Literature search
From 4987 titles retrieved from the databases, 2 from grey literature, and another 10 from reference lists of potential included studies (see Additional file 3), 476 articles were reviewed for titles and abstract after removing duplicates. Following this selection, the Cohen’s kappa was calculated between the two independent authors (VB and MBF) and was of 0.96, indicating excellent agreement between both reviewers. We then identified 26 articles that met the eligibility criteria (10 cohorts, 16 comparative pre and post cohorts, and 0 RCTs). None of studies from the grey literature or reference lists were included.. A PRISMA flow diagram with motives for exclusion of 178 studies is represented in Fig. 1, and details of excluded studies are in Additional file 4. Twenty-six studies that reported outcomes for podiatric interventions in an MDT context were included in this systematic review. Of these, 3 sets of articles were from the same group of authors, [35,36,37,38] and [39, 40]. The decision was made to exclude the oldest ones, based on the fact that the same data set was used. Therefore, 23 studies were included in this systematic review and only 12 for the meta-analysis, considering that 11 studies did not meet the eligibility criteria after full-text reading and analysis. One included cohort study had 4 substudies [41] and another, 2 substudies [38]. Reasons for exclusion (consensus between authors) were: mixed data when reporting primary outcome [42], eminent difference of basic population [43,44,45], podiatric interventions pre and post cohort [46] and incomplete data for pooling the outcomes [40, 47,48,49,50,51].
Description of included articles
Characteristics of the studies included for meta-analysis (n = 12), such as study design and information concerning length of follow-up, setting, source of data, participants, interventions and description of the MDT, comparison, outcomes, and risk stratification are presented in Additional file 5. All 23 studies included in the systematic review were in English. We identified 6 studies from the United States [40, 45, 47, 48, 54, 55], 2 from Canada [44, 55] and 10 from Europe, of which 5 were from the United Kingdom [41, 43, 50, 56, 57], 2 from Spain [36, 38], 1 from Sweden [51], 1 from the Netherlands [58] and 1 from Italy [49]. There were also 3 articles from Asia, of which 2 were from China [42, 59] and 1 from Singapore [60]. One publication was respectively from Australia [61] and another from New Zealand [46]. Publication years were from 1990 to 2019. Four articles were published before 2000 [50,51,52, 57], and 3 articles were from 2000 to 2009 [38, 43, 49], while the majority (16 articles) was published between 2010 and 2019 [36, 40,41,42, 45,46,47, 54,55,56, 58, 60,61,62,63]. Lengths of follow-ups were between 1 and 14 years, with a median of 3.8 years and a mean of 3.6 years. Study settings were mostly in tertiary care [36, 38, 40, 43,44,45,46,47, 50, 54, 55, 57,58,59, 61]. There were 4 studies based in primary care [42, 48, 49, 63], 3 in secondary care settings [51, 56, 60] and 1 unknown [41]. Three articles collected prospective data [38, 50, 57]; all other analyses were carried out using retrospective data (electronic medical records, medical charts, databases with coding). The 12 articles which were combined for meta-analysis accounted for 545,829 patients. The participants’ characteristics at baseline were heterogeneous. According to our stratification system of choice for the population (SIGN) [30], 21 studies had a population stratification categorised as high risk. This is explained by the fact that the population included in the studies could either have a DFU or a history of DFU [45, 47, 50, 55, 57, 58, 60, 61], an amputation or a history of amputation [36, 40, 43,44,45, 48, 51, 61], peripheral vascular disease (PVD) [45, 56], or diabetic foot infection [52, 61, 62]. Stratification of the population with PVD, neuropathy, cellulitis, osteomyelitis or Charcot foot is also categorized as a moderate to high-risk population [41]. Four articles included both categories (high and low risk) [38, 42, 49, 54] and 4 articles had a system of classification of their population or DFUs: surgery classification [47], LEAs risk with King’s classification [60], Wagner’s classification for ulcers [59], and Texas University classification for DFUs [54].
The specific podiatric interventions were all poorly described (without information concerning nature, intensity, duration, frequency) and very heterogeneous. In the 12 included studies, podiatric interventions are stated as contact with podiatry [36, 40,41,42, 45, 49, 51, 55, 58, 59, 61, 63]. Thus, we classified the podiatric interventions as educational strategies [38, 43, 50, 54, 57, 60], foot care strategies [38, 43, 46, 50, 54, 56, 57, 60], offloading strategies [43, 46, 48, 55,56,57], wound care and infection control strategies [44, 48, 54], surgical strategies [44, 47, 54], and stratification [38, 42, 49]. Only a few studies had defined exposure to the interventions as a weekly exposure to podiatry [56, 60], a regular follow-up in podiatry or monthly appointments [38, 43, 50] or at least every 3 months [57]. Concerning the role of the podiatrist, we decided a posteriori to distinguish their role according to their implication in the MDT. With this in mind, the podiatrist intervenes in a primary role in 8 articles (leading role or core of the team) [36, 43, 44, 47, 48, 54, 55, 59]. Specifically, in these articles, the podiatrist formed the core of the team with endocrinologists [36, 59], nurses [43, 55], and vascular surgeons [44, 47, 54]. Podiatrists are sole leaders in one article [48]. In 8 articles, they had a secondary role (support to the MDT but without a leading role) [45, 46, 49, 51, 56, 58, 60, 61] and in 2 articles, they had a tertiary role (external consultation when needed) [38, 42]. Podiatrists’ role was similar to other team members in two articles [50, 57] . Finally, in 3 articles, it was impossible to determine the level of the podiatrist’s implication in the MDT because no description of the team was provided. In one article [52], it was a podiatry-established critical pathway and in the two others, it was with other lower-extremity specialists [40, 41]. The MDTs composition was also variable; some MDTs showed care management in 2 levels of team members’ implication [36, 42, 47, 49]. Finally, funding and conflict of interest in the included articles were clearly mentioned in the full text of 14 out of 23 articles [36, 41,42,43,44, 47, 48, 51, 54, 55, 57, 60,61,62].
Primary outcomes
All the studies included in the meta-analysis (n = 12) reported favourable data for people with diabetes in an MDT management that included podiatry. Therefore, we retrieved data related to our pre-defined outcomes about DFUs and LEAs as stated in Table 1. All included articles had data concerning primary outcomes: LEAs [36, 38, 41, 54,55,56,57,58, 60, 62, 65] and DFUs [38, 54, 55, 57, 58]. With regard to the 11 studies excluded for the meta-analysis, but included in the systematic review (n = 23), 10 out of 11 studies reported data in favour of MDTs including podiatry [40, 42, 44,45,46,47, 49,50,51, 53] and one article reported no effect of the interventions [43]. That led us to conduct two separate meta-analyses based on study design (see Fig. 1). Main results are shown in Fig. 2 from available data pooled together, which respects criteria of heterogeneity. For total LEAs as the primary outcome, the random effect model was applied and a significant result was found in favour of MDTs with podiatry (RR: 0.69, 95% CI 0.54–0.89, I2 = 64%, P = 0.002). For major LEAs (level defined as above knee amputation and/or below knee amputation), results were also in favour of MDTs with podiatry and still significant (RR: 0.45, 95% CI 0.23–0.90, I2 = 67%, P < 0.02). The result was not significant for minor LEAs (level defined as amputations at any level of the foot) (RR: 0.93, 95% CI 0.59–1.40, I2 = 55%, P = 0.76). We succeeded in pooling results from 2 pre and post cohorts’ with cohort study analysis, which increased the number of studies included to 8 for meta-analysis. Raw data from these 2 studies allowed us to calculate the prevalence of LEAs per year per period pre-and post-intervention from a sample size based on government census data in the area. Therefore, events of LEAs from exposed group to MDTs and non-exposed group to MDTs were calculated [36, 56]. For the remaining pre and post cohort (n = 4) [55, 58,59,60], because of the significant heterogeneity between studies, we decided not to pool the data with association measure. Pre and post cohort MDTs have reported significant results in favour of MDTs to improve DFU healing rate [55, 58] and reduce total LEA [58, 59] and major LEA [58,59,60]. Visual inspection of the funnel plot for the included cohort studies for total LEAs has demonstrated no strong evidence of publication bias of the studies in favour of the interventions (Fig. 3). The heterogeneity in DFU data has not allowed meta-analysis for cohort studies.
Secondary outcomes
According to our predefined secondary outcomes (Table 1), data was available for mortality/survival [36, 41, 43, 54, 58], recurrence [43, 54, 57], other complications [54, 61, 64], and healthcare data [49, 54, 56, 57, 59, 61, 63, 64]. Meta-analyses were performed for some studies, but heterogeneity was over 75%. No articles reported data concerning patients’ satisfaction with care provided by MDTs.
Risk of bias assessment of included studies
In relation to the critical appraisal of quality and experimental designs, bias analyses for cohorts have shown that none of the included studies fulfilled all parameters for low risk of bias, but the majority of the studies included (4/6) had a low risk of bias for the following parameters: population, confounders identified, outcomes measured, follow-up time, and appropriate statistical analysis. High risk of bias was present concerning the baseline population (those who were not free of LEAs or DFUs at the beginning of the study) and the exposure (valid and reliable method to measure MDTs contact and intervention) (see Fig. 4a). Bias analysis for pre and post cohorts have also shown the same trend of high risk of bias in included studies. None of the included studies fulfilled all parameters for low risk of bias, but the majority of the studies included (4/6) had a low risk of bias for 2 parameters: outcome measurements and appropriate statistical analysis. In almost all studies, there is confusion about the cause and effect variables (5/6) and difference about follow-up time between pre and post cohorts (4/6). Exposition to intervention was a low risk of bias for only 2 study out of 6. Few studies had a control group (2/6) (see Fig. 4b).