We identified five reviews [1, 13,14,15,16], none of which were fulfilling our inclusion criteria. The review by Shariatmaghani et al.  was not a systematic review. Adler  studied effects of TNF inhibitors on sarcoidosis, but there were no patients with musculoskeletal manifestations. Another (non-systematic) review by Baughman  proposed a treatment scheme but did not report treatment effects. A fourth review by Bechman  was not a systematic review and did not report treatment effects. Finally, the article by Drent  was not a systematic review.
The combined search for systematic reviews and primary studies found 1425 unique records after removal of duplicates. The flow chart (Fig. 1) shows the inclusion process. We obtained 41 articles in full text and included 11 articles [4, 17,18,19,20,21,22,23,24,25,26], which are listed in Table 1. One article  was in Spanish and was therefore excluded. The 30 excluded articles [5, 9, 27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54] are listed in Appendix 2.
Characteristics of included studies Table 1 shows characteristics of the included studies. They were published between 1984 and 2020. Three studies were from France, three were from USA, two from India, and the remainder from Norway, Spain, and Iceland. No studies had a control group, and none reported effects of treatment. The total number of patients was 329. The number of patients in individual studies varied between 5 and 117. The mean per cent males were 54.1 (the mean of the means in each study), and ranged from 22 to 100. Mean age in the studies was 41.5 (standard deviation: 7.4) years. The aims of the studies were mostly to describe occurrence of different diagnoses and symptoms and/or describe treatments given to patients. One study (Garg ) tried to improve on the classification to Löfgren’s syndrome or Ponchet’s disease. Banse  studied the effects of TNF inhibitors. Glennås  explored seasonal variations of disease onset.
Methodological quality appraisal. Only two of the studies [21, 23] were assessed as fulfilling all the quality items in the quality appraisal, whilst six studies had serious shortcomings. The most common shortcomings were lack of consecutive inclusion, unclear reporting of outcomes or follow-up, and unclear reporting of the presenting site(s)/clinic(s) demographic information. Appendix 3 is a description of how we assessed the included studies on methodological quality.
Description of the studies
Arthritis in Sarcoidosis Group 2018 recruited patients from 11 centres in India. The medical records from these centres were reviewed to locate patients with sarcoidosis who had arthritis from 2005 onward. Patients were categorised into acute sarcoid arthritis (≤ 6 months) or chronic sarcoid arthritis (> 6 months). The authors compared acute and chronic sarcoid arthritis groups for differences in demographic profile. Of 117 patients, 88 had acute and 29 had chronic sarcoid arthritis. Forty-five patients in the acute group had Löfgren’s syndrome, and one patient in the chronic group had Heerfordt’s syndrome. Treatments given: All patients had received non-steroidal anti‐inflammatory drugs. For extra‐articular manifestations, 35 received corticosteroids, 17 received weekly methotrexate, 12 were on hydroxychloroquine and three on azathioprine. Prognosis: About four‐fifths of the 49 patients with acute sarcoid arthritis followed up for a median of about 2 years had attained complete remission with non‐steroidal anti‐inflammatory drugs, with corticosteroids and other DMARDs used for extra‐articular features. Acute sarcoid arthritis is mostly self‐limiting. Conclusions: The authors concluded that ankles, knees and wrists are the most commonly involved joints, with oligo-articular involvement more prominent in acute arthritis as opposed to similar oligo‐ and poly-articular involvement in chronic arthritis. Both acute and chronic articular sarcoidosis have greater propensity for hilar involvement, ILD and erythema nodosum. However, uveitis and peripheral adenopathy are more common in chronic sarcoid arthritis.
Banse 2013 evaluated the efficacy and safety of three TNF inhibitors to treat joint manifestations of sarcoidosis. Included patients were refractory to conventional therapy (NSAIDs, corticosteroids, and/or disease-modifying antirheumatic drugs (DMARDs). Amongst the ten patients, five had arthralgias without swelling and five had arthritis, including two mono-, one oligo-, and two polyarthritis. Before initiation of TNF inhibitors, according to disease activity (DAS28-CRP) scores, six patients had low articular disease activity (DAS28 < 3.2), two had moderate (DAS28 3.2–5.1), and two had high disease activity (DAS28 > 5.1). Treatments given: All patients received TNF inhibitors, and the results were reported after 3, 6, and 12 months. The total duration of TNF inhibitor exposure was 17.6 patient-years, which was started a median of 3 (0.33–17) years after sarcoidosis diagnosis. Infliximab, adalimumab and etanercept was first- and second-/third-line choices, respectively. Conclusions: The authors concluded that no significant impact of a TNF inhibitor on articular manifestations (numbers of painful and swollen joints, DAS28 with ESR or CRP, global VAS score), extra-articular involvement (pulmonary, ocular, cardiac, muscular), or biological indicators of inflammatory syndrome were observed. However, TNF inhibitor use obtained significant, albeit moderate, corticosteroid sparing effects. Moreover, their safety seems good, with no severe adverse events occurring under treatment.
Cacciatore 2020 included 39 patients with sarcoid arthropathy in a retrospective study. They contrasted 19 patients with acute disease (Löfgren’s syndrome) with 20 chronic patients. Treatments given: Amongst 20 patients with chronic sarcoidosis, treatment was used in 17 cases, and they used either NSAIDs alone (n = 5), steroids alone (n = 5), hydroxychloroquine (n = 2), methotrexate (n = 3), or TNF inhibitors (n = 2). Conclusions: Sarcoid arthropathies have different clinical phenotypes in acute and chronic forms and various treatment regimens, such as hydroxychloroquine and methotrexate, could be used in chronic forms.
Fayad 2006 performed a retrospective study from hospital charts over the period 1985–2001 in two academic French rheumatology centres. Five patients with muscle sarcoidosis were included. Treatments given: All patients were given prednisolone and two patients used hydroxychloroquine. Conclusion: Symptomatic muscle involvement may be an initial feature of chronic, and usually the systemic form of, sarcoidosis. It responds to corticosteroid therapy, but relapse seems to be frequent.
Garg 2009 examined patients with acute inflammatory ankle arthritis to establish whether they had Löfgren’s syndrome (acute presentation of sarcoidosis) or Ponchet’s disease (reactive arthritis due to tuberculosis infection). They also presented an algorithm for differential diagnosis of such patients. Of 18 patients, 10 were classified with Löfgren’s syndrome, and all of them had a negative Mantoux test (computerised tomography). The remaining patients were classified with Ponchet’s disease, all of them had a positive Mantoux test. Treatments received: The patients also received drug treatment (glucocorticoids and glucocorticoid-sparing drug methotrexate and hydroxychloroquine in patients with Löfgren’s syndrome and anti-tuberculosis drugs in Ponchet’s disease. The numbers treated with each drug were not reported.). Prognosis: All patients with Löfgren’s syndrome responded rapidly to the drugs and became symptom-free over a period of 8–12 weeks. The authors reported that all the patients with Ponchet’s disease responded satisfactorily with complete clinical as well as radiological response. Conclusions: The algorithm was successful at distinguishing between Löfgren’s syndrome and Ponchet’s disease.
Glennås 1995 followed 186 patients presenting with acute arthritis and with suspicion for reactive arthritis for 2 years and sought to classify according to diagnosis. The number of new cases of sarcoid arthritis (SA) per year in Oslo was 2.9/100 000 persons between 18 and 60 years of age. The authors found a clustered onset of SA in the spring (February-June, p = 0.01). All 17 cases of SA had complete remission of arthritis at the 104-week follow-up. Treatments received: Eight of eleven patients received corticosteroids, and 13 of 17 received NSAIDs. Conclusions: The authors concluded that the outcome of acute sarcoid arthritis appeared favourable.
Loupasakis 2015 described a case series with sarcoid arthritis in eleven firefighters in New York who worked at the World Trade Center (WTC) site on September 11, 2001. Nine of 60 firefighters who developed sarcoidosis after this date presented with poly-articular arthritis. Two others diagnosed pre-9/11/2001 developed sarcoid arthritis post-WTC-exposure. All eleven were never cigarette smokers and all performed rescue/recovery at the WTC-site within three days of the attacks. All had biopsy-proven pulmonary sarcoidosis. Treatments received: All required additional disease-modifying anti-rheumatic drugs (DMARDs) for adequate control (stepwise progression from hydroxychloroquine to methotrexate to TNF inhibitors) of their joint manifestations. Conclusions: The authors concluded that chronic inflammatory polyarthritis appears to be an important manifestation of sarcoidosis in firefighters with sarcoidosis and WTC-exposure. Their arthritis is chronic, and unlike arthritis in non-WTC-exposed sarcoid patients, inadequately responsive to conventional oral DMARDs, often requiring TNF inhibitors.
Mana 2003 studied recurrence of sarcoidosis following complete remission in 17 patients from Barcelona, Spain. Sixteen of the patients were women, and they experienced a total of 24 recurrences. The mean follow-up time was 143 ± 80 months. Löfgren’s syndrome was present in 16 patients at onset and in all patients at recurrence. Treatments given: Five patients received corticosteroids. Conclusions: The authors concluded that acute sarcoidosis, and particularly Löfgren’s syndrome, may recur many years after complete remission, and, in general, still has a good outcome.
Miller 2019 retrospectively followed the musculoskeletal and pulmonary outcomes of 24 patients with osseous sarcoidosis. They collected 1-year follow-up and last follow-up outcomes after diagnosis. The authors constructed a composite outcome consisting of (1) osseous sarcoidosis symptoms, (2) musculoskeletal imaging of affected bone, (3) chest imaging, and (4) pulmonary function testing. Treatments given: Three patients were already being treated with DMARDs or glucocorticoids for other sarcoid manifestations at the time of the osseous diagnosis. Miller et al. noted that current DMARD or glucocorticoid use at baseline was associated with a lower proportion of patients with positive worsening composite outcome (22% vs. 60%, p = 0.10). Conclusions: Most patients had a favourable outcome according to symptoms, musculoskeletal/chest imaging, and PFTs, even though only a minority were treated with glucocorticoids or DMARDs.
Perruquet 1984 conducted a retrospective review of all records of patients with a diagnosis of sarcoidosis who were seen at one centre between 1972 and 1982. Thirty-two (21%) of 150 patients with sarcoidosis had articular symptoms. Treatments given: Patients were given salicylates, NSAIDs, prednisone, and colchicine. Conclusion: Joint symptoms were generally transient. Acute sarcoid arthritis has a favourable prognosis.
Petursdottir 2007 used data from the Icelandic Sarcoidosis Study, which contains all tissue-verified cases of sarcoidosis in Iceland since 1981. Their aim was to elucidate the prevalence, clinical manifestations and long-term prognosis of sarcoid arthritis in a nationwide cohort. Forty-seven of 234 patients reported skeletal symptoms. Treatments given: The authors reported that patients were taking NSAIDs, DMARDs and glucocorticosteroids. Of 39 participants, 10 had used physiotherapy or occupational therapy. Three patients had used chiropractic treatment, and one had used acupuncture. Conclusion: Around a fifth of all sarcoidosis patients develop joint symptoms, most frequently in the ankle. The prognosis is mostly favourable, but a subgroup of female patients may develop chronic polyarthritis.
Because the studies were heterogeneous and the methodological quality was so low, we decided to not perform meta-analyses, but present the results as forest plots.
Treatments received in each study.
Figure 2 is a forest plot of the proportion of patients receiving corticosteroids and NSAIDs in each study. The upper panel shows that between 23 and 100 per cent received corticosteroids in the studies. The lower panel shows that between 0 and 100% received NSAIDs. Hence, the figures show substantial variation amongst studies.
Figure 3 shows the use of two DMARDs (hydroxychloroquine and methotrexate) in the same way. It is estimated that between 5 and 100% received hydroxychloroquine in the studies and that between 12 and 100% received methotrexate.
Figure 4 shows the use of TNF inhibitors (upper panel) and of azathioprine (lower panel). It is estimated that between 0 and 100% received TNF inhibitors, and that between 3 and 4% received azathioprine. There were, however, only two studies with a total of four patients that used azathioprine.
Other pharmacological treatments
Petursdottir 2007  reported that six participants had used DMARDs following diagnosis, whilst five had used these medications at the time of follow-up. The specific DMARDs used were not reported. Eleven participants in the study by Perruquet 1984  had used salicylates, and two had used colchicine.
Only the study by Petursdottir  reported use of various non-pharmacological treatments in altogether 10 patients.