Outcomes of Solid Organ Transplant Recipients With Preexisting Malignancies in Remission

A Systematic Review and Meta-Analysis

Abstract

Background 

Solid organ transplant recipients (SOTR) with a pretransplant malignancy (PTM) are at increased risk for cancer recurrence. However, it is unclear whether differences in survival and incidence of posttransplant de novo malignancies exist between recipients with PTM and those without PTM. We designed a systematic review to synthesize all available evidence assessing these outcomes.

Methods 

A systematic search was performed in MEDLINE, EMBASE, and Cochrane Library to identify studies comparing the following outcomes in SOTR by PTM status: (1) all-cause mortality, (2) cancer-specific mortality, and (3) incidence of posttransplant de novo malignancy. Risk of bias was assessed using the Newcastle-Ottawa Scale.

Results 

Thirty-two cohort studies were included. Recipients with PTM were at increased risk of all-cause mortality compared to recipients without PTM (pooled hazard ratio [HR], 1.51; 95% confidence interval [CI], 1.27-1.81). Similarly, recipients with PTM were 3 times more likely to die of cancer (pooled HR, 3.13; 95% CI, 2.29-4.27). The pooled HR for developing posttransplant de novo malignancy was also increased (HR, 1.92; 95% CI, 1.52-2.42). The association of all-cause mortality and SOTR with PTM did not vary by transplanted organ.

Conclusions 

Pretransplant malignancy is associated with increased risk of all cause-mortality, cancer-specific mortality and of developing de novo malignancies after transplantation compared with those without PTM. These results reaffirm the need for careful selection of transplant recipients with PTM. Tailored screening and management strategies should be developed for this group of patients.

In Brief

This systematic review demonstrates that a history of a pretransplant malignancy is associated with increased risk of all-cause mortality, cancer-specific mortality and of developing de novo cancer after transplantation, and that such a history should be considered carefully in the pretransplant evaluation process and posttransplant immunosuppression choices.

Pretransplant malignancy (PTM) in remission has long been considered a relative contraindication for solid organ transplantation due to the concern that immunosuppression required to prevent graft rejection would allow the growth of dormant malignant cells in these patients.^1-4 Historical data from the Israel Penn International Transplant Tumor Registry, a voluntary registry, found the overall rate of recurrence for these patients after transplant was 21%.⁵ However, more contemporary population-based registries have reported lower recurrence rates.^6,7

Recommendations for the listing of transplant candidates with PTM have centered on the recurrence risk. However, recently published studies report that differences in overall survival and incidence of posttransplant de novo malignancies exist between recipients with PTM and those without a previous cancer diagnosis. An increased incidence of posttransplant de novo malignancies is not unexpected in patients with PTM because patients with a history of cancer have an increased incidence of a second primary when compared with patients without a previous malignancy diagnosis.^8,9

In recent years, cancer has emerged as an important cause of death in solid organ transplant recipients (SOTR) and is anticipated to overtake cardiovascular disease as the leading cause of death in transplant recipients within the next 10 years.^10-12 Additionally, because older patients are now being accepted for transplantation, the number of transplant candidates with preexisting malignancies continues to grow. A more thorough evaluation of the impact of PTM on the outcomes of transplant recipients is therefore timely. We designed a systematic review to synthesize all available evidence assessing the risks of all-cause mortality, cancer-specific mortality, and posttransplant de novo malignancy in SOTR with PTM when compared with recipients without such history.

MATERIALS AND METHODS

Search Strategy

The protocol for this systematic review was prospectively registered with PROSPERO (registration no. 42014009155), and reported using the Preferred Reporting Items for Systematic Reviews and Meta-Analysis statement.¹³ A comprehensive literature search strategy was developed in consultation with senior information specialists to identify eligible studies in MEDLINE, EMBASE, and the Cochrane Library (each from its inception to week 3 of December 2015). The full electronic search strategies are included in Appendix 1, SDC, https://links.lww.com/TP/B260. Thesis dissertations and grey literature were also searched to identify any relevant articles (see Appendix 2, SDC, https://links.lww.com/TP/B260). Additionally, citation tracking from eligible studies was performed to identify additional potential studies. Finally, authors were contacted to obtain missing information.

Inclusion and Exclusion Criteria

Studies were included if they compared SOTR with PTM to transplant recipients without history of malignancy for any of the following outcomes: (1) all-cause mortality; (2) cancer-specific mortality; or (3) incidence of posttransplant de novo malignancies. Studies were limited to cohort studies including at least 5 patients with a PTM and to those published in English, French, Spanish, or Portuguese. Studies describing only patients with active or incidentally diagnosed malignancies at the time of transplantation, and those who underwent transplantation as a treatment for malignancy, such as liver transplantation for the treatment of hepatocellular carcinoma, were excluded.

Study Selection and Data Extraction

Two investigators (S.A., J.H.) independently assessed all titles and abstracts for eligible studies, and then reviewed full-text articles for all included citations. In the case of disagreement, consensus was achieved by discussion with a third adjudicator (C.D.). Data extraction was carried out using electronic standardized data extraction forms. We extracted study-level information including study design, publication characteristics, patient inclusion/exclusion criteria, patient demographics of importance to determine the clinical similarity of populations across studies (age, sex, type of transplant, comorbidities), detailed prior malignancy history information (type of malignancy, time in remission), length of follow-up or person-years at risk, and the outcomes of interest. For studies with overlapping patient populations, we included data from only the most recent or comprehensive study.

Outcomes

Measures of effect for the outcomes of interest included: number of events; risk ratio; odds ratio; incidence rate; and/or hazard ratio (HR) with 95% confidence interval (CI), when available. Hazard ratio, the most appropriate statistic for meta-analysis of time-to-event outcomes, was directly used in the meta-analyses when available; when not available, they were estimated, provided that sufficient statistical information was included (eg, Kaplan-Meier survival curves, P values for log-rank test) using established methods described by Parmar et al.^14-16 For studies that only presented Kaplan-Meier survival curves, survival endpoints were extracted using DigitizeIt visualization software¹⁷ and used to estimate the HR. For the analysis of posttransplant de novo malignancies, only incidence rate ratios per person-years of follow-up and incidence density ratios could be estimated in some cases. Although modeled under different assumptions, incidence rate ratios can be considered approximations of hazard ratios¹⁸ and have been previously included in meta-analyses using HR.¹⁵ For this study, incidence density ratios were considered equivalent to HR, assuming the hazards of posttransplant de novo malignancy were constant over time. Cases per person-years of follow-up were calculated for each study that provided the number of observed cases and the length or person-years of follow-up. When follow-up time for recipients with and without preexisting malignancies was not reported separately, we assumed the follow-up time for both groups was the same. When studies included adjusted and unadjusted estimates, both estimates were collected, as well as the covariates included in the models.

Quality Assessment

The methodological quality of all included studies was assessed independently by 2 authors (S.A., J.H.) using a modified score system based on the Newcastle-Ottawa Scale for Cohort Studies Quality Assessment¹⁹ (see Appendix 3, SDC, https://links.lww.com/TP/B260). Studies with scores of 6 or greater were considered good quality, studies with scores of 4 or greater and less than 6 were considered fair quality, and studies with scores less than 4 were considered poor quality. Reviewers were not blinded to the journal or authors' information. Disagreements between reviewers were resolved by discussion between the 2 reviewers. When no consensus was achieved, a third party resolved the dispute (C.D.).

Data Synthesis and Statistical Analyses

Only studies from which hazard ratios could be estimated were included in the meta-analyses. Studies that included only a particular age group or assessed only 1 type of cancer were excluded from pooled analyses. An approximate standard error of the log rate ratio was calculated using the generic inverse-variance method. A correction of 0.5 was added to each count in the case of zero events. Studies that separately presented results by type of PTM were combined using a fixed-effects model before incorporating the outcomes into the meta-analyses as previously described and reported in the literature.^20-23 Heterogeneity of the data was evaluated visually using forest plots. Between-study heterogeneity was assessed with the Cochran Q test and quantified using the I² statistic.²⁴ Pooled HR were determined using weights from random effects models and reported with their corresponding 95% CI. Analyses were stratified by transplanted organ and type of malignancy when possible. Sensitivity analysis was performed by excluding studies that reported only unadjusted estimates or that used nonpopulation-based cohorts. For meta-analyses including 10 or more studies, publication bias and small studies effect were assessed using funnel plots.²⁵ Statistical analyses were performed using Review Manager 5.0 (Cochrane Collaboration, Copenhagen, Denmark).

RESULTS

Study Selection

The literature search identified 9113 articles collectively from MEDLINE, EMBASE, and Cochrane Central, citation tracking and grey literature search (see Appendix 2, SDC, https://links.lww.com/TP/B260). After excluding duplicates, of the 4482 unique articles identified, 4370 were excluded based on title and abstract screening. A full-text review of the remaining 402 articles led to further exclusion of 369 articles, primarily because they did not compare recipients with PTM to SOTR without a previous cancer diagnosis, they included recipients with incidentally discovered malignancies, or they included recipients whose indication for transplantation was cancer. A total of 33 articles^6,26-57 which examined at least one of the outcomes for patients with PTM remained after the full text review. Figure 1 presents the Preferred Reporting Items for Systematic Reviews and Meta-Analysis flow chart of article selection. Meta-analyses for all-cause mortality, cancer-specific mortality, and posttransplant de novo malignancy were performed on 10, 3, and 7 studies, respectively.

FIGURE 1: Study selection PRISMA flow chart. PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analysis.

Quality Assessment and Publication Bias

The quality assessment scores of the included studies are presented in Table 1. Four of the 33 studies were conference abstracts that provided sufficient information for meta-analysis, but insufficient details for quality assessment, and therefore did not receive a score. Out of the 14 studies that reported all-cause mortality and could be assessed, eleven were of good quality and the remaining three studies were of fair quality. Only 8 of the 14 studies provided estimates adjusted for confounding variables. All 5 studies that reported cancer-specific mortality and could be assessed were of good quality, although 2 did not adjust for confounding variables.^43,44 One conference abstract and 5 studies reported cancer-specific mortality, out of which 3 adjusted for confounding variables.^27,45,58 Of the 14 studies that reported the incidence of posttransplant de novo malignancy and were not conference abstracts, 11 studies were of good quality, 2 studies were of fair quality and 1 of poor quality (not included in the meta-analysis because the HR could not be estimated).⁶ Only 8 of the 14 studies provided estimates adjusted for confounding variables.

TABLE 1: Study quality and risk of bias assessment using the newcastle-ottawa quality assessment scale for all included studies

Publication bias was not assessed for the meta-analyses of incidence of posttransplant de novo malignancy and de novo nonmelanoma skin cancer (NMSC) as both included fewer than 10 studies.²⁵ Although some asymmetry was present in the funnel plot of studies included in the all-cause mortality meta-analysis, the interpretation of this finding is difficult given the small number of studies included. There is a possibility that smaller studies showing no association could have been missed (see Figure S1, SDC, https://links.lww.com/TP/B260).

All-Cause Mortality

A total of 17 studies compared all-cause mortality in SOTR with and without PTM (Table 2). In 13 studies, patients with a single type of transplanted organ were included (kidney, 7; heart, 5); the remainder reported on outcomes for patients with various types of solid organ transplants (N = 4). Studies were performed in North America, Europe, and Asia. Eleven studies were conducted in a population-based setting, while the remaining 6 were single-centered. All studies but one included any malignancy. A study that included only patients with pretransplant melanoma³⁹ and 2 studies that included specific age groups (pediatric heart transplant recipients,³¹ and elderly kidney transplant recipients³⁴) were not included in the meta-analysis. Of the 14 remaining studies, 11 provided HR for long-term all-cause mortality or statistical information that allowed the estimation of HR (Figure 2). The pooled HR for all-cause mortality in SOTR with PTM was 1.51 compared with those without PTM (95% CI, 1.28-1.80; I², 74%). The pooled HR was not significantly different from the overall estimate when we conducted the following sensitivity analyses: (a) by including studies with a reported (not estimated) HR only; (b) including studies that reported adjusted HR only; (c) by restricting the analysis to population-based studies to explore reasons for statistical heterogeneity; or (d) when removing studies with unknown length of follow-up (see Figure S2, SDC, https://links.lww.com/TP/B260). Stratified analyses suggested the observed hazard were similar for kidney (HR, 1.53; 95% CI, 1.22-1.92) and nonkidney (HR, 1.61; 95% CI, 1.17-2.22) transplant recipients (see Figure S3, SDC, https://links.lww.com/TP/B260).

TABLE 2: Characteristics of all included studies that reported all-cause mortality in SOTR with PTM

FIGURE 2: Meta-analysis of all-cause mortality. Pooled hazard ratio of long-term all cause-mortality in solid organ transplant recipients with pretransplant malignancies compared to recipients without pretransplant malignancies using a random effects model. For Santos et al^40,41 A and B refer to 2 different nonoverlapping studies by the same author and published the same year. For B, the number of patients with and without PTM was not presented. IV, inverse variance.

Of the 5 studies not included in the pooled HR calculation, 3 demonstrated no difference,^31,33,35 whereas 2 demonstrated statistically significantly worse overall survival for recipients with PTM.^36,39 In a small study of cardiac transplant recipients, the 2-year overall survival of patients with PTM was comparable to that of patients without a preexisting cancer diagnosis.³³ Similarly, in the 2003 report of the Registry of the International Society for Heart and Lung Transplantation, PTM was not significantly associated with mortality at 1 year and 5 years.³⁵ When looking at specific age groups using the United Network for Organ Sharing database, Shah et al³¹ identified 7169 pediatric heart transplant recipients; 107 patients had PTM. Survival was similar between children with versus without PTM (estimated HR, 0.91; 95% CI, 0.68-1.21). On the contrary, Doyle et al³⁴ evaluated the outcomes of kidney transplant recipients 60 years or older. In a multivariable analysis, pretransplant history of nonskin malignancies was associated with decreased overall survival (HR, 5.0; 95% CI, 1.92-13.04).³⁶ The study by Arron et al³⁹ evaluated the outcomes of 336 SOTR recipients with pretransplant melanoma compared to 191 471 recipients with no prior cancer history using data from the Transplant Cancer Match Study. Recipients with pretransplant melanoma had an increased risk of all-cause death (HR, 1.26; 95% CI, 1.04-1.5).

Cancer-Specific Mortality

Cancer-specific mortality in transplant recipients with PTM was evaluated in 6 studies (Table 3). Of these, 1 study evaluated melanoma-associated mortality only,³⁹ and another only examined NMSC-specific mortality⁴⁴; the other studies evaluated any cancer-related mortality. Three studies included patients undergoing kidney transplants, whereas 3 studies included patients undergoing different solid organ transplants. The studies included populations in North America, Europe, and Australia. One study was single-centered,⁴³ the rest were population-based.^{27,38,39,44,45} Meta-analysis for cancer-specific mortality was performed on 3 studies. The pooled analysis showed that patients with PTM were at greater risk of cancer-specific mortality compared with those without PTM (HR, 3.13; 95% CI, 2.29-4.27; I², 71%) (Figure 3).

TABLE 3: Characteristics of included all studies that reported cancer-specific mortality in SOTR with pretransplant malignancies

FIGURE 3: Meta-analysis of cancer-specific mortality. Pooled hazard ratio for cancer-specific mortality in solid organ transplant recipients with pretransplant malignancies compared to recipients without pretransplant malignancies using a random effects model.

Although all studies included in the meta-analysis reported significantly increased HR for cancer-specific mortality, the heterogeneity of the analysis was moderately high. Of the 3 studies that were not included in the pooled HR, a retrospective study of 334 lung and heart-lung transplant recipients by Metcalfe et al⁴³ reported no difference in cancer-specific mortality between the 21 (6.3%) SOTR with PTM and the 334 SOTR without (estimated odds ratio, 1.57; 95% CI, 0.19-12.82). Two studies evaluated the mortality risk of skin malignancy. Arron et al³⁹ reported that patients with pretransplant melanoma had an increased risk of melanoma-specific death (HR, 27; 95% CI, 11.2-65.2). In contrast, Bavinck et al⁴⁴ reported that 46 kidney recipients with pretransplant NMSC did not experience a greater risk of NMSC-specific mortality.

Incidence of Posttransplant De Novo Malignancy

A total of 16 studies explored the incidence of posttransplant de novo malignancy in patients with PTM (Table 4). The study populations were kidney (N = 6), heart (N = 5), lung (N = 2), liver (N = 1), and various multiple solid organ transplants (N = 3) performed in North America, Europe, and Australia. Eight studies were conducted in a population-based setting and 8 were single-centered. In 7 studies, HR for all posttransplant de novo malignancy in patients with PTM was reported or could be estimated from the available data. In these 7 studies, the pooled estimate showed that SOTR with PTM were more likely to develop posttransplant de novo malignancy compared with those without (HR, 1.92; 95% CI, 1.52-2.42; I², 30%) (Figure 4). Pooled HR did not depend on any single outlying study. Excluding studies that required estimation of an HR did not alter our findings. Moreover, the exclusion of unadjusted estimates did not have a significant effect on the pooled HR (see Figure S4, SDC, https://links.lww.com/TP/B260). Stratified analyses suggested the observed hazard for kidney transplant recipients (HR, 2.03; 95% CI, 1.39-2.98) was similar to nonkidney transplant recipients (HR, 1.86; 95% CI, 1.36–2.55) (see Figure S5, SDC, https://links.lww.com/TP/B260). Of the 3 studies that could not be included in the pooled HR calculation, 2 reported consistent findings with the meta-analysis.^6,31 In a single-center audit of 474 heart transplant recipients, PTM was not associated with increased incidence of posttransplant de novo malignancies in multivariable analyses.⁵¹

TABLE 4: Characteristics of all included studies that reported incidence of posttransplant de novo malignancies in SOTR with PTM

FIGURE 4: Meta-analysis of posttransplant de novo malignancy incidence. Pooled hazard ratio for incidence of posttransplant de novo malignancy in solid organ transplant recipients with pretransplant malignancies compared to recipients without prior malignancies using a random effects model. *Pooled HR from the reported unadjusted hazard ratio for posttransplant nonmelanoma skin cancer and solid malignancies using a fixed effect model as described in the methods. † Incidence rate density was assumed to be equivalent to the hazard ratio.

Several studies explored the effect of PTM on site-specific posttransplant de novo malignancies. Pretransplant malignancies were significantly associated with an increased risk of posttransplant NMSC (pooled HR, 4.64; 95% CI, 2.54-8.49; I², 74%; Figure 5). Sensitivity analysis using the leave-one-study-out technique revealed that the study by Modaresi Esfeh et al significantly changed the pooled estimate (data not shown). Excluding this study reduced the HR and heterogeneity (see Figure S6, SDC, https://links.lww.com/TP/B260). One study evaluated patients with pretransplant melanoma. Arron et al³⁹ reported that 336 recipients with pretransplant melanoma had an increased incidence of melanoma after transplant (HR, 5.38; 95% CI, 2.9-9.8) compared with 191 471 subjects without PTM. The 5-year cumulative incidence of incident primary melanoma was 0.75% in patients with PTM compared with 0.14% in subjects without PTM, an absolute risk difference of 0.61%. Additionally, 1 study evaluated the effect of PTM on the incidence of posttransplant lymphoproliferative disorders (PTLD).⁵³ Data from the United States Renal Data System (USRDS) were linked to health administrative databases to identify 25 127 kidney transplant recipients and their incidence of PTLD. In this study, PTM was an independent risk factor for the development of PTLD (adjusted HR, 3.54; 95% CI, 2.31-5.43), but not for death after PTLD (adjusted HR, 2.04; CI, 0.96-4.3).⁵³

FIGURE 5: Meta-analysis of NMSC posttransplant de novo malignancy incidence. Pooled hazard ratio for incidence of posttransplant de novo nonmelanoma skin cancer in solid organ transplant recipients with pretransplant malignancies compared to recipients without pretransplant malignancies using a random effects model.

DISCUSSION

This systematic review reports on the risk of all-cause mortality, cancer-specific mortality, and transplant de novo malignancy in 10 856, 1544, and 4302 SOTR with PTM, respectively. Our analysis indicated that for patients who received SOTR, PTM is associated with increased risk of all cause-mortality, cancer-specific mortality and of developing de novo malignancies after transplantation, compared with those without PTM. The association of all-cause mortality and SOTR with PTM did not vary by the type of transplanted organ.

The critical shortage of suitable organs for transplantation means that clinicians and policy makers must ensure scarce deceased donor organs are allocated to the patients who would benefit from them the most. Although transplant recipients with PTM experience worse outcomes, increasing the current waiting times between successful cancer treatment and transplantation will probably not improve outcomes in this group of patients. Current recommendations for listing transplant candidates with PTM have been based on the high risk of cancer recurrence reported by the Israel Penn International Transplant Tumor Registry and the natural history of the malignancy.^1,2,5,59 However, more contemporary population-based registries, such as the Australia and New Zealand Dialysis and Transplant Registry and the United Network for Organ Sharing in the United States have reported lower recurrence rates, similar to those observed in the general population.^6,7

Prolonging the requirement for remission times may not improve overall survival. Solid organ transplant recipients with PTM are more likely to receive organs from expanded criteria donors and to have prolonged time on dialysis before transplantation, factors associated with increased risks of cardiovascular mortality and graft failure.^60,61 Importantly, simply prolonging mandatory remission times before transplantation may not have an impact on survival outcomes in patients with PTM. In the study by Brattstrom et al,²⁷ the lower cancer-specific mortality observed in SOTR with longer intervals between cancer and transplantation did not translate into a reduction in all-cause mortality, possibly because an increase in noncancer mortality. Lastly, although prolonging remission times may reduce the risk of cancer recurrence, patients with PTM will remain at increased risk for de novo malignancies.

Although there is solid evidence for the risk of posttransplant de novo malignancy and cancer mortality after transplantation,^{10,12,45,58,62} little is known about the outcomes of transplant recipients with PTM. This study provides a systematic evaluation of the literature, including large numbers of patients with PTM. The findings of this review highlight the need to better understand and mitigate the cancer risk in recipients with PTM and to improve their overall survival. Future research is needed to explore whether the association of all-cause mortality in SOTR with PTM is explained solely by the increased cancer mortality or by other factors, such as cardiovascular disease and graft failure. Other aspects worth investigating include effects of immunosuppression in cancer recurrence, identification of pretransplant factors associated with worse outcomes, the effectiveness of enhanced posttransplant screening strategies, and the use of alternative immunosuppressive regimens in this high-risk group.

As the mean age of transplant candidates rises, the number of SOTR with PTM is expected to increase. Although the proportion of patients with PTM ranges from 0.4% to 5.4% in population-based cohorts, these patients accounted for as many as 7.4% of the SOTR in Ontario in 2010.¹⁰ Our synthesis of the literature has important clinical implications as it provides prognostic information for SOTR with PTM and the knowledge base to improve the outcomes of these patients. Our finding of an increased risk of second malignancies in this group is also important because there are currently no specific recommendations on screening for second malignancies in transplant recipients with PTM. Although not evaluated in this study, modification of immunosuppression regimens may have an impact on cancer outcomes in this population. Our findings highlight the need for further studies in this high-risk population.

Some limitations must be considered when interpreting these findings. Our search strategy was limited to select languages, which may have resulted in a language or cultural bias. Nonetheless, our expansive search across several databases has incorporated numerous studies conducted outside of the English-speaking world.^{27-29,32,37,47,49,51,52} We also acknowledge the presence of heterogeneity between studies reporting all-cause and cancer-specific mortality. Statistical heterogeneity was high between studies examining all-cause mortality; however, we used a random effect model for all our meta-analyses and investigated potential sources of heterogeneity. We explored if the heterogeneity was due to variation in organ types transplanted, methodological quality, or study design (population-based vs. nonpopulation based). These factors did not explain the heterogeneity. The sources of the observed heterogeneity are likely multifactorial. For example, differences in selection of transplant candidates with PTM across jurisdictions (0.4% of kidney transplant recipients had a PTM in the United Kingdom vs 2.4% of in the United States),^6,58 and large variations in the comprehensiveness of risk-adjustment across studies may explain some of the residual heterogeneity. Lastly, there is a possibility that smaller studies showing no association between PTM and all-cause mortality could have been missed. Despite these limitations, the vast majority of point estimates were associated with increased risks for all-cause mortality, cancer-specific mortality, and incidence of posttransplant de novo malignancies. These findings strengthen our conclusions.

In conclusion, our findings indicate that SOTR with PTM have worse outcomes in terms of cancer mortality, all-cause mortality, and posttransplant de novo malignancies than those without PTM. As a risk factor for long-term all-cause mortality, PTM confers a similar risk to that of pretransplant cardiac events in kidney and heart recipients.^63,64 These results reaffirm the need for careful selection of transplant recipients with PTM, tailored screening and management strategies, and justify further research to explore risk factors for adverse outcomes in this population.

ACKNOWLEDGMENTS

The authors would like to thank the Information Specialists, Teruko Kishibe and Christine Neilson, at the Scotia Bank Health Sciences Library, Li Ki Shing Knowledge Institute, St. Michael's Hospital for performing the search presented in this systematic review. We would also like to thank Dr. Sarah Arron (UCSF) for her help in providing data used in this article.