Population Level Contribution of Clonal Hematopoiesis of Indeterminate Potential to Myeloid Neoplasm Development: Pooled Analysis and Population Attributable Risk Percent Determination from Case-Control and Cohort Studies

Background. Clonal hematopoiesis of indeterminate potential (CHIP) represents a precursor state of myeloid leukemogenesis. Several studies thus far have convincingly demonstrated that presence of CHIP is associated with increased risks of myeloid neoplasm (MN) development. These findings have important intrinsic utility; however, there has been lack of attempts to translate this knowledge into estimates of disease burden suitable for determining its potential at population level. Therefore, the aim of this analysis was to estimate the population attributable fraction (PAF) for MNs associated with CHIP using risk ratio estimates derived from pooled analysis of observational studies.

Methods. Eligible studies were case-control (n=3) and cohort (n=3) studies with dichotomous outcome (MN or no MN) and the presence or absence of the risk factor (CHIP). Mantel-Haenszel method for calculating the weighted pooled odds ratio (OR) under the fixed effects model was used, using software RevMan 5.3. OR being the key statistic for case-control studies was the preferred risk estimate to measure association between CHIP and MN in the selected case-control studies. We used OR as risk estimate for cohort studies as well, since MN occurred in less than 10% of the CHIP negative population, and therefore, OR provided a reasonable approximation of the relative risk. From the pooled OR, estimate of population attributable risk% (PAR%) was calculated using the formula: 100*(Px*(OR-1))/(1+(Px*(OR-1))), where Px represented estimate of population exposure (proportion of CHIP positive cases in the population).

Results. 23,983 participants with 345 MN cases in six studies were included in the pooled analyses (Figure 1). In the CHIP positive population, there were total 230 MN cases of the total 2304 participants. There were 115 MN cases in the CHIP negative population (n= 21,679). For all studies, the pooled OR for those with CHIP compared with those without was 5.45 (95% CI, 4.06-7.30) with low overall heterogeneity (I²=2%, Fig. 1). Estimate of population exposure (Px) was 0.088 (2074/23638). PAR% calculated using formula 100*(Px*(OR-1))/(1+(Px*(OR-1))), was 28.14%, suggesting that of all the MN cases, including those with or without CHIP, 28.14% could be attributed to CHIP.

Conclusion. CHIP’s strong association with myeloid neoplasm development was confirmed in this pooled analysis. Individuals with MN were 5.45 times more likely to be CHIP positive than those without MN. Further, in this analysis, we estimate population level contribution of CHIP to MN development. We estimate that 28% of MNs were attributable to positive CHIP status, thus representing 28% of the cases in the total population that can be prevented by intervening against the exposure- CHIP. Determining PAR% is an especially useful and often underutilized first step in designing public health interventions directed against harmful exposures. These findings also brings forth an important consideration, that CHIP though an important risk factor is not obligatory for MN development. Majority of MNs can not attributed to positive CHIP status. Therefore, importance of traditionally known risk factors such as smoking, chemotherapy and radiation exposure need appropriate quantification, especially as they can be entirely preventable and are established risk factors for CHIP positivity as well.