joint pain, intra-articular injection of steroids has no specific therapeutic effect. The act of injection may accord some benefit to some patients, but using steroids confers no greater benefit.

Indication: Zygapophysial Joint Pain

Only one study has assessed the outcome of intra-articular injection of steroids in patients diagnosed as having lumbar zygapophysial joint pain. In the study of Carette et al. [22], patients were selected for treatment on the basis of prior diagnostic blocks of the putatively painful joints. To be eligible for treatment, they had to report at least 50% relief of their pain when the targeted joint had been anesthetized.

Eligible patients were randomized for treatment with intra-articular injection of either steroids or normal saline. There was no difference in outcome between the two groups (Table 3). At one month after treatment, pain scores were not significantly different statistically. Nor were the proportions of patients who achieved marked or very marked improvement statistically different at one month or three months. (Data on pain scores were not provided for three months).

At six months, differences emerged between the two groups. Pain scores were slightly, but significantly, better in the steroid group, and a significantly greater proportion of patients reported marked improvement by that time. The differences in pain scores reflect a slight deterioration in the saline group coupled with a slight improvement in the steroid group (Figure 4). The data on marked improvement, however, are revealing. The proportions at six months included patients, in both groups, who reported late responses. They had not reported improvement at earlier nodes. When only those patients are considered who had marked improvement at one month, the proportions with continued good responses at six months were not significantly different in the two groups (Figure 5). This indicates that the difference in responses at six months, evident in the raw data, is spurious and not attributable to a sustained effect of steroids.

Figure 4

Pain scores after lumbar intra-articular (IA) injection of steroids or normal saline obtained in the study of Carette et al. [22].

Figure 5

Proportions of patients with marked or very marked improvement of pain following lumbar intra-articular (IA) injection of steroids or normal saline, obtained in the study of Carette et al. [22]. The filled symbols represent the raw data. The open symbols relate to those patients who had a good response at one month that persisted to six months.

Like Lilius et al. [20], Carette et al. [22] did not report the proportions of patients who obtained complete relief of their pain. They referred only to marked or very marked relief. Even under these conditions, the proportions of patients with sustained relief were modest, but similar in both groups. Furthermore, there was no difference between the two groups, at any time, in scores on the Sickness Impact Profile, or for limitations in activity. As with the study of Lilius et al. [20], Carette et al. [22] found no attributable effect for intra-articular steroids.


The history of intra-articular injection of steroids for low back pain is one of a treatment driven by hope, enthusiasm, and conviction, not by science. Early advocates hoped or believed that intra-articular steroids would work, or should work, for low back pain, just as they were believed to work for other painful conditions. Accordingly, advocates reported results enthusiastically, and attracted others, who urged others to adopt this treatment. The momentum seems to have been driven by claims of good results, but in studies that provided no quantitative data on pain-relief, and no data on other outcome measures.

There were no data, even from observational studies, that could be construed as even remotely compelling. The attractively large figures pertain to the proportions of patients who obtained relief from intra-articular steroids immediately after treatment. Immediate responses, however, do not attest to a therapeutic effect. At best, they might only indicate the prevalence of lumbar zygapophysial joint pain.

In that regard, neither the advocates of lumbar intra-articular injection of steroids nor its critics have noted the dissonance between immediate response rates and the prevalence of lumbar zygapophysial joint pain. No study of the prevalence of lumbar zygapophysial joint pain has encountered as many as 68%, 65%, 53%, or even 42% of patients obtaining complete relief of pain. These extraordinarily high figures, compared with those obtained in controlled studies of prevalence, suggest that the results of observational studies were inflated by false-positive responses because investigators did not use controlled blocks.

Despite high, initial yields, less than half of these patients had sustained, therapeutic responses. However, for lack of controls, even these modest outcomes could have been confounded by placebo responses or reporting bias. Meanwhile, the controlled trials demonstrate that prolonged responses cannot be attributed to the injection of steroids. The same outcomes are achieved by injections of inert agents or by diagnostic blocks.

Protagonists of the intervention point to flaws in the study of Carette et al. [49,50]. Correctly, they note that Carette et al. [22] did not use controlled blocks to select their patients, and that those who were selected were required only to obtain 50% relief from diagnostic blocks. Consequently, the study of Carette et al. [22] could have been confounded by a substantial proportion of patients who did not truly have zygapophysial joint pain, and whose outcomes would have diluted the success rates.

That criticism, however, is not balanced with opposing scientific evidence. It is based solely on the presumption that intra-articular steroids should work if lumbar zygapophysial joint pain is correctly diagnosed. However, no advocate of intra-articular steroids has shown, even in an observational study, that steroids work exclusively, or best, in patients first shown to have lumbar zygapophysial joint pain, using controlled diagnostic blocks. As reflected by the literature, no one in practice screens patients with controlled diagnostic blocks in order to establish a diagnosis of lumbar zygapophysial joint pain before using intra-articular steroids. Instead, they use intra-articular steroids presumptively. Under those conditions, the criticism of Carette et al. [22] becomes hollow and disingenuous, for the critics demand of Carette et al. standards of performance to which they themselves do not adhere in practice. Meanwhile, the manner in which intra-articular steroids are commonly used has been addressed by the studies of Lilius et al. [20]. Their results show that when intra-articular steroids are used without a prior diagnosis, they have no attributable effect.

The literature does not provide an explanation of why advocates are so convinced of the efficacy of intra-articular steroids. The scientific evidence from controlled studies defies this conviction. Analysis of even the observational studies seriously calls their reputed efficacy into doubt. Lacking a derived, scientific basis, the conviction about intra-articular steroids must have some other basis.

A conciliatory interpretation might be that, in the absence of any proven treatment for low back pain, practitioners are drawn to use any intervention that reputedly might work. Although this practice might be honorable in intent, in the case of intra-articular steroids, it is plainly dissonant with the available evidence, which shows that the treatment has no attributable effect.

A more cynical interpretation is that intra-articular injections of steroids are attractive because they empower the practitioner. They make the practitioner feel good and look good. The practitioner feels good because he/she has something to offer to the patient. The practitioner looks good because he/she provides a treatment that is technological. It requires special facilities, and special skills. In executing intra-articular injections under fluoroscopic control, the practitioner emulates a surgeon performing a complex procedure. If practitioners were to admit that intra-articular steroids do not work, they would be denied the opportunity to appear to be competent and to have a technological fix for their patients.

To the patient, the procedure sounds and looks impressive; the practitioner looks like someone who is confident in what he/she is doing. Furthermore, technology is used. This satisfies all the ingredients for a placebo response. There is nothing in the available literature to indicate that successful outcomes following intra-articular steroids are anything more than placebo responses.

Nevertheless, studies to date have not been incisive. No one has tested whether intra-articular steroids might be of genuine benefit to patients who definitely have pain from their lumbar zygapophysial joints. The few patients, whom advocates report anecdotally as having outstanding success from the treatment, might be evidence of a subgroup of patients for whom the treatment is indicated. The singular way of establishing this is to conduct a controlled trial of patients proven to have zygapophysial joint pain.

That obligation falls on advocates of the treatment. The paradigm is that there is a subgroup of patients who have zygapophysial joint pain and for whom intra-articular steroids should work. Those patients should be found, and the treatment tested in them. The required study, however, will be demanding. If the prevalence of lumbar zygapophysial joint pain is 15%, over 330 patients would need to be screened in order to find 50 who genuinely have zygapophysial joint pain. An even greater number would need to be screened in order to find 50 patients willing to participate in a controlled trial. These numbers rise greatly if the prevalence of true zygapophysial joint pain is lower than 15%.

In the meantime, there is no justification for the wholesale use of intra-articular steroids to treat low back pain. There is sufficient scientific evidence to draw the conclusion that intra-articular steroids do not work under those conditions.

Advocates may argue that all patients should be allowed to undergo a trial of therapy in the interests of finding those few who do obtain gratifying and long-lasting relief. Unfortunately, although such patients are reputed anecdotally to exist, their prevalence is unknown, even from observational studies. If they number but a few, the resulting number needed to treat (NNT) becomes exorbitantly high; and there is still no evidence that the observed benefit is either specific for steroids or due explicitly to their injection into the lumbar zygapophysial joints. Nor is there any evidence that intra-articular steroids achieve anything more than relief of pain. No study has shown that they restore normal activities of daily living or permit return to work, on a lasting basis.

Compounding the problem is the high cost of facility fees incurred for fluoroscopically guided procedures [51]. While so long as the cost for a trial of therapy amounted to only $100 or $150, perhaps the argument for allowing such a trial could be sustained; but when $1,000 or $2,000 dollars is added for the facility fee, the total cost for an anecdotal success becomes prohibitive to some observers. If the NNT is 10, the effective cost of one anecdotal success becomes at least $10,000.

If fortuitous, stunning results are the objective, expensive, fluoroscopically guided procedures are not required to achieve them. Prolotherapy is a simple office-procedure in which tender areas of the back are injected with sclerosing agents. A recent controlled trial showed that one in five patients were rendered pain-free, for over 12 months, if such injections are performed, and a further 20% reduced their pain by at least half [52]. But such outcomes were achieved even if placebo injections of normal saline were used. This study establishes a competing benchmark. Placebo injections can achieve outcomes that, in quantity and in duration, exceed those attributed to intra-articular injections of steroids.


The author wishes to thank Ms. Phillipa Powis, Research Librarian, Department of Clinical Research, Royal Newcastle Hospital, for conducting the electronic search of the literature.


View Abstract