A total 764 citations were identified by the database search process (Additional file 3), of which 23 studies were included in the review (Additional file 4). The exclusion grounds for articles rejected after full-text assessment are available in Additional file 5.
Quality index scores ranged from 29 to 80% (mean = 55%) demonstrating moderate overall methodological quality (Additional file 6). The majority of studies provided inadequate descriptions of control group source populations and characteristics, and only eight studies [12, 13, 20–25] applied a blinding technique to the image assessor. Additionally, 11 studies [12, 13, 22–24, 26–31] included data from both feet of control participants, and 11 studies [12, 13, 21, 22, 24, 26–29, 31, 32] included data from both feet of participants with bilateral heel pain. As statistical tests assume that each data point represents a truly independent observation, inclusion of both feet may result in an artificially inflated sample size and decreased data variability, thereby increasing the risk of Type I error [33]. Despite these limitations, most studies reported predetermined outcome variables, and clearly described imaging equipment settings and measurement techniques.
Thickness of the proximal plantar fascia
The thickness of the proximal plantar fascia was reported in 15 studies, 12 of which were measured by ultrasonography alone [13, 14, 20, 22, 25, 28–31, 34–36], one by ultrasonography and magnetic resonance imaging (MRI) [23], one by MRI alone [27], and one by plain film x-ray [21]. A factor considered important for this outcome was the prevalence of diabetes mellitus within each group, as research has shown a thickening of the plantar fascia in people with diabetes [37]. Only two studies [21, 22] considered diabetes as a specific exclusion criterion for condition groups, and only one study [22] for the control group.
Ultrasonography
The 13 studies reporting plantar fascia thickness by ultrasonography had a mean quality index score of 56%. Five studies [13, 20, 22, 23, 25] applied a blinding technique to the image assessor. A description of the methodological variability between studies is available in Additional file 7.
As the protocols and participant characteristics of the studies reporting this outcome were found to be similar, meta-analysis was considered appropriate. However, two studies could not be included: one study [31] reported data separately for the medial, central and lateral components of the plantar fascia, and another study [30] did not report the standard deviation of the mean plantar fascia thickness values.
Eleven studies with a total 379 CPHP participants and 434 control participants were included in this analysis. Statistical heterogeneity between studies was very high (I2 = 95%; Chi2 = 199.84, df = 10, P < 0.001), therefore meta-analysis was undertaken using the random-effect method. The mean difference between groups was statistically significant (P < 0.001), with the proximal plantar fascia of CPHP participants 2.16 mm thicker than control participants (95% CI = 1.60 to 2.71 mm) (Figure 1). Sub-group analysis revealed a more conservative pooled estimate by studies that applied a blinding technique to the image assessor. The mean difference between groups for blinded studies was 1.82 mm (95% CI = 1.00 to 2.65 mm, P < 0.001) and for non-blinded studies was 2.47 mm (95% CI = 1.94 to 3.00 mm, P < 0.001) (Figure 2). Funnel plot inspection revealed that studies were absent from the lower left corner of the plot, suggesting that smaller studies reporting less difference between groups had not been published. However, this distribution was explained by identifying studies on the plot that applied a blinding technique to the image assessor (Figure 3). Non-blinded studies had smaller sample sizes than the majority of blinded studies, and as a result appeared lower on the plot. Therefore, the funnel plot distribution illustrates that non-blinded studies reported larger mean differences between groups than the majority of blinded studies, indicating an overestimation of the thickness of the plantar fascia in CPHP groups.
Two studies included in the analysis above [13, 29] also reported the proportion of participants in each group with plantar fascia thickness values > 4.0 mm (i.e. the thickness values for participants were dichotomised). Additionally, one study included in the analysis above [25] reported the individual thickness values for each participant, allowing dichotomisation for the purpose of this review. These studies included a total 161 CPHP participants and 116 control participants. Statistical heterogeneity between studies was very high (I2 = 85%; Chi2 = 13.22, df = 2, P = 0.001), therefore meta-analysis was undertaken using the random-effect method. The mean difference between groups was statistically significant (P = 0.01) with CPHP participants over 100 times more likely than control participants to have plantar fascia thickness values > 4.0 mm (OR = 105.11, 95% CI = 3.09 to 3577.28) (Figure 4).
MRI
Two studies measured the thickness of the proximal plantar fascia by MRI. One study [23] applied a blinding technique to the image assessor and had a quality index score of 70%, the other [27] did not blind the image assessor and had a quality index score of 43%. A description of the methodological variability between studies is available in Additional file 8. As the protocols and participant characteristics of the studies reporting this outcome were found to be similar, meta-analysis was considered appropriate.
Two studies with a total 78 CPHP participants and 163 control participants were included in this analysis. Statistical heterogeneity between studies was very high (I2 = 93%; Chi2 = 13.50, df = 1, P < 0.001), therefore meta-analysis was undertaken using the random-effect method. The mean difference between groups was statistically significant (P < 0.001), with the proximal plantar fascia of CPHP participants 3.35 mm thicker than control participants (95% CI = 1.80 to 4.89 mm) (Figure 5). The blinded study reported a more conservative difference between groups (2.60 mm, 95% CI = 2.28 to 2.92 mm) than the non-blinded study (4.18 mm, 95% CI = 3.40 to 4.96 mm).
Plain film x-ray
One study [21] measured the thickness of the proximal plantar fascia by plain film x-ray. This study had a quality index score of 68% and applied a blinding technique to the image assessor. The sagittal thickness of the plantar fascia was measured from a lateral non-weight bearing radiograph within 5.0 mm of the calcaneal insertion. This study reports a statistically significant mean difference between groups, with the plantar fascia of CPHP participants 2.4 mm thicker than control participants (P < 0.001). The 95% CI for the difference between groups was not reported.
Ultrasound echogenicity and MRI signal intensity of the proximal plantar fascia
Four studies reported the echogenicity (presence or absence of fluid collection) of the proximal plantar fascia [13, 20, 29, 36]. The mean quality index score was 60%, and two studies [13, 20] applied a blinding technique to the image assessor. A description of the methodological variability between studies is available in Additional file 9. As the protocols and participant characteristics of the studies reporting this outcome were found to be similar, meta-analysis was considered appropriate.
Four studies with a total 209 CPHP participants and 146 control participants were included in this analysis. Statistical heterogeneity between studies was low (I2 = 0%; Chi2 = 0.20, df = 3, P = 0.98), therefore meta-analysis was undertaken using the fixed-effect method. The mean difference between groups was statistically significant (P < 0.001) with CPHP participants over 200 times more likely to demonstrate hypoechogenicity of the proximal plantar fascia than control participants (OR = 204.12, 95% CI = 52.00 to 801.28) (Figure 6). Sensitivity analysis revealed an increased pooled estimate after exclusion of the two [29, 36] non-blinded studies (OR = 211.87, 95% CI = 28.53 to 1573.54, P < 0.001).
One study [27] reported the MRI signal intensity (presence or absence of fluid collection) of the proximal plantar fascia. This study had a quality index score of 43% and did not apply a blinding technique to the image assessor. Increased signal intensity was observed in the region of fascia thickening for CPHP participants, compared with homogenous low signal intensity of the plantar fascia in all control participants.
Evidence of plantar calcaneal spur
Seven studies reported evidence of plantar calcaneal spur by plain film x-ray [12, 21, 24, 26, 38–40]. The mean quality index score was 58% and only three studies [12, 21, 24] applied a blinding technique to the image assessor. A description of the methodological variability between studies is available in Additional file 10. As the protocols and participant characteristics of the studies reporting this outcome were found to be similar, meta-analysis was considered appropriate.
Seven studies with a total 322 CPHP participants and 749 control participants were included in this analysis. Statistical heterogeneity between studies was high (I2 = 74%; Chi2 = 23.25, df = 6, P < 0.001), therefore meta-analysis was undertaken using the random-effect method. The mean difference between groups was statistically significant (P < 0.001) with CPHP participants over 8 times more likely to show evidence of subcalcaneal spur than control participants (OR = 8.52, 95% CI = 4.08 to 17.77) (Figure 7). Sensitivity analysis revealed an increased pooled estimate after exclusion of the four [26, 38–40] non-blinded studies (OR = 16.11, 95% CI = 7.09 to 36.60, P < 0.001).
One study [30] reported evidence of subcalcaneal spur by ultrasonography. This study had a quality index score of 43% and did not apply a blinding technique to the image assessor. A variable frequency (5-10 MHz) linear array transducer was used to assess the heels of 190 CPHP and 48 control participants. The presence of subcalcaneal spur was a subjective observation found in 45% of CPHP participants and only 2% of control participants.
Radioisotope uptake and vascular perfusion of the proximal plantar fascia
Three studies reported the presence of increased radioisotope uptake within the subcalcaneal region in participants with CPHP [12, 40, 41]. The mean quality index score was 45% and one study [12] applied a blinding technique to the image assessor. Meta-analysis of data from these studies was not found to be appropriate as one study did not report the control group sample size [12], and another did not report the phase (early or delayed) in which scintigraphic images were assessed [40]. Participant characteristics of condition [41] and control [12] groups were also poorly reported. All three studies reported increased subcalcaneal uptake of technetium-99 m methylene diphosphonate in participants with CPHP compared to control groups (Figure 8), though no statistical comparisons were made.
One study described the degree of vascular perfusion within the proximal plantar fascia by use of power doppler ultrasound [36]. This study had a quality index score of 47% and did not apply a blinding technique to the image assessor. Doppler ultrasound with a pulse-repetition frequency of 1102 Hz was used to grade the colour signal of the proximal plantar fascia. This study reported moderate to marked hyperaemia of the proximal plantar fascia in 8 of 20 CPHP participants, and only mild hyperaemia in 1 of 20 control participants. The difference between groups was not statistically analysed, however the authors report a statistically significant correlation between hyperaemia and symptom duration of less than six months (Spearman r = -0.68, P < 0.05).