Literature search
The process of the literature search is illustrated in Fig. 1, a total of 604 potentially relevant articles were identified from 6 databases. 311 duplicated articles were excluded by using Endnotes. After a title and abstract screened, 274 articles were excluded. 46 records considered potentially eligible and full text reviewed, but 37 were excluded for different reasons (27 articles lack of crucial data, 6 articles follow-up less than 6 months, and 4 articles sample sizes less than 10 cases). Finally, 9 articles were reflected in the present meta-analysis[15-23].
Study characteristics
The basic characteristics of included studies are shown in Table 1. The study included 409 patients with an average age of approximately43.24 (18–69)years, a pulse range from 800 to 6000, a female ratio of approximately 29.83%,an energy flux density range from 0.12-0.51 mJ/mm2 and an average follow-up of 54.3 (6–228) months. 9 studies included 1 RCT, 3 case control studies,and 5 case series studies. The RCT study was assessed according to modified jaded scale and got 4 points.5 case series was applied in the NICE case series scoring criteria for quality evaluation and got an average point of 5.4.We used NOS scales to evaluate 3 case controls and got an average score of 8.2.

The HHS of patients before and after ESWT
HHS scored patients according to four parts including pain, function, deformity and range of motion. More score patients got, the better function patients had. The pooled results of HHS with eight studies of 337 hips show that ESWT achieved higher Harris hip score after treatment (MD= -20.03;95%CI,-25.56, -14.49;Fig.2). The differences were statistically significant(p<0.01).However, the heterogeneity analysis show an excessive heterogeneity(I2=91%).according to the language of articles, we divided articles into the Chinese group and the English group and performed a subgroups analysis(Fig 3). In subgroups analysis, the heterogeneity of the English group has been eliminated. ESWT also achieved higher scores compared to baseline scores.
The VAS of patients before and after ESWT
VAS is the level of pain expressed by eleven figures from 0 to 10. 0 means no pain and 10 means excessive pain. Patients chose one figure to express their degree of pain. The pooled results of VAS with seven studies of 253 hips demonstrate that ESWT achieved lower VAS scores after treatment(MD= 2.77;95%CI,1.88, 3.65;Fig.4), the differences were statistically significant(p<0.01).However, the heterogeneity analysis shows an excessive heterogeneity(I2=86%). We performed subgroups analysis(Fig 5). Studies were divided into the low score group and the high score group according to the degree of the pain score decreased. In subgroups analysis, the heterogeneity of low score group has been eliminated. All group of ESWT have achieved lower scores compared to baseline scores, the differences were statistically significant(p<0.01).
Changes of lesion area on MRI before and after ESWT
6 studies were include in this meta-analysis. We selected SMD as Consolidated statistics due to Different concepts of lesions in the 4 articles. The pooled results of lesion of MRI with 164 hips show that ESWT decreased the lesion area of MRI after treatment(SMD=1.03,CI,0.75,1.30,Fig.6). the differences were statistically significant(p<0.01).However, CI of the four studies included 0,it can be referred that ESWT have no efficient effect on changing the lesion area of ONFH. the Heterogeneity analysis also show an Excessive heterogeneity(I2=97%).
Effect of ESWT compare to surgical therapy in HHS and VAS
Two studies compared ESWT to surgical method (core decompression or core decompression with bone grafting)in HHS and VAS[16,21]. We used degree of changed score as effect size to evaluate the effect of different intervention method. We conducted a calculation to get compared data by using raw data. The polled result of two studies of 180 patients(218hips)showed that ESWT have a better improvement in HHS(MD = 13.64; 95% CI, 10.60,16.65; Fig.7) compare to surgical therapy, the differences were statistically significant(p<0.01).And ESWT have a better effect on pain relief compare to surgical method(SMD = 1.25; 95% CI, 0.95,1.54; Fig.8).
Etiologic factors
A total of 5 studies (170 patients) recorded the etiologic factors and the number of patients. alcohol was the most common etiologic factor (66 patients, 39%), followed by corticosteroid (43 patients, 25%)and idiopathic (42 patients, 25%).
Change of ARCO stage before and after ESWT
ARCO staging is more systematic, more comprehensive and more practical compared to the previous staging of ONFH. Two articles recorded the ARCO stage of patients before and after ESWT. It is shown in Fig.9.Considering 3 patients were lost to follow-up, there is no significant change of ARCO stage before and after ESWT (p>0.05).
Rate of total hip arthroplasty
Three articles (104hips) recorded the rate of THA (total hip arthroplasty).About 19.2% hips developed a THA,and 80.8% hips had survived approximately. The duration of follow-up reached 8-9years in one study with a rate of THA was 24%.However,hip arthroplasty rate of surgical group (core decompression) reached to 64%[16].
Sensitive analysis
In the comparison of VAS score before and after ESWT,Heterogeneity (I2=54%) still existed in the high score group after a subgroups analysis (Fig 5). So a sensitive analysis was performed. When the study of Wu 2017 had been omitted, Heterogeneity (I2=24%) was significantly reduced. So we read this article again to find the reason. Relatively small number of pulses may be one reason for this heterogeneity. This article type is case-control and lack of blinding method may be another reason.
publication bias
We performed an egger test to evaluate the publication bias of the change of HHS sore (table 2). It can be referred that there is no publication bias by the p-value. The value of p is 0.477 and more than 0.05.