Interventions for foot disease in rheumatoid arthritis: A systematic review




To systematically review medical and surgical foot intervention studies in rheumatoid arthritis (RA), focusing on clinical efficacy, study quality, and risk of harm.


We searched appropriate databases using a combination of the terms “rheumatoid arthritis” and “foot” against terms indicating treatment; we also hand-searched references. We selected articles in English (1968–2003) comprising randomized controlled trials (RCTs), controlled clinical trials (CCTs), prospective observational studies, and large retrospective observational surgical studies (>50 cases). RCT quality was examined using Jadad scoring; other designs were assessed qualitatively.


Inclusion criteria were met by 33 of 894 identified studies, comprising 5 RCTs and 1 CCT (all nonsurgical), 15 prospective observational studies (8 nonsurgical, 7 surgical), and 12 large retrospective studies (all surgical). Functional, custom-designed and semirigid orthoses and extra-depth shoes were effective in single RCTs of variable quality; no comparative studies have been conducted. This finding was supported by a CCT and prospective observational studies. There was no evidence of harm. There were no controlled trials of surgery. Prospective observational studies suggest that forefoot arthroplasty and first metatarsophalangeal joint implants, but not plantar callous debridement, are effective. Comparative retrospective analyses suggest that some procedure variants may be better, and surgery may relieve pain better than orthoses. Infection was the main risk.


RCT evidence shows that orthoses and special shoes are likely to be beneficial in patients with RA. The only evidence of benefit from surgery comes from observational studies, because no RCTs have been conducted. Further RCT evidence is needed, although well-designed observational studies may be helpful.


Clinically-relevant foot problems are common in rheumatoid arthritis (RA) (1). Pain is the most common presentation (2), but foot disease may also cause deformity and walking difficulties. In early RA, clinical studies have shown that the foot is the initial site of involvement in 16–36% of patients with RA (3); this has been confirmed using plain radiographs (4) and magnetic resonance imaging (5). In established RA, >85% of patients have clinically significant foot involvement (6). In late RA, severe metatarsophalangeal (MTP) joint damage occurs in >25% of patients (7), many of whom ultimately undergo surgery. RA causes both forefoot and rearfoot problems; however, the balance of available evidence suggests that forefoot disease is more prominent and is consequently the major focus of the therapeutic studies described in this review. Associated problems, such as foot ulcers, which are important clinical aspects of lower limb function in patients with RA, are beyond the scope of the present review.

Systemic therapy with antiinflammatory and disease-modifying drugs improves foot disease in RA. However, there is a range of medical and surgical treatments specifically aimed at foot problems. Although there has been a Cochrane systematic review of orthoses and splints in RA, which included information about the efficacy of those treatments in foot RA (8), there has been no systematic review of the effectiveness of the full range of foot interventions in patients with RA. Such evidence is crucial for optimizing patient care and planning future rheumatology, orthopedic, and podiatry services.

Therefore, we undertook a systematic review of medical and surgical foot interventions in RA focusing on evidence of efficacy, the quality of available studies, and potential risk of harm. Both randomized trials and prospective observational studies were included.


Search strategy

All relevant available databases were searched, specifically Medline, Embase, the Cochrane Database, and the Cochrane Musculoskeletal Group Register. References from all articles identified from the databases were also hand searched.

The terms “rheumatoid arthritis” and “foot” were combined and searched in conjunction with the following treatment terms: treatment, management, therapy, surgery, ultrasound, injection, massage, laser therapy, orthotics, shoes, acupuncture, physiotherapy, hydrotherapy, and homeopathy.

Selection criteria.

Types of studies included.

Studies from 1968 to 2003 that examined clinical benefits from interventions specifically involving the foot in patients with RA were identified. Studies not written in English and those described only in abstract form were excluded.

Randomized controlled trials (RCTs), controlled clinical trials (CCTs), and observational studies (mainly case-control and cohort designs) were initially reviewed, and a subset of these was selected for detailed analysis. For nonsurgical interventions, all RCTs, CCTs, and prospective observational studies were selected. For surgical interventions, because there were no RCTs or CCTs, all prospective observational studies were selected; in addition, because of the lack of evidence, large retrospective observational studies including >50 patients were also included. The cut-off point of 50 patients was chosen to exclude the very large number of small, surgical, retrospective case series that were uniformly of low quality, primarily due to their ill-defined inclusion criteria, their poor description of the population from which patients were recruited, their inadequate size, and the absence of defined objective outcome measures.

Quality of studies

The quality of RCTs was judged using the scoring instrument developed by Jadad et al (9), which was developed and validated to measure the likelihood of study bias in RCTs. Three items (each scored as 1 for yes and 0 for no) included in the instrument are whether the study was described as randomized (including words such as randomly, random, and randomization), whether the study was described as double blind, and whether there was a description of subject withdrawals or dropouts. Additional points were given if appropriate methods of randomization and/or double blinding were described, and points were deducted if inappropriate methods were described. The maximum score was 5.

Non-RCT studies were evaluated qualitatively based on the following design factors: presence of a description of the population from which study patients were recruited, including the study setting and consideration of possible referral bias; use of defined study inclusion criteria from inception of the cohort to be studied (with information about patients excluded, withdrawals, and dropouts); presence of information about patients included in the study (diagnostic criteria, disease severity, comorbidity, and drug therapy); adequate study size and followup rate; use of defined objective and reproducible outcome measures; and evidence of collection of standardized information about the risks of harm.

Types of participants

Only studies that reported patients with RA were included, although a few of these reports additionally described results in participants with other rheumatologic diagnoses.

Types of interventions

Studies of all types of foot interventions were selected for further analysis. Interventions included orthoses, special shoes, intraarticular radiation, and orthopedic surgical procedures.

Types of outcome

Studies of all types of outcome were selected for further analysis. Outcome measures included the Foot Function Index (10), the Health Assessment Questionnaire, dynamic and static foot measures, and a range of patient satisfaction measures (both standardized and descriptive); in descriptions of satisfaction in which there was a clarification of the extent of satisfaction (e.g., satisfied and satisfied with minor reservations), only satisfied (unclarified) was recorded. Outcomes indicating risks of harm from treatment were also examined.


Studies included and excluded

The initial search identified 894 studies; 861 did not meet the inclusion criteria (including 130 that were not written in English and 32 that were retrospective studies of surgical procedures of <50 cases). The remaining 33 studies that were selected for further analysis comprised 5 RCTs, 1 CCT, 15 prospective observational studies, and 12 retrospective studies, all of which were surgical reports involving >50 cases.

We analyzed the efficacy of these studies assessed in controlled trials and observational studies, discussed study quality, and examined the potential risks of harm from treatment.

Information on clinical efficacy from controlled trials.

Randomized controlled trials.

Four RCTs, all of which evaluated orthoses, were identified, 3 with parallel group design and 1 with a crossover design; details are shown in Table 1. In terms of trial quality, only 1 RCT (reported in 2 studies examining different aspects by Budiman-Mak et al &lsqbr;11&rsqbr; and Conrad et al &lsqbr;12&rsqbr;) was of high quality, with a Jadad score of 5. All the other RCTs were of relatively low quality, with Jadad scores of 1 for one study (Woodburn et al &lsqbr;13&rsqbr;) and 2 for the remaining 2 studies (Fransen and Edmonds &lsqbr;14&rsqbr; and Chalmers et al &lsqbr;15&rsqbr;).

Table 1. Randomized controlled trials of non-surgical interventions (orthoses and shoes)*
RefYearNo. entered/completed studyNo. F/MInterventionDurationJadad scoreDesign analysisPrimary outcomeOther outcomes
  • *

    ITT = intention to treat; FFI = foot function index; VAS = visual analog scale; DAS = disease activity score; HAQ = health assessment questionnaire.

  • This single study was reported in 2 articles (11, 12) each examining different aspects.

11,121995, 1996102/880/102Functional foot orthoses3 years5Parallel group ITT, completer- ITT analysis of progression of hallux abductus angle - 5/50 (10%) treatment group - 12/48 (25%) controls - Between group P = 0.05- Total FFI and joint counts in completers - Total FFI similar in treated cases and controls - Total foot joint counts similar in treated cases and controls
14199730/2822/8Extra- depth shoes2 months2Parallel group completerNone stated- Intra-group changes in completers
       - Treated cases reduced VAS pain walking (mean 17.2; P = 0.001)
         - Treated cases reduced VAS pain climbing stairs (mean 18.4; P = 0.001)
         - No significant changes in controls
15200028/2421/3a) Supportive shoes alone b) with soft approach orthoses c) with semi-rigid orthoses3 months each approach2Cross-over completerCompleter analysis changes in mean VAS foot pain - Shoes alone = 0.05 - Soft orthoses = 0.01 - Rigid orthoses = 1.87 Between group P = 0.027- Completers showed no difference between groups in foot joint counts, 50-foot walking time, Robinson Bashall functional assessment, Toronto Activities of Daily Living
13200298/8165/33Custom orthoses30 months1Parallel group ITTMean area under curve total FFI - orthoses = 241 - control = 23 Between group P = 0.026- ITT analysis between groups showed reduced eversion in treatment group (P = 0.009) - No differences in global pain, DAS, HAQ or Larsen scores

Two of these RCTs compared custom-designed (11, 12) or functional (13) foot orthoses with placebo orthoses. A third RCT (14) compared extra-depth with normal shoes, and a fourth (15) examined supportive extra-depth shoes with or without soft or semirigid orthoses. Because these trials used diverse interventions, dissimilar patients, and different outcome measures, their results could not be pooled. Therefore, they are described individually below and in Table 1.

The first RCT (11, 12), comparing functional foot orthoses with thin shoe inserts over a 3-year period, showed no clinically significant differences in pain or function, although there was less progression of the abnormal hallux abductus angle in subjects using the functional orthoses.

The second RCT (13) compared, over a 30-month period, patients continuously using custom-designed rigid foot orthoses with controls who received orthoses only if prescribed through normal medical care. The custom orthosis group showed a significant reduction in pain (despite initial short-term discomfort) and improvement in the total FFI and its pain and disability subscales. These changes were immediate, and peaked at 12 months.

The third RCT (14), comparing extra-depth, off-the-shelf footwear with normal shoes over a 2-month period, showed that pain on weight bearing and HAQ scores improved significantly in patients using extra-depth shoes, whereas those wearing normal shoes showed no change.

The final RCT (15) used a crossover design over a 12-week period to compare supportive extra-depth shoes with or without soft or semirigid orthoses. After 6 weeks of treatment, patients showed significantly less metatarsalgia when wearing supportive extra-depth shoes with semirigid orthoses than when wearing shoes alone or with soft orthoses. Twenty-two of 24 patients preferred wearing orthoses. Interestingly, despite the greater improvement in metatarsalgia with semirigid orthoses, preferences were equally divided between soft and semirigid orthoses.

In summary, RCT evidence suggests that functional (11, 12), custom-designed (13), and semirigid orthoses (15) in RA are likely to be beneficial, although only single studies of each orthosis exist and 3 of the 4 studies have low Jadad scores (Table 1). Extra-depth shoes also appear to be effective (14), although their benefit is greater if combined with orthoses (15). The different nonstandard orthoses have not been compared with each other.

Nonrandomized controlled trials (CCTs)

The search identified 1 CCT by Hodge et al (16). They compared, in 11 volunteer patients with RA, a “shoe only” control with 4 types of orthoses (prefabricated, standard custom molded, custom with a metatarsal bar, and custom with a metatarsal dome) worn for 5 minutes. As assessed by an in-shoe sensor system, all orthoses significantly reduced pressure beneath the first and second metatarsal heads; pressure under the second metatarsal head was shown to correlate with pain ratings (r = 0.562). In terms of reducing subjective walking and standing pain ratings, the custom-molded orthosis with a metatarsal dome was the most effective.

As well as demonstrating a measurement technology that may be of benefit in assessing foot interventions, this CCT supports the use of foot orthoses in RA. However, the reliance placed upon it must be limited by methodologic deficiencies, notably its small size and the lack of a defined recruitment strategy.

Information on clinical efficacy from observational studies

Given the paucity of controlled trials and their complete absence in studies of surgical treatment, prospective and large (n > 50) retrospective observational studies were also selected for evaluation as described in the Materials and Methods. The major rationale for examining these studies was the widespread use of the treatments they investigate despite the lack of controlled trial evidence. Unfortunately, even allowing for their uncontrolled nature, the conclusions that can be drawn are further limited by other more avoidable methodologic defects as revealed by qualitative assessment of study design (see Materials and Methods for details).

One set of problems concerned cohort definition and inclusion in the study. Difficulties included lack of detailed information about patient populations and the study setting, lack of consideration of possible referral bias, inadequate description of recruitment and trial entry strategies, small study size, and inadequate followup, especially in longer-term studies. Information on patients who were excluded or withdrawn and those who dropped out was also lacking. These defects were present, to a greater or lesser extent, in most of the prospective studies. Retrospective analyses are even more subject to cohort definition and inclusion biases because it is almost impossible, in retrospect, to ascertain whether all relevant patients have been included.

Problems with outcome measures were also apparent. Prospective nonsurgical studies did tend to use quantifiable objective criteria, as did some prospective surgical studies; in contrast, retrospective surgical studies generally used nonstandardized global outcome descriptors such as good, satisfactory, or poor. Furthermore, very limited information was provided on harm assessment, particularly in the retrospective studies; this amplifies the underlying difficulty of generalizing procedure risk using evidence from observational studies that do not involve defined inception cohorts.

Prospective observational studies of orthoses and surgical shoes

Eight uncontrolled observational studies of special shoes and orthoses were identified that evaluated custom-built orthoses (17), conventional surgical and seamless shoes (18), experimental sandals (19), surgical footwear with or without insoles (20), the Cherwell splint (support for valgus foot inside normal shoes) (21), heat-molded shoes (22), and custom-molded orthoses in various types of shoes (23, 24). Similar to the RCTs, these studies could not be pooled because they used diverse interventions, patient groups, and outcome measures; they are therefore described individually below and in Table 2.

Table 2. Observational studies of non-surgical interventions (orthoses and shoes)*
(Ref) yearNo. patientsInterventionsDurationMethodsResults
  • *

    UCBL = University of California at Berkeley; VAS = visual analog scale; FFI = foot function index.

(18) 196818Surgical shoe; seamless shoe6 monthsShoe preference14 preferred seamless shoe
(19) 197625Experimental sandal6 monthsLickert pain scaleGenerally improved (no details given)
(20) 198171Surgical footwear +/− insolesVariablePostal survey of metatarsalgia63% relieved; >90% significantly ameliorated
    Footwear acceptability36 of patients considered footwear acceptable
(21) 198350Cherwell splintVariableDuration of wearing orthoses>2 years (16 patients)
     6–2 years (14 patients)
     <6 months (7 patients)
(17) 198410UCBL extended orthosis1 montha) Lickert pain scalea) Decreased in 9 cases
    b) Stride velocity c) Single limb support timeb) Increased 36–50 m/minute (barefoot versus orthoses; P < 0.01)
     c) Increased 63–79% of normal (barefoot versus orthoses; P < 0.01)
(22) 199025Heat-molded shoes3 monthsa) Lickert walking ability scalea) 80% of patients improved
    b) Frequency of shoe useb) 80% of patients used all day and 20% some of the day
(23) 19998Custom orthoses6 monthsa) VAS Comfort Scorea) All patients showed improved scores
    b) Stride lengthb) Mean score increased 4.36cm (P < 0.05)
    c) Gait velocity and cadencec) Both showed non-significant increases
(24) 200318Custom orthoses3 monthsa) FFI Pain Scorea) Mean score decreased 61–43 mm (P < 0.05)
    b) Stride lengthb) Mean score increased 76–102 cm (P < 0.05)
    c) 15 meter walkway energy expenditurec) Mean score decreased 0.3–0.2 (P < 0.05)

A biomechanical observational study (17) was primarily intended to compare normal and RA ankle/subtalar motion and stride characteristics using biomechanical assessment techniques. However, an uncontrolled subgroup analysis examined how biomechanical variables and symptoms changed in patients with RA who were given rigid orthoses for 1 month. As well as showing that such insoles increased stride velocity and single-limb support time, the authors noted that 9 of 10 patients with RA wearing insoles reported a decrease in ankle and hindfoot pain.

A crossover comparison (18) of seamless and conventional shoes showed that most patients found seamless shoes to be more comfortable, although they had cosmetic concerns and the shoes were heavy and hot. An observational study (19) over 6 months of an experimental sandal found that the sandal improved pain and functional capacity and reduced pressure areas and forefoot callouses. Another study (20) examining surgical footwear with or without insoles, worn for variable periods of time, found that the intervention relieved metatarsalgia in 63% of patients and improved it in 90%, but the footwear was acceptable to only 50% of the patients.

The Cherwell splint (21), which supports a collapsing valgus foot in normal shoes, was studied over a 3-year period in 50 patients, 30 of whom wore it for ≥6 months. Only 5 patients reported no benefit, but the splint appeared most useful in RA foot disease with a valgus deformity. Heat-molded shoes (22), with or without custom-made orthoses and/or shoe stabilization, were studied over a minimum of 3 months in 25 patients, 80% of whom wore the shoes all day and 20% for a portion of the day, with 80% of patients reporting that they were able to walk better with the heat-moldable shoes.

A study (23) assessing the benefit of wearing custom-molded ethyl vinyl acetate foot orthoses for up to 6 months placed in patient-selected footwear (chosen with podiatric advice) showed that most patients found the footwear comfortable; there was a statistically significant increase in average stride length with orthoses in situ. A final study (24) evaluated customized orthoses (specifically designed for each patient and type of foot problem), which significantly reduced pain, improved stride length, and decreased physiologic cost index over 3 months.

In summary, these nonsurgical observational studies further support the use of orthoses and special shoes in RA; however, their value is limited, not only because they are uncontrolled, but also because of the other methodologic limitations previously discussed.

Prospective observational surgical studies

Seven reports of 4 prospective observational studies were identified. These evaluated minor surgery (plantar callous debridement) (25), intraarticular radiation to the ankle (26), forefoot arthroplasty involving the great toe (4 reports of the same patient group with different followup times and outcome measures) (27–30), and first MTP joint silicone implantation (31). Almost all studies, apart from the study on intraarticular radiation, involved only the forefoot. Once again, because of the diversity in interventions, patient groups, and outcome measures, the results could not be pooled and are described individually below and in Table 3.

Table 3. Prospective studies of surgical interventions*
(Ref) YearNo. patientsDurationInterventionOutcome measuresResults
  • *

    MTPJ, metatarsophalangeal joint; Preop = preoperative; postop = postoperative; VAS = visual analog scale.

(31) 1992326 yearsFirst MTPJ silicone implanta) Patient satisfaction b) Pain c) Walking d) Function e) Footwear fittinga) 84% of cases completely satisfied; 3% dissatisfied b) Reduced from 3.4–7.6 points (P < 0.001) c) Non-significant increased (6.4–7.1 points)
     d) Non-significant increase (5.1–6.0 points)
     e) Improved (94% difficulty preop, 53% postop; P < 0.05)
(26) 199484 months–4.5 yearsIntra-articular radiation (ankle)Overall results judged by patient and surgeon6 of 8 had good results
(25) 20008Repeated at 28-day intervalsPlantar callous debridementa) Forefoot pain VAS b) Global arthritis pain VASa) Mean 48% improvement (P = 0.01) postop; lost by 7 days b) No significant change
(27) 1997352 yearsGreat toe forefoot arthroplastya) Static and dynamic foot pressurea) Marked reduction under central metatarsals (P = 0.001)
(28) 1990353 yearsGreat toe forefoot arthroplastya) Patient's satisfactiona) 91% satisfied
   b) Painb) 70% pain free on exercise
    c) Mobilityc) 66% had improved mobility
    d) Hallux valgus e) Footweard) 73% postoperatively (all corrected intra-operatively)
     e) 80% standard shoes
(29) 1989284 yearsGreat toe forefoot arthroplastya) Patient's satisfactiona) 89% satisfied
   b) Painb) 75% pain-free on walking
(30) 19881411 yearsGreat toe forefoot arthroplastya) Patient's satisfactiona) 80% satisfied
   b) Painb) 84% reduced pain
    c) Mobilityc) 71% had improved mobility
    d) Footweard) 64% normal shoes
    e) Cosmesise) 56% satisfied

Plantar callous debridement was studied at baseline in 8 patients and was repeated at 28-day intervals (25). Although pain scores improved by an average of 48% immediately following debridement, this effect was short lived and was lost 7 days after treatment.

Intraarticular radiation to the ankle in 8 patients (26) was assessed globally by both the patient and the surgeon as good, fair, or poor. Overall, 6 of the 8 patients (75%) reported a good result.

Four reports (27–30) described the results of forefoot arthroplasty involving the great toe in the same patient cohort at different times using a variety of outcome measures. The longest period of observation (14 patients, average followup 11 years) (27) demonstrated that >80% of patients were satisfied with the results, 84% had reduced pain, 71% had increased mobility, 64% could wear normal shoes with or without insoles, and >50% were satisfied with cosmesis. Larger groups of patients, evaluated postoperatively, showed better clinical results; at a mean of 3 years after surgery, 91% of 35 patients were satisfied (29) and at a mean of 4 years, 89% of 28 patients were satisfied and 75% noted painless walking (28). A study examining static and dynamic foot pressures at a mean of 2 years after surgery (30) showed that such surgery led to a marked reduction of pressure under the central metatarsals.

The final procedure evaluated in a prospective surgical study was a first MTP joint silicone implant (31), studied in 32 patients followed for an average of 6 years. Overall, 84% of patients were completely satisfied and pain was significantly reduced. Footwear fitting improved; only 53% reported difficulty postoperatively compared with 94% preoperatively.

In summary, these prospective observational studies suggest that certain surgical foot interventions, including forefoot arthroplasty and first MTP silicone joint implant, may be helpful in patients with RA. In contrast, plantar callous debridement produced <1 week's improvement and is unlikely to be clinically effective. The data on intraarticular radiation to the ankle is too limited to comment on. Again, the reliance that can be placed on this evidence is limited by methodologic concerns.

Retrospective observational surgical studies

There were 12 retrospective observational surgical studies (Table 4) involving >50 patients, the number required by our inclusion criteria; all of these studies evaluated forefoot arthroplasties. The studies could be divided into 2 groups: in 1 group of 6 studies, surgery invariably involved the great toe (32–37); in the other group of 6 studies, not all patients had surgery to the great toe (38–43). Of the 6 studies that always involved the great toe, 3 studies evaluated identical great toe procedures and 3 evaluated different great toe interventions. Once again, the differences in patients recruited, procedures undertaken, and outcome measures used precluded any pooled analysis.

Table 4. Restrospective studies of surgical interventions
Interventions(Ref) YearCasesDurationMain outcomesResults
All cases had great toe surgery(32) 19751285 yearsSatisfaction85% excellent/good
 (33) 1988 68 feetNot statedSatisfactionMost satisfied
 (34) 19974706 yearsSatisfaction> 60% satisfied
 (35) 1997103 feet3–15 yearsPain relief72% improved
 (36) 1999 802–3 yearsSatisfaction30–54% satisfied
 (37) 20011888 yearsa) Satisfactiona) 49–67% satisfied
    b) Pain improvementb) 33–52% improved
Not all cases had great toe surgery(38) 1968138 feet2 yearsNo specific measure75% good
(39) 1971 882 yearsSatisfaction88% good/satisfactory
 (40) 1971 772 yearsPain relief80% unlimited pain-free walking
 (41) 19801005 yearsa) Satisfactiona) 51% good
    b) Pain reliefb) 67% good
 (42) 1982 655–9 yearsPain relief55% at 5–9 years
 (43) 2003 796 yearsa) Satisfactiona) 67% very satisfied
    b) Pain reliefb) 56% complete pain relief

Various outcome measures were assessed; the main parameters measured were overall outcome, overall satisfaction, pain relief, walking ability, shoe wear, and cosmesis. Overall outcome varied from excellent or good in 60–85% (32, 34), good in 75% (38), good or satisfactory in 88% (39), and fair or good in 93% (41) of cases. Satisfaction with the operative procedure varied between 25% and 67% (36, 37, 43). Pain relief ranged from 33% to 100% (35, 37, 41–43).

Walking ability was reduced in only 1 study (35) and not improved in another (37). Improvements in walking ability ranged from 67% to 80% in the remaining 4 studies (36, 40). Footwear improvement was noted in 25% of patients (36) and ordinary shoe wear was possible in 45–80% (40, 41) of patients postoperatively. Cosmesis was satisfactory in 75% of patients in one study (43).

Several studies compared different interventions. Two compared first MTP interventions and concluded that arthrodesis had worse outcomes than either first metatarsal head excision or first MTP arthroplasty (33, 36). Stabilizing the great toe was thought to lead to better outcomes than instability (43). Finally, surgical outcome was concluded to be better in terms of pain relief than conservative (orthosis) management in a further study (42).

In summary, these retrospective surgical observational studies suggest that some surgical interventions may be effective in RA and make some attempt to compare different interventions. However, the methodologic limitations discussed at the beginning of the observational studies section, which apply even more to retrospective compared with prospective observational studies, indicate that these data can only be used as a pointer towards true clinical efficacy.

Evidence of potential risks of harm from treatment

Little or no evidence of risk of harm was identified from the nonsurgical studies other than the possibility of treatment failure (described previously).

Surgical treatment clearly has potential local and systemic risks including infection, tissue damage, delayed wound healing, bleeding, venous thrombosis, and anesthetic complications. Assessment of the level of risk is difficult because the surgical studies evaluated were neither controlled trials nor standardized inception cohorts. In addition, many of the studies included only limited or no discussion of potential harm; in particular, risks were not reviewed in a standardized manner and often were not quantified.

The main risk of harm reported was infection, which was specifically reported in 8 surgical observational studies (29, 31, 32, 35, 39, 40, 42, 43); the frequency of infection in these reports averaged 7.5% (range 1.4–16%). Individual reports also described delayed wound healing (31, 35), skin necrosis (41), and implant fragmentation (31). Other recognized risks, notably deep venous thrombosis and pulmonary embolus (44), were not identified in the studies included in the systematic review but were reported in some of the smaller retrospective surgical series, which were excluded from our final evaluation.


This systematic review of foot interventions in RA has identified RCT evidence suggesting that functional (11, 12), custom-designed (13), and semirigid orthoses (15) and extra-depth shoes (14) (especially if combined with orthoses &lsqbr;15&rsqbr;) are likely to be beneficial; the various nonstandard orthoses have not been compared with each other. The clinical efficacy of these nonsurgical interventions is further supported by a single CCT and by prospective observational studies. In contrast, there is no RCT or CCT evidence investigating surgical procedures; the best evidence comes from prospective observational studies (supported by retrospective studies), which suggest that forefoot arthroplasty and first MTP silicone joint implant may be helpful. The only comparative surgical evidence comes from large retrospective studies that suggest surgery may provide better pain relief than orthoses, and certain surgical procedure variants may be better than others. Finally, significant methodologic limitations for virtually all the studies in the review indicate that definitive conclusions cannot be drawn about optimal treatment, and further research is needed.

The four RCTs identified (Table 1) all assessed nonsurgical interventions; only one (11, 12) was of high quality, with a Jadad (9) score of 5, whereas the others were of low quality, with Jadad scores of 2 in 2 cases (14, 15) and a score of 1 for the remaining study (13). Because scoring is based on reports of relevant design factors in the published studies, we cannot exclude the possibility that the actual studies were of higher quality. However, this seems unlikely because Jadad scores of 1 or 2 imply either that 1 of the 3 major assessed items was omitted or that the randomization or blinding techniques reported were inappropriate. Other study defects included the use of completer rather than intention-to-treat analyses in 2 studies (14, 15) (though completion rates were high at 93% &lsqbr;14&rsqbr; and 86% &lsqbr;15&rsqbr;), small size in 2 studies (30 &lsqbr;14&rsqbr; and 28 &lsqbr;15&rsqbr; patients), and no stated primary outcome measure in 1 study (14). On a positive note, all RCTs used objective outcome measures and had satisfactory completion rates (83–93%). These defects, together with the fact that each intervention was only examined in a single RCT, suggest that caution is needed, hence our conclusion that orthoses are likely to be beneficial.

Because we found no surgical RCTs, we cannot make a definitive statement about the efficacy of surgical interventions. There are few RCTs of any orthopedic intervention in patients with arthritis, even for common procedures such as knee replacement (45). Many surgeons have concerns about RCTs in surgery, because outcomes are operator dependent and appropriate controls (e.g., “sham” operations) are problematic. In addition, surgeons often believe they know the best treatment for a patient; in the absence of genuine equipoise concerning 2 treatments, it is ethically difficult to enter a patient into a study. Initiating trials of surgical procedures is therefore difficult (46, 47), emphasizing the need for collaborations such as the UK Musculoskeletal Clinical Research Collaboration involving the British Orthopaedic Association and British Society for Rheumatology with the Arthritis Research Campaign and Medical Research Council (48).

Much of the evidence in this systematic review was from non-RCT studies. Such observational studies cannot provide definitive statements on efficacy but give insights into treatment benefits. We included observational studies in our review, recognizing their inherent problems outlined by von Elm and Egger (49) and Pocock et al (50). These authors focused on epidemiologic research, but their conclusions are general.

The quality of non-RCT studies was assessed using a series of “best practice” design factors, and many deficiencies were found. Studies were often small (only 1 prospective study entered more than 35 patients) and, especially in retrospective analyses, little information was provided about patient populations, inclusions, withdrawals, and dropouts. Most surprisingly, all retrospective and many prospective observational studies (Tables 2–4 relied on global outcome descriptors (“good results,” “satisfied,” and “improved”) rather than using validated, objective, quantifiable, reproducible outcome measures, even though these existed.

Another area where surgical studies were deficient was reporting harm. Unlike orthoses and special shoes where harm is not a problem, surgery has well-known risks including infection, tissue damage, delayed wound healing, bleeding, venous thrombosis, and complications of anesthesia. These problems were not highlighted in the surgical studies we evaluated, and often there was no systematic approach to identifying harm. Furthermore, generalizable harm risk cannot be assessed from observational studies unless they are prospective, formally defined patient cohorts. An additional complexity in assessing harm is that patients may be harmed as a consequence of medical inactivity; although usually ignored, the risks of not treating a patient are one aspect of weighing risks and benefits.

One simple way to improve the quality of evidence available to the clinician would be to encourage good practice in designing and conducting observational studies. Such an approach is being undertaken for epidemiology (49) through the Standards for Reporting of Observational Studies in Epidemiology statement (available at, which is being developed to guide the reporting of cohort, case-control, and cross-sectional studies. Similar guidance for interventional observational studies would be helpful.

In conclusion, current RCT evidence shows that orthoses and special shoes are likely to be beneficial. Though the effectiveness of surgery cannot be definitively established without RCTs, observational studies support a number of procedures. Further evidence of efficacy and risk of harm is needed, especially for surgery. This requires high-quality RCTs preferably, but well-designed prospective observational studies, particularly of surgical procedures, will enhance understanding of this important component of care.

  • 1

    Drs. Farrow and Kingsley contributed equally to this work and should be considered as joint first authors; they have been listed alphabetically.