Ultrasound parameters to identify and diagnose carpal tunnel syndrome. A review of the literature

Abstract

Numerous works of literature have assessed the use of ultrasound to detect carpal tunnel syndrome, suggesting various techniques and cut‐off values. Currently, an effective parameter and cut‐off value are still debated. The aim of this review is to determine if these parameters have sufficient rigour to allow their use in clinical practice. Twenty‐one studies using sonographic parameters to identify carpal tunnel syndrome in comparison with electrodiagnostic testing (EDx) were selected for review. Methodological differences were found between studies in the use of EDx criteria, scanning and recruitment protocols, with participant biometrics often not reported. Parameters including the cross‐sectional area of the median nerve at the level of the pisiform bone in addition to the wrist‐to‐forearm difference demonstrated high diagnostic utility for set cut‐off values. Doppler techniques and mobility are promising, and further research is required to understand the effectiveness of these techniques.

Introduction

Carpal tunnel syndrome (CTS) defines the symptomatic compression of the median nerve at the carpal tunnel.¹ The condition accounts for 90% of peripheral entrapment neuropathies affecting up to 5% of the population.², ³ Affected individuals experience paraesthesia and motor sensory loss in the distribution of the median nerve with thenar muscle atrophy, risking increasing morbidity without timely diagnosis.³, ⁴, ⁵ Currently, a strong consensus exists for clinical diagnosis based on the patient history and physical examination in conjunction with electrodiagnostic testing (EDx).⁶, ⁷, ⁸, ⁹, ¹⁰

Electrodiagnostic testing is considered by authors to be an accurate test, with a sensitivity and specificity of 78–95% and 80–95%, respectively.¹, ¹¹, ¹², ¹³, ¹⁴ However, other authors have reported varying false‐negative rates of 10–34%.², ⁶, ⁷, ⁸, ⁹ The test assesses for physiological delays within the median nerve through slowed electrical conduction due to compression, representing the degree of inflammation and axon loss.¹⁰, ¹¹, ¹⁵ This enables the detection of CTS with grading of the severity.¹¹, ¹⁴ However, EDx requires an invasive approach and an extended examination time, with limited and often cost‐prohibitive availability.², ⁵, ⁷ Additionally, EDx lacks the ability to diagnose underlying anatomical causes of compressive neuropathy.¹⁵, ¹⁶, ¹⁷ As such, with lack of a universal gold standard test, a growing trend is seen for additional diagnostic methods in medical imaging such as ultrasound.⁶, ⁹

The use of ultrasound in the diagnosis of CTS has increased over the last two decades, owing to its accessible, inexpensive and non‐invasive nature.¹, ³, ⁴, ⁵ The high‐resolution real‐time imaging capabilities demonstrate altered median nerve characteristics along with anomalous muscles, ganglions, and tendon pathology.⁵, ⁹, ¹⁶, ¹⁸ Multiple works have investigated differing techniques for quantifying the degree of neural thickening in affected wrists. Examples include the inlet and outlet cross‐sectional area, inlet‐outlet ratio, wrist‐to‐forearm ratio, and the cross‐sectional area difference between the wrist and forearm.¹, ³, ⁴, ⁵, ⁶, ⁷, ⁸, ⁹, ¹⁰, ¹¹, ¹², ¹³, ¹⁴, ¹⁵, ¹⁶, ¹⁷, ¹⁸, ¹⁹, ²⁰, ²¹ In addition, a flattening ratio and supplementary parameters of colour Doppler and median nerve mobility are also reported in the literature.², ⁴, ⁶, ⁷, ¹¹, ²⁰

Many studies have compared these ultrasound parameters using differing study designs with varying results.¹, ², ³, ⁴, ⁵, ⁶, ⁷, ⁸, ⁹, ¹⁰, ¹¹, ¹², ¹³, ¹⁴, ¹⁵, ¹⁶, ¹⁷, ¹⁸, ¹⁹, ²⁰, ²¹ Factors that account for variability between studies include participant characteristics (e.g. race, age, gender, height, weight, and body mass index) and anthropometrics (e.g. wrist circumference).¹⁵, ¹⁸, ¹⁹ The variations in sample size, with differing inclusion and exclusion criteria of case/control groups, combined with no established gold standard for diagnosis of CTS, cause further variation between studies.⁶, ⁹ The non‐standardisation of measuring protocols and equipment differences leads to further variability.¹⁸, ¹⁹ Furthermore, interrater reliability, operator experience, and blinding are often unaccounted for, potentially reducing the impact of results.⁶, ¹⁰, ¹⁷ Consequently, meaningful analysis of the literature is challenging due to differences between study designs, thus increasing the difficulty of determining the diagnostic utility of ultrasound in CTS.², ¹⁴ While not a perfect test, the use of EDx may provide a basis for comparison for the diagnostic utility of ultrasound in CTS.

The aim of this review is to identify the sonographic parameters used to diagnose CTS in comparison with EDx and determine if these sonographic parameters have sufficient rigour to allow their use in clinical practice. Previously published articles will be examined for participant numbers, demographics and biometrics, reference standards, inclusion and exclusion criteria, equipment, scanning protocol and reliability in addition to the sensitivity and specificity of each parameter.

Methodology

Search Strategy

Online databases were accessed through an institution. PubMed, MEDLINE, Springer, and Science Direct were searched using a Boolean search strategy utilising the operators ‘AND’, ‘OR’, and ‘NOT’, in conjunction with the keywords ‘carpal tunnel syndrome’, ‘ultrasound’, ‘ultrasonography’, ‘sonography’ in titles; and ‘electrodiagnostic’, ‘electrophysiologic’, and ‘cross‐sectional area’. Figure 1 demonstrates the selection process of retrieved articles. For articles to be considered for inclusion, abstracts needed to contain at least one of the additional keywords: wrist‐to‐forearm ratio, wrist‐to‐forearm difference, inlet, outlet, flattening ratio, colour or power Doppler, dynamic, mobility, transverse displacement, sensitivity, or specificity.

Figure 1.

Study selection flow chart.

Inclusion criteria

For inclusion, articles were required to be prospective, have measured either the cross‐sectional area at the carpal tunnel inlet or outlet, wrist‐to‐forearm ratio, wrist‐to‐forearm difference, flattening ratio, or other complimentary findings including colour Doppler and nerve mobility. The sensitivity and specificity of the above parameters needed to be specified, utilising EDx as the reference standard. A research design of a significant sample size (>30 participants) with clearly specified exclusion criteria and having been published in English between 2008 and 2018 were also required.

Exclusion criteria

Exclusion criteria were using clinical diagnosis or other methods as a reference standard.

Results

The 21 articles included in the review utilised either a case‐control or observational design to assess the sonographic parameters. The main characteristics of these studies are summarised in Tables 1, 2, 3.

Table 1.

Control testing and demographic data of case and control groups.

Authors	Total participants (n)	Case group male/female (n) mean age (SD)	Control group male/female (n) mean age (SD)	Case group height cm (SD) weight kg (SD) BMI (SD)	Control group height cm (SD) weight kg (SD) BMI (SD)	Control testing
Mohamed et al. (2018)1	75	14/36 53.3 (8.4)	9/16 51.6 (8.8)	Not stated	Not stated	EDx
Kuo et al. (2016)2	72	Not stated	Not stated	Not stated	Not stated	EDx
Azami et al. (2014)3	180	5/55 56.8 (10.6)	2/28 54.8 (7.8)	162.5 (5.8) 77.8 (4.7) 29.5 (2.4)*	162.9 (4.9) 71.2 (4.9) 26.8 (4.9)	Clinical
Maha et al. (2012)4	41	10/31 47 (5.62)/49.1 (6.5)	No control	Not stated	No control	No Control
Ackar et al. (2010)5	42	16/26 45.7 (9.7)	11/22 42 (12)	Not stated	Not stated	Clinical
Ooi et al. (2014)6	66	9/42 55	3/12 53	Not stated	Not stated	Clinical
Ng et al. (2018)7	41	4/20 57 (5.4)	4/13 55.5 (8.7)	Not stated	Not stated	Clinical
Kim et al. (2014)8	154	30/216 53	3/27 41.7	Not stated	Not stated	Clinical
Borire et al. (2016)9	121	Not stated 55.0 (11.8)	Not stated 51.1 (8.2)	165.9 (9.08) 83.7 (21) 30.4 (7.5)	173 (10.3) 87.3 (16.2) 29.0 (5.0)	Not stated
Mhoon et al. (2012)10	125	42/58 55 (12.3)	11/14 45 (13.2)	167.6 (11.2) 90.1 (23.1) (Not stated)	168.4 (9.9) 80.8 (20.5) (Not stated)	EDx
Kutlar et al. (2017)11	121	12/59 48.6 (10.9)	10/40 48.3 (9.2)	Not stated	Not stated	Clinical
Kang et al. (2011)12	140	10/100 53.18	2/36 53.18 (6.20)	Not stated	Not stated	Not stated
Hobson‐Web et al. (2008)13	62	10/44 52.4 (12.8)	10/18 39.2 (14.2)	165.6 (9.4) Not stated Not stated	173.7 (10.4)* Not stated Not stated	Clinical
Azman et al. (2017)14	136	28/58 53.8 (12.3)	18/32 53.8 (14.4)	Not stated Not stated 29.4 (5.17)	Not stated Not stated 26.8 (4.12)	EDx
Hunderfund et al. (2011)15	104	23/32 59 (15)	23/26 6 (16)	166.6 (9.8) 86.4 (20.8) 30.9 (5.9)*	169.7 (11.2) 80.1 (21.9) 27.4 (5.5)	EDx
Ulasli et al. (2012)16	82	4/51 47.3 (11.9)	4/23 40.5 (9.6)	162 (5) 79.5 (12.5) 36.6 (5.93)*	162 (6) 70.6 (10) 32.3 (5.25)	EDx
Fu et al. (2015)17	90	16/30 45.3 (19.8)	18/26 50.4 (13.9)	163 (8) 58.2 (11.5) 21.7 (3.33)	165 (7) 57.8 (11.3) 21.1 (2.7)	EDx
Klauser et al. (2009)18	126	16/52 (57.9)	16/42 (55.1)	Not stated	Not stated	Clinical
Roll et al. (2011)19	91	10/37 45.6 (10.6)	14/30 40 (12.1)	166 (8.3) 88 (21.2) 32 (7.4)	168.8 (8.5) 78 (19.7) 27.4 (7)	Clinical
Ghasemi‐Esfe et al. (2011)20	156	23/78 50.9 (1.29)	14/41 48.51 (1.7)	No stated	Not stated	EDx
Dejaco et al. (2013)21	135	36/99 51.0 (14.5)	5/18 50.4 (6.7)	Not stated Not Stated 26.8 (4.3)	Not stated Not stated Not stated	Clinical

BMI, Body Mass Index; cm, centimeters; EDx, electrodiagnostic testing; kg, kilograms; n, number; SD, standard deviation.

Significant difference.

Table 2.

Cut‐off values for electrodiagnostic testing and operator variables.

Authors	Electrodiagnostic value for recruitment of CTS cases	Ultrasound machine/Tx frequency	Measuring technique	Operator/experience	Interrater reliability/blinding
Mohamed et al. (2018)1	Not stated	Medison Accuvix IX, Seoul, Korea/12 MHz	Continuous trace	Radiologist/>12 years	Not stated
Kuo et al. (2016)2	SVC < 40 m/s DSL > 2.5 ms DML > 4.0 ms	Model t3000, Terason, MA/10 MHz	Transverse displacement	Musculoskeletal physician/>10 years	Not stated
Azami et al. (2014)3	SNCV < 40 m/s DML > 4.5 ms	A MEDISON V20 ultrasound machine/14 MHz	Not stated	Radiologist/not stated	Not stated/Blinded
Maha et al. (2012)4	DSL > 2.9 ms DML > 4.2 ms	Philips HD 10/12 MHz	Ellipse method	Radiologist/>5 years	Not stated
Ackar et al. (2010)5	SNCV < 50 m/s DML > 4 ms MUDS 0.4 ms	Toshiba SSA 770 A/7.5 MHz	Continuous trace	Radiologist/12 years	Not stated
Ooi et al. (2014)6	Not stated	Philips iU22 machine/17–5 MHz	Continuous trace	2 Sonographers/>5 years	k = 0.71–0.89 (CSA)/Blinded
Ng et al. (2018).7	DML > 4 ms MUDS > 0.4 ms	Philips iU22 machine/12–17 MHz	Continuous trace	Radiologist/>15 years	Not stated/Not blinded
Kim et al. (2014).8	Not stated	Sonix RP/not specified	Continuous trace	Not stated	Not stated
Borire et al. (2016).9	CMAP (Value not specified) DML (value not specified) SNAP (Value not specified)	MyLabOne system/10 –18 MHz	Continuous trace	Not stated	Not stated
Mhoon et al. (2012).10	DML> 4.3 ms MUDS> 0.4 ms Median mixed nerve latency > 2.2 ms	Biosound Esaote MyLab 70/18‐MHz	Continuous trace	Neurologist/Not stated	r = 0.9845/0.9641 (CSA/WFR)/Blinded
Kutlar et al. (2017).11	DML (value not specified) SNAPs (value not specified) MUDS ≥ 0.5 ms	Aplio 500 A/7–14 MHz	Continuous trace	Radiologist/>10 years	Not stated/Blinded
Kang et al. (2011).12	DML latency > 4.2 ms CMAP amplitude: 4.1–4.5 mV SNAP amplitude: 16–20 μV	M2540A Envisor Ultrasound Imaging System/not specified	Continuous trace	Not stated	Not stated
Hobson‐Web et al. (2008)13	DML ≥ 4.4 ms Median mixed nerve latency ≥ 2.2	Phillips HDI 5000/15–7 MHz	Continuous trace	Sonographer/Not stated	Not stated/Blinded
Azman et al. (2017)14	DML ≥ 4.3 ms MUDS ≥ 0.4 ms	ProSound a7 Premium US system/5–13 Mhz	Continuous trace and ellipse	Sonographer/Not stated	Not stated/Blinded
Hunderfund et al. (2011)15	DML (value not specified) SNAP (Absence) CMAP (value not specified)	GE logiq e device/5–13 MHz	Continuous trace	Imaging by sonographers, measurements by unspecified investigator	Not stated/unblinded Imaging, blinded measurement
Ulasli et al. (2012)16	NCVs (values not specified) SNAP (values not specified) CMAP (Values not specified)	GE Healthcare Logiq S6/10–14 MHz	Continuous trace	Radiologist and physiatrist/not stated	Not stated
Fu et al. (2015)17	MUDS > 0.4 ms DML > 4.0 ms	Aplio 500 Ultrasound System/5–13 MHz	Continuous trace	Radiologist/>10 years	k = 0.79/0.88 (inlet/outlet CSA)/blinded
Klauser et al. (2009)18	Padua et al. scoring system (values not specified)	Mylab90/14–8 MHz	Continuous trace	Radiologist/>5 years	Not stated/blinded
Roll et al. (2011)19	SNCV < 50 m/s SNAP < 10 μV DML > 4.2 ms CMAP < 4.0 mV	GE Healthcare LOGIQ i ultrasound system/12 MHz	Continuous trace	Sonographer/not stated	Value not stated/Blinded to EDx
Ghasemi‐Esfe et al. (2011)20	Prolonged median sensory latency 3.5 ms DML > 4.2 ms	Siemens Sonoline Antares/9–13 MHz	Continuous trace	Radiologist/>5 years	Value not stated/Blinded
Dejaco et al. (2013)21	DML 4.2 ms CMAP 5 mV	GE Logiq E9 ultrasound device/6–15 MHz	Continuous trace	2 Rheumatologist/>5 years	Not stated/Blinded

CMAP, Compound muscle action potential; CTS, Carpel Tunnel Syndrome; DML, Distal Motor Latency; DSL, Distal Sensory Latency; k, Kappa score; m/s, meters per second; MHz, Megahertz; ms, millisecond; MUDS, Median‐Ulnar Differences for Sensory Latencies; mV, millivolt; NCV, Nerve Conduction Velocity; r, Pearson's Correlation Coefficient; SNAP, Sensory Nerve Action Potential; SNCV, Sensory Nerve Conduction Velocity; Tx, transducer; μV, microvolt.

Table 3.

Sensitivity and specificity of diagnostic values for sonographic parameters.

Authors	Parameter	Cut‐off value	Sensitivity	Specificity	Case wrists (n)	Control wrists (n)
Mohamed et al. (2018)1	CSA inlet (pisiform) Outlet (hamate)	9 mm2 9 mm2	95% 96%	100% 100%	50	25
Kuo et al. (2016)2	Mobility (transverse displacement)	–	85%	96.9%	40	32
Azami et al. (2014)3	CSA inlet (proximal edge retinaculum) CSA outlet (distal edge retinaculum) FR (outlet)	9.15 mm2 8.15 mm2 0.93	99.2% 96.7% 99.2%	88.3% 55% 75%	120	60
Maha et al. (2012)4	CSA inlet (not defined) Vascularity	11 mm2 Power Doppler	94% 98%	40% 60%	53	0
Ackar et al. (2010)5	CSA inlet (pisiform) WFD (pisiform/hamate) FR (outlet‐hamate)	11 mm2 3.65 mm2 2.57	76% 83% 66%	77% 82% 64%	62	33
Ooi et al. (2014)6	CSA inlet (pisiform) CSA Outlet (Hamate) FR (outlet hamate) Mobility Vascularity	9.8 mm2 10 mm2 3.4 Transverse displacement Colour Doppler	92% 65% 65% 58% 69%	90% 85% 60% 86% 95%	95	30
Ng et al. (2018)7	CSA Inlet (not defined) CSA Outlet (not defined) FR (inlet – pisiform) Vascularity	14 mm2 14 mm2 2.51 Colour Doppler	75% 63.6 68% 38.1%	87.5% 100% 45% 87.5	44	32
Kim et al. (2014)8	CSA Inlet (pisiform) FR (inlet – pisiform)	10 mm2 3.4	88.5% 77.8%	90% 50%	246	30
Borire et al. (2016)9	CSA inlet (pisiform) WFR (pisiform/mid‐forearm)	11.4 mm2 1.53	71% 60%	97.5% 92.5%	131	40
Mhoon et al. (2012)10	CSA Inlet (distal wrist crease) WFR (distal wrist crease/mid‐forearm)	9 mm2 >1.4	99% 97%	22% 29%	192	50
Kutlar et al. (2017)11	CSA Inlet (pisiform) Vascularity	10 mm2 Power Doppler	90.9% 93.4%	94% 90%	121	50
Kang et al. (2011)12	CSA Inlet (distal wrist crease) WFR (distal wrist crease/forearm 12 cm proximal to wrist crease)	9.5 mm2 1.34	96.4% 99.9%	92.1% 100%	110	38
Hobson‐Web (2008)13	WFR (pisiform/forearm 12 cm proximal to wrist crease)	>1.4	100%	Not stated	Not stated	Not stated
Azman et al. (2017)14	CSA Inlet (pisiform) CSA outlet (hamate) WFR (not specified) IOR	10 mm2 11 mm2 1.7 –	94.6% 74.1% 77.8% –	82.9% 92.5% 91.4% –	135	93
Hunderfund et al. (2011)15	CSA Inlet (proximal edge retinaculum) WFR (pisiform/mid‐forearm) WFD (pisiform/mid‐forearm)	11 mm2 ≥2.4 >6 mm2	84% 67% 82%	84% 86% 84%	Not stated	Not stated
Ulasli et al. (2012)16	CSA Inlet (proximal edge retinaculum) CSA Oulet (hamate) WFR (proximal edge retinaculum/forearm 12 cm proximal to wrist crease)	9.5 mm2 9.5 mm2 1.93	95% 86% 95%	71% 88% 98%	95	48
Fu et al. (2015)17	IOR	1.3	91%	93%	46	44
Klauser et al. (2009)18	CSA (proximal to distal tunnel) WFD (Wrist/forearm level of pronator quadratus)	12 mm2 2 mm2	94% 96%	95% 100%	100	93
Roll et al. (2011)19	CSA Inlet (pisiform) WFR (pisiform/distal radius) WFD (pisiform/distal radius)	10.3 mm2 1.7 4.16 mm2	80.4% 80.4% 82.4%	90.6% 81.2% 87.5%	83	83
Ghasemi‐Esfe et al. (2011)20	Vascularity CSA Inlet (pisiform)	Colour/power Doppler 10.5 mm2	83% 71%	89% 91%	101	Not stated
Dejaco et al. (2013)21	CSA Inlet (proximal edge retinaculum) CSA largest (Inlet to outlet) WFR (wrist/proximal third of pronator quadratus) WFD (CSA largest to proximal third of pronator quadratus) Vascularity	9.8 mm2 9.8 mm2 1.3 2.5 mm2 Power Doppler	90.9% 91.8% 90.9% 93.6% 47.4%	61.2% 60% 50.6% 55.3% 89.8%	111	40

CSA, Cross‐sectional area; FR, Flattening ratio; IOR, Inlet‐to‐outlet ratio; mm, millimetre; n, number; WFD, Wrist‐to‐forearm difference; WFR, Wrist‐to‐forearm ratio.

Participant variables of the study design

Table 1 demonstrates all studies recruited participants over 18 years of age.¹, ², ³, ⁴, ⁵, ⁶, ⁷, ⁸, ⁹, ¹⁰, ¹¹, ¹², ¹³, ¹⁴, ¹⁵, ¹⁶, ¹⁷, ¹⁸, ¹⁹, ²⁰, ²¹ Each study had a sample volume ranging from 41 to 180 total participants.³, ⁴ Case groups were composed of volunteers with clinical signs of CTS and positive EDx. In 10 studies, the control groups were asymptomatic volunteers screened through questioning of clinical backgrounds.³, ⁵, ⁶, ⁷, ⁸, ¹¹, ¹³, ¹⁸, ¹⁹, ²¹ In addition, eight studies assessed the control groups with EDx,¹, ², ¹⁰, ¹⁴, ¹⁵, ¹⁶, ¹⁷, ²⁰ two studies did not state the recruitment methods of controls,⁹, ¹² and one study had no control group.⁴

Many studies used bilateral observations of single participants, treating each wrist individually in the statistical analysis.³, ⁵, ¹⁰, ¹², ¹³, ¹⁴, ¹⁶, ¹⁷, ¹⁸, ¹⁹ The participant characteristics (age, height, weight, body mass index, and gender ratio) of all 21 studies are summarised in Table 1. Many studies specified the mean age and male‐to‐female ratio of both case and control groups, with matching of case/controls in 13 studies.¹, ³, ⁵, ⁶, ⁷, ⁹, ¹¹, ¹², ¹⁴, ¹⁵, ¹⁸, ²⁰, ²¹ Moreover, the height, weight, and BMI of case/control participants were recorded in six studies³, ⁹, ¹⁵, ¹⁶, ¹⁷, ¹⁹ with three studies only measuring the height,¹³ BMI,¹⁴ and both weight and height.¹⁰ Fu et al.¹⁷ had matched case and control groups stating no significant differences. Conversely, BMI was increased in case groups across many studies.³, ¹⁴, ¹⁵, ¹⁶, ¹⁹ Increasing BMI had significant correlations with increased wrist CSA measurements in two studies.¹⁷, ¹⁹ Weight was increased in four case groups³, ¹⁰, ¹⁷, ¹⁹ and had significant correlations with increasing neural thickness at the forearm¹⁶ and wrist.¹⁷ Additionally, two studies found that increased height correlated with a larger CSA within the forearm in control groups.¹³, ¹⁶ However, one study had found no significant correlation between height, weight, BMI, and the median nerve thickness.¹⁵

Technical factors of the study design

Summarised in Table 2, EDx criteria for recruitment of participants differed between many studies using differing techniques and cut‐off values for diagnosing CTS. The electrodiagnostic values used for inclusion of case groups were stated by 11 studies.³, ⁴, ⁵, ⁷, ¹⁰, ¹², ¹³, ¹⁴, ¹⁷, ¹⁹, ²⁰ Six studies did not specify the diagnostic values of these techniques,⁹, ¹¹, ¹⁵, ¹⁶, ¹⁸, ²¹ and four studies had not specified any information regarding the techniques and diagnostic values used.¹, ², ⁶, ⁸

The exclusion criteria for case and control groups varied between studies. Thirteen studies clearly specified the exclusion of systemic disorders including either diabetes,¹, ², ⁴, ⁵, ⁶, ⁸, ¹⁰, ¹², ¹⁵, ¹⁶, ¹⁸, ²⁰ hypothyroidism,², ⁴, ⁵, ⁶, ⁸, ¹⁶, ¹⁸, ²⁰ chronic renal failure,¹, ², ⁴, ⁵, ⁶, ¹⁸, ²⁰ or rheumatoid arthritis¹, ², ⁴, ⁵, ⁷, ⁸, ¹⁶ in addition to other factors of traumatic injury¹, ², ³, ⁴, ⁵, ⁶, ⁷, ⁸, ⁹, ¹⁰, ¹⁶, ¹⁷, ¹⁹, ²⁰ and previous wrist surgery.³, ⁴, ⁵, ⁶, ⁹, ¹¹, ¹³, ¹⁴, ¹⁵, ¹⁶, ¹⁷, ¹⁹, ²¹ Furthermore, cervical radiculopathy and polyneuropathy were clearly excluded by 10 studies.², ³, ⁴, ⁶, ¹², ¹⁴, ¹⁶, ¹⁷, ²⁰, ²¹ Contrary to this, many studies did not specify the inclusion or exclusion of systemic diseases³, ⁷, ⁹, ¹³, ¹⁴, ¹⁷ or pregnancy¹, ², ⁵, ⁷, ⁸, ⁹, ¹¹, ¹², ¹⁸ while three studies included participants with either diabetes or thyroid disease.¹⁰, ¹⁹, ²¹ Anatomical variations including bifid median nerves were excluded by 10 studies to further standardise participants.¹, ⁴, ⁵, ⁶, ¹¹, ¹⁴, ¹⁶, ¹⁷, ¹⁹, ²⁰ Additionally, four studies had included these variations for more variability within participants.⁹, ¹⁰, ¹⁵, ²¹ However, seven studies did not state whether these anatomical variations were included or excluded.², ³, ⁷, ⁸, ¹², ¹³, ¹⁸ Persisting median arteries were included by three studies,⁹, ¹⁰, ¹⁹ and excluded by three others.⁶, ¹⁵, ²⁰

Table 2 summarises the specified equipment and operator used in the assessment of the median nerve between studies. The transducer frequency varied across studies ranging from 7 MHz to 18 MHz. Many of the recently published articles used a higher resolution 14 or 18 MHz linear array transducer.³, ⁶, ⁷, ⁹, ¹⁰, ¹¹, ¹⁶, ²¹ The median nerve cross‐sectional area was measured through continuous tracing of the inner hyperechoic margin in 18 studies.¹, ⁵, ⁶, ⁷, ⁸, ⁹, ¹⁰, ¹¹, ¹², ¹³, ¹⁵, ¹⁶, ¹⁷, ¹⁸, ¹⁹, ²⁰, ²¹ One study used an ellipse method calculated by the machine,⁴ with an additional study utilising both methods.¹⁴ In addition, the operator experience varied considerably between studies. In six studies, a radiologist with over 10 years’ experience conducted the ultrasound¹, ², ⁵, ⁷, ¹¹, ¹⁷ and five years’ experience in three other studies.⁴, ¹⁸, ²⁰ In the remaining studies, the ultrasound was performed by either a sonographer, neurologist, or rheumatologist with varying levels of experience.², ⁶, ¹⁰, ¹³, ¹⁴, ¹⁵, ¹⁹ Operators were blinded to participant symptoms and EDx in 12 studies,³, ⁶, ¹⁰, ¹¹, ¹³, ¹⁴, ¹⁵, ¹⁷, ¹⁸, ¹⁹, ²⁰, ²¹ and while only a few studies assessed the interrater reliability of cross‐sectional area measurements, the results showed a good level of agreement (k = 0.71–0.89,⁶, ¹⁷ r = 0.96–0.98).¹⁰

Diagnostic value of parameters

The main parameters and associated diagnostic values from the 21 selected articles are summarised in Table 3. Overall, the sensitivity and specificity of ultrasound were good for most parameters. The inlet cross‐sectional area was the most commonly assessed parameter. The sensitivity and specificity ranged from 71 to 99% and from 22 to 100%. Ten studies stated the sensitivity and specificity to be greater than 80%, reaching up to 99% and 100%, respectively.¹, ², ³, ⁶, ⁸, ¹¹, ¹², ¹⁴, ¹⁵, ¹⁹ However, the associated optimal cut‐off values used to diagnose CTS differed between studies, ranging from 9 to 14 mm². Additionally, the definition of the tunnel inlet varied across studies, with authors using the pisiform,¹, ⁵, ⁶, ⁸, ⁹, ¹¹, ¹⁴, ¹⁹, ²⁰ proximal margin of the retinaculum,³, ¹⁵, ¹⁶, ²¹ and distal wrist crease¹⁰, ¹² as landmarks. The tunnel outlet cross‐sectional area was investigated by six studies and commonly defined by the level of the hook of the hamate bone.¹, ⁶, ¹⁴, ¹⁶ The outlet sensitivity and specificity ranged from 64 to 97% and from 55 to 100%, respectively. Three studies found a sensitivity of >86%¹, ³, ¹⁶ with five studies determining a specificity of >85%.¹, ⁶, ⁷, ¹⁴, ¹⁶ However, cut‐off values varied from 8.1 to 14 mm².¹, ³, ⁶, ⁷, ¹⁴, ¹⁶

Many studies assessed the application of measuring the CSA at two locations using the wrist inlet, outlet or forearm. The wrist‐to‐forearm ratio sensitivity and specificity ranged from 60 to 100% and from 29 to 100%, respectively. The optimal cut‐off ratios ranged from 1.3 to 1.9 in eight studies⁹, ¹⁰, ¹², ¹³, ¹⁴, ¹⁶, ¹⁹, ²¹ and 2.4 by an additional study.¹⁵ Ratios were calculated by using the CSA at the inlet and forearm. Described locations of forearm measurements differed between studies; involving the mid‐forearm,⁹, ¹⁰, ¹⁵ 12 cm proximal to the wrist crease,¹², ¹³, ¹⁶ level of the pronator quadratus muscle,²¹ distal forearm,¹⁴ and distal radius.¹⁹ Additionally, the wrist‐to‐forearm difference was subject to similar variations with the sensitivity and specificity of the difference ranging from 82 to 96% and from 55 to 100% and optimal cut‐off values of 2–6 mm² reported.¹⁵, ¹⁸, ¹⁹, ²¹ In addition, the inlet‐outlet ratio described by Fu et al.¹⁷ was reported to be 91% sensitive and 93% specific with a ratio of 1.3. However, Azman et al.¹⁴ had excluded this parameter from further analysis due to insignificant differences between case and control groups.¹⁴

Seven studies had assessed the flattening ratio, using differing locations at the wrist including the level of the hook of hamate,⁵, ⁶ pisiform,⁷, ⁸, ¹⁹ distal flexor retinaculum,³ and mid‐carpal tunnel.¹⁴ The sensitivity and specificity ranged from 65 to 78% and from 45 to 75%, with optimal cut‐off values ranging from 2.5 to 3.4.⁵, ⁶, ⁷, ⁸ One study had reported a sensitivity and specificity of 92% and 75% at a cut‐off of 0.93.³ However, two studies had excluded this parameter from further assessment due to insignificant values between cases and controls.¹⁴, ¹⁹

The vascularity of the median nerve was assessed by six studies, using either colour⁶, ⁷, ²⁰ or power Doppler.⁴, ¹¹, ²⁰, ²¹ The sensitivity and specificity varied between studies ranging from 38 to 98% and from 60 to 95%, respectively. In addition, the mobility of the median nerve defined by degree of the transverse displacement in two studies during flexion/extension of the fingers had a sensitivity and specificity of 69–85% and 95–97%, respectively.², ⁶

Discussion

For the diagnosis of CTS, the CSA of the median nerve at the level of the pisiform bone in addition to the wrist‐to‐forearm difference demonstrated good diagnostic utility for set cut‐off values of 10 mm² and >2 mm², respectively.⁶, ⁸, ¹¹, ¹⁴, ¹⁵, ¹⁸, ¹⁹, ²⁰, ²¹ In addition, a wrist‐to‐forearm ratio of 1.3 and 1.4 demonstrated a high sensitivity.¹⁰, ¹², ¹³, ²¹ The use of ultrasound for diagnosing CTS is progressively increasing. Over the last two decades, many studies have assessed the diagnostic utility of ultrasound for detecting CTS. However, derived optimal cut‐off values in addition to the sensitivity and specificity vary significantly for each ultrasound parameter. The results of this literature review outline significant differences in study designs that account for variability between published articles. Consequently, the diagnostic values suggested by these articles are confounded by a lack of standardisation. This emphasises the need to improve the rigour of study designs to identify effective sonographic parameters for detecting CTS.

Study design limitations

Studies included in the review differed significantly through reference standards and scanning protocols. Currently, there is no established gold standard for diagnosing CTS.⁹, ¹⁰, ¹¹ Electrodiagnostic testing is considered a credible reference standard due to the reasonable sensitivity and specificity in addition to the capabilities of grading the severity of CTS.², ⁷, ¹¹, ¹⁴ Conversely, false‐negative rates of EDx have been reported ranging from 10 to 34%.², ⁶, ⁷, ⁸, ⁹, ²⁰ This may impact the allocation of both case and control groups. Participants with mild symptomatic CTS may potentially be misdiagnosed with EDx as a false negative, as neural damage may not have occurred in the early stages.⁹ In addition, as outlined in Table 2, although many studies had specified the diagnostic values of EDx used to recruit CTS cases, these values varied, especially the distal median nerve latency (DML) ranging from >4 to >4.5 ms.³, ⁴, ⁵ This was also apparent for the sensory nerve conduction velocity (SNCV) and median to ulnar nerve difference (MUDS).³, ⁵, ¹¹, ¹⁷ These differences in criteria may lead to misrepresentation through the inclusion and exclusion of affected participants in both case and control groups. Furthermore, many authors did not assess control groups using EDx.³, ⁵, ⁶, ⁷, ⁸, ¹¹, ¹³, ¹⁸, ¹⁹, ²¹ This reduces rigour, as case/control group selection criteria differed through the reference standard.⁹, ¹⁵ One study by Azman et al.¹⁴ had excluded a control volunteer who tested positive in the electrodiagnostic test. By testing control groups, participants are more standardised through the selection process.¹⁵

The scanning technique varied between studies, utilising different methods of calculating the cross‐sectional area. Most authors used direct tracing of the inner hyperechoic border of the median nerve.¹, ⁵, ⁶, ⁷, ⁸, ⁹, ¹⁰, ¹¹, ¹², ¹³, ¹⁴, ¹⁵, ¹⁶, ¹⁷, ¹⁸, ¹⁹, ²⁰, ²¹ This method is suggested to be superior to linear measures as the median nerve may become displaced by the adjacent tissues resulting in an irregular shape.¹⁹, ²¹ In contrast, one study used an ellipse formula to calculate the cross‐sectional area, potentially limiting the diagnostic accuracy of the measurements.⁴ Additionally, the variability of measurements was only accounted for by seven studies.⁶, ⁹, ¹¹, ¹², ¹⁹, ²⁰, ²¹ These studies calculated a mean using two – three measurements to determine the cross‐sectional area, improving the diagnostic accuracy by accounting for operator related error.⁶, ⁹, ¹¹, ¹², ¹⁹, ²⁰, ²¹ Furthermore, studies varied in the machines used (Table 2). However, probe frequencies were rather similar for many studies ranging from 12 to 18 MHz, with few studies using <10 MHz (Table 2). This indicates similarities of high‐resolution imaging between studies, resulting in improved diagnostic performance.

Summarised in Table 2, 12 studies accounted for blinding of the operator³, ⁶, ¹⁰, ¹¹, ¹³, ¹⁴, ¹⁵, ¹⁷, ¹⁸, ¹⁹, ²⁰, ²¹ with many studies not reporting¹, ², ⁴, ⁵, ⁸, ⁹, ¹², ¹⁶ or attempting operator blinding.⁷ This poses limitations, as knowing participant symptoms and EDx values introduces potential bias into the results.⁷, ¹⁴, ¹⁵ Moreover, while many operators were experienced specialists with over five years’ experience in ultrasound, very limited data were available on interrater reliability.⁴, ⁷, ¹¹ Consequently, this limits the diagnostic value of many studies and impacts the diagnostic value of the results.⁶, ⁷, ¹⁰, ¹⁷ The few studies that accounted for reliability found good agreement with correlation values ranging from 0.71 to 0.98⁶, ⁷, ¹⁰, ¹⁷ An additional two studies had claimed good reliability in previous research but did not state the values.¹⁹, ²⁰ Future research needs to focus on standardisation of study design, measurements and participant case and control enrolment characteristics and reporting reproducibility and reliability of the measures.

Variations in cut‐off values in addition to the range in sensitivity and specificity may relate to participant variables. In total, nine studies had measured a participant biometric (Table 1). The impact of these variables across these studies was varied. While BMI was increased in many case groups³, ¹⁴, ¹⁵, ¹⁶, ¹⁹ only two studies had stated a significant correlation with increased wrist CSA measurements.¹⁷, ¹⁹ Conversely, Roll et al.¹⁹ state that this relationship is not clear, as increased CSA measurements in participants with larger BMIs may be attributed to either a pathological result of increased BMI or increased anthropometric composition. Two studies described increased weight and height with larger forearm CSA measurements in control groups.¹³, ¹⁶ This may predispose variations of normal CSA measurements between studies.¹³, ¹⁶ Furthermore, one study described increased weight with larger inlet CSA measurements.¹⁷ However, one study stated no significant correlations between height, weight and BMI and the median nerve thickness.¹⁵ The significance of biometrics to potentially influence differing diagnostic cut‐off values remains varied and unclear, however, dual‐level parameters account for these variables.¹⁶, ¹⁹

The age and gender may increase risk of CTS.¹⁰, ¹⁶ Considering this, study designs that accounted for these variables through case/control matching may be of increased validity through comparable controls.¹⁹ In addition, the clinical background of participants is known to affect the median nerve.⁴ This is evident through hypothyroidism, diabetes, chronic renal failure, and rheumatoid arthritis.², ³, ⁶, ¹⁰ Additionally, pregnancy and post‐surgical changes may lead to further variability between participants.⁷, ¹³, ¹⁵ This emphasises the need to implement clear and rigorous exclusion criteria for both case and control groups.¹, ², ³, ⁴, ⁶, ⁸, ¹², ¹³, ¹⁴, ¹⁵, ¹⁶, ¹⁷, ¹⁹, ²⁰, ²¹ By standardising the participants for analysis, the study design is strengthened. As disease status and pregnancy were not specified in the exclusion criteria of several studies, the reliability of the results may be reduced.¹, ², ³, ⁵, ⁷, ⁸, ⁹, ¹⁰, ¹¹, ¹², ¹³, ¹⁴, ¹⁷, ¹⁸

Four studies included participants with bifid median nerves in both case and control groups using the summated CSA of both branches.⁹, ¹⁰, ¹⁵, ²¹ Hunderfund et al.¹⁵ and Borire et al.⁹ tested for differences in results when excluding these variations from the analysis and found no statistical change. While many studies did not specify the inclusion or exclusion of this variation, results are likely to remain consistent.

Diagnostic utility of parameters

Single‐level cross‐sectional area measurements are the most widely assessed parameter. Nineteen studies summarised in Table 3 assessed the diagnostic utility with varying anatomical descriptions used to define the level of measurement locations. The pisiform bone was the most commonly used anatomical reference.¹, ⁵, ⁶, ⁸, ⁹, ¹¹, ¹⁴, ¹⁹, ²⁰ Other authors considered the proximal margin of the retinaculum³, ¹⁵, ¹⁶, ²¹ and fewer authors specified the distal wrist crease for inlet measurements.¹⁰, ¹² Cross‐sectional area measurements at the level of the pisiform appeared to consistently result in a cut‐off value of 10 mm² with a high sensitivity and specificity range of 71–95% and 83–94%, respectively.⁶, ⁸, ¹¹, ¹⁴, ¹⁹, ²⁰ This may be due to ease of measuring at this level, using a bony landmark as the reference point. Contrary to this, the outlet was commonly defined by the level of the hamate bone in four out of six studies.¹, ⁶, ¹⁴, ¹⁶ However, derived cut‐off values varied from 9 to 11 mm² with a sensitivity and specificity ranging from 65 to 96% and from 85 to 100%, respectively. One author described a cut‐off value of 8.15 mm² with a high sensitivity of 97% and low specificity of 55% at the distal margin of the flexor retinaculum.³ The variability may be attributed to the challenge of obtaining clear images at this region, as poor signal quality in addition to palm curvature and oblique coursing of the median nerve limits accurate measurements.¹, ³, ⁶, ⁷, ¹⁴, ¹⁶ The distal tunnel is known for increased pressure and measuring at this location (distal margin of flexor retinaculum) may explain the smaller cut‐off value opposed to the level of the hamate bone.³ Variations in results may reflect the median nerve swelling occurring at differing locations in the wrist. Using the largest cross‐sectional area at the wrist, two studies found a sensitivity of 91–94% and specificity of 61–95%.¹⁸, ²¹ Specificity differences may be attributed to a smaller cut‐off value stated in Table 3. Furthermore, the singular measurement design is potentially vulnerable to differences in participant variables, as variations in normal nerve thickness secondary to demographics and biometrics may lead to false‐positive results.¹⁶, ¹⁹

Multilevel parameters address variations in participant demographics and biometrics. The wrist‐to‐forearm ratio and wrist‐to‐forearm difference account for these factors through incorporating a normal segment of the median nerve within the forearm as a comparison to the wrist.⁹, ¹⁰, ¹⁵ By using the participant as the control for testing, false‐positive rates in larger participants are potentially decreased.¹², ¹⁹ Seven studies assessed the diagnostic utility with promising results.⁹, ¹⁰, ¹³, ¹⁴, ¹⁵, ¹⁹, ²¹ The sensitivity and specificity of the wrist‐to‐forearm ratio for cut‐off values of 1.3–2.4 ranged from 71 to 100% and from 29 to 100%, respectively.⁹, ¹⁰, ¹³, ¹⁴, ¹⁵, ¹⁹, ²¹ However, ratios of 1.3–1.4 were highly sensitive providing values from 91 to 100%.¹⁰, ¹², ¹³, ²¹ The larger cut‐off values may be related to differences in measurement locations between authors.⁹, ¹³, ¹⁹ Moreover, differences between study designs may contribute to cut‐off variations through poor case/control matching and differences in scanning technique, measuring at differing locations.¹⁵ Future studies need to be designed to ensure predictive validity is optimised.

In addition, the wrist‐to‐forearm difference quantifies the differences of the cross‐sectional area between the wrist and forearm. This method consistently demonstrated high sensitivity and specificity in four studies ranging from 82 to 99% and from 82 to 100% for values of 2–6 mm².⁵, ¹⁵, ¹⁸, ¹⁹ An additional study derived a lower specificity of 54%; however, differed in study design through exclusion of healthy controls in the ROC analysis.²¹ These studies were subject to similar limitations of the previous parameter through differing measurement locations. This was evident through the difference between the cut‐off values of Hunderfund et al.¹⁵ and Klauser et al.¹⁸ who found a difference of 6 and 2 mm², respectively. Hunderfund et al.¹⁵ explained that the CSA at the distal forearm was often found to be thickened in CTS cases, while measurements in the mid‐forearm resulted in no significant differences between cases and controls. This may account for the smaller value of 2 mm² by Klauser et al.¹⁸ as similarities in a thickened segment reduce the difference reported.

The flattening ratio quantifies the degree of neural compression at the carpal tunnel by measuring the width and height of the median nerve.⁵, ⁶, ⁷, ⁸ This parameter consistently resulted in lower sensitivity and specificity in six studies ranging from 65 to 78% and from 45 to 75%, respectively, for ratios of 2.5–3.4.⁵, ⁶, ⁷, ⁸ Additionally, two studies excluded the parameter from further analysis due to insignificant results between case and control groups.¹⁴, ¹⁹ One study reported a lower ratio of 0.93, resulting in 92% sensitivity.³ Such variations may be the product of a lack of standardised measuring protocols across the studies, with authors measuring at differing anatomical landmarks in the wrist. However, the diagnostic utility of this parameter was questioned by studies owing to the large variability in ratio cut‐offs and lower diagnostic values.¹⁴, ¹⁹

Intraneural vascularity as assessed by colour or power Doppler is a subjective parameter that has been shown to match the sensitivity and specificity of EDx.⁶, ²⁰ The parameter correlates with the severity of CTS and the degree of neural thickening.⁶, ¹¹ Additionally, the technique may detect participants in occult stages before thickening occurs.²⁰ This is reflected by the sensitivity and specificity of four studies ranging from 69 to 98% and from 60 to 90%, respectively, using either colour⁶, ⁷, ²⁰ or power Doppler⁴, ¹¹, ²⁰, ²¹ techniques. However, two studies found a low sensitivity of <47% with a high specificity of >88%.⁷, ²¹ These differences may be attributed to differing methods of quantifying the degree of vascularity varied between studies, as two studies used the presence of intraneural vascularity,⁴, ⁶ one study assigned a score of 0–3 based on the number of vessels seen²¹ and two studies used software programs. One quantified the number of coloured pixels²⁰ and the other measured the largest area of colour filling.¹¹ Furthermore, differences in participant severity, differing equipment (Table 2), and optimisation may also contribute to this.⁷

The mobility of the median nerve was assessed by two studies, correlating reduced transverse displacement with increased severity of CTS.², ⁶ The sensitivity and specificity were 58–86% and 85–97%, respectively.², ⁶ Kuo et al.² assessed mobility using a quantitative speckle tracking algorithm, enabling displacement values to be quantified. Conversely, Ooi et al.⁶ determined the degree of displacement more subjectively through operator observation at the carpal tunnel using a transverse plane with flexion and extension of the fingers. This presents a limitation of the study design and may account for differences in sensitivity.² Both studies found nerve mobility effective in CTS assessment.², ⁶ The combination of these subjective parameters was found to increase the diagnostic utility of ultrasound.⁶ Future studies should incorporate nerve mobility and colour Doppler assessment for CTS.

Limitations and suggestions for future research

This review is predisposed to bias as sonographic parameters were subjectively assessed based on the study design and diagnostic values. Furthermore, the validity of the research methodology is decreased by using a basic protocol for the selection of studies without critical appraisal assessments.

Future prospective study designs should focus on implementing more standardised criteria for EDx. Additionally, standardisation of anatomical locations for sonographic data collection is essential to improve research consistency for future systematic reviews and meta‐analysis. Limited research has assessed the relationship between the cross‐sectional area and participant biometrics and further investigation to determine the impact on diagnostic cut‐off values would be beneficial. Comparisons of the cross‐sectional area at the wrist and forearm are potentially more diagnostic than single‐level parameters and should be a focus in future research. Studies assessing the use of intraneural vascularity and median nerve mobility are limited and their use should be further investigated, as these parameters may provide additional diagnostic value to cross‐sectional area measurements.

Conclusion

To date, an effective standardised sonographic protocol is yet to be established for the diagnosis of CTS. Many studies have published differing parameters with varying diagnostic values. Significant differences in study design have led to inconsistencies between studies, limiting the determination of effective sonographic parameters in previous research. From the studies evaluated; ultrasound holds significant diagnostic utility in the assessment of CTS. The cross‐sectional area at the pisiform, wrist‐to‐forearm ratio, and difference consistently demonstrated high diagnostic value. The assessment of nerve mobility and intraneural vascularity is likely to increase this further. Variations in cut‐off criteria exist between studies and may be attributed to measurements at differing anatomical locations and characteristics of selection criteria. New research should focus to minimise variations in study designs to enable more comparable results.

Authorship statement

The authors acknowledge that this article conforms with the journal’s authorship policy, and all authors agree with the content of the submitted manuscript.

Funding

No funding information is provided.

Conflicts of interest

None of the authors have any conflict of interest to disclose.

Author contributions

Callum Linehan: Methodology (equal); Writing‐original draft (lead). Jessie Childs: Methodology (supporting); Writing‐original draft (supporting); Writing‐review & editing (equal). Ann Quinton: Writing‐original draft (supporting); Writing‐review & editing (equal). Aamer Aziz: Methodology (supporting); Writing‐review & editing (supporting).