A flow diagram of the study search is shown in Fig. 1. Our search through PubMed identified 462 potentially relevant studies. Of these, 416 studies were excluded by reviewing the title and abstract. The remaining 46 studies were assessed for eligibility by full-text review, and eight were excluded. Finally, 38 studies that met the inclusion criteria were included in this study. The annual trends in the number of studies identified through the database search have increased since 2013. Also, the annual number of included studies that address prognosis prediction of cancer patients by lipidomics using MS analysis has been increasing since 2016 (Fig. 2).

Flow diagram of the study search. Abbreviation: MS, mass spectrometry. *There were studies that meet multiple exclusion criteria

The annual trends in number of studies identified through the database search was on an increasing trend from 2013 (blue line). The annual number of published studies that address prognosis prediction of cancer patients by lipid profiling using mass spectrometry analysis is on an increasing trend from 2016 (orange bar)

*The published articles in 2023 are limited from January to September

The results of the QUADAS-2 assessment of the included studies are shown in Fig. 3. In evaluating the risk of bias (Fig. 3a), 76% of the studies were evaluated to have a ‘high’ risk of bias regarding the ‘patient selection’ mainly attributable to retrospective patient selection and adopted case-control design. 92% of the studies had a ‘high’ risk of bias with respect to the ‘index test’: this is due to the necessity of using diagnostic results based on existing diagnostic methods as the reference standard for evaluating the results of the index test, and impossibility to pre-specify the threshold of the index test due to the prevalence of exploratory studies without validation cohorts. All of the studies had a ‘low’ risk of bias regarding ‘reference standard’ and ‘flow and timing’. In evaluating concerns regarding applicability (Fig. 3b), all of the studies had ‘low’ concerns for ‘patient selection’, ‘index test’, and ‘reference standard’ as these matters were addressed with a study design that matched the review question. The detailed evaluation results are presented in the Supplemental file.

Graph showing the proportion of the included articles with varying risk of bias (a) and applicability (b) of the QUADAS-2 domains

Study characteristics of the 38 included studies are summarized in Table 1. Below, we describe the results for each extracted study characteristic.

The included studies were performed on 16 cancer types. The number of studies reported for each cancer type is as follows (Fig. 4a): there were six studies (16%) each on colorectal [24,25,26,27,28,29] and prostate cancer [30,31,32,33,34,35], five (13%) on breast cancer [36,37,38,39,40], three (8%) each on lung [41,42,43] and ovarian cancer [44,45,46], two (5%) on bladder cancer [47, 48], intrahepatic cholangiocarcinoma [49, 50], pancreatic cancer [51, 52], and renal cell carcinoma [53, 54], one (3%) each from cervical cancer [55], chordoma [56], endometrial cancer [57], liver cancer [58], malignant mesothelioma [59], nasopharyngeal carcinoma [60], and oral cancer [61].

Diagrams showing the proportion of studied cancer types (a), countries (b), sample and MS types (c), and molecular identification methods (d)

* Others included cervical cancer, chordoma, endometrial cancer, liver cancer, malignant mesothelioma, nasopharyngeal carcinoma, and oral cancer; each reported number was one (3%), respectively

** Others included Argentina, Canada, Czech, Finland, Italy, Netherlands, Portugal, Romania, Singapore, and Spain; each reported number was one (3%), respectively

*** Analysis platform was LC-MS.

IC, intrahepatic cholangiocarcinoma; IMS, imaging mass spectrometry; LC-MS, liquid chromatography - mass spectrometry, MS, mass spectrometry; PC, pancreatic cancer; RCC, renal cell carcinoma

The included studies were reported from 16 countries. The number of studies reported for each country is as follows (Fig. 4b): 10 studies (26%) were reported from China [24, 26, 28, 45, 46, 50, 52, 55, 59, 60], eight (21%) from Japan [31, 37,38,39, 41,42,43, 53], four (11%) from the USA [30, 33, 47, 48], two (5%) each from Australia [32, 34], Brazil [29, 61], and Germany [25, 40], one (3%) each from Argentina [54], Canada [57], Czech [51], Finland [44], Italy [56], Netherlands [49], Portugal [36], Romania [27], Singapore [58], and Spain [35].

The sample types used in the studies were tissue (19 cases; 47%) [24, 25, 30, 31, 36,37,38, 41,42,43, 47,48,49,50, 53, 56, 58, 60, 61], serum (20 cases; 50%) [26,27,28,29, 32,33,34, 39, 40, 44,45,46, 51, 52, 54, 55, 57,58,59, 61], and urine (one case; 3%) [35]. In two cases of serum and one urine, extracellular vesicle (EV) was used as the measurement materials [35,