Abstract
Background: Corticosteroid injections are commonly used in the treatment of flexor tenosynovitis in adults. The present study was performed in an attempt to identify prognostic indicators of symptom recurrence one year after corticosteroid injection for the treatment of trigger digits.
Methods: One hundred and thirty consecutive patients with trigger digits treated with corticosteroid injection were prospectively enrolled. Exclusion criteria were prior treatment and inflammatory arthritis. Demographic data and information on existing comorbidities were identified with a questionnaire. Patients were contacted at one year after treatment to determine symptom recurrence. Kaplan-Meier analysis and the Cox regression model were used to estimate recurrence rates and identify predictors.
Results: One hundred and twenty-four trigger digits in 119 patients (average age, 62.3 years) were included. The most commonly involved digits were the thumb (35% of the digits), ring finger (31%), and long finger (23%). Seventy digits (56%) had a recurrence of symptoms at a median of 5.6 months (range, 0.5 to 13.1 months) after the injection. Twenty-two digits (18%) underwent surgical release at a median of 7.4 months after the injection. According to the Kaplan-Meier analysis, the estimated rate of freedom from symptom recurrence was 70% (95% confidence interval, 63% to 77%) at six months and 45% (95% confidence interval, 36% to 54%) at twelve months and the estimated rate of freedom from surgical release was 95% (95% confidence interval, 92% to 98%) at six months and 83% (95% confidence interval, 77% to 89%) at twelve months. Insulin-dependent diabetes mellitus was identified as a strong predictor of symptom recurrence (p < 0.01). Younger age (p < 0.01), involvement of other digits prior to presentation (p < 0.01), and a history of other tendinopathies of the upper extremity (p = 0.02) were all independent predictors of a surgical release. The duration and severity of symptoms were not predictive of poor outcomes following injections.
Conclusions: At one year following injection, 56% of the digits had a recurrence of symptoms. Younger age, insulin-dependent diabetes mellitus, involvement of multiple digits, and a history of other tendinopathies of the upper extremity were associated with a higher rate of treatment failure. Symptoms often recurred several months after the injection.
Level of Evidence: Prognostic Level I. See Instructions to Authors for a complete description of levels of evidence.
Trigger finger, or stenosing tenosynovitis, characterized by inflammatory changes in the retinacular sheath and peritendinous tissue rather than in the tenosynovium itself, is a common diagnosis in the adult population. Although activity modification, splinting, and nonsteroidal anti-inflammatory medication are useful in the management of this condition, injections of intermediate-acting cortisone (triamcinolone) or long-acting cortisone (betamethasone) are often utilized as the first line of treatment. Complete resolution of symptoms has been reported in 60% to 90% of patients treated in this fashion1,2.
Studies have demonstrated a diminished response to injection in patients with long-standing symptoms, those with diabetes, and those receiving multiple injections3-5. Despite this evidence, few investigators have analyzed the failure rate of corticosteroid injections in the treatment of stenosing tenosynovitis. In particular, little information is available regarding prognostic indicators of symptom recurrence after injection. The purpose of this study was to identify independent indicators of recurrence one year after treatment with corticosteroid injection and to further define the time course of symptom recurrence following such intervention.
One hundred and thirty consecutive adult patients presenting to the investigators' practices were prospectively enrolled in this study. Inclusion criteria were a first-time diagnosis of trigger finger treated with a corticosteroid injection in the clinic. Patients who had undergone prior treatment of the digit in question or who had inflammatory arthritis were excluded from the study. If a patient presented with symptoms in more than one digit, only the digits that had been treated with an injection at the time of the initial presentation were enrolled in the study. Institutional review board approval was obtained prior to the initiation of the study, and all patients signed an informed consent form prior to participation.
The diagnosis of stenosing tenosynovitis was made on the basis of a history of triggering and the findings of the clinical examination. Tenderness over the A1 pulley, pain along the flexor tendon, pain with passive stretch, and locking were all considered indicative of active tenosynovitis.
At the time of the initial presentation, the treating physician administered a one-page questionnaire that collected detailed demographic data (age, sex, hand dominance, body mass index, digits involved at the time of presentation, and involvement of other digits) and addressed the duration and severity of symptoms, including the presence of a locked digit (defined as a clasped digit that the patient was unable to extend without manual assistance from the examiner). Patients were also asked whether they had had other symptomatic digits in the past and/or whether they had them at the time of presentation. Comorbidities (osteoarthritis involving the hand and elsewhere, non-insulin-dependent diabetes, insulin-dependent diabetes, hypothyroidism, hyperthyroidism, coronary artery disease, and peripheral vascular disease) and other hand disorders (carpal tunnel syndrome, Dupuytren disease, and inflammatory tendinopathies of the upper extremity) were also noted. The diagnosis of comorbid conditions was typically made by the patients' primary care physicians and documented in their medical records. The presence of other hand disorders was confirmed by the treating surgeon.
All patients were initially treated with a single palmar injection of 1 mL of 0.5% plain lidocaine mixed with 1 mL of 40 mg/mL triamcinolone (a long-acting steroid) following sterile preparation with an alcohol wipe. Injections were performed with a 25-gauge, 1.5-in (3.8-cm) needle inserted at a 30° angle over the palmar aspect of the metacarpal head. The needle was advanced through the flexor tendon and into bone with the bevel held parallel to the tendon fibers. The needle was then withdrawn a few millimeters, and the preparation was injected into the subtendinous space. Palpation at the tip of the digit confirmed placement in the flexor tendon sheath. Following injection, a Band-Aid was applied, and patients were instructed to resume normal activity. No adjuvant splinting or occupational therapy was employed.
Patients were initially followed at six weeks after the injection to determine whether the injection had been successful. Treatment success or failure was defined for each individually treated digit. Patients in whom the symptoms had decreased but not resolved, which was defined as >50% improvement on a visual analog scale, were offered a second cortisone injection. The visual analog scale consisted of a straight line with the left representing no pain and the right representing the worst possible pain. Patients were asked to mark the line where they perceived their level of pain to be. Patients with <50% improvement and those who refused a second injection were referred for surgical release of the A1 pulley. If a patient had received injections in multiple digits that were enrolled at the initial presentation, each digit was individually subjected to this algorithm. Patients who were symptom-free were instructed to return for follow-up if their symptoms recurred.
At one year after the initial treatment, telephone interviews were conducted with all patients to determine the rate of symptom recurrence. Symptom recurrence was defined as a report of continued pain at the palmar base of the digit, pain with flexion, pain with passive extension, clicking or locking, or a combination of those symptoms in each individually treated digit. Digits with a persistent flexion contracture but no pain were also categorized as treatment failures. These physical findings were demonstrated to patients at the time of the initial injection to facilitate the telephone follow-up. Additional information that was recorded included the duration of any symptom-free interval as well as any secondary treatment with repeat injection or surgical release. Patients were also asked whether other digits had become symptomatic since the time of enrollment. The inpatient and outpatient records of all patients were reviewed to validate the above information.
Of the initial 130 patients, eleven were lost to follow-up. The remaining 119 patients form the basis of this report.
Statistical Analysis
Ninety-five percent confidence intervals for a single binary proportion were calculated with use of the Wilson procedure6. To account for censoring (i.e., to account for patients who remained symptom-free throughout the study period), the Kaplan-Meier product-limit method was used to estimate the rates of freedom from (1) recurrence of symptoms, (2) repeat injection, and (3) surgical release7. The median survivorship time was determined, and the Greenwood formula, a common method with which to construct confidence intervals around survivorship curves, was applied for that purpose8. Multivariate modeling was performed with the Cox proportional-hazards regression model (a model used to adjust for confounding variables in the analysis of time-dependent data) in order to adjust for confounding variables while identifying independent predictors of each outcome. Convergence was set on the basis of a p value of 0.10 for entry and 0.05 for retention in the model with use of a backward-selection maximum-likelihood procedure in order to exclude variables with weak associations from the final equations9. Twenty covariates were evaluated with multivariate analysis for each of the three outcomes; twenty was considered an appropriate number to test on the basis of the sample size of patients10. Hazard ratios and 95% confidence intervals were calculated for significant predictors to estimate the monthly risk of each outcome. The log-rank test was used to compare survivorship curves11. P < 0.05 (two-tailed) was considered significant. Power analysis indicated that a sample size of 124 trigger digits would provide 80% power for estimating freedom from recurrence at twelve months with a precision of 10%.
Demographics
One hundred and twenty-four trigger digits (eighty-six digits in women and thirty-eight in men) in 119 patients with an average age of 62.3 years (range, 23.9 to 100.6 years) were included. The average duration of follow-up (and standard deviation) was 12.6 ± 0.3 months (range, 12.1 to 13.2 months). (Two patients who were followed at thirteen months were included in the analysis.) The most commonly involved digits were the thumb (forty-three; 35%), ring finger (thirty-eight; 31%), and long finger (twenty-eight; 23%). Seventy-one (57%) of the cases of triggering involved the dominant hand. Forty-eight (39%) of the digits were symptomatic for less than three months prior to presentation; fifty-five (44%), for three to six months; and twenty-one (17%), for more than six months. Thirty-nine patients reported other symptomatic digits prior to enrollment in the study, twenty-seven had other symptomatic digits at the time of presentation, and fifteen patients had other digits that became symptomatic after the original presentation. The most common associated conditions were osteoarthritis (40%), carpal tunnel syndrome (29%), other tendinopathies of the upper extremity (20%), and non-insulin-dependent diabetes (16%). Less common were peripheral vascular disease (7%), coronary artery disease (6%), insulin-dependent diabetes (5%), hypothyroidism (5%), Dupuytren disease (3%), and hyperthyroidism (1%). These results are summarized in Table I.
Symptom Recurrence
All patients reported initial relief of symptoms after the injection. Symptoms recurred in seventy digits (56%) at a median of 5.6 months (range, 0.5 to 13.1 months) after the injection. One case with a recurrence at 13.1 months was included in the analysis as a failure of treatment at one year following injection. The 95% confidence interval for the recurrence rate at twelve months was 48% to 65%. Recurrence was associated, although not significantly so (p = 0.35), with an increased duration of symptoms; there was a recurrence in 71% of the digits with symptoms for more than six months, 55% of those with symptoms for three to six months, and 54% with symptoms for less than three months. Twenty-nine (23%) of the digits underwent a second injection at a median of 7.0 months (range, 1.5 to 13.1 months) after the initial treatment, and twenty-two (18%) underwent surgical release at a median of 7.4 months (range, 2.6 to 12.3 months). Twenty-three digits with recurrence of symptoms underwent a second injection but did not undergo surgical release. Also, some patients with symptom recurrence declined a second injection or surgical release and chose to continue expectant management.
Fifty-four digits were in patients who had either a history of multiple trigger digits or other symptomatic digits at the time of presentation. Thirteen (24%) of these digits underwent surgical release compared with nine (13%) of the seventy digits in patients with a single trigger digit (p = 0.008).
Duration of Symptom-Free Interval
According to the Kaplan-Meier analysis, the estimated rate of freedom from symptom recurrence was 70% at six months (95% confidence interval, 63% to 77%) and 45% at twelve months (95% confidence interval, 36% to 54%) (Fig. 1). The log-rank test indicated that the area under the survivorship curve was significantly smaller for patients with insulin-dependent diabetes mellitus, revealing a faster rate of recurrence in these patients than in those without insulin-dependent diabetes (median, 2.1 months [95% confidence interval, 0.5 to 3.7 months] compared with 10.6 months [95% confidence interval, 8.2 to 13.1 months]; log-rank = 16.41, p < 0.001). Kaplan-Meier analysis demonstrated an estimated rate of freedom from a repeat injection of 90% (95% confidence interval, 86% to 94%) at six months and 77% (95% confidence interval, 70% to 84%) at twelve months (Fig. 2) and an estimated rate of freedom from surgical release of 95% (95% confidence interval, 92% to 98%) at six months and 83% (95% confidence interval, 77% to 89%) at twelve months (Fig. 3).
Patients Receiving a Second Injection
Twenty-nine digits with recurrence of symptoms received a second injection. Of these, six (21%) eventually required surgical release compared with twenty-two (23%) of ninety-five digits that did not undergo a second injection (p = 0.59). Comparison of the Kaplan-Meier survival rates between digits that had undergone a second injection and those that had not revealed a log-rank value of 0.321 (p = 0.57), suggesting no difference in the time to surgical release between these two groups.
Independent Predictors of Recurrence, Second Injection, and Surgical Release
The Cox regression model identified insulin-dependent diabetes as a strong independent predictor of symptom recurrence (p = 0.003). Younger age (defined as a continuous variable) (p < 0.01) and involvement of multiple digits at the time of presentation (p = 0.04) were both predictors of a repeat corticosteroid injection. Younger age (p = 0.001), prior involvement of other digits (p = 0.002), insulin-dependent diabetes (p = 0.02), and other tendinopathies of the upper extremity (p = 0.02) were independent predictors of a surgical release. With the numbers studied, there was no significant relationship between surgical release and carpal tunnel syndrome (p = 0.12), non-insulin-dependent diabetes (p = 0.06), hypothyroidism (p = 0.54), or hyperthyroidism (p = 0.86). Patients who underwent surgical release were an average of eight years younger than patients who did not (p = 0.009). We could not identify a relationship between a locked digit at the time of presentation and symptom recurrence (p = 0.62) or between the digit involved (including the thumb) (p = 0.46) and symptom recurrence. We also could not identify a relationship between a repeat injection and surgical release (p = 0.23). Results of the multivariate analysis, including hazard ratios to estimate the monthly risk, are summarized in Table II.
Since their introduction in the 1950s12, controversy has surrounded the use of local corticosteroids for the treatment of musculoskeletal conditions. Several studies have demonstrated that most primary trigger digits can be treated successfully with cortisone injections with few complications. Kolind-Sorensen reported resolution of symptoms in 67% of 106 digits so treated13, and subsequent studies demonstrated successful outcomes in 50% to 90% of patients1-5,14,15. In our prospective series of 124 digits, we found that although all patients reported initial symptom relief after the injection, only 44% of the digits continued to be symptom-free one year later. This recurrence rate is similar to that previously published in the literature15. In addition, 23% of the digits in our series underwent a repeat injection, and 18% had a surgical release.
Although the rate of symptom recurrence in trigger digits has previously been established15, our literature search did not reveal any studies in which the rate of symptom recurrence was specifically investigated in patients treated with corticosteroid injections. The Kaplan-Meier analysis performed in our study predicted that 70% of the digits (95% confidence interval, 63% to 77%) would be symptom-free at six months after the injection, whereas only 45% (95% confidence interval, 36% to 54%) would remain symptom-free at twelve months after the injection. Ten percent of the digits that underwent a repeat injection did so within the first six months, and the remainder did so between six and twelve months. Similarly, only 5% of the releases were performed in the first six months after the injection. These findings demonstrate that a substantial number of recurrences develop after the initial six-month period following injection and patients should be advised accordingly.
Authors have reported conflicting evidence regarding factors associated with poor outcomes following injection for trigger finger. Many of these investigators have tried to establish a relationship between patient characteristics and symptom recurrence by comparing proportions. Our study differs in that we attempted to examine associations through a multivariate regression model, thereby allowing more definitive identification of independent predictors of symptom recurrence while simultaneously controlling for confounding variables.
In a prospective series, Rhoades et al. found that involvement of multiple digits and a symptom duration of more than four months were associated with unsatisfactory results3. Newport et al. similarly reported that single trigger digits with symptoms for less than six months had more favorable outcomes16. Other authors have found no association between the duration of symptoms and the effectiveness of an injection17. Although recurrence was more common in our patients with longer durations of symptoms, our analysis did not identify an independent relationship between the duration of symptoms and recurrence after corticosteroid injection. We also found that more severe symptoms at the time of presentation (in the form of a locked digit) did not predict a poor outcome following injection.
Involvement of multiple digits was identified as a predictor of both a repeat injection and surgical release. In addition, younger patient age was found to predict both a subsequent injection and surgical release. Possible explanations for these findings include more severe symptoms at presentation or the patients' reluctance to live with continued symptoms in the younger and more active age group.
Several studies18,19 have demonstrated an association between trigger finger and other medical conditions, including carpal tunnel syndrome, de Quervain tenosynovitis, osteoarthritis, hypothyroidism, and diabetes. Of these relationships, the one between trigger finger and diabetes has been the most extensively studied4,17. Griggs et al. described a recurrence rate of 50% in diabetic patients, with patients who had insulin-dependent diabetes having a significantly greater chance of requiring surgical release than those with non-insulin-dependent diabetes mellitus (p = 0.02)4. Benson and Ptaszek also found a higher rate of recurrence in patients with insulin-dependent diabetes when compared with the general population17. Our Cox regression model showed that non-insulin-dependent diabetes was not a predictor of symptom recurrence, repeat injection, or surgical release. In contrast, insulin-dependent diabetes was a strong independent predictor of both symptom recurrence and surgical release. All six patients with insulin-dependent diabetes in this study had a recurrence of symptoms and underwent surgical release of the A1 pulley. In addition, symptoms recurred sooner in patients with insulin-dependent diabetes (at an average of 2.1 months compared with 10.6 months). The presence of other tendinopathies of the upper extremity also predicted a future surgical release. The reason for the poor response to injection among patients with insulin-dependent diabetes mellitus and those with other tendinopathies is unknown. Perhaps these patients have a more diffuse type of disease or an abnormal response at the cellular level with deposition and proliferation of fibrous tissue in the tendon sheath20.
This study had several limitations. We arbitrarily chose a 40 mg/mL preparation of triamcinolone for injection. A review of the literature, however, reveals that no particular preparation of steroid has been found to be more effective and surgeons use a variety of steroids in different concentrations1-5. Some surgeons prefer to use lower steroid concentrations and to perform multiple injections. Our chosen preparation, containing 1 mL of 0.5% lidocaine and 1 mL of 40 mg/mL of triamcinolone, has previously been reported to achieve good results with minimal side effects15. It is possible that multiple injections at lower concentrations would result in lower rates of recurrence and/or surgical release, although this has yet to be demonstrated in the literature.
Another possible limitation of our study is that patients were contacted by telephone at one year after the injection to determine symptom recurrence. A physical examination by the treating physician could have been useful to validate the patients' stated symptoms. Nevertheless, we found an excellent concordance between patient-reported recurrence of symptoms and their accounts of additional treatment as well as the documentation in their medical records. We thus believe that a telephone interview was sufficient to determine if and when symptoms recurred.
The small number of patients with insulin-dependent diabetes raises the possibility of a type-I (false-positive) error. Multivariate analysis, however, is geared toward identifying variables associated with outcomes despite certain uncommon patient characteristics.
Finally, the number of predictor variables tested by multivariate analysis is of importance as it affects the validity of the logistic model. This can become a problem when the ratio of the number of variables analyzed to the number of events becomes large. When this occurs, there may be concerns about the precision or bias of the regression coefficients and tests of significance, which may produce a multivariate model with misleading associations. Thus, caution needs to be exercised when interpreting models that are based on a low number of events per variable studied. In our study, we tried to reduce the number of independent variables by excluding redundant variables but keeping variables that we thought were of potential importance as predictors or confounders. The usual rule of thumb, based on Monte Carlo simulation, suggests that ten events per variable analyzed are appropriate21. However, a smaller number of variables, a larger sample size, or a larger number of events would have ensured that our model would be more easily validated in future studies.
While the efficacy of cortisone injections has already been established by good evidence-based papers, we believe that this study provides new information in that it identifies independent predictors of recurrence and determines the time course of recurrence. To our knowledge, neither of these issues has been addressed previously. Although analyses of proportions had established associations between diabetes and trigger digits, and between carpal tunnel syndrome and trigger digits, none had been definitive. We used multivariate regression to control for confounding variables and therefore were able to identify predictors of failure of treatment for trigger digits with a much greater degree of certainty. We thus believe that this study provides useful information for patients (i.e., what preexisting factors can affect success rates and when to expect symptom recurrence) and orthopaedic surgeons treating this common condition. Although we believe that corticosteroid injections should continue to be used as first-line agents in the management of trigger digits, younger patients and those with insulin-dependent diabetes mellitus have a higher treatment failure rate than the average patient population. In addition, patients should be informed that symptoms often recur several months after the injection. 
Note: The authors acknowledge Gary Newman, PA, for his contribution to this manuscript.
Freiberg A, Mulholland RS, Levine R. Nonoperative treatment of trigger fingers and thumbs. J Hand Surg [Am].1989;14:553-8.14553
1989
[PubMed][CrossRef]
Anderson B, Kaye S. Treatment of flexor tenosynovitis of the hand (‘trigger finger’) with corticosteroids. A prospective study of the response to local injection. Arch Intern Med.1991;151:153-6.151153
1991
[CrossRef]
Rhoades CE, Gelberman RH, Manjarris JF. Stenosing tenosynovitis of the fingers and thumb. Results of a prospective trial of steroid injection and splinting. Clin Orthop Relat Res.1984;190:236-8.190236
1984
Griggs SM, Weiss AP, Lane LB, Schwenker C, Akelman E, Sachar K. Treatment of trigger finger in patients with diabetes mellitus. J Hand Surg [Am].1995;20:787-9.20787
1995
[CrossRef]
Buch-Jaeger N, Foucher G, Ehrler S, Sammut D. The results of conservative management of trigger finger. A series of 169 patients. Ann Chir Main Memb Super.1992;11:189-93.11189
1992
[CrossRef]
Newcombe RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med.1998;17:857-72.17857
1998
[CrossRef]
Kocher MS, Zurakowski D. Clinical epidemiology and biostatistics: a primer for orthopaedic surgeons. J Bone Joint Surg Am.2004;86:607-20.86607
2004
Katz MH. Multivariable analysis. A practical guide for clinicians. 2nd ed. Cambridge, England: Cambridge University Press; 2006. Performing the analysis; p 96-116.
2006
Mickey RM, Greenland S. The impact of confounder selection criteria on effect estimation. Am J Epidemiol.1989;129:125-37. Erratum in: Am J Epidemiol. 1989;130:1066.129125
1989
Kalbfleisch JD, Prentice RL. The statistical analysis of failure time data. Hoboken, NJ: John Wiley & Sons; 1980. p 10-23.
1980
Machin D, Cheung YB, Parmar MKB. Survival analysis: a practical approach. 2nd ed. West Sussex, England: John Wiley & Sons; 2006. p 40-8.
2006
Hollander JL, Brown EM Jr, Jessar RA, Brown CY. Hydrocortisone and cortisone injected into arthritic joints; comparative effects of and use of hydrocortisone as a local antiarthritic agent. J Am Med Assoc.1951;147:1629-35.1471629
1951
Kolind-Sorensen V. Treatment of trigger fingers. Acta Orthop Scand.1970;41:428-32.41428
1970
[CrossRef]
Marks MR, Gunther SF. Efficacy of cortisone injection in treatment of trigger fingers and thumbs. J Hand Surg [Am].1989;14:722-7.14722
1989
[CrossRef]
Fleisch SB, Spindler KP, Lee DH. Corticosteroid injections in the treatment of trigger finger: a level I and II systematic review. J Am Acad Orthop Surg.2007;15:166-71.15166
2007
Newport ML, Lane LB, Stuchin SA. Treatment of trigger finger by steroid injection. J Hand Surg [Am].1990;15:748-50.15748
1990
[CrossRef]
Benson LS, Ptaszek AJ. Injection versus surgery in the treatment of trigger finger. J Hand Surg [Am].1997;22:138-44.22138
1997
[CrossRef]
Ryzewicz M, Wolf JM. Trigger digits: principles, management, and complications. J Hand Surg [Am].2006;31:135-46.31135
2006
[CrossRef]
Lipscomb PR. Tenosynovitis of the hand and the wrist: carpal tunnel syndrome, de Quervain's disease, trigger digit. Clin Orthop Relat Res.1959;13:164-81.13164
1959
Stahl S, Kanter Y, Karnielli E. Outcome of trigger finger treatment in diabetes. J Diabetes Complications.1997;11:287-90.11287
1997
[CrossRef]
Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol.1996;49:1373-9.491373
1996
[CrossRef]