Page 126 - 2021_02-Haematologica-web

T. Raskovalova et al.
activity and CCL18 concentration and derived geometric mean ratios along with 95% Confidence Intervals (CI), in order to account for skewed distributions.24,25 The comparative accuracy of CCL18 concentration relative to chitotriosidase activity in dis- criminating patients with the outcomes of interest was quantified by the difference in the areas under the receiver operating charac- teristic (AUC-ROC) curves. Data synthesis was performed with one- and two-stage approaches.26,27
Results
Study selection and obtained IPD
Overall, 2,636 records were identified by database searching (Figure 1). Two additional records were found by contacting field experts and searching clinical trial reg- istries, respectively. After removing 13 duplicates, the titles and abstracts of 2,625 records were screened for eli- gibility. Of these, 318 records were identified as being potentially relevant and full-text articles were retrieved for a more thorough review. After excluding 295 records based on the full-text article, our systematic review included 23 primary studies from which IPD were sought.
IPD were available only for nine of these studies totaling 463 patients with type I GD (Figure 1). The sponsor of four trials of ERT with velaglucerase28-31 and the principal investigators of three observational studies17,32,33 supplied computerized IPD upon request. The IPD of the other two randomized controlled trials on ERT with taliglucerase were extracted from the published articles34,35 (Online Supplementary Appendix S4). Our meta-analytical sample consisted of 1,109 observations from 334 patients with type I GD, after excluding 111 patients with undocumen- ted values for chitotriosidase activity and/or serum CCL18 concentration, and 18 patients with deficient chitotriosi- dase activity (Figure 1). The median number of observa- tions per patient was 2 (range: 1-14).
IPD were not available for the other 14 eligible primary studies that included 11 academic observational stu- dies8,9,14,36-43 and three industry-sponsored clinical trials of ERT or substrate reduction therapy44-46 (Online Supplementary Appendix S5). Overall, these studies enrolled 565 patients with GD, although overlap in study popula- tions could not be excluded for five studies carried out in the Netherlands8,9,14,37,42 (201 patients) and four studies per- formed in Spain36,39,40,41 (203 patients).
Study characteristics
Among the nine studies that supplied IPD, six were international industry-funded clinical trials of ERT28-31,34,35 and three were observational in design17,32,33 (Online Supplementary Appendix S4). The study enrollment periods extended from 2003 to 2015, and the length of the follow- up ranged from 12 to 132 months for longitudinal studies. The median number of primary study participants and observations contributing to the meta-analysis was 32 (range: 9 -98) and 117 (range: 20-224), respectively. All but one study recruited adult or mixed populations. The exception was a clinical trial performed in a pediatric set- ting.35 Children younger than 16 years of age accounted for 13% of all participants.
Chitotriosidase activity was measured using the 4MU- deoxy-chitobiose28-31,33,35 or 4MU-chitotriose17,32 fluorogenic substrates, in compliance with the published methods. Serum CCL18 concentration was assayed using DELFIA or
ELISA in five and three studies, respectively (Online Supplementary Appendix S6). Information on chitotriosidase activity and serum CCL18 assays was not available for one study.34 The median values ranged from 1,340-14,809 nmol/mL/h for chitotriosidase activity, and from 237-1,113 ng/mL for serum CCL18 concentration. In five clinical tri- als,28-31,35 chitotriosidase activity and serum CCL18 concen- tration were both measured at a single central core labora- tory (i.e., the Academic Medical Center in Amsterdam, the Netherlands). Liver and spleen volumes were quantified using magnetic resonance imaging in eight studies, assessed by independent blinded reviewers in five studies, and undocumented in one study. The median values in pri- mary studies ranged from 0.8-1.7 MN for liver volume, 2.7- 14.1 MN for spleen volume, 11.7-14.0 g/dL for hemoglobin concentration, and 82-260 x109/L for platelet count (Online Supplementary Appendix S6).
Study quality assessment
Six of the eight studies that evaluated the primary com- posite outcome fulfilled five or more QUADAS-2 tool cri- teria (Online Supplementary Appendix S7). The other two studies met four QUADAS-2 tool criteria. As data were collected by retrospective chart review, the Yale’s National Gaucher Disease Treatment Center study33 was consid- ered at high risk of bias for index tests and pre-specified surrogate outcome assessment. This study was also con- sidered at high risk of flow bias due to undocumented chi- totriosidase activity or serum CCL18 concentration in 68% (113 of 167) of participants. In a randomized con- trolled trial that enrolled only treatment-naive patients, the applicability concern was high and it was not possible to formally determine whether biomaker assessment was blinded to the pre-specified outcomes and which analyti- cal methods were used.34
Chitotriosidase activity and serum CCL18 concentration according to the outcomes
In one-stage meta-analysis involving 492 observations nested within 177 participants from eight primary studies (Figure 1), the primary composite outcome was associated with increased geometric mean of chitotriosidase activity (7,623 vs. 1,478 nmol/mL/h; geometric mean ratio: 5.29, 95% CI: 4.24-6.61, P<0.001) and CCL18 concentration (679 vs. 198 ng/mL; geometric mean ratio: 3.04, 95% CI: 2.57-3.58, P<0.001) compared with the absence of out- come (Table 1). Despite substantial between-study hetero- geneity, the two-stage meta-analysis yielded comparable results, using random-effect models (Online Supplementary Appendices S8-9).
The effect size was quite homogeneous among the indi- vidual components of the primary composite outcome, with geometric mean ratio point estimates ranging from 2.96-5.43 for chitotriosidase activity and from 2.28-3.22 for serum CCL18 concentration (Table 1). Similar results were obtained for the secondary composite outcome and its individual components (Table 1), except for serum CCL18 concentration, which did not differ according to severe anemia (445 vs. 666 ng/mL, geometric mean ratio, 1.39, 95% CI: 0.82-2.37, P=0.22).
The geometric means of chitotriosidase activity (3,556 vs. 1,618 nmol/mL/h; geometric mean ratio: 1.47, 95% CI: 0.78-2.79, P=0.24) and serum CCL18 concentration (786 vs. 449 ng/mL; geometric mean ratio: 1.22, 95% CI: 0.81- 1.81, P=0.34) did not differ according to symptomatic
440
haematologica | 2021; 106(2)