Influence of injection volume on rate of subsequent intervention in Carpal Tunnel Syndrome over 1 year follow-up

Abstract

Purpose

The optimal volume and dose of corticosteroid injections for treatment of carpal tunnel syndrome (CTS) has not yet been established. It is unknown whether volume of injectate influences the outcome of carpal tunnel injection. The purpose of our study was to assess whether there is an association between volume of injectate and subsequent intervention in the treatment of CTS.

Methods

This study evaluated residents of Olmsted County, MN treated with a corticosteroid injection for CTS between 2001 and 2010. Failure of treatment was the primary outcome and defined as a subsequent intervention: either a second injection or carpal tunnel release within one year of initial injection. General estimating equations logistic regression was used to assess the association between injectate volume and rate of treatment failure, adjusting for age, sex, effective dose of steroid, type of steroid injected, electrodiagnostic (EDX) severity, and the presence of comorbidities such as rheumatoid arthritis, diabetes mellitus, peripheral neuropathy and radiculopathy.

Results

There were 856 affected hands in 651 patients. Fifty-six percent (N=484) of the treated hands received subsequent treatment within 1 year. Multivariable analysis showed that larger injectate volume was significantly associated with reduced rate of treatment failure within one year. Rheumatoid arthritis, ultrasound-guided procedures, and severe EDX results were also associated with increased rate of failure.

Conclusions

This study shows that a larger volume of corticosteroid injection is associated with reduced odds of subsequent intervention after a single corticosteroid injection in CTS. Further research is needed to determine the optimal volume for steroid injections in the treatment of CTS.

Type of study

retrospective case series, prognostic level IV

Introduction

For patients with CTS, local corticosteroid injection provides greater clinical improvement in symptom relief one month after the injection in comparison to a placebo injection¹. An injection for CTS commonly involves a combination of steroid and local anesthetic^2,3.

A systematic review has reported that the site of injection, dose of steroid, type of steroid and/or local anesthetic, and concomitant splinting after injection do not alter the underlying efficacy of a steroid injection for CTS⁴. However, it is important to note that most of the recorded interventions involve injections of relatively low volumes of total injectate, most commonly 1–3 mL.

While corticosteroid injection provides short term benefit to many patients with CTS and longer term benefit to some, the optimal volume of corticosteroid injections for CTS has not yet been determined, and it is unknown whether volume of injectate influences the outcome of carpal tunnel injection. It is the opinion of some practitioners that larger volumes may be associated with a reduced risk of symptom relapse, due to the effects of the added volume to ‘hydrodissect’ and mobilize the nerve^5,6. In addition, higher volume injections could be related to larger area of fluid dispersion/distribution, greater contact area with soft tissues, or extravasation of fluid outside the canal^7–9. Therefore, the aim of our retrospective study was to assess whether there is an association between volume of injection and subsequent intervention after a single corticosteroid injection in patients with CTS.

Materials and methods

Data collection

The cohort of patients receiving an injection for CTS was identified from residents of Olmsted County, Minnesota using the medical records-linkage system of the Rochester Epidemiology Project (REP)¹⁰. The Institutional Review Board approved the study. The Rochester Epidemiology Project (REP) is a research infrastructure system that links together nearly all of the medical records (1966-present) of the residents of Olmsted County, MN. The REP consists of demographic data, diagnostic codes and surgical procedure codes organized in electronic indexes that can be searched. Multiple medical records for the same patient are linked within and across institutions to create a comprehensive record, irrespective of where a county resident is seen. Participating health care providers include not only those within Olmsted County but also those in the surrounding region. Studies have shown that the database includes nearly all care provided to nearly all (i.e., >90%) county residents¹¹ . Thus, there is a high degree of confidence that treatment failures will be captured, even if subsequent treatment is provided at a different facility. The study subjects were selected based on an International Classification of Disease diagnosis code of carpal tunnel syndrome (ICD-9 354.0) and a Current Procedural Terminology code for carpal tunnel injection (CPT 20526) between 2001 and 2010.

Failure of treatment, defined as any subsequent corticosteroid injection or carpal tunnel release surgery performed on the injected hand within one year of initial injection was the primary outcome of interest. Inclusion criteria included a diagnosis of primary CTS, no previous invasive intervention (injection or Carpal Tunnel Release) for CTS in that hand, a therapeutic corticosteroid treatment for CTS, age at least 18 years old, and affirmative research authorization (over 90% of county residents have provided such authorization for record review research). Subjects who had less than 1 year of follow-up in their medical record were excluded from the analysis. If a patient did not have at least one visit to any REP health care provider of any specialty at or after one year past their date of injection, we considered the patient as having less than one year of follow-up. All the steroids were standardized to be able to compare the relative anti-inflammatory potency. Dose of steroid was standardized to an equivalent effective dose of triamcinolone, which had the highest use in the cohort and was converted according to Leversee (1986)¹². For example, the equipotent dose of 0.6 mg betamethasone was 4.0 mg triamcinolone. Subsequently, the standardized dose (mg/mL) was multiplied by the volume of the steroid in mL. The volume of the steroid was documented in mL in the medical charts, ranging from 0.5 – 3.0 with an interval of 0.5 mL. The amount of anesthetic was separately documented.

Study data were collected and managed by three physicians (SE, AB, TS) using Research Electronic Data Capture (RedCAP) software. Age, sex, laterality, comorbidities including diabetes, rheumatoid arthritis, peripheral neuropathy/cervical radiculopathy, and other relevant comorbidities such as previous trauma to the affected hand or wrist, Kienbock’s disease, or ganglion cyst were recorded. In addition, information on the diagnosis for CTS was abstracted from the medical charts. We required a clinical diagnosis of CTS to be entered in the medical record, based on the clinical impression of the treating physician. We also collected the type and dose of steroid injected, type and volume of anesthetic injected, number of injections and surgical intervention (CPT code for surgical cases 29848, 64721). Furthermore, the severity of CTS was assessed by documenting available EDX data, with severity graded as recommended by Stevens¹³. Our study adhered to the STROBE guidelines for observational studies¹⁴.

Statistical analysis

For the purposes of power calculations we assumed that injectate volume would follow a normal distribution (N(3,1)) and that the proportion of treatment failure would be 50%. For a logistic regression model, we would have 80% power at the alpha=0.05 level to detect an odds ratio of 1.3 for 1 mL unit change in injectate volume with N=600 patients.

Summary statistics for demographics and clinical characteristics are shown as N (%) or mean ± standard deviation (SD) and are divided into characteristics at the patient level and hand level. Patient-level covariates include sex and previous history of comorbidities. Hand-level covariates included type and volume of anesthetic, type and dose of steroid, total injectate volume, age at treatment, the use of US guidance and EDX severity at time of injection. Confidence intervals for failure proportions by volume use degrees-of-freedom adjusted effective sample sizes to account for clustered observations.

General estimating equations (GEE) logistic regression was used to model suspected risk factors' association with treatment failure^15,16. General estimating equations are a type of generalized linear model, which account for correlated repeated measurements within individuals. To appropriately calculate variance for repeated measures analyses, the GEE method requires a correlation structure be specified. Model fit and choice of correlation structure were assessed using the quasi-likelihood under the independence model information criterion (QIC)¹⁷. GEE methods are typically robust to misspecification of correlation structure, nevertheless several possible correlation structures were fit for each model and the best selected by QIC. Ultimately, an exchangeable correlation structure, i.e. one in which any pair of observations for a subject will have the same correlation, was used.

Prior to model inclusion, dose of steroid was converted to effective dose of triamcinolone as previously specified. Factors assessed included age, sex, total injectate volume (combined steroid and anesthetic volume), effective dose of steroid, type of steroid injected, type of anesthetic injected, volume of anesthetic injected, ultrasound-guidance, history of diabetes, diagnosis of peripheral neuropathy or cervical radiculopathy, diagnosis of rheumatoid arthritis, diagnosis of other relevant comorbidities, and EDX severity at time of injection. Four categories of EDX severity were originally assigned: normal, mild, moderate, and severe¹³. Subjects on whom EDX study was not performed or EDX severity was not known were assigned to a fifth category. Variables were reviewed for evidence of collinearity by means of variance inflation factors and principal component analysis prior to model inclusion.

Sensitivity analyses for different parameterizations of variables (e.g. groupings of EDX severity) and other model diagnostics were performed. First-order interactions between all variables were assessed during model fitting but were excluded from final model if non-significant. Odds ratios (OR) and 95% Wald confidence intervals (CI) were reported. P-values less than 0.05 were considered significant.

Results

A total of 988 eligible patients were identified. Patients without at least 1 year of follow-up (N=21), diagnosed with pregnancy-related CTS (N=20), and observations with missing injectate volume or steroid dose (N=91) were excluded from the analysis (Figure 1). After exclusions, the cohort consisted of 856 observations in 651 distinct patients. Patients were treated by plastic surgeons, orthopedic surgeons, PMR physicians, family physicians, rheumatologists, radiologists and internists.

Figure 1.

Subject selection flow chart.

Demographic, clinical, and procedural characteristics are shown in Table 1. The patients were 30% male and had a mean (SD) age at treatment of 51 years (13.7). Injections had a mean total injectate volume of 3.54 mL (1.17) and effective steroid dose of 41 mg (21.9). Over 96% of injections used 1% lidocaine as the anesthetic, with an additional 2% having an undocumented type. Given this distribution, type of anesthetic was not included as a predictor in the final model. At the hand level, 57% (N=484) received subsequent treatment within 1 year. Of those, 30% were treated with a second injection (N=143) and 70% (N=341) underwent carpal tunnel release. Of the patients who experienced failure of treatment within one year, the median time to a second injection was 161 days and the median time to surgery was 128 days.

Table 1.

Demographic, clinical, and procedural characteristics of corticosteroid injection for treatment of carpal tunnel syndrome cohort. Electrodiagnostic (EDX) severity.

	Subject-level (N=651) Hand-level* (N=856)
Age (years)* Mean (SD)	51 (13.7)
Sex (Male)	194 (30%)
EDX severity*
Untested	129 (15%)
Normal	81 (9%)
Mild	244 (29%)
Moderate	346 (40%)
Severe	56 (7%)
Injectate volume (mL)* Mean (SD)	3.54 (1.17)
Type of anesthetic
1% Lidocaine	824 (96%)
2% Lidocaine	3 (<1%)
0.25% Bupivacaine	6 (1%)
0.5% Bupivacaine	5 (1%)
None/Other/Unknown	18 (2%)
Effective steroid dose (mg)* Mean (SD)	41 (21.9)
Type of steroid
Triamcinolone	488 (57%)
Betamethasone	304 (36%)
Methylprednisolone	62 (7%)
Rheumatoid arthritis	44 (7%)
Diabetes mellitus	56 (9%)
Peripheral neuropathy or Cervical radiculopathy	48 (7%)
Other comorbidity	18 (3%)

^*Indicates hand-level characteristics.

Figure 2A shows the number and proportion of failure of injections grouped by volume of injectate in the cohort unadjusted for other covariates. Overall, there is a downward trend observed for proportion of failures with higher injectate volumes (Figure 2B). Given data sparsity at high and low levels of injectate volume, models were also fit excluding extreme values of injectate volume (<1mL and <6mL) to test model sensitivity; as no change in sign or significance of injectate volume was found, results presented are from the model fit on the full set of data (i.e. without any exclusions due to injectate volume). A model using injectate as a binary variable (<=3 mL, >3 mL) was also assessed. Although this model showed poorer overall model fit, we have included the results from the binary representation of volume in addition to the one based on a numeric representation of volume. Based on model fit criteria, EDX severity was collapsed to two categories (severe vs. not severe, which included untested, normal, mild and moderate). Sensitivity analyses were performed and changes in how the levels of EDX severity were grouped did not affect sign or significance of other variables included in the model. No first order interactions were found to be significant.

Figure 2.

A. Proportion of failure of injections subdivided by volume of injectate into categories by 1mL volumes, adjusted to account for subjects with multiple observations. Error bars represent 95% confidence intervals. B. Percentage of corticosteroid injections failures by injectate volume, with observations treated as independent. Solid line represents a natural cubic spline fit with 3 knots and the shaded area the 95% confidence interval for the estimate.

Odds ratios and 95% Wald confidence intervals from the final multivariable model fit are shown in Table 2. Higher injectate volume was significantly associated with reduced treatment failure within one year (adjusted OR 0.777 [0.671, 0.901], p <0.001). Rheumatoid arthritis and ultrasound-guided procedures were also associated with reduced treatment failure (RA adjusted OR 0.432 [0.22–0.846], p=0.01; US-guided adjusted OR 0.447 [0.247–0.806], p=0.01) and severe EMG diagnosis was associated with increased failure within one year (adjusted OR 1.82 [1.07–3.07], p=0.03). Effective dose of steroid was not significant (adjusted OR 0.998 [0.990, 1.01], p = 0.38) nor was type of steroid injected (p=0.65). Table 3 shows the estimates from the binary representation of volume (<=3, >3).

Table 2.

Demographics, clinical characteristics and procedural characteristics and their association with re-intervention (either second corticosteroid injection or carpal tunnel release) within one year using a General estimating equations (GEE) logistic regression model.

Variable	OR	Lower	Upper CI	P-value
Sex (Male)	0.761	0.543	1.07	0.11
Age (/10 years)	0.984	0.877	1.10	0.78
Diabetes	0.947	0.545	1.64	0.85
Peripheral Neuropathy or Cervical Radiculopathy	0.811	0.448	1.47	0.49
Rheumatoid Arthritis	0.432	0.221	0.846	0.01
Other Comorbidity	1.40	0.569	3.44	0.46
EDX severity (Severe vs. all other severities)	1.82	1.07	3.07	0.03
Injectate volume	0.777	0.671	0.901	<0.001
Effective steroid dose	0.998	0.990	1.01	0.68
US-Guided	0.447	0.247	0.806	0.01
Type of steroid injected (compared to Triamcinolone)				0.65
Betamethasone	0.855	0.594	1.23
Methylprednisolone	0.817	0.387	1.73

Odds ratios (OR), 95% confidence intervals (CI), and model p-values are presented.

Table 3.

Results from model with binary volumes (<=3, >3), showing significant results only:

	OR	95% low	95% high	p-value
Rheumatoid Arthritis	0.465	0.237	0.913	0.026
Severe EMG	1.80	1.06	3.05	0.029
US-guided	0.435	0.241	0.785	0.006
High volume (>3mL)	0.509	0.35	0.741	<.001

Discussion

This retrospective population-based study showed that larger volume of injection is associated with reduced odds of subsequent intervention within one year after a single corticosteroid injection in patients with CTS. There is currently no consensus regarding the optimal volume of injection for CTS. The effectiveness of corticosteroid injections has been studied and there is strong evidence for the benefits in the short-term compared to placebo¹⁸. Armstrong et al. evaluated 43 patients given 1 mL of 6 mg betamethasone and 1mL lidocaine and compared them to 38 patients who received 1 mL of saline placebo combined with 1 mL lidocaine¹. The primary outcome measures were satisfaction and symptom relief. Thirty patients (70%) in the steroid-treated group were satisfied or highly satisfied compared with 13 (34%) of placebo- treated patients. Reported rates of subsequent treatment vary from 10% –81% within one year^18–23. However, most of the studies are based on relatively low injectate volumes (1–3 mL) only. As opposed to dose-response, the volume-response relationship in corticosteroid injections for CTS has not been studied. A placebo controlled trial comparing injection of 40mg methylprednisolone, 80mg methylprednisolone and placebo in patients with idiopathic CTS did not find a difference between 40mg and 80mg methylprednisolone injections (both 3mL mixtures) at 10 weeks, but compared with patients who received placebo, those who received 80 mg of methylprednisolone were less likely to have surgery at 1 year follow-up¹⁸. In our study, we did not find an association between effective dose of steroid and proportion of subsequent intervention. Karadas et al. compared local injection of 40 mg triamcinolone to procaine (4 mL) in a randomized placebo-controlled trial.²⁴. At 6 months follow-up, the steroid and procaine both had a significant improvement in clinical outcome and electrophysiological findings, with no difference between groups. The 1 mL saline (placebo) group did not improve. This result indicates that the mechanism of action of an injection in CTS might not solely depend on the effect of a corticosteroid drug.

Our finding that large volume injections lead to a reduced rate of subsequent intervention might be due to a greater distribution of the injectate. Cadaveric studies have shown that there is a wide variability in the distribution of injectate in the carpal tunnel^7–9. Jariwala et al. investigated the diffusion pattern of a 3.2 mL injectate of local anesthetic, steroid and dye into the carpal tunnel in a cadaveric model using a commonly used technique described by Green in 1984^7,25. Their study showed three dye distribution patterns: free distribution within the carpal tunnel (60%), distribution in the tendon sheath and a mixed distribution (40%). The variability in the distribution suggests that, especially with smaller injectate volumes, the injectate might have had little or no contact with the median nerve, and little distribution within the carpal tunnel synovium. This may be important, as recent studies have suggested that fibrosis of the carpal tunnel subsynovial connective tissue (SSCT) is an important part of the pathophysiology of CTS^26,27. Larger volume injectates might lead to greater distribution within the synovium, and a consequently greater effect on this tissue.

Increased pressure within the carpal tunnel is a characteristic finding of CTS²⁸. Therefore, it might be argued that there should be concern with using a large volume injectate, considering that the average carpal tunnel volume is approximately 5–6 mL^29,30 and injecting a large volume solution might increase the pressure within the carpal tunnel even further. However, there is no anatomical evidence that the carpal tunnel is a closed compartment³¹, thus extravasation of the injectate outside the tunnel will likely occur. Moreover, to the extent that SSCT fibrosis may create a closed compartment where none existed normally, the fluid volume might serve to mechanically disrupt this, through the mechanism of hydrodissection^5,6.

The study has several important limitations. First, the review was retrospective in nature. This introduces the potential for errors in documentation of inclusion/exclusion criteria, relevant comorbidities and other risk factors, as well as missing outcomes. For example, over 9% of potential subjects were excluded due to lack of data on volume and steroid dose and 2% were excluded due to lack of follow-up and it is unknown what effect inclusion of these subjects may have had. The lack of randomization makes it more likely that unmeasured confounders may have had an effect. For example, injectate volume and type of steroid was physician-dependent and so individual variations in injection technique, or even patient selection, might be confounders⁸. Although we have accounted for ultrasound-guided procedures, we could not rule out that injection technique has influenced the outcome.

The review was powered to assess for the effect of a change in injectate volume on retreatment. While other risk factors were included as adjustors in the multivariable model and first-order interactions were tested, the study was not powered to assess potential interactions. This may have led to non-significance of risk factors previously shown to be associated with increased failure in CTS patients or exclusion of relevant interactions from the model. In addition, while this study showed an association between increased volume of injectate and lower risk for subsequent treatment, it is difficult to state what the optimal volume of injectate might be. The relationship between injectate volume and outcome might change at volumes outside the reported range.

Our outcome as defined (seeking retreatment with second injection or carpal tunnel surgery within one year) is only a proxy for overall clinical outcome. While a relatively small percentage of observations were excluded due to insufficient follow-up (2.1%), the outcome likely underestimates the true rate of treatment failure, both for patients who still have clinically relevant symptoms but who choose not to seek re-injection or surgery for other reasons and for subjects who receive treatment that is not captured in the available medical record.

Gelfman et al. stated in their study on long-terms trends in CTS using data from the REP, that approximately 80% of the administrative diagnostic data met the symptom criteria for CTS³². Since our original data was derived from administrative diagnostic codes of the REP, we confirmed the diagnosis of CTS through chart review.

In conclusion, this cohort showed an association between larger injectate volumes and reduced rates of subsequent intervention after a single corticosteroid injection in the treatment of CTS. Further research is necessary to define an optimal volume, as well as the mechanism for this effect. A prospective study with a low and high volume injection group and to reassess outcomes would be ideal.