Predictive and Prognostic Value of ALK Gene Rearrangement in Non- Small Cell Lung Cancer

Anaplastic Lymphoma Kinase (ALK) is relatively new in terms of an oncogenic driver and a drug target in Non-Small Cell Lung Cancer (NSCLC); however, little is known about the “natural history” of ALKrearranged NSCLC. Some investigators have speculated that it may represent a more indolent disease [1,2] or be an independent positive prognostic factor [3]. Others have suggested that ALK rearrangement may be a negative prognostic factor when controlling for known factors such as age, sex, smoking status, stage/grade, and histology [4-6]. With the advent of ALK-specific therapies and crossover in clinical trials, it is unlikely that the natural history of ALK-rearranged (ALKpositive) NSCLC can be examined in an unbiased manner moving forward. However, a handful of retrospective studies examining the outcomes with conventional therapy in ALK-positive NSCLC have been published or presented at scientific meetings. Here we review data from these retrospective studies, exclusive of those involving ALK inhibitor therapy, with the goal to evaluate historical survival outcomes and treatment outcomes from chemotherapy, EGFR Tyrosine Kinase Inhibitor (TKI) therapy, surgical therapy, and thoracic radiotherapy in ALK-positive NSCLC.

We searched published literature in English in peer- reviewed journals indexed in Pub Med, Google Scholar and presentations at conferences from July 2007 to Nov 2013 that had an observational study design assessing the predictive and prognostic value of ALK in NSCLC, and that tested for ALK status using various diagnostic tests including fluorescent in situ hybridization (FISH), Immunohistochemistry (IHC), or polymerase chain reaction (PCR). A total of 26 publications were identified and 8 were excluded, where two were reported from the same cohort [7,8]. Five studies reported only the outcome or gave a conclusion but did not have enough study description or data details [3,9-13]. Two studies had no confirmed ALK-negative comparator groups [1,14]. Aggregate data are summarized, comparing survival outcomes between ALK-positive versus ALK-negative NSCLC patients followed by an evaluation of responses with current non-ALK-targeted therapies (Figure 1). Clinical outcomes considered were Overall Survival (OS), Progression-Free Survival (PFS), Recurrence-Free Survival (RFS), Disease-Free Survival (DFS), Time to Progression (TTP), and Objective Response Rate (ORR). Studies were reviewed with a focus on the use of techniques within the study to control by study design and/or adjust with statistical methods for confounding factors that could impact the outcomes being investigated.

Figure 1: Flowchart of analysis.

ALK gene rearrangement as a prognostic biomarker in NSCLC

Summary of studies with no reported control of or adjustment for confounding factors: Six studies have examined OS in ALKpositive compared with ALK-negative, EGFR wild type (WT/WT) cases and two other studies examined OS in ALK-positive versus ALKnegative cases in which EGFR status was unknown. With a median follow-up time of 13 months at the time of analysis, Shaw et al. reported a median OS of 20 months in ALK-positive cases and 16 months in WT/WT cases (p=0.152; Table 1) [15]. In another study, which was an indirect comparison of OS between ALK-positive, crizotinib-naive cases and WT/WT controls, Shaw et al., reported the median OS from time of metastatic diagnosis as similar for the ALK-positive and WT/ WT cases at 20 and 15 months, respectively (unadjusted HR 0.77; p=0.244) [2]. Although an additional OS subset analysis conducted on cases ≤ 60 years old who were never- or light-smokers accounted to some degree for age and smoking status and showed median OS of 20 versus 24 months for ALK-positive and WT/WT cases, respectively (HR=1.01; p=0.978) [2], it was not formally controlled for all potential confounding variables including histology and treatment. With a median follow-up time of 10.8 months, Takeda et al., reported a median OS of 15.7 months in ALK-positive cases and 15.2 months in WT/ WT cases (HR=0.83, p=0.591; Table 1) [16]. Wang et al., reported a median OS of 19.27 months in 9 ALK-positive cases and 18.93 months in 45 WT/WT cases (p=0.481, Table 1) [17]. Martinez et al., reported a median OS of 4.5 months for WT/WT cases (n=65) versus a median OS not reached for 7 ALK-positive cases (p=0.103) and 15.7 months in 13 EGFR mutant cases (p=0.018) [8]. No statistically significant differences were found in median OS between groups in all these studies.

Table 1: Overall Survival.

Hayashi et al., reported a median OS in locally advanced adenocarcinoma patients of 7.7 months in 3 ALK-positive cases and 42.6 months in 23 WT/WT cases (p=0.007; Table 1) [18]. Fukui et al. selected adenocarcinoma cases who underwent pulmonary resection and reported the 5-year OS rate for ALK-positive patients was 81%; whereas, the ALK-negative (EGFR status unknown) was 77% (p=0.76) [19,20].

In early stage lung cancer patients, Paik et al. reported a median OS in stage I-III NSCLC patients of 97.7 months in ALK-positive cases and 78.9 months in ALK-negative (EGFR status unknown) cases (p=0.10) [19].

Summary of studies with reported control of or adjustment for confounding factors: Four studies to date have, a priori, matched or controlled for important independent prognostic factors. Three of them suggest or clearly demonstrate a shorter OS or DFS for ALK-positive versus ALK-negative cases and one stated a prolonged OS in ALKpositive cases. The case-matched analysis by JK Lee et al., reported a median OS in stage IIIb-IV cases of 12.23 months in ALK-positive (n=23), 29.63 months in EGFR mutant (n=46) and 19.33 months in WT/ WT (n=46) cases (p=0.001 versus EGFR mutant; p=0.127 versus WT/ WT) [6]. Yang et al., with selection of never-smoker, adenocarcinoma cases and control for age, sex, stage, and treatment, showed more than a 2-fold greater risk of recurrence or progression within 5 years of diagnosis in ALK-positive (n=22) versus ALK-negative (EGFR status unknown) cases (n=274; p=0.004; Table 1) [4]. In the same study, a higher rate of extra-thoracic metastasis was observed among ALKpositive cases compared with ALK-negative cases, (HR=2.44, p=0.03); albeit, the number of later stage patients in this analysis was limited (n=13). With selection of never-smokers and comparator groups which were balanced in terms of age, sex, histology, stage and performance status (PS), Kim et al., reported a shorter median OS of 14.3 months in ALK-positive cases compared with 33.3 months in ALK-negative, EGFR WT and KRAS WT (triple WT), and 37.2 months in EGFR mutant (ALK-negative) cases (p=0.016 for ALK-positive versus triple WT) [5]. In the same study, in multivariate analysis, ALK-positivity was associated with a lower OS in patients with resected NSCLC (adjusted HR, 4.162; p=0.005). The authors suggested that ALK-positivity may be a negative prognostic factor for early stage NSCLC. Wu et al., examined survival outcomes in lung adenocarcinoma patients with malignant pleural effusions and wild type EGFR in which ALK-positive and ALKnegative comparator groups were balanced in terms of age, sex, smoking history, PS, and treatment, and reported a longer median OS of 14.7 months in ALK-positive cases (n=39) compared with 10.3 months in WT/ WT cases (n=77, HR=0.53, p=0.011) [21]. The authors concluded that ALK translocation is associated with longer overall survival in lung adenocarcinoma EGFR–WT patients.

ALK gene rearrangement as a predictive biomarker

Six studies published to date report response to platinum-based chemotherapy in ALK-positive NSCLC (Table 2) [5,6,15-17,22]. Two of these studies controlled for or matched cases on potential confounding factors, while the other four did not. Four other studies, representing varying degrees of balance or control for confounding factors, reported response to pemetrexed either as a single agent [23,24], or in combination (Table 3) [25,26]. Five studies to date reported the efficacy of EGFR TKI therapy in ALK-positive NSCLC and four of them showed a 0% response rate [5,6,15,22]. Four studies reported survival outcomes in ALK-positive compared with ALK-negative patients who underwent surgical resection [5,19,20,27]. One of these studies balanced on clinically relevant factors [5].

Table 2: Median PFS or TTP with Chemotherapy and EGFR TKI Regimens.

Table 3: Response to pemetrexed.

Summary of studies with no reported control of or adjustment for confounding factors

Analyses by Koh et al. [22] and Shaw et al. [15] showed similar ORR, PFS or TTP on chemotherapy in retrospectively-identified cohorts of patients with ALK-positive versus WT/WT NSCLC cases (Table 2). Koh et al. [22] reported median PFS with first-line platinumbased doublet therapy of 6.2 months in ALK-positive (n=32) versus 7.3 months in WT/WT cases (n=57), with ORR of 18.8% and 40.4%, respectively. Both PFS and ORR were reported as statistically nonsignificant. This analysis was not balanced for age, with patients in the ALK-positive cohort being statistically significantly younger than WT/WT patients (median 49 versus 61 years, respectively; p<0.001). Similarly, Shaw et al., reported a median TTP with first-line platinum-based chemotherapy in the range of 8-10 months for patients with ALKpositive, EGFR mutant, and WT/WT disease [15]. The chemotherapy ORR in this study was 25% for ALK-positive and 35% for WT/WT cases with no statistically significant difference between these groups (p=0.723). This cohort was not balanced for age, sex, smoking history or exposure to ALK inhibitor therapy, which would have had an impact on clinical outcome. In addition to these two studies, Takeda et al reported the median PFS with first-line platinum-based chemotherapy of 6.5 months in 18 ALK-positive versus 4.3 months in 151 WT/WT cases (p=0.437), with an ORR of 44% and 39%, respectively [16]. Both PFS and ORR were reported as statistically non-significant. Similarly, this analysis was not balanced for age, with patients in the ALK-positive cohort being statistically significantly younger than WT/WT patients (median 46 versus 64 years, respectively; p<0.001). The regimens were also not balanced between the comparator groups. The proportion of patients in ALK-positive and WT/WT groups treated with platinum plus pemetrexed were 39% and 17%, respectively (p=0.049). These would have had an impact on clinical outcome. Wang et al reported median PFS with first-line and second-line platinum-based doublet therapy of 8.3 months in ALK-positive versus 4.9 months in WT/ WT cases (p=0.25) [17]. The chemotherapy ORRs were 25.0% and 32.4%, respectively (p=0.762). Both ORR and PFS were not statistically significant. This study was not balanced in age, smoking status and other important prognostic factors between ALK-positive and WT/ WT groups.

Response to pemetrexed-based therapy: Camidg et al., Lee et al. and Lee et al. each retrospectively studied ALK-positive NSCLC response to pemetrexed-based therapy, yielding similar results (Table 3) [23-25]. Lee et al. examined single-agent pemetrexed response in second-line therapy or beyond and reported an overall median TTP of 9.2 months (95% CI 4.65-13.74 months) in ALK-positive (n=15) versus 2.9 months (95% CI of 0.51- 5.28 months) for WT/WT (n=37) and 1.4 months (95% CI of 1.27-1.52 months) in EGFR mutant cases (n=43, p=0.001) [23]. In the treatment-line stratified multivariate analysis, ALK-positivity was an independently significant predictor for TTP (HR=0.44; p=0.005). However, there were imbalances in the ALKpositive versus WT/WT groups in this cohort with respect to: (1) the majority (60%) of ALK-positive cases were in their third line of therapy or beyond compared with 29.7% of WT/WT cases and (2) the median age of ALK-positive cases was the youngest at 52 years. The authors did take the line of therapy into consideration by stratifying TTP by second or ≥ 3 lines of therapy; however, this step reduced the ALK-positive sample sizes to 6 and 9 patients, respectively [23].

Camidge et al. assessed response to pemetrexed as single agent or in combination in metastatic NSCLC across all lines of therapy. A median PFS of 9 months (95% CI 3-12 months) in 19 ALK-positive cases compared with 4 months (95% CI 3-5 months) in 37 triple WT cases was observed [25]. Similar to the analysis of Lee et al., the only significant variable associated with prolonged PFS in a multivariate analysis was ALK-positivity (HR=0.36; p=0.0051). Important imbalances in the comparator groups in this analysis are depicted in Figure 2, Camidge et al. were careful to state that multiple other confounding factors may contribute to the longer PFS observed in the ALK-positive group [25].

Figure 2: Progression-free survival of patients with NSCLC treated with pemetrexed.

Lee et al. examined single-agent pemetrexed response in secondline therapy or beyond and reported a median PFS of 4.0 months (95% CI 2.2-5.8 months) in 32 ALK-positive cases compared with 1.6 months (95% CI 1.0-2.2 months) in WT/WT cases [24]. Compared with the ALK-negative patients, ALK-positive patients were significantly younger (median age, 62 versus 50 years, respectively), with a higher proportion of never or light smokers (32.9% versus 73.9%, respectively; p=0.002). The authors stated that the difference in PFS between the ALK-positive and WT/WT cases should be interpreted cautiously.

Recently, Shaw et al., studied ALK-positive NSCLC response to pemetrexed-based therapy stratified for treatment line, age, smoking status, and different regimen [26]. The authors assessed response to pemetrexed-platinum combination regimen and reported the median PFS of 7.3 months in ALK-positive versus 5.9 months in WT/WT cases (p=0.182). Although an additional PFS subgroup analysis was conducted on first line treatment patients age ≤ 65 years old with never/ light smoking history accounted to some degree for confounders; it was not formally controlled for all potential confounding variables in the same sub group (Table 3). Only in the subset of patients who received pemetrexed–platinum combination as first line treatment, was there a statistically significant difference in median PFS among ALK-positive and WT/WT patients (p=0.018). All other subsets had no statistically significant differences between ALK-positive and WT/WT groups. The authors also examined single-agent pemetrexed or no-platinum combination response and reported a median PFS of 5.5 months in ALK-positive cases compared with 3.9 months in WT/WT cases (p=0.409). An additional PFS subgroup analysis was conducted on second or third line treatment, age ≤ 65 years old, never/light smoking history and pemetrexed alone with never/light smoking history (Table 3). In all the subsets of patients who received single-agent pemetrexed orno-platinum combination, no statistically significant differences were found in median PFS between ALK-positive and WT/WT groups.

Response to EGFR TKI therapy: Koh et al. reported a median PFS with EGFR TKI therapy of 4.3 weeks in ALK-positive (n=16) versus 6.0 weeks in WT/WT (n=29) and 19.6 weeks in EGFR mutant (n=18) cases (p<0.001) [22]. Shaw et al. measured median TTP as 5 months in ALK-positive, 6 months in WT/WT, and 16 months in EGFR mutant cases (p=0.004 for ALK-positive versus EGFR mutant; Table 2) [15]. Wang et al. reported a median PFS of 2.1 months in ALK-positive versus 2.2 months in WT/WT and 8.80 months in EGFR mutant cases (p=0.696 for ALK-positive versus WT/WT; p=0.032 for ALK-positive versus EGFR mutant) [17]. As mentioned above, these three studies had imbalances relative to age and smoking status, which may have confounded median PFS/TTP estimates. Additionally, in the study of Wang et al, there were only 9ALK-positive cases among the patients received EGFR TKI therapy

Outcome from surgical therapy: With selection of surgically resected stage I-III NSCLC patients, Paik et al. reported a median DFS of 76.4 months in ALK-positive and 71.3 months in ALK-negative (EGFR status unknown) cases (p=0.66) [19]. The authors suggested ALK-positivity may not be a predictive factor in the early (surgically resectable) stages of NSCLC. However, this analysis was not balanced for gender, age, smoking status and histology. Fukui et al. selected adenocarcinoma patients with primary lung cancer who underwent pulmonary resection and reported the 5-year DFS rate for ALKpositive patients as 74%; whereas, the DFS rate for ALK-negative (EGFR status unknown) was 68% (p=0.45) [20]. No significant difference was observed between the ALK-positive and ALK-negative groups. The analysis was not balanced for age and smoking history. Zhou et al. analyzed outcomes of patients with NSCLC who underwent radical surgical resection stratified into specific clinical stages [27]. In stage IA, ALK-positive cases had significantly longer DFS than ALK-negative (EGFR status unknown) cases (5-year DFS rate, 100% versus 29.5%, p=0.04). In stage IIIA, ALK-positive patients had poorer DFS than ALKnegative patients (median DFS, 6 months versus 16 months, p=0.0057, Table 4). In a multivariate analysis, the ALK-positivity was the only significant variable associated with poor survival in stage IIIA NSCLC (HR=4.0, p<0.001). Although stratified subset analysis accounted for stage to some degree; it was not controlled for age and gender.

Table 4: Median DFS or RFS with Surgical Therapy.

Response to thoracic radiotherapy: Hayashi et al. reported outcomes of patients with locally advanced adenocarcinoma NSCLC who underwent thoracic radiotherapy (TRT) alone or together with chemotherapy. The median PFS with TRT in 3 ALK-positive cases was 4.2 months compared with 18.6 months in 23 WT/WT and 13.1 months in 11 EGFR mutant cases (p=0.037 for ALK-positive versus WT/WT) [18]. The authors concluded the ALK ALK-positive patients had a poorer outcome after TRT treatment. However, this study was not matched or controlled for important prognostic factors. Additionally, there were only 3 ALK-positive patients who treated with TRT.

Summary of studies with control of or adjustment for confounding factors: In order to account for imbalances in clinical or patient characteristics between patient subgroups, a few investigators matched cases or applied statistical adjustment (control) in their survival analyses, accounting for age, sex, smoking status, histology, and stage of disease (Table 2).

Response to chemotherapy: Lee et al. retrospectively identified stage IIIb-IV cases of non-squamous histology and created a case cohort of ALK-positive cases matched 2:1 to both EGFR mutant and WT/WT cases on age at diagnosis, sex, stage, and smoking status. They observed a median PFS with first-line chemotherapy (none of which included pemetrexed) of 3.87 months in ALK-positive versus 4.93 months in EGFR mutant (p=0.825), and 3.73 months in WT/WT NSCLC cases (p=0.474) [6]. The first-line chemotherapy ORRs in this study were 28.6%, 32.4%, and 35.1% for ALK-positive, EGFR mutant, and WT/ WT groups, respectively (p=0.857 versus EGFR mutant; p=0.695 versus WT/WT).

Kim et al. [5] examined outcomes in NSCLC patients who were never-smokers, controlling for age, sex, histology, and PS in a multivariate analysis. In this analysis, the ORR to first-line platinumbased chemotherapy was 0% for ALK-positive (n=12) and 23% for triple WT cases (n=61). The difference in ORRs between groups was not statistically significant, nor was the median PFS estimate of 5.0 months for ALK-positive and 5.9 months for triple WT cases. The comparator groups, particularly the ALK-positive and triple WT cases, in this study were well-balanced in terms of age, sex, histology, PS, stage, and smoking status. Thus, even though unadjusted statistically, the Kaplan–Meier survival curves in this study represent well-balanced comparisons.

Response to EGFR TKI therapy: The matched case cohort analysis by Lee et al. [6] demonstrated a median PFS of 1.37 months in ALKpositive versus 2.07 months in WT/WT and 9.80 months in EGFR mutant cases (p=0.037 for ALK-positive versus WT/WT; p<0.001 for ALK-positive versus EGFR mutant).In the analysis by Kim et al. [5], median PFS with EGFR TKI therapy in ALK-positive cases was 1.6 months compared with 6.3 months in WT/WT and 12.8 months in EGFR mutant cases (p<0.001 for ALK-positive versus EGFR mutant; p=0.001 for ALK-positive versus WT/WT; Table 2). Thus, these results corroborate the findings of Shaw et al.; Koh et al., and Wang et al., and show worse response to EGFR TKI therapy in ALK-positive NSCLC cases.

Outcome from surgical therapy: Kim et al. examined outcomes in surgically resected NSCLC patients who were never-smokers, controlling for age, sex, histology, stage, and PS in comparator groups [5]. The median RFS with radical surgery in ALK-positive cases was 20.0 months compared with 26.8 months in WT/WT and 39.7 months in EGFR mutant cases (p=0.344). Although in a multivariate analysis, ALK-positivity was associated with a lower OS in patients with resected NSCLC (adjusted HR, 4.162; p=0.005) and the authors suggested that ALK-positivity may be a negative prognostic factor for early stage NSCLC.

Lessons learned from the ALK NSCLC experience emphasize the importance of early molecular and clinical epidemiology research needed to understand the prognostic and predictive value of candidate biomarkers or drug targets. Overall, studies that controlled for potential confounding factors either by study design or in the analyses suggest worse or equivalent prognosis for ALK-positive NSCLC cases. Only one analysis, studied by Wu et al., concluded that ALK rearrangement is a favorable predictive factor for OS in ALK-positive NSCLC [21]. One observation in this study was the percentage of cases who received platinum-based doublet chemotherapy, second- or subsequent-line therapy, pemetrexed treatment and EGFR TKI therapy were all higher in ALK-positive group than in the ALK-negative group even though individually there was no significant treatment difference between ALK-positive cases and ALK-negative cases. These multiple favorable factors, when combined, in the ALK-positive group could skew the overall results and in addition, the absence of significant differences could have possibly been due to too small a sample size.

Both non-balanced and balanced/controlled analyses demonstrated statistically significantly shorter PFS or TTP amongst ALK-positive (including EGFR WT or EGFR unknown but EGFR TKI resistant) cases compared with EGFR mutant and WT/WT cases treated with EGFR TKI therapy. All studies except one reported a 0% ORR to EGFR TKI therapy in ALK-positive cases (Table 2). Although a small total number of patients were studied, these data suggest that ALK is a negative predictive factor for EGFR TKI therapy outcomes in ALKpositive NSCLC. All studies included in this review, regardless of whether they were balanced/controlled or not, suggested that PFS with platinum-based chemotherapy might not be significantly influenced by ALK status in patients with NSCLC. Although in two of the studies the patients received pemetrexed as a platinum partner, the difference between ALK-positive and WT/WT were not found to be statistically significant. Therefore, ALK rearrangements might not be a predictive marker for PFS with conventional platinum-based chemotherapy. It is important to note though, that sample size which could influence statistical significance of comparisons is a limitation in most of these studies.

The analysis by Camidge et al. exploring PFS with pemetrexedbased therapy in groups defined by ALK, EGFR, and KRAS status has important imbalances to consider. Specifically, the ALK-positive and triple WT groups represented considerably different treatment populations, with 63% of the former being 1st-line-treated patients versus 38% of the latter. There were also third-line treatment patients in the triple WT group (16%) but none in the ALK-positive group in addition to differences in smoking status and age. Thus, the median PFS difference observed of 9 months versus 4 months for ALK-positive versus triple WT cases can, at least in part, be explained by a differential prognosis of the two groups based upon the above-mentioned factors. The study by Lee et al.; however, did demonstrate some degree of balance between patient and clinical characteristics for ALK-positive and WT/WT cases treated with single-agent pemetrexed; however, there were still important imbalances for age and line of therapy, with more ALK-positive patients being younger and in a third- or greater line of therapy [23]. With stratifying by important clinical characteristics such as line of therapy, age, smoking status, and different regimen, the study by Shaw et al stated similar PFS between ALK-positive and WT/ WT groups with pemetrexed-based therapy [26]. None of the studies examining response to pemetrexed matched cases or controlled for known confounding factors when comparing the survival curves. Thus, despite the multivariate analyses identifying ALK as the only significant predictive variable in response to pemetrexed, the point estimates of median TTP or PFS and the accompanying Kaplan–Meier curves do not account for these same potentially confounding variables used in the multivariate analyses.

Retrospective studies to date on ALK-positive NSCLC highlight the importance of controlling for already known independent prognostic or predictive factors when trying to evaluate the impact of a new NSCLC biomarker. For ALK-positive NSCLC, opposing conclusions can be drawn from aggregate balanced/controlled versus non-balanced analysis. Smoking status can be an important confounding factor. For example, across several previous studies, the hazard ratio (HR) for survival is consistently lower for never-smokers versus smokers (HR across studies approximately 0.8) [28-32]. Thus, smoking status is a critical factor to consider in analyses assessing the effect of a new biomarker on NSCLC treatment response or survival.

Combining all published analyses of observational data to date, with the understanding of limitations of small sample sizes, degree of control of confounding factors, and retrospective study designs, ALK rearrangement in NSCLC appears to be (1) not predictive of improved outcomes with standard chemotherapy, (2) predictive of poor response to EGFR TKI therapy, and (3) not a favorable prognostic factor in NSCLC [2,4-6,15,22]. In fact, the majority of controlled or case matched analyses suggest that ALK positivity is a negative prognostic factor in NSCLC [4-6]. Thus, the example of ALK translocation serves as a good model of the need for attention to careful control or adjustment for clinically relevant confounding factors when evaluating the prognostic and predictive value of newly identified candidate biomarkers in NSCLC.

Conflict of interest statement

Kimary Kulig is a former employee of Pfizer Inc. Shrividya Iyer is a current employee of Pfizer. Yi Wang and Ping Yang have declared no conflicts of interest.

Editorial assistance was provided by Martin Quinn at ACUMED® (Tytherington, UK) and funded by Pfizer Inc.

We thank Susan Ernst at Mayo Clinic (Rochester, MN, USA) for her technical assistance with the manuscript.

Contributors: Kimary Kulig and Yi Wang contributed the conception, design, literature search, and review. Kimary Kulig, Yi Wang, Shrividya Iyer, and Ping Yang contributed to data analysis interpretation, manuscript writing, and final approval of the manuscript.