Long-Term Outcomes Following Manipulation Under Anaesthetic for Patients with Primary and Secondary Frozen Shoulder

Abstract

Background

Frozen Shoulder (FS) is a common, debilitating condition for which manipulation under anaesthetic (MUA) is a non-invasive and effective treatment option. Current literature evaluates short to medium-term outcomes, but there is a paucity of long-term (>10 years) studies. Knowledge of long-term outcomes is also needed to evaluate whether FS or its treatment pre-disposes to other shoulder pathology in the long-term.

Methods

A retrospective analysis of 398 shoulders undergoing MUA for FS between Jan 1999 and Jan 2010; 240 complete datasets were obtained. Outcomes were Oxford Shoulder Score (OSS), recurrence and development of other shoulder pathology (arthritis or rotator cuff tear).

Results

At long-term follow-up (mean 13.2 years), 71.3% had no symptoms (OSS 48), 16.6% had minor symptoms (OSS 42–47) and 12.1% had significant symptoms (OSS < 42). There were 4/240 (1.7%) self-reported recurrences > 5 years after initial MUA and 2/240 (0.8%) repeat MUAs. In the long-term 6.7% developed rotator cuff pathology and 3.8% shoulder OA.

Discussion

This study suggests that long-term outcome after MUA for FS is favourable. Late recurrence of FS is uncommon and the development of OA or rotator cuff pathology is no greater than that of the general population.

Introduction

Frozen Shoulder (FS) is a common and debilitation condition characterised by pain and a reduction in both active and passive movement of the glenohumeral joint. FS classically has three distinct phases: painful, stiffening and resolution. FS is believed to be caused by inflammation and contraction of myofibroblasts in the shoulder capsule resulting in tightening and a reduced range of movement (ROM).^1–3 It is unknown why this effect resolves. FS can be categorised into primary (spontaneously occurring) or secondary (to a known precipitating factor, such as soft tissue injury to the shoulder, upper limb fracture or cardiothoracic/breast surgery). FS was first described by Codman in 1934 ⁴ and later refined by Lundberg in 1969 ⁵ who first proposed a distinction between primary and secondary causes.

The incidence of FS is estimated at 2–5%, but most notably higher (10–20%) in patients with diabetes mellitus, although other endocrine disorder such as thyroid disease are also known to be associated with higher incidences.^5,6

Whilst it is commonly believed that FS will eventually go onto resolution, this assumption has been questioned. ⁶ The time to recovery is highly variable and often incomplete.^1,7–9 FS has been observed to last for around 18 months in non-diabetic patients, but can occur for longer (3–5 years) in diabetic patients^9–10 with variable recurrence rates.^11,12

In 1872 Duplay first described the use of manipulation under anaesthetic (MUA) to treat a condition then known as “péri-arthrite scapula-humérale” which has since gone on to be a first-line treatment after non-surgical management (analgesia, corticosteroid injections, exercise programmes and physiotherapy).^13–16 MUA for FS provides relief of symptoms and improvement in ROM.^9,11,17–21 A recent multicentre trial (UKFROST) concluded that it was the most cost-effective treatment in comparison with physiotherapy or arthroscopic release when assessed at twelve months from treatment. ²² There is a paucity of literature on the long-term (> 10 years) outcomes, in particular the late recurrence rate or the development of other shoulder pathology such as osteoarthritis or rotator cuff disease. Higher recurrence rates have been reported in comparison with alternative treatments for FS, namely hydrodilatation ²³ or arthroscopic release.^24,25

The aim of this study was to determine the long-term outcomes (> 10 years) for FS following MUA using a patient reported functional outcome score (OSS), rates of recurrence and the development of other shoulder pathology at long-term follow-up (mean 13.2 years) as our outcome measures.

Materials and methods

This study had ethical approval from the Medical Advisory Committee of our institution. A retrospective analysis of consecutive patients undergoing MUA plus steroid injection for FS between January 1999 and January 2010 under the care of a single surgeon was performed. A diagnosis of FS was made using the following criteria:

1. a history of increasing shoulder pain and stiffness

2. loss of passive and active movement of the shoulder in one or more of the anatomical planes

3. normal plain radiography of the glenohumeral joint

Further imaging (USS/ MRI) was requested if there was history suggestive of other pathology and patients excluded if an alternative diagnosis was made. Patients were classified as primary (idiopathic) or secondary (to a known precipitating factor) FS. Surgery was only considered once patients had trialled non-surgical methods in the community.

All participants underwent an MUA of the shoulder as soon as was practical after the diagnosis was made. ²¹ MUA was performed under general anaesthetic in the supine position, baseline glenohumeral ROM was assessed whilst ensuring scapula fixation. The glenohumeral joint was then manipulated through all anatomical planes of motion. A glenohumeral intra-articular injection of 10 ml 0.5% bupivacaine with 80 mg depomedrone (Pharmacia Ltd, Sandwich, UK) was then given via an anterior approach. Post-operatively patients received a course of physiotherapy until no further improvement in ROM was made. Patients were excluded from the study if they were found to have a normal ROM at the time of procedure.

Retrospective data collected included age, sex, primary or secondary FS, ROM and OSS ²⁶ recorded prior to manipulation, and OSS at the first post-MUA follow up (mean 4 weeks). At a minimum of 10 years after initial MUA a postal questionnaire (see Appendix 1) was sent to all eligible patients. Data collected at this long-term follow-up on the affected shoulder included:

• Return of FS symptoms (“self-reported recurrence”) and the timing

• Subsequent diagnoses of rotator cuff disease or shoulder OA

• Any further imaging, operations or procedures to their shoulder.

• Current OSS

Participants were also asked about current health conditions. If the patient had any subsequent diagnoses, imaging or procedures the dates, locations, condition and specific treatment delivered was confirmed where possible.

When analysing OSS we used the validated Minimal Clinically Important Difference (MCID) of six points^27,28 to set 42/48 as the threshold at which a patient was deemed not to have clinically significant symptoms. Patients scoring < 42 at long-term follow-up were considered to be clinically symptomatic, 42–47 were considered to have minor symptoms and those scoring 48 were considered asymptomatic. Long-term ROM data was not feasible to collect in this study so self-reported return of FS symptoms and repeat MUA were used as surrogate markers of recurrence. Timing of recurrence after initial MUA was categorised as: early (< 1 year), medium (1–5 years) and late (> 5 years).

Statistical methods

The distribution of the OSS data for pre-MUA, four-weeks post MUA and at long-term follow-up was checked using the Kolmogorov-Smirnov test for normality.^29,30 The four-week post and the long-term follow up data were found to be non-parametric. The Friedman Test was used as a validated method for non-parametric linked data with more than two datasets. The Kruskal-Wallis Test was used for non-linked samples with more than two datasets and the Mann-Whitney U test used when two datasets were compared.^29,30

Results

Population

There were 398 shoulders in the initial cohort with a final dataset of 240 shoulders (60.3%), exclusions are shown in Figure 1.

Figure 1.

Participant flowchart.

Of those that responded, average age at first MUA was 52 years (range 31–80) with a mean interval to follow up of 13.2 years (range 10–20 years). The demographic data of the patient shoulders are shown in Table 1 and the causes of secondary FS in Table 2.

Table 1.

Demographic data of participant shoulders.

Demographic	Participants
Type 1 diabetes mellitus	10
Type 2 diabetes mellitus	23
Primary	211
Secondary	29

Table 2.

Causes of secondary frozen shoulder.

Secondary Frozen Shoulder Cause	Participants
Soft tissue injury to the shoulder (excluding cuff tear)	15
Upper limb fracture	10
Recent breast surgery	2
Recent upper limb surgery	2

Outcomes

OSS

The mean OSS at long-term follow-up post MUA was 45.7 (Interquartile Range (IQR): 47–48). The distributions of OSS pre MUA, four weeks post MUA and at long-term follow-up are shown in Figure 2. Of our cohort, 211/240 (87.9%) reported a long-term OSS > 42 and 171/240 (71.2%) an OSS of 48. Twenty-nine shoulders (12.1%) had a long-term OSS < 42 and were considered symptomatic. Mean OSS at long-term follow-up post MUA was 46.0 for primary FS and 43.7 for secondary (p = 0.0465). Mean long-term OSS was 45.9 for non-diabetic patients, 42.8 for T1DM and 45.5 for T2DM (p = 0.0345). Excluding shoulders who self-reported symptom recurrence, underwent a repeat MUA, suffered rotator cuff pathology or shoulder OA the mean long-term OSS was 47.0.

Figure 2.

Box & whisker plots (IQ range) of OSS for pre, 4-weeks post and long-term follow-up (mean 13.2 years) post initial-MUA, (p < 0.00001).

Recurrence: self-reported & repeat MUA

We recorded both subjective self-reported recurrence of FS symptoms and repeat MUA rates. Table 3/Figure 3 shows the proportion of participants reporting a return of FS symptoms and the timing in which they occurred. Table 4/Figure 4 demonstrates the proportion of participants who underwent a repeat MUA. Overall, 75/240 (31.3%) participants reported a recurrence of their FS symptoms, with 4/240 (1.7%) occurring the long-term (> 5 years post initial MUA). In comparison, a total of 46/240 (19.2%) underwent a repeat MUA for recurrence of FS, with 2/240 (0.8%) performed > 5 years post initial MUA.

Table 3.

Percentage of shoulders self-reporting recurrence of their FS symptoms.

	Total	Primary	Secondary	Non-Diabetic	T1DM	T2DM
None	68.8%	69.7%	62.1%	69.1%	40.0%	78.3%
< 1 year	25.4%	25.1%	27.6%	25.6%	30.0%	21.7%
1-5 years	4.2%	3.3%	10.3%	3.4%	30.0%	0.0%
≥ 5 years	1.7%	1.9%	0.0%	1.9%	0.0%	0.0%
Total No.	240	211	29	207	10	23

Figure 3.

Percentage (total number) of shoulders self-reporting recurrence of their FS symptoms.

Table 4.

Percentage of shoulders undergoing repeat MUA for FS.

	Total	Primary	Secondary	Non-Diabetic	T1DM	T2DM
None	80.8%	82.0%	72.4%	81.2%	70.0%	82.6%
< 1 year	16.3%	14.7%	27.6%	15.5%	30.0%	17.4%
1-5 years	2.1%	2.4%	0.0%	2.4%	0.0%	0.0%
≥ 5 years	0.8%	0.9%	0.0%	1.0%	0.0%	0.0%
Total No.	240	211	29	207	10	23

Figure 4.

Percentage (total number) of shoulders undergoing repeat MUA for FS.

Mean long-term OSS was 43.8 for shoulders that experienced an early self-reported recurrence of symptoms, 42.8 for the medium-term group and 40.5 for the late recurrence group. In shoulders that did not experience any recurrence the long-term OSS was 46.8. Of the seventy-five shoulders that experienced a recurrence, nineteen were still symptomatic at long-term follow-up with an OSS < 42 (15 early, 3 medium, 1 late).

Development of other shoulder pathology

Sixteen patients (6.7%) reported a proven diagnosis of an ipsilateral rotator cuff tear (15 primary, 1 secondary; 14 non-diabetic, 0 T1DM, 2 T2DM), eight of which had been confirmed on USS, MRI or at operative intervention. Of these sixteen, three underwent rotator cuff repair. The mean long-term OSS for shoulders that developed rotator cuff disease on the ipsilateral side was 44.9.

Nine (3.75%) patients reported a subsequent diagnosis of proven shoulder osteoarthritis (OA) (8 primary, 1 secondary; 7 non-diabetic, 0 T1DM, 2 T2DM); four had confirmatory imaging and three underwent total shoulder replacement. Five patients who developed shoulder OA had pre-existing OA of other joints. The mean long-term OSS for shoulders that developed ipsilateral shoulder OA was 38.4. There were no patients who developed symptoms of instability or required surgical stabilisation in this cohort.

When we excluded the twenty-seven shoulders that had developed rotator cuff disease or shoulder OA, the mean OSS at long-term follow-up was 46.1. Of the twenty-nine shoulders that were symptomatic at long-term follow-up (OSS < 42), four had developed rotator cuff disease and six had developed shoulder OA.

Evaluation of symptomatic shoulders (OSS < 42)

Twenty-nine (12.1%) shoulders had a OSS of < 42 at long-term follow-up (23 primary, 6 secondary; 21 non-diabetic, 4 T1DM, 4 T2DM). Ten (4.1%) of these patients reported no recurrence of FS (8 primary, 2 secondary; 6 non-diabetic, 1 T1DM, 3 T2DM). Of these ten symptomatic shoulders that reported no symptom recurrence; two had developed rotator cuff disease (2 primary, 0 secondary; 1 non-diabetic, 0 T2DM, 1 T1DM) and three had developed shoulder OA (2 primary, 1 secondary; 1 non-diabetic, 0 T1DM, 2 T2DM). Excluding shoulders that suffered from a recurrence, rotator cuff pathology or shoulder OA, 6/240 (2.5%) had an OSS < 42 at long-term follow-up (5 primary, 1 secondary; 4 non-diabetic, 1 T1DM, 1 T2DM).

Discussion

This study demonstrates that the majority of patients have a favourable outcome after MUA with minimal ongoing symptoms in the long-term. Within this cohort secondary FS was associated with a lower OSS in the long-term and higher rates of early/medium-term recurrence but not late recurrence. Despite the more severe disease classically associated with diabetes, there appeared to be no consistent difference in their long-term OSS or recurrence rates in our cohort.

The data collection has been robust, with only a small number of patients excluded due to incomplete data. Although our loss to follow-up was 40%, this is classed as good for an epidemiology study at > 10 years. ³¹ Retrospective study design, questionnaire format, lack of objective ROM data to confirm true recurrence and an uncertainty about the exact diagnosis in six (2.5%) patients who were symptomatic (OSS < 42) at long-term follow-up without explantion are limitations of the study.

We know of no previous publications looking at this length of follow up with as large a patient cohort. One study by Hand et al. ⁸ followed 223 patients for an average of 4.4 years (2–20 years) but gives no further breakdown of follow-up duration. They found 59% had normal or near normal function, 38.5% had ongoing mild pain and 2.5% had significant pain or functional loss. A further small study followed 19 shoulders with a follow-up range of 8.1–20.6 years. Sixteen shoulders had minimal or no symptoms and three had moderate or severe pain. ³² Despite the limitations of these studies, both present similar findings that long-term functional outcome for FS post MUA is favourable.

There is currently limited data looking at long-term recurrence rates post-MUA. Hand et al. reported no recurrences during their follow-up of 2-20 years, but they did not define their criteria for “recurrence” and provide no explanation for the ongoing symptoms in 41%, 6% of whom classified this as severe. ⁸ A larger cohort of 792 shoulders recorded short-term recurrence in 18% rising to 38% in T1DM shoulders.¹² Recurrence is this study was defined as requiring repeat MUA so is likely to underestimate true recurrence. These studies do not elaborate on the timing of recurrence.

Reported recurrence rates after alternative treatment include a 7.6% reintervention rate after hydrodilatation in a series of 2432 cases followed for an average of 5.4 years ²³ and an 11% recurrence rate following arthroscopic release in 54 shoulders followed for twelve months. ²⁴ In the hydrodilatation study, the rate of recurrence in diabetic patients was found to be 14%. Although the mean time to repeat hydrodilatation was nine months, there is no data on the length of time before recurrence occurred. ²³

True recurrence of FS can only be confirmed by recording a reduced ROM at the time of symptom return. In our study we used two estimates of recurrence (self-reported return of symptoms and repeat MUA) but were unable to collect long-term ROM data in order to confirm true recurrence. Patients may not present nor have access to a clinic at this point, so it is common practice for recurrence to be defined using alternative criteria: from self-reported return of symptoms to requiring further treatment (including repeat MUA). ³³ The variable and surrogate definitions of recurrence may account for variation in reported recurrence rates. Recurrence measures are further confounded by a number of factors, including the expectation of patients, tolerance of minor symptoms, access back to the clinic following discharge and the threshold of the treating physician to offer further treatment. In our results the recurrence estimated by self-reporting is unsurprisingly higher than the percentage requiring repeat MUA, particularly in the long-term. The true pathological recurrence of FS is likely to lie in between our two surrogate measures.

It is as yet unknown why the intervention of MUA, arthroscopic release or hydrodilatation reverses the triad of pain, inflammation and capsular tightness in most patients but in some, a short-term recurrence occurs. It has been proposed that this may be due to a persistence of the inflammation and contracture of the capsule once the immediate effect of the MUA and cortisone injection has worn off. Our data shows that whilst short-term recurrence does seem to occur, recurrence in the long-term is rare and does not appear to occur more frequently in secondary FS nor diabetic patients.

The incidence of symptomatic rotator cuff disease is known to be age dependent, rising from under 10% of the population aged under 40 years to over 60% at age 80 years. ³⁴ Given that our study looked at patients with a mean age of 52 years at the time of MUA and followed them for a mean of 13.2 years, our patients now have a mean age of over 65 years so we would expect a rate of symptomatic rotator cuff disease to be a minimum of 15%. ³⁵ Our reported rate of rotator cuff disease was low at 6.7% but even if we assume that all of our 12.2% with significant pain (OSS < 42) had undiagnosed rotator cuff disease, this would still be within the expected prevalence for an age-matched population. This implies MUA for FS does not predispose to rotator cuff pathology. A similar picture is seen for the development of shoulder OA. The UK incidence of OA of the shoulder is estimated at 2–26% rising with age,^36,37 our prevalence of 3.8% suggests that this group is not more at risk of developing glenohumeral OA compared to the general population.

Conclusion

In this study of the long-term outcomes after MUA for FS (mean 13.2 years) the majority of patients (71.3%) were asymptomatic (OSS = 48) and 16.6% had only minimal symptoms (OSS 42–47). Clinically symptomatic shoulders (OSS < 42) remained in the long-term in 12.1%, falling to 2.5% when excluding those suffering a recurrence, rotator cuff pathology or shoulder OA. Self-reported recurrence > 1 year was low at 5.8% and only occurred > 5 years in 1.7%. Repeat MUA was performed at > 1 year in 2.9% and > 5 years in 0.8%. All recurrences occurring at > 5 years (self-reported and repeat MUA) were in primary, non-diabetic shoulders. MUA for FS does not appear to predispose the patient to other shoulder pathology in the long-term, with rotator cuff disease reported in 6.7% and shoulder arthritis in 3.8%. At an average age of 65, both of these percentages are comparable to that of an age-matched population.