We identified a total of 4,479 records from seven databases. We retrieved 97 of them for full-text screening and 17 studies met the inclusion criteria. Fifteen studies were used for our quantitative analysis [7, 9,10,11,12,13,14,15,16,17,18,19,20,21,22]. The database search for the qualitative studies identified three papers for full-text review (n = 2 qualitative studies [48, 49] and n = 1 the qualitative component of a mixed methods study [21].

Most commonly, studies were excluded due to being the wrong publication type (i.e., thesis, editorals, secondary studies (such as reviews and meta-analysis), conference papers, discussion papers with no original data, and grey literature) and did not meet the definition of a VLCD or LCD, or failed to report both pre and post intervention results relevant to this review (i.e., HbA1c). Inter-rater reliability agreement using Cohen’s kappa (k) resulted in almost perfect agreement between reviewers (k = 0.81 (95%CI 0.67 to 0.95), P < 0.05. The PRISMA flow diagram is shown in Fig. 1.

Included quantitative studies were published from 1992 to 2023 [19, 22] and were conducted in the USA (n = 3), [9, 14, 17], UK (n = 1), [10] Europe (n = 6), [7, 11,12,13, 16, 20] Australia (n = 4), [15, 19, 21, 22], and New Zealand (n = 1) [18]. There were three RCTs, [7, 16, 18], three quasi-experimental [12, 19, 20], three case series (studies that grouped together similar case studies/reports) [14, 17, 22], five case reports (studies that included one participant) [9,10,11, 13, 15], and one mixed methods study [21]. The sample sizes in the quantitative studies varied from one to 48 adults [9,10,11, 13, 15, 17, 20], with a total of 188 participants included. Participants mean age was 38 years (20–52 years), with a diabetes duration of 21 years (14–41 years), and a baseline mean HbA1c of 8.6% (6.4–16.8%) [70 mmol/mol (46-160 mmol/mol)]. Diet intervention duration ranged from one week to four years and ten months [15, 16]. Dietary intervention characteristics included nine studies which examined a VLCD [9,10,11,12,13,14,15,16,17] and six LCD [7, 18,19,20,21,22]. The study baseline characteristics of included quantitative studies are shown in Table 1.

Macronutrient distribution (i.e., carbohydrate, protein, and fat) varied across the studies. Dietary support was provided to participants by either nutrition education, meal plans, or food supplies (i.e., nine out of seventeen studies) [7, 14, 16,17,18,19,20,21,22] (Additional file 5).

In addition, the included qualitative studies were published from 2015 to 2022 [21, 48, 49]. The three studies (n = 2 qualitative studies [48, 49] and n = 1 mixed methods [21]) analysed data thematically. The qualitative studies were conducted in Canada (n = 1) [48], and New Zealand (n = 1) [49]. The mixed methods study was undertaken in Australia [21]. The sample of the qualitative studies included a total of 33 adults ranging from three to 22 participants per study [21, 48, 49]. The demographic data for two of the three studies was incomplete, preventing the aggregate reporting of participant mean age, diabetes duration, and baseline HbA1c [48, 49]. The qualitative studies (n = 2) included one that explored the use of a VLCD [48] and one a LCD [49]. The mixed methods study examined a LCD approach [21]. These study characteristics are presented in Table 2.

Using the JBI critical appraisal checklists, Additional File 6 provides a summary of the critical evaluation and assessment of the methodological quality of the included studies. RCTs scores ranged from eight to nine out of ten, quasi-experimental studies scores ranged from six and a half to seven and a half out of ten, case reports scores ranged from six and half to eight out of eight, case series scores ranged from nine to ten out of ten, and qualitative studies score ranged from seven to seven and a half out of ten. These scores reflect how well each study has addressed the possibility of bias in its design, conduct and analysis [43]. The greater the score, the better the methodological quality [43].

Examination of the primary outcome HbA1c for the quantitative studies showed an average mean difference (improvement) pre- to post-intervention for VLCD and LCD studies that was 2.9% [11,12,13,14,15,16,17, 19, 20] and 0.4% [9, 10, 18,