In 1976, Peter C. Nowell (1) laid the foundation for the clonal evolution model of cancer. He hypothesized that tumors arise from a founder cell and that genetic instability gives the tumor the potential to produce multiple “sublines.” In 1977, Fidler et al. (2) inoculated different parts of tumor cells of melanoma cells into mice and produced a variety of lung metastases numbers and proposed that there are distinct “clones” in tumors. In recent years, with the rapid development of next-generation sequencing technology, intratumor heterogeneity (ITH) reflects the subclone information of solid tumors and is used to study tumor metastatic patterns (3–9). In addition, ITH has the potential to be a new clinical prognostic indicator (10, 11). However, the results of previous studies did not provide consistent conclusion regarding the association between ITH and prognosis.

ITH is the genetic diversity reserve of tumors in the face of natural selection, which is largely responsible for the failure of targeted therapies (12). However, a key question is whether this diversity has important impact on a tumor’s natural history or a patient’s outcome (especially those who do not receive targeted therapy). To clarify these issues, we conducted a meta-analysis to explore the relationship between ITH and the prognosis of patients with solid tumors and evaluate the ability of ITH in predicting outcome by stratified confounding factors.

We performed a systematic review and meta-analysis following PRISMA guidelines (registration numbers CRD42020193878 and INPLASY202060107).

We retrieved studies from PubMed, Embase, Cochrane, and Web of Science databases up to October 10, 2020, using ITH and prognosis as the main keywords. We also screened EMSO, ASCO, and WCLC of recent years to avoid missing updated data. The type of study or language of the study was not limited in the study collection.

A total of 17,239 relevant studies were retrieved. After removing duplicates, we screened 483 potentially relevant articles by scanning the titles and abstracts. We reviewed the full text and screened the candidate studies according to the inclusion criteria, and 437 studies with no prognostic information were excluded. Of the remaining 46 studies, 25 were excluded using the exclusion criteria. In the end, 21 studies were included (15–33). These studies contain 38 pieces of comparison information that were extracted. Details of the flow chart for study identification are shown in Figure 1 .

Quality evaluation: The assessment includes study participation, study attrition, prognostic factor measurement, outcome measurement, confounding measurement and account, and analysis. Sixteen studies were not at risk in the quality assessment. Two studies were considered as high risk in the study participation, and four studies did not describe confounding measurement and account. No study had more than two risks ( Supplementary Table S1 ).

Characteristics of all cohorts: The final study population included 9,804 participants. All studies are retrospective studies published between 2013 and 2020 and involved various tumor types ( Supplementary Table S2 ).

This meta-analysis was an exploratory study probing the relationship between ITH and prognosis. Previous findings that examined the relation of ITH on prognosis have been controversial. The same is true for the prognostic implications of ITH for different tumor types. We demonstrated through a meta-analysis that a high ITH was associated with a poorer prognosis in patients with solid tumors. We stratified results by different solid tumors and found that the relation of ITH with prognosis was different for different tumor types. We also stratified results independently according to various confounding factors and found that various confounding factors did not affect the prognostic relation of ITH. Our findings have deepened the understanding of the field and suggest that ITH is a very promising indicator for clinical applications.

Overall, our analysis shows that OS, PFS, and DFS results indicate that the relationship between high ITH and worse prognosis is forceful. Seventeen solid tumors were included in these analyses. ITH was correlated with prognosis in 12 solid tumors. ITH may be the effect of time accumulation (36). Taking COADREAD as an example, the gradual progression from adenoma to adenocarcinoma contributes to the accumulation of ITH (37, 38), and the relation of ITH on prognosis may be more easily reflected in this cancer type. However, ITH was not associated with prognosis in patients with SKCM, KIRC, LUAD, LUSC, and OV in OS. Some tumors metastasize in the early stages (39, 40), and highly malignant tumors may have a shorter evolution time (41). ITH may not have enough time to develop to detect when tumor is diagnosed (41). Besides, because of the limitations of sequencing depth and purity of tumors, a large number of subclones may not be detected (42). In addition, various solid tumors can be further classified by pathological types. Although more research is needed to clarify the heterogeneity between tumor types, the available results are encouraging.

The challenge in ITH research is that there is no unified standard to calculate ITH, which greatly limits the clinical application. However, the vast majority of studies showed that there was no significant difference in the relationship between different ITH assessment methods and prognosis, regardless of single-region sampling or multi-region sampling. For example, in LUAD, different ITH evaluation methods [clone numbers (15), MATH (20), and subclonal populations (23)] show that ITH is not associated with prognosis. This may be because the principle of calculating the ITH is similar, which is based on SNV and CNV, and reflects the numbers of subpopulation in tumors. Of course, efforts have been made to develop more reasonable ITH algorithms, and a promising approach to evaluating ITH has been recently reported (43), Stefan et al. made a contribution to characterize ITH across cancer types based on the PCAWG dataset, including SNVs, indels, SVs, and CNVs, as well as subclonal drivers, subclonal selection, and mutation signatures. The assessment of prognosis based on the new method should be expected.

In the included studies, regardless of whether participants had distant metastasis, most of the samples used for sequencing were taken from the primary lesion. The correlation between ITH and prognosis was confirmed in tumors without distant metastasis. The increased number of subclones is associated with a greater possibility of distant metastasis, recurrence, and drug resistance, thus affecting the prognosis (44). However, for tumors in which distant metastases progress and more mutations appear, the results from primary tumor may not be sufficient. For patients with metastasis at the time of diagnosis, there may be differences in subclone composition between the metastatic lesion and primary lesion (45). Kim et al. proposed that the genetic distance between the metastatic and primary lesion may affect the prognosis (46). However, Reiter et al. (47) analyzed the ITH of distant metastases and found that CRC distant metastases are derived from subclones of the primary tumor. Anyway, so far, the results show that that detection in the primary lesion is sufficient to evaluate the role of ITH in prognosis.

Our results showed that single-site sampling and multi-region sampling were similar in terms of prognosis assessment. There was no evidence to suggest which sampling type was better because the subgroups did not come from the same cohorts. Huang et al. found that deep targeted sequencing of a single tumor specimen seems to be sufficient to evaluate ITH (48). However, tumor subclones have different spatial distribution (49). Zhang et al. reported that single-region sampling only reflects about 76% of the total mutation in multi-region sampling (50). Therefore, at present, when evaluating the ITH, sampling a single region may be sufficient in terms of validity or cost. Of course, if conditions permit, evaluating a diversified ITH provides more complete information.

It should be noted that these condition factors may not be perfectly matched. We analyzed the influence of different staging and sampling methods on the results based on all the information available in the original study, although this may somewhat damage the confidence level. The results presented by the available data are already suggestive; of course, more evidence is needed to verify our conclusions on staging and sampling methods in the future.

(1) Although the number of included cohorts and participants was large, the results did not cover all solid tumor types (2). Multiple cancer types and participant differences in the number of participants between TCGA cohort and small cohort studies lead to heterogeneity and publication bias risk in this study (3). For some tumor types, few cohorts were included, which was not enough to draw a positive or negative conclusion (4). The duration of follow-up varied from study to study; however, a minimum follow-up of 4 years is acceptable (5). Evaluation of ITH still requires a more uniform approach that could be better used in the clinic (6). Multiple confounders were analyzed in our meta-analysis. However, more strictly matching data are needed in the future to verify our conclusions (7). Some cancers (e.g., lung cancer) have shown that ITH is not prognostic and should be given more attention (8). For the moment, widespread clinical application could be a challenge. However, as tumor-based sequencing technologies become more widely available and ITH-related modeling algorithms continue to be developed, this problem is expected to be solved in the future.

High ITH is associated with worse prognosis in many solid tumors in general, although this association was absent for some cancers. ITH is expected to be a promising clinical prognostic factor for the improvement of assessment, treatment, and surveillance strategy.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding author.

TY: conceptualization and writing—original draft. XG: conceptualization and visualization, and writing—review and editing. SZ, ZZ, XZ, and CL: data curation. GL: conceptualization, project administration, supervision, and writing—review and editing. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.