Exploratory anchor-derived six-month improvement thresholds for sleep quality and pain after reverse total shoulder arthroplasty

Abstract

Background

Sleep disturbance is a common patient-reported problem in patients undergoing reverse total shoulder arthroplasty (rTSA). However, clinically interpretable improvement thresholds for sleep quality after rTSA remain unclear. This study aimed to derive exploratory anchor-based 6-month improvement thresholds for sleep quality, assessed using the Pittsburgh Sleep Quality Index (PSQI), and pain, assessed using the visual analog scale (VAS), after rTSA.

Methods

Sixty-one patients undergoing primary rTSA with complete pre-operative and 6-month post-operative PSQI, VAS, and global rating of change (GRC) data were retrospectively analyzed. Exploratory receiver-operating characteristic (ROC) curve analyses were performed using the GRC anchor, and Youden-derived thresholds were calculated. Anchor validity was assessed using Spearman correlation analysis between ordinal GRC responses and changes in PSQI and VAS.

Results

rTSA was associated with significant 6-month improvements in both PSQI and VAS scores (both P < .001). Based on the GRC anchor, 54 patients were classified as improved and 7 as not improved, indicating marked imbalance between groups. The GRC anchor demonstrated a weak, nonsignificant association with PSQI change and a moderate association with VAS change. Exploratory ROC analysis yielded approximate improvement thresholds of 7 points for PSQI and 4 points for VAS pain.

Conclusion

rTSA was associated with substantial 6-month improvements in sleep quality and pain. In this small retrospective cohort, exploratory anchor-based ROC analyses generated preliminary improvement thresholds of approximately 7 points for PSQI and 4 points for VAS pain. However, the weak PSQI anchor correlation, severe imbalance in anchor-defined groups, and modest ROC discrimination indicate that these estimates should be considered hypothesis-generating rather than definitive Minimal Clinically Important Difference values.

Sleep disturbance is a highly prevalent and clinically significant but frequently under-recognized problem in patients with degenerative shoulder disorders, with pain-related nocturnal symptoms representing a major determinant of impaired quality of life.^14,19,24 Pre-operative sleep impairment has been shown to correlate strongly with pain severity, functional limitation, and psychosocial burden in patients with rotator cuff pathology and glenohumeral arthritis.^22,23 Importantly, disturbed sleep is not merely a secondary symptom but an independent contributor to worse post-operative pain perception and delayed functional recovery following shoulder surgery.^13,23

Multiple studies have demonstrated that surgical treatment of shoulder pathology leads to significant post-operative improvements in sleep quality. Arthroscopic rotator cuff repair has been shown to produce marked reductions in Pittsburgh Sleep Quality Index (PSQI) scores within the first post-operative year, with sustained improvements over midterm follow-up.^11,13 Similar beneficial effects on sleep disturbance have been reported following anatomic and reverse total shoulder arthroplasty (rTSA), suggesting that pain relief and restoration of shoulder function may contribute to improved sleep quality.^20,22 More recently, O'Donnell et al (2025) demonstrated that patients undergoing rTSA experience substantial improvements in PSQI-derived sleep quality, although recovery trajectories may differ from those observed after rotator cuff repair or anatomic arthroplasty.¹⁶

Despite this growing body of evidence, the clinical interpretability of post-operative sleep improvement remains limited, as most studies report statistical changes without defining thresholds that represent a Minimal Clinically Important Difference (MCID) from the patient's perspective.^10,11,13 Although MCID and patient acceptable symptom state (PASS) values for PSQI have been proposed in rotator cuff repair populations, these thresholds have not been clearly validated for patients undergoing rTSA, whose pathology severity, biomechanical alterations, and pain mechanisms differ substantially.^8,10,11,13 Furthermore, while several studies suggest that baseline sleep quality may influence post-operative symptom recovery, data evaluating factors associated with patient-perceived sleep improvement after rTSA remain limited.^8,13,23

Despite the well-documented improvements in pain and function following rTSA, clinically interpretable thresholds for patient-perceived improvement in sleep quality remain insufficiently defined in this population. Although MCID and PASS values for PSQI have been proposed in rotator cuff repair cohorts, these values may not be directly generalizable to patients undergoing rTSA because of differences in pathology severity, baseline symptom burden, and recovery trajectories. Therefore, the primary aim of this study was to derive exploratory anchor-based 6-month improvement thresholds for PSQI-derived sleep quality and visual analog scale (VAS) pain after rTSA. The secondary aim was to explore whether baseline sleep quality was associated with anchor-defined improvement. We hypothesized that rTSA would be associated with substantial improvements in sleep quality and pain and that exploratory anchor-based receiver-operating characteristic (ROC) analysis could provide hypothesis-generating improvement thresholds requiring external validation.

Materials and methods

This study was designed as a single-center retrospective observational cohort study conducted at a tertiary academic orthopedic referral hospital. The study protocol was approved by the local institutional review board, and all procedures were performed in accordance with the Declaration of Helsinki. Owing to the retrospective nature of the study, the requirement for written informed consent was waived.

During the study period, all patients undergoing primary rTSA were screened for eligibility. Patients were included if they had complete pre-operative and 6-month post-operative PSQI, VAS, and global rating of change (GRC) data. All otherwise eligible patients during the study period had complete outcome and anchor data; therefore, no patient was excluded because of missing PSQI, VAS, or GRC assessment. The final cohort represented a consecutive institutional series of primary rTSA patients with complete 6-month follow-up data.

All procedures were performed by senior shoulder surgeons using a standardized deltopectoral approach. After subscapularis tenotomy, humeral canal preparation was performed with sequential broaching to determine optimal stem size. Glenoid preparation involved concentric reaming followed by implantation of a baseplate and glenosphere. A cemented humeral stem was used in all cases. Final polyethylene thickness was selected to optimize soft tissue tension and joint stability.

Post-operatively, the operated arm was immobilized in a shoulder sling for 3 weeks. Passive range of motion exercises were initiated during the first post-operative week. Active-assisted range of motion exercises were commenced after the third post-operative week, and progressive strengthening exercises were introduced after 6 weeks under physiotherapist supervision.

Demographic and clinical data were retrospectively extracted from electronic medical records. The following variables were systematically reviewed:

• •Age, sex, height, weight, and body mass index
• •Operated side
• •Primary surgical indication
• •Smoking status
• •Presence of medical comorbidities (hypertension, diabetes mellitus, coronary artery disease, chronic pulmonary disease)
• •Pre-operative and 6-month post-operative PSQI and VAS scores
• •Patient-reported GRC anchor responses

All data were independently collected by 2 investigators, and discrepancies were resolved by consensus.

Pre-operative diagnosis was recorded and reported descriptively. Because the cohort was predominantly composed of patients with cuff tear arthropathy and the number of patients with other indications was limited, diagnosis-stratified ROC analyses were not performed. Such subgroup analyses would have been statistically unstable and could have generated misleading diagnosis-specific improvement thresholds. Accordingly, the reported thresholds should be interpreted as preliminary exploratory estimates for a predominantly cuff tear arthropathy rTSA population and require validation in larger diagnosis-stratified cohorts.

As outcome measures, sleep quality was assessed using the PSQI, a validated 19-item questionnaire generating a global score ranging from 0 to 21. Higher scores indicate poorer sleep quality. Pain intensity was measured using a 10-cm VAS for overall shoulder pain during the preceding week, where 0 indicated no pain and 10 indicated the worst imaginable pain.

Clinical improvement was assessed using a GRC anchor. At the 6-month post-operative visit, patients completed a shoulder-specific GRC question: “Compared with your condition before surgery, how would you rate the overall change in your shoulder-related symptoms, including pain and sleep disturbance?” Responses were recorded on a 5-point ordinal scale: “much worse,” “slightly worse,” “unchanged,” “slightly improved,” and “much improved.” Patients who answered “slightly improved” or “much improved” were classified as anchor-defined improved, whereas those reporting “unchanged,” “slightly worse,” or “much worse” were classified as not improved. The distribution of responses across all 5 GRC categories was recorded to allow transparent assessment of anchor imbalance.

Statistical analysis

All analyses were performed using IBM SPSS Statistics version 26.0 (IBM Corp., Armonk, NY, USA). Data normality was assessed using the Shapiro–Wilk test and Q–Q plots. Continuous variables were reported as mean ± standard deviation or median (interquartile range), and categorical variables were reported as frequencies and percentages. Pre-operative and 6-month post-operative PSQI and VAS scores were compared using the paired samples t-test or Wilcoxon signed-rank test, as appropriate. Within-patient effect sizes were calculated using Cohen's dz. Patients were classified as improved or not improved according to the GRC anchor. Between-group comparisons of change scores were performed using the independent samples t-test or Mann–Whitney U test, as appropriate. Exploratory anchor-based ROC analyses were performed to derive preliminary improvement thresholds for PSQI and VAS, with cutoff points derived using the Youden index. Discriminative ability was assessed using the area under the curve (AUC). To evaluate anchor validity, associations between ordinal GRC responses and changes in PSQI and VAS were examined using Spearman correlation coefficients. Given the small and imbalanced anchor-defined groups, ROC-derived thresholds were interpreted as hypothesis-generating only. Because only 7 patients were classified as not improved, multivariable logistic regression was not performed because of the high risk of model overfitting. Therefore, predictor analysis was limited to exploratory univariable logistic regression evaluating the association between baseline PSQI and anchor-defined improvement. A 2-tailed P value < .05 was considered statistically significant.

Results

A total of 61 patients undergoing primary rTSA with complete pre-operative and 6-month post-operative assessments were included. The cohort was predominantly female (78.7%), with a mean age of 69.4 ± 5.6 years and a mean body mass index of 30.4 ± 3.2 kg/m². The operated side was right in 72.1%, and the most common indication was cuff tear arthropathy (77.0%). At least one comorbidity was present in 91.8%, and 26.2% were active smokers (Table I).

Table I.

Demographic and clinical characteristics of the patients.

Variable	Mean ± SD/median (IQR) or n (%)
Age (yr)	69.4 ± 5.6/69.0 (65.0-73.0)
Height (m)	1.66 ± 0.08/1.66 (1.60-1.70)
Weight (kg)	83.8 ± 8.4/83.0 (78.0-90.0)
BMI (kg/m2)	30.4 ± 3.2/30.1 (28.7-32.3)
Sex
Female	48 (78.7%)
Male	13 (21.3%)
Operated side
Right	44 (72.1%)
Left	17 (27.9%)
Primary diagnosis
Cuff tear arthropathy	47 (77.0%)
Glenohumeral osteoarthritis	12 (19.7%)
Other	2 (3.3%)
Presence of comorbidities, yes	56 (91.8%)
Current smoker, yes	16 (26.2%)

BMI, body mass index; IQR, interquartile range; SD, standard deviation.

Data are presented as mean ± SD/median (IQR) or n (%).

At 6 months, rTSA was associated with marked improvements in both sleep quality (PSQI) and pain (VAS), with statistically significant pre–post differences (P < .001 for both) (Table II).

Table II.

Comparison of pre-operative and 6-month post-operative PSQI and VAS pain scores.

Outcome measure	Pre-operative		6-mo post-operative		Absolute improvement		P value
	Mean ± SD	Median (IQR)	Mean ± SD	Median (IQR)	Mean ± SD	Median (IQR)
PSQI score	12.30 ± 2.80	13.0 (11.0-14.0)	3.54 ± 1.90	3.0 (2.0-4.0)	8.75 ± 2.84	9.0 (7.0-11.0)	<.001
VAS pain score	8.03 ± 1.03	8.0 (7.0-9.0)	1.92 ± 1.32	2.0 (1.0-2.0)	6.11 ± 1.77	6.0 (5.0-7.0)	<.001

PSQI, Pittsburgh Sleep Quality Index; VAS, visual analog scale; IQR, interquartile range; SD, standard deviation.

Absolute improvement was calculated as pre-operative score minus post-operative score; positive values indicate improvement. The Wilcoxon signed-rank test was used. Bold values indicate statistical significance at P < .05.

PSQI decreased from 12.30 ± 2.80 pre-operatively to 3.54 ± 1.90 at 6 months post-operatively, yielding an absolute improvement of 8.75 ± 2.84 points (median, 9.0 [7.0-11.0]; P < .001). This corresponded to an approximate 71% reduction from baseline PSQI and a very large within-patient standardized change (Cohen's dz = 3.08). VAS pain decreased from 8.03 ± 1.03 pre-operatively to 1.92 ± 1.32 at 6 months post-operatively, yielding an absolute improvement of 6.11 ± 1.77 points (median, 6.0 [5.0-7.0]; P < .001). This corresponded to an approximate 76% reduction from baseline VAS and a very large within-patient standardized change (Cohen's dz = 3.45).

Using the GRC anchor, 54 patients (88.5%) were classified as improved and 7 patients (11.5%) as not improved. The distribution of GRC responses was as follows: much worse, n = 0; slightly worse, n = 0; unchanged, n = 7; slightly improved, n = 10; and much improved, n = 44. Anchor-validation analysis showed that the ordinal GRC anchor demonstrated a weak, nonsignificant association with PSQI change (Spearman's rho = 0.18, P = .164) and a moderate association with VAS change (Spearman's rho = 0.32, P = .012). These findings suggest that the GRC anchor reflected patient-perceived pain improvement more consistently than sleep-quality improvement. Therefore, the PSQI-derived value should be interpreted as an exploratory anchor-derived improvement threshold rather than a definitive MCID. Bootstrap analysis supported similar exploratory thresholds of approximately 7 points for PSQI and 4 points for VAS, with threshold 95% confidence intervals of 6-8 points and 3-5 points, respectively.

Between-group comparisons showed numerically greater VAS pain absolute improvement in the GRC-improved group (6.35 ± 1.47) than in the GRC-not-improved group (4.29 ± 2.81), although this difference did not reach statistical significance (P = .078). PSQI absolute improvement similarly did not differ significantly between groups (P = .326) (Table III).

Table III.

Comparison of absolute improvement scores according to GRC-defined improvement status.

Outcome measure	GRC-improved group (n = 54)		GRC-not-improved group (n = 7)		P value
	Mean ± SD	Median (IQR)	Mean ± SD	Median (IQR)
PSQI absolute improvement	8.96 ± 2.58	9.0 (8.0-11.0)	7.14 ± 4.34	6.0 (4.0-10.5)	.326
VAS pain absolute improvement	6.35 ± 1.47	6.0 (5.0-7.0)	4.29 ± 2.81	6.0 (3.0-6.0)	.078

GRC, global rating of change; PSQI, Pittsburgh Sleep Quality Index; VAS, visual analog scale; IQR, interquartile range; SD, standard deviation.

Absolute improvement was calculated as pre-operative score minus post-operative score; positive values indicate improvement. The Mann–Whitney U test was used.

Exploratory ROC analyses generated preliminary anchor-derived improvement thresholds for both outcomes (Table IV, Fig. 1):

• •PSQI sleep quality: exploratory improvement threshold of approximately 7 points, with AUC = 0.615, sensitivity 79.6%, and specificity 57.1%.
• •VAS pain: exploratory improvement threshold of approximately 4 points, with AUC = 0.702, sensitivity 100.0%, and specificity 42.9%.

Table IV.

Exploratory Youden-derived improvement thresholds for PSQI and VAS pain.

Variable	AUC	Exploratory threshold	Threshold 95% CI	Sensitivity	Specificity
PSQI absolute improvement	0.615	Approximately 7 points	6-8 points	79.6%	57.1%
VAS pain absolute improvement	0.702	Approximately 4 points	3-5 points	100.0%	42.9%

AUC, area under the curve; CI, confidence interval; PSQI, Pittsburgh Sleep Quality Index; VAS, visual analog scale; MCID, Minimal Clinically Important Difference.

Thresholds were derived using the Youden index based on absolute improvement scores and should be interpreted as exploratory and hypothesis-generating rather than definitive MCID values.

Figure 1.

Exploratory anchor-based receiver-operating characteristic (ROC) curve analyses for deriving preliminary improvement thresholds after reverse total shoulder arthroplasty. (Left) ROC curve for Pittsburgh Sleep Quality Index (PSQI) absolute improvement scores, demonstrating an approximate Youden-derived threshold of 7 points (AUC = 0.615). (Right) ROC curve for visual analog scale (VAS) pain absolute improvement scores, demonstrating an approximate Youden-derived threshold of 4 points (AUC = 0.702). The red circles indicate the Youden-derived thresholds. AUC, area under the curve.

Because only 7 patients were classified as not improved, multivariable logistic regression was not performed. In exploratory univariable logistic regression, pre-operative PSQI was not significantly associated with anchor-defined improvement (odds ratio, 1.13; 95% confidence interval, 0.86-1.49; P = .384) (Table V). This finding should be interpreted cautiously because of the small number of nonimproved patients and the exploratory nature of the analysis.

Table V.

Exploratory univariable logistic regression analysis of baseline PSQI and GRC-defined improvement.

Variable	OR	95% CI	P value
Pre-operative PSQI score	1.13	0.86-1.49	.384

CI, confidence interval; GRC, global rating of change; OR, odds ratio; PSQI, Pittsburgh Sleep Quality Index.

Because only 7 patients were classified as not improved, multivariable logistic regression was not performed. This exploratory univariable analysis should not be interpreted as identifying an independent predictor.

Discussion

The principal finding of this study was that rTSA was associated with substantial 6-month improvements in both sleep quality and pain. Exploratory GRC-anchored ROC analyses generated approximate improvement thresholds of 7 points for PSQI and 4 points for VAS pain. However, these values should not be interpreted as validated MCID thresholds, particularly for PSQI. The GRC anchor demonstrated only a weak and nonsignificant association with PSQI change; the anchor-defined groups were markedly imbalanced, and ROC discrimination was modest, particularly for PSQI. Therefore, the primary contribution of this study is the provision of hypothesis-generating, anchor-derived improvement thresholds rather than validated clinical benchmarks.

Sleep disturbance is increasingly recognized as a clinically relevant symptom in patients with shoulder disorders. PSQI is a widely used instrument for quantifying sleep quality, and previous studies have shown that shoulder pain, nocturnal symptoms, and functional disability are associated with impaired sleep quality.^3,14,19 Systematic reviews of shoulder surgery have further demonstrated that sleep outcomes often improve after operative treatment, supporting the concept that sleep disturbance is a modifiable component of shoulder-related symptom burden rather than an incidental complaint.^2,18 Prospective 2-year rotator cuff repair data also indicate that sleep disturbance has been evaluated beyond the early post-operative period, underscoring the importance of longitudinal assessment of sleep recovery after shoulder surgery.¹ In shoulder arthroplasty populations, both anatomic and reverse shoulder arthroplasty have been associated with post-operative improvements in sleep disturbance, suggesting that pain relief and restoration of shoulder mechanics may translate into improved sleep quality.^20,22 More recent rTSA-specific and mixed shoulder surgery studies have similarly reported improvements in PSQI-derived sleep quality after rTSA, although recovery trajectories may differ between arthroplasty and rotator cuff repair populations.^8,16 The present findings are consistent with this literature, as both PSQI and VAS pain scores improved substantially by 6 months after rTSA.

The interpretation of the PSQI threshold requires particular caution. In rotator cuff repair populations, anchor- and distribution-based approaches have yielded PSQI MCID or PASS values, but these values may not be directly transferable to rTSA patients because of differences in age, diagnosis, baseline symptom severity, shoulder biomechanics, and recovery trajectory.^10,11 Additional studies of rotator cuff repair have also suggested that post-operative sleep quality may be influenced by pain, function, narcotic use, comorbidities, and other patient-related factors, emphasizing that sleep recovery is multifactorial.^13,24 In the present cohort, the approximate 7-point PSQI threshold was numerically larger than previously reported values in rotator cuff repair populations. This may reflect greater baseline sleep impairment in rTSA patients, but it may also reflect limitations of the anchor and the small number of nonimproved patients. Because the correlation between the GRC anchor and PSQI change was weak, the PSQI value should be considered an exploratory anchor-derived improvement threshold rather than a true MCID.

The VAS pain threshold should also be interpreted as preliminary. The approximate 4-point VAS pain threshold was larger than some previously reported MCID estimates for VAS pain after shoulder arthroplasty, highlighting that threshold values are sensitive to methodology, anchor selection, baseline symptom severity, and follow-up timing.¹⁷ In the early post-operative period after shoulder arthroplasty, pain, sleep, and activities of daily living may improve along different trajectories, supporting the need to interpret pain improvement together with sleep-specific outcomes rather than in isolation.⁹ Mixed shoulder surgery data have also shown that sleep quality improves alongside other patient-reported outcomes, reinforcing the importance of including sleep as a separate patient-centered outcome domain.¹⁶ In the present study, the GRC anchor correlated more strongly with VAS change than with PSQI change, suggesting that patients may have based their global perception of improvement more on pain relief than sleep quality. This finding reinforces the need for sleep-specific anchors in future studies designed to validate PSQI thresholds after rTSA.

The exploratory predictor analysis did not identify a significant association between baseline PSQI and anchor-defined improvement. This finding should not be interpreted as evidence against the clinical relevance of baseline sleep disturbance, but rather as a reflection of the limited statistical stability of the present dataset. Only 7 patients were classified as not improved, which precluded reliable multivariable modeling and limited the ability to identify independent predictors. Previous studies have suggested that pre-operative sleep disturbance, psychological burden, inflammatory markers, tobacco use, opioid exposure, comorbidity burden, and pre-operative nocturnal pain may influence post-operative pain or sleep-related recovery after shoulder surgery.^6,12,13,15,21, 22, 23 In addition, pathway-level interventions targeting sleep and analgesia after shoulder arthroplasty suggest that sleep may represent an actionable perioperative endpoint.⁴ Mechanistic data linking nocturnal pain and melatonin signaling, as well as rTSA studies evaluating sleep disturbance in cuff tear arthropathy, further support the biologic and clinical relevance of assessing sleep outcomes in this population.^5,7 However, the present study was not powered to determine whether these factors independently influence sleep recovery after rTSA.

Several limitations should be acknowledged. First, this was a retrospective single-center study with a relatively small sample size. Second, only 7 patients were classified as not improved according to the GRC anchor, resulting in severe imbalance between anchor-defined groups and limiting the stability of ROC-derived thresholds and specificity estimates. Third, the weak correlation between the GRC anchor and PSQI change limits the validity of interpreting the PSQI threshold as a true MCID. Fourth, outcomes were assessed only at 6 months, preventing evaluation of earlier threshold achievement or long-term durability of sleep and pain improvement. Prior studies have shown that sleep improvement may continue or stabilize over longer follow-up after shoulder surgery, indicating that thresholds may vary according to assessment timing.^1,8,16 Fifth, objective sleep measures such as actigraphy or polysomnography were not available. Finally, unmeasured factors such as psychosocial status, sleep medication use, analgesic protocols, comorbidity burden, and diagnosis-related heterogeneity may have influenced patient-reported sleep and pain outcomes.

The clinical implication of this study is that sleep quality should be considered an important patient-reported outcome after rTSA, alongside pain and function. The observed improvements suggest that rTSA may provide symptomatic benefit in patients with substantial pre-operative sleep disturbance. Nevertheless, the proposed PSQI and VAS pain thresholds should not be used as definitive decision-making benchmarks. Instead, they may serve as preliminary reference values for designing future prospective studies and for generating hypotheses regarding sleep recovery after rTSA. Larger prospective, multicenter, diagnosis-stratified studies with multiple post-operative time points, objective sleep assessment, and sleep-specific anchors are required to validate these exploratory thresholds and determine their clinical applicability across different rTSA populations.

Conclusion

rTSA was associated with substantial 6-month improvements in sleep quality and pain. In this small retrospective cohort, exploratory anchor-based ROC analyses generated preliminary improvement thresholds of approximately 7 points for PSQI and 4 points for VAS pain. However, the PSQI anchor correlation was weak, the anchor-defined groups were highly imbalanced, and ROC discrimination was modest. These estimates should therefore be considered hypothesis-generating rather than definitive MCID values and require validation in larger, prospective, diagnosis-stratified cohorts with multiple follow-up time points and objective sleep assessment.

Disclaimers

Funding: No funding was disclosed by the authors.

Conflicts of interest: The authors, their immediate families, and any research foundations with which they are affiliated have not received any financial payments or other benefits from any commercial entity related to the subject of this article.