What Are the Most Clinically Effective Nonoperative Interventions for Thumb Carpometacarpal Osteoarthritis? An Up-to-date Systematic Review and Network Meta-analysis

Abstract

Background

Thumb carpometacarpal osteoarthritis (CMC-1 OA) is a common and debilitating condition, particularly among older adults and women. With the aging population, the prevalence of CMC-1 OA is expected to rise, emphasizing the need to find effective nonoperative strategies. So far, for determining the most effective nonoperative interventions in CMC-1 OA, two network meta-analyses (NMAs) have been published. However, these NMAs were limited to specific intervention types: one comparing multiple splints and the other comparing different intraarticular injections. Therefore, an NMA that compared all nonoperative intervention types is urgently needed.

Questions/purposes

This study aimed to assess and compare the effectiveness of available nonoperative interventions (both nonpharmacologic and pharmacologic) for CMC-1 OA to establish which nonoperative options are more effective than control in terms of (1) pain, (2) function, and (3) grip strength.

Methods

We adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) NMA guidelines (PROSPERO: CRD2021272247) and conducted a comprehensive search across Medline, Embase, CENTRAL, and CINAHL up to March 2023. We included randomized controlled trials (RCTs) and quasi-RCTs evaluating nonoperative interventions for symptomatic CMC-1 OA, excluding inflammatory or posttraumatic arthritis. Studies comparing ≥ 2 interventions or against a control, focusing on pain reduction, functional improvement, and grip strength, were selected. We assessed methodologic quality using the modified Coleman Methodology Score, including only studies scoring > 70. Risk of bias was evaluated with the Risk of Bias 2.0 tool, and evidence quality with Confidence in Network Meta-Analysis (CINeMA). Of 29 screened studies, 22 (21 RCTs and one quasi-RCT) were included, involving 1631 women and 331 men. We analyzed eight different nonoperative interventions, including splints, hand exercises, injections, and multimodal treatment (≥ 2 nonpharmacologic interventions or nonpharmacologic with a pharmacologic intervention). Six studies had a low risk of bias, eight had a high risk, and the remainder were moderate. We extracted mean and SD scores, and NMA and pairwise analyses were performed at short- (≤ 3 months) and medium-term (> 3 to ≤ 12 months) time points. Standardized mean differences were re-expressed into common units for interpretation, which were the VAS (range 0 to 10) for pain, the DASH test (range to 100) for function, and pounds for grip strength. Clinical recommendations were considered strong if the mean differences exceeded the minimum clinically important difference—1.4 points for VAS, 10 points for DASH, and 14 pounds for grip strength—and were supported by moderate or high confidence in the evidence, as assessed using CINeMA methodology.

Results

Our NMA (based on moderate or high confidence) showed a clinically important reduction in pain at the short-term time point for multimodal treatment and hand exercises versus control (mean difference VAS score -5.3 [95% confidence interval (CI) -7.6 to -3.0] and -5.0 [95% CI -8.5 to -1.5]). At the medium-term time point, only the rigid carpometacarpal-metacarpophalangeal (CMC-MCP) splint was superior to control (mean difference VAS score -1.9 [95% CI -3.1 to -0.6]) and demonstrated clinical importance. For function, only the rigid CMC-MCP splint demonstrated a clinically important improvement at the medium-term time point versus control (mean difference DASH score -11 [95% CI -21 to -1]). Hand exercises resulted in a clinically important improvement in short-term grip strength versus control (mean difference 21 pounds [95% CI 11 to 31]).

Conclusion

This systematic review and NMA show that multimodal treatment and hand exercises reduce short-term pain and improve grip strength, while a rigid CMC-MCP splint enhances medium-term function. Future research should evaluate long-term efficacy.

Level of Evidence

Level I, therapeutic study.

Introduction

Thumb carpometacarpal osteoarthritis (CMC-1 OA) is common and often debilitating, especially among older people. It can cause pain and reduce functional ability, impacting tasks that involve gripping and pinching [21]. Nonoperative interventions for CMC-1 OA can be divided into nonpharmacologic and pharmacologic. Nonpharmacologic interventions include joint protection advice (guidance on reducing stress on the CMC-1 joint), hand exercises, and splints. Pharmacologic options include topical and oral NSAIDs and intraarticular injections. The 2018 update of the European Alliance of Associations for Rheumatology (EULAR) recommendations for hand OA supports a stepwise approach, beginning with nonpharmacologic interventions before considering pharmacologic interventions [30]. More recently, recent guidelines from the British Society for Surgery of the Hand Evidence for Surgical Treatment (BEST) have suggested that all patients with symptomatic CMC-1 OA should first be offered noninvasive interventions consisting of a multimodal treatment package. Following this, splints should be considered for patients who do not respond, followed by intraarticular corticosteroid injections [20]. Although surgery is reserved for patients with persistent symptoms that do not respond to nonoperative interventions, many patients accommodate and manage their condition nonoperatively [15].

Despite the widespread use of nonoperative interventions for CMC-1 OA, there is no consensus on which interventions are the most effective. Prior studies have often focused on individual interventions or have not directly compared different interventions [43], leading to gaps in knowledge. Given the high prevalence of CMC-1 OA and its potential to cause disability, high-quality evidence synthesis is essential for clinicians and hand therapists to understand which interventions are most effective in improving patients’ quality of life. Due to this need for robust evidence, network meta-analysis (NMA) has emerged as an excellent tool for comparing multiple interventions that have not been directly compared in head-to-head clinical trials [44]. Two NMAs have been previously conducted; however, these NMAs were limited to specific types of interventions only, with one comparing multiple types of splints and the other comparing different intraarticular injections (for example, corticosteroid versus hyaluronic acid versus platelet-rich plasma) [33, 49]. No comprehensive NMA has yet compared all nonoperative interventions (nonpharmacologic and pharmacologic) for CMC-1 OA in one analysis.

This study, therefore, aimed to assess and compare the effectiveness of available nonoperative intervention options (both nonpharmacologic and pharmacologic) for CMC-1 OA to establish which nonoperative options are more effective than control in terms of (1) pain, (2) function, and (3) grip strength.

Materials and Methods

Study Design

We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) NMA guidelines [27]. The protocol was registered with the internal Prospective Register of Systematic Reviews (PROSPERO: CRD2021272247).

Identification of Studies

We identified eligible trials through searches on Medline and Embase via Ovid, Cochrane Central Register of Controlled Trials (CENTRAL), and Cumulated Index in Nursing and Allied Health Literature (CINAHL) via EBSCO from their inception to March 2023. We also manually searched the reference lists of retrieved systematic reviews and clinical guidelines. The search strategy was devised following the Peer Review of Electronic Search Strategies (PRESS) framework by our senior librarian (PD) [45].

Study Eligibility

We considered randomized controlled trials (RCTs), crossover trials (participants switch between interventions after completion), and quasi-RCTs (intervention groups for participants using a nonrandom method). We did not limit studies by the year of publication. We excluded non-RCTs (including comparative case series), review articles, short reports, research letters, conference abstracts, and publications using duplicated data. Only studies published in English were included.

The following inclusion criteria for study eligibility were also adopted based on the PICO framework. Population (P): Patients with symptomatic CMC-1 OA (diagnosed on either clinical or radiographic examination or both) of all severity. Those with rheumatoid, inflammatory, and posttraumatic arthritis were excluded. Interventions (I): Studies comparing ≥ 2 nonoperative interventions. This included two types of commonly used splints: the long carpometacarpal-metacarpophalangeal (CMC-MCP) splint, which immobilizes the wrist and extends to the interphalangeal joint of the thumb, and the short CMC splint, which is palm based, covering up to or just beyond the first MCP joint. These splints can be made from custom-molded thermoplastic (rigid) or prefabricated neoprene (soft) materials. Hand exercises, which included static and dynamic movements, were performed individually or were physiotherapy guided. Pharmacologic interventions included intraarticular injections with corticosteroid or hyaluronic acid. Last, a multimodal treatment package could consist of combinations of nonpharmacologic interventions or a pharmacologic alongside a nonpharmacologic intervention (for example, hand exercises and splint, or splint and intraarticular injection). Comparator (C): Comparisons were between the same type of nonoperative interventions (such as two types of splints or two types of injections) or between different nonoperative interventions. Alternatively, we included studies that compared an intervention with a control group, where the control received no intervention, placebo sham treatment, or minimal care such as joint protection advice (guidance on reducing stress on the CMC-1 joint). Outcomes (O): Prespecified outcomes were a reduction in pain and/or an improvement in hand function (measured using patient-reported outcome measures [PROMs]) and/or grip strength. Studies with follow-up time points < 4 weeks were excluded to ensure that outcome data included in the NMA reflected more stable and meaningful improvements following the intervention or comparator.

Selection of Studies

Two reviewers (AT, JPR) independently evaluated each study based on the title, abstracts, and full-paper screening. When disagreements occurred, reviewers reached a consensus through discussion with senior authors (NJ, JD).

Study Methodology Assessment and Data Extraction

Selected studies underwent assessment for methodologic quality using the modified Coleman Methodology score (mCMS, scored out of 100, with higher scores indicating superior study design and reporting quality) by two reviewers (AT, JPR) independently [12]. This scoring system was originally developed to assess the methods of clinical studies on patellar and Achilles tendinopathy and has been modified for the assessment of upper limb disorders [38]. Scores were corroborated, and after discussion with senior authors (NJ, JD), a cutoff of 70 was applied to exclude studies of poorer quality. For any missing data, we attempted to contact the corresponding authors of published studies. Twenty-nine studies were selected for Coleman screening, of which 22 were given scores of ≥ 70 and underwent subsequent data extraction [1-6, 9, 11, 13, 16, 22, 24, 25, 34, 35, 40, 42, 46, 47, 51-53]. Coleman scores ranged from 70 to 84, with the majority of studies losing points due to the total number of patients recruited, short follow-up times, and not measuring or stating compliance to a particular intervention or comparator used in the study (Supplemental Table 1; http://links.lww.com/CORR/B353).

Outcome data for all available follow-up time points were extracted. However, for our NMA, predefined time points were set at short (≤ 3 months), medium (> 3 to ≤ 12 months), and, if available, long term (> 12 months). If a study reported two follow-up results within the same time frame, we analyzed data that were assessed at the time point closest to the upper limit of the respective category.

Risk of Bias Assessment

Two reviewers (AT, JPR) judged all studies using the Cochrane Risk of Bias tool (RoB 2.0) [48]. Disagreements in bias assessments were discussed with senior authors (NJ, JD), if necessary. Six studies were deemed to have an overall low risk of bias [2, 13, 22, 24, 42, 53], eight studies had a high risk of bias [6, 9, 11, 16, 34, 35, 40, 47], and the remaining studies had a moderate risk of bias [1, 3-5, 25, 46, 51, 52] (Supplemental Fig. 1; http://links.lww.com/CORR/B353).

Evidence Quality and Confidence Ratings

Evidence quality for each comparison was evaluated by two reviewers (AT, JPR) and was assessed using an adapted version of the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) methodology using Confidence in Network Meta-Analysis (CINeMA). CINeMA (versus GRADE) improves transparency and avoids the selective use of evidence when forming judgments, thus limiting subjectivity in the process [37].

For the network meta-analysis, we evaluated the credibility of intervention effects by considering within-study bias, reporting bias, indirectness, imprecision, heterogeneity, and incoherence based on CINeMA recommendations [37]. We then graded each domain by combining judgments about direct evidence with their statistical contribution to NMA results. We summarized across six domains to obtain confidence in each comparison (pairwise effect estimate), rated as high, moderate, low, and very low.

For pairwise meta-analyses, we assessed confidence ratings based on five domains: within-study bias, reporting bias, indirectness, imprecision, and heterogeneity, adapting the recommendations from CINeMA for pairwise analyses. Within-study bias was taken from the RoB 2.0 ratings for each study, with the highest risk of bias chosen. For imprecision, the optimal information size (minimum number of overall participants combined in each meta-analysis for sufficient precision) was used (n > 20). For heterogeneity, we used the I² statistic. Where this was found to be high (I² = 50% to 80%), the confidence rating was downgraded by one level; where it was found to be substantial (I² > 80%), two levels were downgraded. In those cases, subgroup analysis was performed to explain this heterogeneity. The result of each comparison of interventions and overall confidence rating was assigned one of high, moderate, low, or very low confidence.

Overall, for the NMA, confidence across the comparisons (point estimates) generated for pain, function, and grip strength (Supplemental Tables 2-7; http://links.lww.com/CORR/B353) varied from high to very low. Downgrading was mainly due to within-study bias, imprecision, and heterogeneity [37]. For the pairwise meta-analyses, downgrading was mainly due to within-study bias and heterogeneity, with only one pairwise comparison having high confidence, three moderate confidence, and the remainder with low or very low confidence (Supplemental Table 8; http://links.lww.com/CORR/B353).

Protocol Deviations

For studies that used joint protection advice alone as an intervention or comparator, we classified this under the control node due to insufficient connectivity between intervention nodes when analyzed separately. This decision was necessary to maintain the integrity of the NMA and ensure adequate comparisons across interventions.

Studies Included

As per PRISMA flow, our search yielded 2185 potentially relevant studies, with a further 29 identified from reference lists searching. After de-duplicating and title and abstract screening, 51 studies were eligible for full-text screening, of which 29 met our eligibility criteria and underwent further methodologic assessment using the mCMS. A total of 22 papers had a cutoff mCMS of > 70 and were included in the systematic review and NMA (Fig. 1). Twenty-one of 22 included studies were parallel-group RCTs [1-6, 9, 13, 16, 22, 24, 25, 34, 35, 40, 42, 46, 47, 51-53] and one study was a quasi-RCT [11]. The proportion of women to men across the studies was approximately 5:1, with 1631 women and 331 men recruited. The inclusion criteria were highly variable across the studies, with some studies basing their inclusion criteria on age, sex, classification of CMC-1 OA using Eaton and Littler or Kellgren-Lawrence [14, 28], and duration of symptoms (Table 1). Eight different nonoperative interventions were investigated, creating eight different nodes for our NMA, with the addition of a control node (placebo or sham treatment, no intervention, or minimal care) (Table 2). All studies measured pain, with the VAS most commonly used (n = 14). Twenty papers measured functional outcomes using PROMs, with the DASH as the most commonly used scale (n = 5). Fourteen studies of 15 measured grip strength. Follow-up times varied between the studies, with none having follow-up times beyond 12 months (long-term time point) (Table 3).

Fig. 1.

The PRISMA flow diagram for the study selection process.

Table 1.

Characteristics of the selected studies, inclusion criteria, and participant characteristics

Study	Study type	Country	Inclusion criteria	Mean ± SD age of patients in study arm one in years	Mean ± SD age of patients in study arm two in years	Mean ± SD age of patients in study arm three in years	% (n) women
Adams et al. [2]	RCT	England	Clinical CMC-1 OA, age > 30 years, AUSCAN hand pain index score > 5, AUSCAN hand functional disability > 9	62 ± 9.1	64 ± 9.4	NA	79 (274 of 349)
Arazpour et al. [3]	RCT	Iran	Grade 1 or 2 CMC-1 OA, base of thumb pain	50 ± 5.7	52 ± 6.4	NA	88 (22 of 25)
Bahadir et al. [4]	RCT	Turkey	Grade 2 or 3 Eaton-Littler	62 ± 9.1	61 ± 7.3	NA	100 (40 of 40)
Bani et al. [5]	RCT	Iran	Radiologic and clinical diagnosis of CMC-1 OA	53	54	59	71 (25 of 35)
Becker et al. [6]	RCT	USA	Clinical CMC-1 OA, age ≥ 18 years	63 ± 8.1	62 ± 9.4	NA	77 (48 of 62)
Can and Tezel [9]	RCT	Turkey	Grade 1 or 2 Eaton-Littler	56 ± 7.5	57 ± 9.2	NA	90 (57 of 63)
Cantero-Téllez et al. [11]	Quasi-RCTa	Spain	Grade 2 or 3 Eaton-Littler, VAS score > 4 of 10	60 ± 9.6	61 ± 9.8	NA	92 (77 of 84)
Gomes Carreira et al. [22]	RCT	Brazil	Grade 2 or 3 Eaton-Littler in dominant hand	63 ± 8.5	65 ± 10.1	NA	95 (38 of 40)
Deveza et al. [13]	RCT	Australia	Grade 2 > Kellgren-Lawrence, age > 40 years, VAS score > 4 of 10, FIHOA score > 6	66 ± 7.8	65 ± 8.5	NA	76 (155 of 204)
Figen Ayhan and Ustün [16]	RCT	Turkey	Radiographic grades 1-4 Eaton-Littler, bilateral symptoms with failed prior treatment, VAS score > 4 of 10	62 ± 6.4	62 ± 6.4	NA	100 (33 of 33)
Hermann et al. [24]	RCT	Norway	American College of Rheumatology classification CMC-1 OA, thumb pain on palpation, understands Norwegian	71 ± 6.7	71± 7.3	NA	98 (58 of 59)
Heyworth et al. [25]	RCT	USA	Symptomatic CMC-1 OA and age > 40 years	60 ± 2.0	65 ± 2.0	64 ± 2.0	87 (52 of 60)
McVeigh et al. [34]	RCT	USA	Patients between 18 and 86 years and a diagnosis of CMC-1 OA based on clinical and radiographic examination	63	66	NA	78 (52 of 67)
Monfort et al. [35]	RCT	Spain	Grades 1-3 Kellgren-Lawrence, symptoms lasting > 90 days requiring analgesic relief	63 ± 8.7b	63 ± 8.7b	NA	88 (77 of 88)
Pisano et al. [40]	RCT	USA	Symptomatic CMC-1 OA diagnosed by an orthopaedic surgeon and/or physician assistant	60 ± 8.8	61 ± 9.6	NA	79 (149 of 189)
Rannou et al. [42]	RCT	France	Radiographic CMC-1 OA, 1 of 2 clinical features of CMC OA, thumb base pain VAS score > 3	63 ± 7.9	63 ± 7.6	NA	90 (101 of 112)
Abdelsabor Sabaah et al. [1]	RCT	Egypt	CMC-1 OA with at least 4 weeks of failed conservative treatment	52 ± 8.3b	52 ± 8.3b	NA	87 (39 of 45)
Sillem et al. [46]	RCT	Canada	Clinical diagnosis of CMC-1 OA, age > 45 years, understands English	64 ± 8.6b	64 ± 8.6b	NA	91 (51 of 56)
Stahl et al. [47]	RCT	Israel	Grade 2 Eaton-Littler, symptomatic CMC-1 OA	62	62	NA	88 (46 of 52)
Tveter et al. [51]	RCT	Norway	CMC-1 OA, understands Norwegian	62 ± 7.5	63 ± 7.8	NA	79 (142 of 180)
Vegt et al. [52]	RCT	Netherlands	CMC-1 OA, age > 18, understands Dutch	61 ± 8.0	59 ± 8.3	NA	75 (44 of 59)
Villafañe et al. [53]	RCT	Spain	Patients age 70–90 years with repetitive hand use and Grade III/IV Eaton-Littler CMC-1 OA diagnosis	82 ± 2.0	83 ± 1.0	NA	85 (51 of 60)

AUSCAN = Australian/Canadian Osteoarthritis Hand Index; NA = not applicable; FIHOA = Functional Index for Hand Osteoarthritis.

^aParticipants were assigned to intervention groups using a nonrandom method.

^bMean for the entire study population.

Table 2.

Treatment nodes, interventions, and definitions used in our network meta-analysis

Treatment node	Description
Comparison node
Control	Placebo/sham intervention, no intervention (such as, education/joint protection advice) or minimal usual care (if no description by study)
Pharmacologic interventions
Corticosteroids	Any preparation of corticosteroids (such as, methylprednisolone acetate, triamcinolone hexacetonide, or betamethasone acetate)
Hyaluronic acid	Either hylan G-F 20 (Synvisc One® [Sanofi]) or sodium hyaluronate
Nonpharmacologic interventions
Rigid CMC splint	Short thumb splint made from thermoplastic material that does not include the metacarpophalangeal joint
Rigid CMC-MCP splint	Long thumb splint made from thermoplastic material that includes the metacarpophalangeal joint
Soft CMC splint	Short thumb splint made from neoprene material that does not include the metacarpophalangeal joint
Soft CMC-MCP splint	Long thumb splint made from neoprene material that includes the metacarpophalangeal joint
Hand exercises	Hand exercises based on biomechanical principles of the CMC joint and the forces that act upon the joint. Both stretching and strengthening exercises are included.
Multimodal treatment	A combination of two or more nonpharmacologic interventions or a nonpharmacologic intervention with a pharmacologic intervention

In this table, each “node” represents a different intervention analyzed in the network meta-analysis. The table also provides definitions and explanations for each intervention.

Table 3.

Interventions, regimen and/or dose outcome measures, assigned nodes, and longest follow-up for selected studies

Study	Intervention 1	Regimen and/or dose	Assigned node	Intervention 2	Regimen and/or dose	Assigned node	Intervention 3	Regimen and/or dose	Assigned node	Pain measure	Function measure	Grip strength	Longest follow-up time in weeks
Adams et al. [2]	Splint and hand exercises	Thumb splint (prefabricated neoprene soft CMC) was to be worn for a minimum of 6 hours per day and hand exercises performed at least 3 times a week for at least 20 minutes each	Multimodal treatment	Hand exercises	Participants were requested to repeat the hand exercises at least 3 times a week for at least 20 minutes each time	Hand exercises				AUSCAN pain	AUSCAN function	No	12
Arazpour et al. [3]	Custom-made rigid CMC thermoplastic splint	Orthosis worn when performing activities of daily living and removed when sleeping and bathing	Rigid CMC	Control group (no splint)	No orthosis, usual care at the discretion of their physician	Control				VAS	MHQ	No	4
Bahadir et al. [4]	Triamcinolone acetonide	One 20-mg injection of triamcinolone acetonide at the start of the trial	Corticosteroid	Hyaluronic acid	Weekly 5-mg injections for 3 weeks	Hyaluronic acid				VAS	DHI	Yes	52
Bani et al. [5]	Custom-made rigid CMC-MCP thermoplastic splint	Splint was worn for 4 weeks during aggravating activities followed by a 2-week washout period without the splint and a subsequent crossover	Rigid CMC-MCP	Prefabricated soft CMC-MCP neoprene splint	Splint was worn for 4 weeks during aggravating activities followed by a 2-week washout period without the splint and a subsequent crossover	Soft CMC-MCP	Not mentioned	Not mentioned	Control	VAS	DASH	Yes	8
Becker et al. [6]	Prefabricated soft CMC-MCP neoprene splint	Orthosis worn for pain relief during activities of daily living and at night	Soft CMC-MCP	Custom-made rigid CMC-MCP thermoplastic splint	Orthosis worn for pain relief during activities of daily living and at night	Rigid CMC-MCP				NRS	DASH	Yes (% of unaffected hand)	9
Can and Tezel [9]	Prefabricated soft CMC-MCP neoprene splint and education program	Patients instructed to wear splint all the time (day and night) as much as possible for first 3 weeks, then only for painful activities for another 3 weeks	Soft CMC-MCP	Education program	Education on causes of thumb OA, symptoms, joint protection, and treatments	Control				AUSCAN pain	AUSCAN function	Yes	6
Cantero-Téllez et al. [11]	Custom-made thermoplastic rigid CMC-MCP thermoplastic splint	Splint used at night and 3-4 hours per day for 3 months	Rigid CMC-MCP	Custom-made rigid CMC thermoplastic splint	Splint used at night and 3-4 hours per day for 3 months	Rigid CMC				VAS	DASH	No	12
Gomes Carreira et al. [22]	Custom-made rigid CMC-MCP thermoplastic splint	Orthosis worn when performing activities of daily living and removed when sleeping and bathing	Rigid CMC-MCP	Control group (no splint)	No splint given from Days 1-90 of study evaluation period, splint given afterward	Control				VAS	DASH	Yes	24
Deveza et al. [13]	Splint, hand exercises, and diclofenac sodium 1% gel	Thumb splint (prefabricated neoprene soft CMC) was to be worn for a minimum of 4 hours per day during activities of daily living and hand exercises performed at least 3 times a week; diclofenac 1% gel applied daily to thumb base 3 times per day	Multimodal treatment	Education program	Education about OA and joint protection	Control				VAS	FIHOA	Yes	12
Figen Ayhan and Ustün [16]	Hyaluronic acid	One 8-mg injection of hyaluronic acid	Hyaluronic acid	Placebo	One 1-mL injection of saline	Control				VAS	FIHOA	No	24
Hermann et al. [24]	Hand exercises and splint	Patients were advised to perform 2 exercise sessions per day with 10 repetitions of each exercise in the study period. Patients were instructed to wear the splint (prefabricated neoprene soft CMC-MCP) as much as they wanted	Multimodal treatment	Hand exercises	Patients were advised to perform 2 exercise sessions per day with 10 repetitions of each exercise in the study period	Hand exercises				AUSCAN pain	AUSCAN function	Yes	8
Heyworth et al. [25]	Betamethasone acetate	Two injections, the first was 1 mL of saline; 1 week later, a 1-mL injection of sodium betamethasone	Corticosteroid	Hyaluronic acid	Two 8-mg injections 1 week apart	Hyaluronic acid	Placebo	Two injections, both 1 mL saline, 1 week apart	Control	VAS	DASH	Yes	24
McVeigh et al. [34]	Custom-made rigid CMC-MCP thermoplastic splint	Orthosis worn when performing activities of daily living and removed when sleeping and bathing	Rigid CMC-MCP	Joint protection advice, activities of daily living modification, adaptive equipment instruction, custom-made splint and hand exercises	Thumb splint was worn for activities of daily living and hand exercises performed at least 3 times a day for 10 repetitions	Multimodal treatment				NRS	Q-DASH	Yes	24
Monfort et al. [35]	Betamethasone acetate	Weekly ultrasound-guided 1.5-mg injections for 3 weeks	Corticosteroid	Hyaluronic acid	Weekly ultrasound-guided 5-mg injections for 3 weeks	Hyaluronic acid				VAS	FIHOA	No	24
Pisano et al. [40]	Custom-made rigid CMC-MCP thermoplastic splint	Orthosis worn when performing activities of daily living and removed when sleeping and bathing	Rigid CMC-MCP	Joint protection advice, activities of daily living modification, adaptive equipment instruction, custom-made splint and hand exercises	Thumb splint was worn for activities of daily living and hand exercises performed at least 3 times a day for 10 repetitions	Multimodal treatment				NRS	Q-DASH	Yes	52
Rannou et al. [42]	Custom-made soft CMC neoprene splint	Splint to be worn at night only	Soft CMC	Usual care	Patients received usual care at the discretion of their physician (general practitioner or rheumatologist)	Control				VAS	CHFS	No	52
Abdelsabor Sabaah et al. [1]	Betamethasone acetate	One 1-mL injection of betamethasone acetate at the start of the trial	Corticosteroid	Hyaluronic acid	One injection containing 730,000 daltons of hyaluronic acid at the start of the trial	Hyaluronic acid				VAS	AUSCAN function	Yes	12
Sillem et al. [46]	Prefabricated soft CMC-MCP neoprene splint	Patients instructed to wear splint when symptomatic, during heavier manual tasks, and at nighttime if desired	Soft CMC-MCP	Custom-made rigid CMC thermoplastic splint	Patients instructed to wear splint when symptomatic, during heavier manual tasks, and at nighttime if desired	Rigid CMC				AUSCAN pain	AUSCAN function	Yes	9
Stahl et al. [47]	Methylprednisolone acetate	One 40-mg injection of methylprednisolone acetate at the beginning of the study	Corticosteroid	Hyaluronic acid	One 15-mg injection of hyaluronic acid	Hyaluronic acid				VAS	No	Yes	24
Tveter et al. [51]	Hand exercises, day and night splints, and five commonly used assistive devices	Patients instructed to exercise at home 3 times per week for 12 weeks, encouraged to use orthoses (thermoplastic custom-made rigid CMC) as often as possible, as well as assistive devices	Multimodal treatment	Usual care	No treatment, usual care at the discretion of their physician	Control				NRS	MAP Hand	Yes	16
Vegt et al. [52]	Custom-made rigid CMC-MCP thermoplastic splint	Patients instructed to wear orthosis for 2 weeks	Rigid CMC-MCP	Prefabricated soft CMC neoprene splint	Patients instructed to wear orthosis for 2 weeks	Soft CMC				VAS	FIHOA	No	4
Villafañe et al. [53]	Joint mobilization, neurodynamic intervention, and hand exercises	Each patient received 3 sessions per week. Joint mobilization according to Kaltenborn technique. Neural mobilization (passive “nerve slider” neurodynamic technique). Exercise as standard regimen	Multimodal treatment	Placebo	Sham (nontherapeutic) pulsed ultrasound with 0 watts/cm2 on the hypothenar area of the dominant hand	Control				VAS	No	Yes	8

MHQ = Michigan Hand Outcomes Questionnaire; DHI = Duruöz Hand Index; NRS = Numeric Rating Scale; Q-DASH = QuickDASH; CHFS = Cochin Hand Function Scale; MAP Hand = Measure of Activity Performance of the Hand.

All selected studies (n = 22) were included in the NMA, which was performed separately for pain, function, and grip strength at short- and medium-term time points. Analysis at the long-term time point was not possible because of a lack of data.

Statistical Analysis

We extracted all relevant mean values and SDs. For grip strength, mean and SD were converted to pounds where possible. In studies where median, range, and sample size had been reported, we used methods described by Wan et al. [54] to estimate the mean and SD. We performed a frequentist NMA using a random-effect meta-regression model through MetaInsight software v6.0.1 using the control node as a reference to generate comparisons. Sensitivity analysis was performed to exclude the quasi-RCT and studies with estimated data. Furthermore, we also performed pairwise meta-analysis where possible when ≥ 2 studies compared the same interventions, measuring the same outcomes at the same predefined time points. We used Review Manager V5.4 (RevMan) software to generate forest plots for pairwise meta-analyses and their accompanying heterogeneity tests and p values using a random-effects model. As studies used different scales for measuring outcomes, pooled standardized mean differences (SMDs) with accompanying 95% confidence intervals (CIs) were calculated for the NMA. For the pairwise meta-analyses, SMDs were calculated only when two studies used different units of measurement, otherwise, mean difference was calculated. Statistical significance was set at two-sided p < 0.05. For interpretation of results across our NMA and pairwise meta-analyses, all SMDs generated for pain, function, and grip strength at the predefined time points were re-expressed in the units of the most commonly used measurement instruments from the included studies (for example, pain measured on the VAS, function measured using the DASH score, and grip strength measured in pounds). This conversion allowed us to calculate scale-specific pooled mean differences with their 95% CIs, which could then be compared with the minimum clinically important difference (MCID) for that specific scale. The re-expression of SMDs was conducted following the methods described by Gallardo-Gómez et al. [19].

Clinical Recommendations

Recommendations for clinical practice were considered strong if: (1) results were clinically important, defined as comparisons with MDs greater than the minimum MCID (according to the research, MCID values are 1.4 points for pain measured on the VAS [range 0 to 10] [41], 10 points for the DASH score [range 0 to 100] for function [18], and 14 pounds for grip strength [8]), and (2) they were based on moderate or high confidence ratings according to CINeMA methodology.

Results

Pain

For reducing pain at the short-term time point (21 studies [1-6, 9, 11, 13, 16, 22, 24, 25, 34, 35, 40, 42, 46, 47, 52, 53] and 9 nodes), forest plot analysis demonstrated superiority in VAS scores for multimodal treatment (mean difference -5.3 [95% CI -7.6 to -3.0]), hand exercises (mean difference -5.0 [95% CI -8.5 to -1.5]), the rigid CMC-MCP splint (mean difference -3.5 [95% CI -5.4 to -1.6]), and the soft CMC-MCP splint (mean difference -2.8 [95% CI -5.0 to -0.6]) versus control (Supplemental Fig. 2; http://links.lww.com/CORR/B353). These results met the MCID, indicating clinically important reductions in pain; however, only multimodal treatment and hand exercises attained moderate confidence in their respective comparisons. From the league table analysis, multimodal treatment, hand exercises, and the rigid CMC-MCP splint demonstrated clinically important reductions in pain versus injections (Supplemental Table 2; http://links.lww.com/CORR/B353); however, only the comparisons involving multimodal treatment and hand exercises were supported by moderate confidence in their respective comparisons. Sensitivity analysis, excluding the quasi-RCT [11] and the study in which mean and SD were estimated, did not alter the results [1]. For reducing pain at the medium-term time point (10 studies [4, 16, 22, 25, 34, 35, 40, 42, 47, 51] and 6 nodes), forest plot analysis demonstrated superiority in VAS scores for the soft CMC splint (mean difference -1.9 [95% CI -3.3 to -0.5]), the rigid CMC-MCP splint (mean difference -1.9 [95% CI -3.1 to -0.6]), and multimodal treatment (mean difference -1.3 [95% CI –2.4 to -0.2]) versus control. While clinically important results were observed for the soft CMC and rigid CMC-MCP splints, only the comparison with the rigid CMC-MCP splint was based on moderate confidence (Supplemental Fig. 3; http://links.lww.com/CORR/B353). League table analysis did not demonstrate any superiority between different interventions (Supplemental Table 3; http://links.lww.com/CORR/B353).

Function

For improving function at the short-term time point (17 studies [1-6, 9, 11, 13, 16, 22, 25, 34, 40, 42, 46, 52] and 9 nodes), forest plot analysis did not demonstrate any superiority of any intervention versus control (Supplemental Fig. 4; http://links.lww.com/CORR/B353). Furthermore, league table analysis did not demonstrate any superiority between any interventions (Supplemental Table 4; http://links.lww.com/CORR/B353). Similarly, as with short-term pain, sensitivity analysis did not alter the results. For medium-term function (8 studies [4, 16, 22, 25, 34, 40, 42, 51] and 6 nodes), forest plot analysis demonstrated superiority in DASH scores for both the rigid CMC-MCP splint (mean difference -11 [95% CI -21 to -1]) and multimodal treatment (mean difference -10 [95% CI -19 to -1]) versus control. Although clinically important results were observed for both, only the comparison involving the rigid CMC-MCP splint was supported by moderate confidence (Supplemental Fig. 5; http://links.lww.com/CORR/B353). League table analysis did not demonstrate any superiority between any interventions (Supplemental Table 5; http://links.lww.com/CORR/B353).

Grip Strength

For improving grip strength at the short-term time point (13 studies [4-6, 9, 13, 22, 24, 25, 34, 40, 46, 47, 53] and 8 nodes), forest plot analysis demonstrated that hand exercises were superior, with a mean difference of 21 pounds (95% CI 11 to 31), indicating a clinically important improvement in grip strength (Supplemental Fig. 6; http://links.lww.com/CORR/B353). League table analysis also demonstrated clinically important improvements in grip strength for hand exercises versus splints (both the rigid and soft CMC and CMC-MCP splint) and injections (corticosteroid and hyaluronic acid), with moderate or high confidence in their respective comparisons (Supplemental Table 6; http://links.lww.com/CORR/B353). At the medium-term time point (5 studies [4, 22, 25, 40, 51] and 5 nodes), forest plot analysis did not demonstrate any superior intervention versus control (Supplemental Fig. 7; http://links.lww.com/CORR/B353). Similarly, league table analysis did not demonstrate any superiority of any intervention (Supplemental Table 7; http://links.lww.com/CORR/B353).

Other Relevant Findings: Pairwise Meta-analyses

Overall, pairwise meta-analyses were possible for eight comparisons (Supplemental Table 8; http://links.lww.com/CORR/B353).

Hyaluronic Acid Versus Control

For reducing pain at the short-term time point, we found no difference in VAS scores between hyaluronic acid and control (mean difference -0.4 [95% CI -2.2 to 1.4]; p = 0.66, low confidence rating) (Supplemental Fig. 8; http://links.lww.com/CORR/B353). At the medium-term point, hyaluronic acid injections demonstrated a clinically important reduction in pain versus control (mean difference VAS score -1.7 [95% CI -3.4 to -0.1]; p = 0.04, low confidence rating) (Supplemental Fig. 9; http://links.lww.com/CORR/B353). For improving function at the short- and medium-term time points, we found no difference in DASH scores between hyaluronic acid and control (mean difference 0.8 [95% CI -4.4 to 5.8]; p = 0.79, low confidence rating, and mean difference -2.8 [95% CI -15.4 to 9.7]; p = 0.66, very low confidence rating, respectively) (Supplemental Figs. 10 and 11; http://links.lww.com/CORR/B353.

Multimodal Treatment Versus Control

At the short-term time point, for reducing pain, we found a clinically important reduction in pain with multimodal treatment versus control (mean difference VAS score -2.0 [95% CI -3.8 to -0.2]; p = 0.03, very low confidence rating) (Supplemental Fig. 12; http://links.lww.com/CORR/B353). For improving grip strength, we found no difference between multimodal treatment and control (mean difference -0.2 pounds [95% CI -1.4 to 1.0]; p = 0.73, moderate confidence rating) (Supplemental Fig. 13; http://links.lww.com/CORR/B353).

Rigid CMC-MCP Splint Versus Control

For reducing pain at the short-term time point, we found a clinically important reduction in pain with the rigid CMC-MCP splint versus control (mean difference VAS score -2.9 [95% CI -5.2 to -0.7]; p = 0.01, very low confidence rating) (Supplemental Fig. 14; http://links.lww.com/CORR/B353). For improving short-term function and grip strength, we found no difference between the rigid CMC-MCP splint and control (mean difference DASH score -0.5 [95% CI -12 to 11]; p = 0.94, very low confidence rating, and mean difference 2.9 pounds [95% -0.1 to 6.0]; p = 0.06, moderate confidence rating, respectively) (Supplemental Figs. 15 and 16; http://links.lww.com/CORR/B353).

Soft CMC-MCP Splint Versus Control

At the short-term time point for pain, the soft CMC-MCP splint demonstrated a clinically important reduction in pain versus control (mean difference VAS score -2.7 [95% CI -4.3 to -1.1]; p = 0.001, very low confidence rating) (Supplemental Fig. 17; http://links.lww.com/CORR/B353). For improving function at the short-term time point, we found no difference between the soft CMC-MCP splint and control (mean difference DASH score -0.3 [95% CI -26 to 25]; p = 0.98, very low confidence rating) (Supplemental Fig. 18; http://links.lww.com/CORR/B353). For improving grip strength at the short-term time point, we found no clinically important difference between the soft CMC-MCP splint and control (mean difference 3.6 pounds [95% 0.3 to 7.0]; p = 0.03, low confidence rating) (Supplemental Fig. 19; http://links.lww.com/CORR/B353).

Hand Exercises Versus Multimodal Treatment

For reducing pain at the short-term time point, we found no difference between hand exercises and multimodal treatment (mean difference VAS score 0.3 [95% CI -0.5 to 1.2]; p = 0.44, high confidence rating) (Supplemental Fig. 20; http://links.lww.com/CORR/B353).

Corticosteroid Versus Hyaluronic Acid

For reducing pain at both short- and medium-term time points, we found no difference in VAS scores between corticosteroid and hyaluronic acid injections (mean difference 0.4 [95% CI -1.5 to 2.2]; p = 0.70, very low confidence rating, and mean difference 0.2 [95% CI -0.9 to 1.2]; p = 0.76, very low confidence rating, respectively) (Supplemental Figs. 21 and 22; http://links.lww.com/CORR/B353). Subgroup analysis by removing the study contributing to the most heterogeneity did not alter the finding for the short-term time point [1]. Similarly, for improving function, we found no difference in DASH scores between corticosteroid and hyaluronic acid injections at the short- and medium-term time points (mean difference 6 [95% CI -16 to 27]; p = 0.62, very low confidence rating, and mean difference -1 [95% CI -9.4 to 8.0]; p = 0.87, moderate confidence rating, respectively) (Supplemental Figs. 23 and 24; http://links.lww.com/CORR/B353), and subgroup analysis did not alter the findings. Last, for improving grip strength at both the short- and medium-term time points, we found no difference between corticosteroid and hyaluronic acid injections (mean difference -2.6 pounds [95% CI -13.3 to 8.2]; p = 0.64, low confidence rating, and mean difference -0.9 pounds [95% CI -13.4 to 11.6]; p = 0.89, low confidence rating, respectively) (Supplemental Figs. 25 and 26; http://links.lww.com/CORR/B353).

Rigid CMC-MCP Versus Soft CMC-MCP Splint

For reducing pain and improving function and grip strength at the short-term time point, we found no difference between the rigid CMC-MCP and soft CMC-MCP splint (mean difference VAS score -0.3 [95% CI -1.2 to 0.5]; p = 0.44, low confidence rating; mean difference DASH score -2.7 [95% CI -7.6 to 2.0]; p = 0.27, low confidence rating; and mean difference 2.6 pounds [95% CI -2.8 to 7.9]; p = 0.34, low confidence rating, respectively) (Supplemental Figs. 27-29; http://links.lww.com/CORR/B353).

Multimodal Treatment Versus Rigid CMC-MCP Splint

At the short-term time point, for reducing pain and improving function and grip strength, we found no difference between multimodal treatment and the rigid CMC-MCP splint alone (mean difference VAS score -1.3 [95% CI -9.0 to 6.5]; p = 0.75, low confidence rating; mean difference DASH score 0 [95% CI -2.9 to 3.0]; p = 0.98, low confidence rating; and mean difference -1.3 pounds [95% CI -8.0 to 5.3]; p = 0.69, low confidence rating, respectively) (Supplemental Figs. 30-32; http://links.lww.com/CORR/B353). Similarly, at the medium-term time point, no difference was found for pain and function (mean difference VAS score 4.0 [95% CI -0.1 to 8.1]; p = 0.06, low confidence rating, and mean difference DASH score 0.8 [95% CI -6.2 to 7.8]; p = 0.82, low confidence rating, respectively) (Supplemental Figs. 33 and 34; http://links.lww.com/CORR/B353).

Discussion

CMC-1 OA is a common, debilitating condition, particularly in older adults, leading to pain and functional limitations [21]. Nonoperative interventions include both nonpharmacologic (joint protection, hand exercises, and splints) and pharmacologic. Guidelines from EULAR and the British Society for Surgery of the Hand recommend a stepwise approach, beginning with noninvasive interventions such as splints, hand exercises, and multimodal treatment packages [20, 30]. However, despite their widespread use, there is still no consensus on the most effective interventions. Based on clinically important results of moderate to high confidence, our NMA demonstrated that multimodal treatment and hand exercises reduced pain at the short-term time point, while the rigid CMC-MCP splint was superior at the medium-term time point. For function, the rigid CMC-MCP splint showed improvement at the medium-term time point, and hand exercises improved grip strength at the short-term time point. Clinicians should prioritize multimodal treatment and/or hand exercises for pain and grip strength management for short-term improvements and consider the rigid CMC-MCP splint beyond this. These results can help guide more effective, personalized treatment plans for patients with CMC-1 OA.

Limitations

A key limitation of this NMA was the dilemma between lumping and splitting treatments [56], especially given the lack of previous NMAs comparing different nonoperative interventions for CMC-1 OA. Lumping combines similar interventions into one group, which simplifies the model but potentially masks specific treatment effects. For instance, joint protection advice, as outlined by the EULAR 2018 guidelines [30], was included in the control node because of insufficient network connectivity, potentially obscuring its individual and collective benefits of reducing pain and improving function. Additionally, placebo injections, which are theorized to generate a stronger placebo effect compared with no intervention or joint protection advice alone [17], could ideally have been analyzed as a separate entity. However, while recent studies demonstrate the noninferiority of placebo injections [31, 43], evidence for their superiority over corticosteroids or hyaluronic acid is still lacking. This justifies their inclusion within the control arm.

Another limitation of this study was the re-expression of SMDs to MDs from the most commonly used scales (VAS, DASH, and pounds). For pain (VAS) and grip strength (in pounds), this method was consistent, as these scales were used in the majority of studies. However, for function, the DASH score was used in only 5 of 20 studies, limiting the precision of the mean difference estimates for functional outcomes. Thus, readers should interpret the functional findings with caution, as the smaller data set may reduce the reliability of our comparisons (point estimates) generated at both the short- and medium-term time points.

We also did not explicitly record how studies handled concomitant treatments unless this was stated as part of the intervention or comparator details. In our NMA, there was heterogeneity in many included studies as to whether patients were explicitly told to stop taking oral or topical NSAIDs. Consequently, a larger cohort of patients could have been included in a multimodal treatment regimen, as defined by our intervention node (Table 2; combinations of nonpharmacologic interventions or a pharmacologic intervention alongside a nonpharmacologic one), rather than being restricted to a single intervention with a distinctly defined node. We feel that this is of key importance as patients are not usually advised to stop taking medications while receiving another intervention.

Last, a previous systematic review related to CMC-1 OA highlighted the difficulties that arise when attempting to conduct statistical comparisons across various interventions mainly due to the lack of homogeneity in both populations, nonspecific effects related to invasiveness of interventions, and PROMs used by different authors [32]. To account for these differences, where we felt necessary, we downgraded by one or two levels for indirectness, as per CINeMA methodology.

Pain

The findings of our NMA on using multimodal treatment for reducing pain as a first-line strategy in patients with symptomatic CMC-1 OA align with the current evidence. Wouters et al. [55] reported that integrating hand exercises with splinting resulted in a larger decrease in VAS scores compared with splinting alone in patients with CMC-1 OA. Similarly, Ye et al. [57] found that combining nonpharmacologic therapies such as physical therapy with NSAIDs resulted in a 40% improvement in pain management. The results of our NMA and the current evidence strengthen our recommendations that clinicians should trial multimodal treatment at first contact, especially given the cost-effectiveness reported at an addition of 0.06 quality-adjusted life-years [50].

Although the results of our pairwise meta-analysis did not show any difference in pain reduction between multimodal treatment and hand exercises alone, our NMA demonstrated that hand exercises alone can be beneficial in treating patients with CMC-1 OA when compared with no intervention or injections. This finding is similar to a recent Cochrane review supporting the use of exercise-based interventions for improving pain in patients with hand OA [39].

Splints, in particular those that extend up to the MCP joint to provide greater support, appear to be superior to their shorter counterpart, with the rigid CMC-MCP splint being better at alleviating pain at the medium-term time point versus control. Our findings for pain agree with a recent NMA by Marotta et al. [33]. This difference in pain relief could reflect the rigidity of the CMC-MCP splint blocking movement at multiple joints. MCP joint stabilization may avert excessive MCP extension and reduce pain by restricting certain movements, types of actions, or overall activities. Additionally, the stabilization of the MCP joint can correct the distribution of thumb loads during activities of daily living, thereby avoiding painful joint hyperextension [10].

Function

In contrast to Marotta et al. [33], our NMA identified medium-term functional improvement with the rigid CMC-MCP splint, a novel finding that contradicts the usual biomechanical understanding. Longer splints, extending to support the MCP joint, are generally believed to restrict hand function. Shorter splints, however, stabilizing only the CMC-1 joint, are thought to offer better mobility and functionality by allowing greater hand and thumb movement, improving the ability to grip or pinch effectively [7]. Therefore, differences in our findings may be due to factors affecting internal validity, such as variations in sample size, splint-wearing schedules, CMC-1 OA stages, hand dominance, concomitant therapy, and the variables measured in studies using the rigid CMC-MCP splint as an intervention. These differences may explain the unexpected functional benefits seen with the longer splints in our analysis.

Grip Strength

Our NMA demonstrated notable efficacy in improving grip strength with hand exercises in the short term for patients with CMC-1 OA. This also supports the findings of recent meta-analyses that have reinforced these short-term benefits of hand exercises. Gutierrez Espinoza et al. [23] demonstrated an improvement in grip strength with moderate certainty of evidence in five trials. More recently, Huang et al. [26] demonstrated improvement across 14 trials in grip strength (SMD 0.21 [95% CI 0.03 to 0.38]) with hand exercises, albeit with a small effect size. Similar to the findings of our NMA, benefits were only present in the short term but were not sustained during medium- to longer-term follow-up. Interestingly, our NMA showed superiority of hand exercises alone over multimodal treatment. Exercise is commonly recommended as a primary intervention for OA due to its effectiveness in reducing pain and stiffness and in improving joint function. It achieves these benefits by strengthening the muscles around the affected joint, enhancing joint lubrication and nourishment, and preserving joint flexibility and stability [29, 36]. In contrast, when exercise is part of a multimodal treatment plan that includes passive interventions such as splints and pharmacologic therapy, its effectiveness may be diminished as these passive interventions can potentially limit the engagement and benefits of active exercises, thereby reducing their overall therapeutic impact.

Conclusion

Our network meta-analysis found that multimodal treatment and hand exercises reduce pain and improve grip strength in the short term for CMC-1 OA. For medium-term functional improvement, a rigid CMC-MCP splint seems more effective. Clinicians should use multimodal treatment and hand exercises for immediate relief and consider a rigid CMC-MCP splint for ongoing function enhancement. Future research should investigate the long-term efficacy of these interventions and explore various multimodal combinations in larger, diverse patient populations to further refine and personalize CMC-1 OA management strategies.