Osteoarthritis (OA) is a major cause of knee disability involving cartilage damage related to an inadequate healing response in the inflammatory milieu [1]. Current non-surgical treatment modalities include physiotherapy, analgesia, non-steroidal anti-inflammatory drugs, and intra-articular injections, such as hyaluronic acid (HA), corticosteroids, or Ozone, with the purpose of reducing symptoms and improving joint function [24].

In the past decade, there has been an increasing interest in the use of autologous growth factors, such as intra-articular injections of platelet-rich plasma (PRP) for treatment of knee OA [5]. PRP is a fraction of whole blood and prepared by the centrifugation of autologous blood, thereby yielding a higher concentration of platelets than baseline values. The regenerative effect and anti-inflammatory potential of PRP in the tissue healing process have led to extensive investigation of PRP as a potential treatment for a variety of musculoskeletal indications, including OA [68].

A number of randomized controlled trials (RCTs) were reported with favourable outcomes of PRP injections [917]; several reviews, including systematic reviews and meta-analysis, have been published with conclusion that PRP was found to be an effective and safe orthobiologic in the treatment of knee OA compared with other intra-articular injections [1828]. However, these reviewers also concluded that more RCTs, in particular high-quality studies, were still needed. Considering that prior reviews either included non-RCTs or only synthesized a small number of RCTs (less than 9) for analysis [1828] and that quite a few more RCTs recently have been published [2935], we believe that it is necessary to perform an updated systematic review and meta-analysis, if appropriate, to evaluate whether the evidence-based support for PRP treatment will be strengthened or compromised. Furthermore, a large number of studies may allow us to fully investigate the temporal effect of PRP specifically on knee pain and physical function.


This systematic review was registered online in PROSPERO (registration number: CRD42016045410) and was performed following the guidelines of the PRISMA statement. The protocol and the PRISMA checklist were provided as Additional files 1 and 2, respectively.

Inclusion and exclusion criteria

All published RCTs evaluating the efficacy and/or safety of PRP (or preparations including autologous platelet concentrate, autologous conditioned plasma, and plasma rich in growth factors) in the treatment of knee OA in human were eligible for inclusion. Only studies that included patients aged 18 years or older with symptomatic knee OA and had a minimum follow-up of 12 weeks were included. All studies had to include at least 1 control group treated by intra-articular agents other than PRP. The studies that PRP was used in combination with operations were excluded. Published abstracts of RCTs without complete data for analysis were also excluded.

Primary and secondary outcomes

For data synthesis across studies, the primary outcome was the Western Ontario and McMaster Universities Arthritis Index (WOMAC) [36]. Specifically, the WOMAC pain subscores, physical function subscores, and total scores at 3, 6, and 12 months after treatment were recorded. The secondary outcome was the number of patients reporting adverse events.

Search strategy

Two investigators performed a systematic search of Pubmed, Embase, Cochrane library, and Scopus independently on July 15, 2016 and updated on November 15, 2016. The search strategy was as follows: (platelet[text word] OR plasma[text word]) AND (knee[text word] OR tibiofemoral[text word] OR patellofemoral[text word]) AND (*arthritis[text word] OR *arthritic[text word] OR cartilage[text word] OR *arthrosis[text word] OR gonarthrosis[text word]) AND random*[text word]. In Scopus, the search field [text word] was replaced with [TITLE-ABS-KEY]. No language or date exclusions were applied (Additional file 3).

Two investigators reviewed all titles and abstracts to remove duplicates and evaluate the relevance according to the inclusion and exclusion criteria. If ambiguity was encountered, the full-text review was performed. Any discrepancy was resolved through panel discussion with a third investigator. The references of prior systematic reviews were also reviewed to find potential eligible studies.

Data extraction

Two reviewers independently performed data extraction using a pre-developed data extraction table. We extracted the basal characteristics of the included studies to form descriptive summaries. In multi-arm trials including more than one PRP treatment groups, only the group treated with at least twice PRP injections was considered as the intervention group, as the regimen of multiple PRP injections was more common and reported to be more efficacious than a single injection [37, 38]. Although data concerning the patients treated with single-PRP injection in those trials were also extracted, they were not used for quantitative synthesis. The extracted data were checked for consistency, and discrepancies were discussed until a consensus was reached. Personal correspondence was attempted to obtain missing data or clarify ambiguous information.

Quality assessment

Two investigators independently assessed the methodological quality of each eligible study using Review Manager 5.3 (The Cochrane Collaboration, Oxford, England) to determine the risk of bias. The following domains were assessed: random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants (performance bias), blinding of personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias), and other bias. The risk of bias for each domain was graded as either low (+), high (−), or unclear (?) [39]. A trial was regarded as low risk of bias only when all domains were scored as low risk of bias; if 1 or 2 domains were scored as high or moderate risk of bias, the trial was regarded as moderate risk of bias; if more than 2 domains were scored as high or moderate risk of bias, then high risk of bias was considered [21]. Differences were settled by panel discussion with a third investigator.

Data analysis

For the continuous variables, the mean difference (MD) with 95% confidence interval (CI) was used, while the relative risk (RR) with 95% CI was adopted for dichotomous variables to express intervention effects. We assumed the presence of heterogeneity a priori and used the random-effects model in all pooled analysis. The I 2 was used to test heterogeneity. As defined previously, a value less than 40% means the heterogeneity might not be important, whereas the value more than 75% means considerable heterogeneity [39]. To detect the effect of individual studies on the pooled effect, sensitivity analysis was conducted. Publication bias was assessed with a funnel plot if there were at least 10 studies in a comparison [39]. Any p value less than 0.05 was considered to be statistically significant. All analysis was undertaken using Review Manager 5.3.


Study characteristics

In total, 14 RCTs [911, 1315, 17, 2935] were included in the analysis published between 2011 and 2016. Details of the literature search were shown in a flowchart (Fig. 1). Search strategy and study selection process could be found in the Additional file 3.

Fig. 1
figure 1

Study flow diagram

A total of 1423 patients were included for randomization (Table 1). The sample size of PRP group ranged from 12 to 96 patients, whereas that of control groups including HA, placebo, ozone, and corticosteroids, ranged from 11 to 96 participants. WOMAC was the most commonly used efficacy outcome, and 9 studies reported WOMAC (8 studies) [911, 14, 15, 29, 34, 35] or normalized WOMAC (1 study) [13] scores. Follow-up intervals and length were variable among studies. The shortest follow-up was 12 weeks [32] and the longest was 12 months [15, 17, 29, 33, 34]. A summary of PRP intervention effect per study demonstrated comparable efficacy between PRP and HA among 215 patients in 2 studies [17, 32] and superior results in PRP-treated patients compared with control among 1208 patients in the rest 12 studies [911, 1315, 2931, 3335].

Table 1 Basic characteristics of included studies

PRP treatment protocols varied among studies in terms of preparation devices, centrifugations, the use of exogenous activators, and the injection regimen of dose, times, and intervals (Table 2).

Table 2 Details of PRP treatment protocols and control

Among the 14 studies, 2 different radiographic OA grading systems were used: the Kellgren Lawrence grading (0–IV) [40] in 12 studies [9, 10, 14, 15, 17, 2935] and the Ahlbäck scale (I–V) [41] in 2 studies [11, 13] (Table 3). According to the distribution of these cases, most participants receiving PRP treatment were at the early or mid-stage of knee OA.

Table 3 Radiographic OA grading

Risk of bias assessment

A summary of risk of bias assessment of all included studies was illustrated in Fig. 2. Four studies [13, 17, 32, 34] achieved a moderate risk of bias, while the rest 10 [911, 14, 15, 2931, 33, 35] obtained a high risk of bias (Table 1). A detailed justification of the evaluation of each domain of bias was described and provided in the Additional file 4.

Fig. 2
figure 2

Risk of bias summary of all included studies. Methodological quality assessment of each study at 8 domains was illustrated. + means low risk of bias, ? means unclear risk of bias, and − means high risk of bias

Knee pain

At 3 months, 3 studies reported WOMAC pain subscores, and a statistically significant difference was found in favor of PRP treatment compared with control (MD, −3.69 [95% CI, −6.87 to −0.51], I 2 = 94%, p = 0.02). At 6 months, the synthesis of 5 studies demonstrated a statistically significant difference in favor of PRP treatment (MD, −3.82 [95% CI, −6.40 to −1.25], I 2 = 96%, p = 0.004). At 12 months, the pooling results of 4 studies still favored PRP treatment (MD, −3.76 [95% CI, −5.36 to −2.16], I 2 = 86%, p < 0.001) (Fig. 3).

Fig. 3
figure 3

Forest plots investigating the effect of PRP on WOMAC pain subscores at 3, 6 and 12 months compared with control. (IV, inverse variance; M-H, Mantel-Haenszel; CI, confidence interval)

Physical function

At 3 months, 3 studies reported WOMAC physical function subscores, and a statistically significant difference was found in favor of PRP treatment compared with control (MD, −14.24 [95% CI, −23.43 to −5.05], I 2 = 91%, p = 0.002). PRP treatment was also found to improve physical function significantly according to the pooling analysis of 5 studies at 6 months (MD, −13.51 [95% CI, −23.77 to −3.26], I 2 = 97%, p = 0.01) and 4 studies at 12 months (MD, −13.96 [95% CI, −18.64 to −9.28], I 2 = 84%, p < 0.001) (Fig. 4).

Fig. 4
figure 4

Forest plots investigating the effect of PRP on WOMAC physical function subscores at 3, 6, and 12 months compared with control. (IV, inverse variance; M-H, Mantel-Haenszel; CI, confidence interval)

Total WOMAC scores

At 3 months, 6 studies reported total WOMAC scores and a statistically significant difference was found in favor of PRP treatment compared with control (MD, −14.53 [95% CI, −21.97 to −7.09], I 2 = 90%, p < 0.001). PRP treatment was also found to improve total WOMAC scores significantly according to the pooling analysis of 8 studies at 6 months (MD, −18.21 [95% CI, −27.84 to −8.59], I 2 = 97%, p < 0.001) and 4 studies at 12 months (MD, −19.45 [95% CI, −26.09 to −12.82], I 2 = 85%, p < 0.001) (Fig. 5).

Fig. 5
figure 5

Forest plots investigating the effect of PRP on total WOMAC scores at 3, 6, and 12 months compared with control. (IV, inverse variance; M-H, Mantel-Haenszel; CI, confidence interval)

Adverse events

A total of 10 studies [911, 1315, 17, 32, 34, 35] recorded adverse events. Excluding the study by Filardo et al. [17], which reported adverse events in a different form, there was no statistically significant difference in the number of patients with adverse events between PRP and HA among the rest 9 studies (RR, 1.40 [95% CI, 0.80 to 2.45], I 2 = 59%, p = 0.24) (Fig. 6). All adverse events were non-specific, the symptoms including pain, stiffness, syncope, dizziness, headache, nausea, gastritis, sweating, and tachycardia. No severe complications were recorded and all the events were self-resolved in days.

Fig. 6
figure 6

Forest plots comparing the risk of adverse events between PRP and control. (M-H, Mantel-Haenszel; CI, confidence interval)


This systematic review included 14 RCTs and assessed the temporal effect of PRP on knee pain and physical function in the treatment of knee OA compared with other intra-articular injections, including saline, HA, ozone, and corticosteroids. Data synthesis consistently showed intra-articular PRP injections significantly reduced knee pain, improved physical function, and total WOMAC scores compared with control. Such superiority was observed at 3, 6, and 12 months after treatment. However, the risk of adverse events in PRP-treated participants was not significantly increased in comparison with other intra-articular injections.

Although previous systematic reviews concluded that PRP was an effective and safe alternative to treat knee OA, such conclusion was reached on the basis of less than 9 RCTs [1828], and thus the temporal effect of PRP injections on knee pain and physical function was not fully investigated. Chang et al. calculated the effect size of PRP treatment from different outcome measurements at 2, 6, and 12 follow-up, but half of the 16 studies included for analysis were case series, and 5 were RCTs [19]. Another systematic review pooled 6 RCTs and found that PRP obtained significantly better WOMAC total scores than HA from 3 to 12 months post-injection, however, only 2 studies reported WOMAC scores at 3 months and another 2 at 12 months [22]. Laudy et al. specifically evaluated the effect of PRP injections on knee pain and physical function at 6 and 12 months post-treatment [21]. Nonetheless, most comparisons included only 1 or 2 studies due to the small number of RCTs pooled for analysis. Another review included 9 RCTs and synthesized the WOMAC pain subscores and physical function subscores to compare the efficacy of PRP with control [23]. Due to the varied follow-up among studies, synthesis of the data at the latest follow-up might not reflect the changes of PRP efficacy. The strength of this study was to assess the effect of PRP treatment on knee pain and physical function at different time-points post-injection based on a larger number of RCTs.

It remains unclear regarding the duration period of the beneficial effect of PRP injections. Our study found that PRP was superior to other intra-articular injections in terms of pain relief and function improvement through 3 to 12 months. Filardo et al. investigated the persistence of the favorable effect of PRP infiltration during a 24-month follow-up [42]. Results show that all the evaluated parameters were significantly reduced at 24 months compared with those at 12 months, but still better than the baseline before treatment. The median duration of the clinical improvement was 9 months. This may explain why all current RCTs followed participants within 12 months. The short-term efficacy of PRP injections indicates that PRP only temporarily influences the joint milieu, without affecting the joint structure or progression of knee OA.

There are a few limitations in this review. The placebo effect was reportedly substantial in the treatment of knee OA, especially in terms of pain relief and self-reported function improvement [43]. Interventions that are recently “hot” or that were administered through needles, such as intra-articular injections, would result in larger placebo effect [44]. Therefore, blinding of participants is critical to minimize the potential placebo effect. Half of the 14 RCTs in this review were believed to have successfully performed blinding of participants [13, 17, 2935] according to the risk of bias assessment. While 2 more studies [11, 15] stated blinding of participants, the difference in injection times between the intervention and control groups actually made it difficult to perform blinding reliably. So future RCTs should be designed as double-blinding, which ought to be performed successfully during the whole trials. Another limitation is the high heterogeneity among studies, which was also common in previous reviews [1828].


Intra-articular PRP injections probably are more efficacious in the treatment of knee OA in terms of pain relief and self-reported function improvement at 3, 6, and 12 months follow-up, compared with other injections, including saline placebo, HA, ozone, and corticosteroids.