Meta-analysis and retrospective pharmacovigilance study of MDS and AML in patients receiving PARP inhibitor treatment

Poly(ADP-ribose) polymerase (PARP) inhibitors have led to improvements in progression-free survival in a range of neoplasms, particularly in ovarian cancers. Fatigue, and hematological and gastrointestinal toxicities are the most commonly reported adverse events (AEs) in randomized controlled trials (RCTs) of PARP inhibitors, however there is concern that patients may also be at increased risk of developing secondary myelodysplastic syndromes (MDS) and acute myeloid leukemia (AML).

Pierre-Marie Morice and colleagues conducted a systematic review and safety meta-analysis (PROSPERO CRD42020175050) to estimate the risk of MDS and AML as delayed AEs related to PARP inhibition in RCTs in patients with cancer. Additionally, they performed an observational, retrospective study of the World Health Organization (WHO) pharmacovigilance database (NCT04326023) to identify cases of MDS and AML in patients treated with PARP inhibitors and identify relevant clinical features. Their findings, which were recently published in Lancet Hematology,¹ are summarized below.

Study design

Safety meta-analysis

The study systematically reviewed RCTs up to May 31, 2020 from MEDLINE, the Cochrane Central Register of Controlled Trials, and the ClinicalTrials.gov registry. The primary outcome was the summary risk of MDS and AML related to PARP inhibition versus placebo in RCTs in patients with cancer. Secondary outcomes were:

• Summary incidence of MDS and AML cases related to PARP inhibition or control treatment in placebo RCTs, non-placebo RCTs, and all RCTs (placebo and non-placebo)
• Summary risk of MDS and AML related to PARP inhibition versus all control treatments (placebo and non-placebo) in RCTs

Retrospective analysis of WHO database

Cases of MDS and AML suspected to be caused by PARP inhibitors were extracted from VigiBase, the WHO’s pharmacovigilance database, on May 3, 2020. Clinical features were collected, including:

• Duration of exposure to PARP inhibitor treatment
• Latency period between first drug exposure and diagnosis
• Outcome, particularly cases resulting in death

Results

Safety meta-analysis

There were 1,617 citations initially identified, of which 31 studies fulfilled the inclusion criteria and 28 of these had available AE data. Study characteristics are given in Table 1. Eighteen studies were placebo and ten were non-placebo RCTs, giving a total of 9,099 enrolled patients and comprising:

• 5,693 patients (62.6%) with exposure to PARP inhibitors
• 3,406 (37.4%) patients in control groups

Table 1. Characteristics of the 28 studies analysed in the safety meta-analysis¹

PARP, poly(ADP-ribose) polymerase.
Characteristic	Studies (N = 28)
Indication being investigated, %
Breast cancer	17.9
Colorectal cancer	3.6
Gastric cancer	7.1
Lung cancer	10.7
Melanoma	3.6
Ovarian cancer	42.9
Pancreatic cancer	7.1
Prostate cancer	7.1
PARP inhibitor used, %
Niraparib	7.1
Olaparib	46.4
Rucaparib	3.6
Talazoparib	3.6
Veliparib	39.2
Range of median follow-up, months	3.8–78.0

Of the 7,307 patients in the 18 placebo RCTs, PARP inhibitors were found to significantly increase the risk of MDS and AML compared with placebo, as shown in Table 2. All cases of MDS and AML were reported in patients with ovarian cancer. The median latency period was 20.3 months (IQR, 18.7–22.1). Furthermore, across all placebo and non-placebo RCTs, the risk of MDS and AML was significantly increased in patients exposed to PARP inhibitors compared with control treatments. No heterogeneity was detected between studies.

Table 2. Analysis of the risk and incidence of MDS and AML compared with placebo and non-placebo in RCTs of PARP inhibitors¹

AML, acute myeloid leukemia; CI, confidence interval; MDS, myelodyplastic syndromes; OR, odds ratio; PARP, poly(ADP-ribose) polymerase; PARPi, PARP inhibitors; RCT, randomized controlled trial. *Fixed-effects meta-analysis used to calculate Peto OR with 95% CI; †Between-study heterogeneity assessed using inconsistency index Ι2 statistic and Χ2 test.
Characteristic	Peto OR (95% CI)*	p value	Incidence, % (95% CI)	Ι2 (%)†	Χ2 p value†	Events/patients
MDS/AML risk with PARPi vs placebo	2.63 (1.13–6.14)	0.026	—	0	0.91	—
Incidence of MDS/AML related to PARP inhibitors
Placebo RCTs	—	—	0.73 (0.50–1.07)	0	0.87	21/4533
Non-placebo RCTs	—	—	1.22 (0.62–2.37)	0	0.59	5/1160
All RCTs	—	—	0.83 (0.59–1.15)	0	0.84	26/5693
MDS/AML risk with PARPi vs all controls	2.25 (1.07–4.75)	0.033	—	—	—	—
Incidence of MDS/AML related to controls
Placebo RCTs	—	—	0.47 (0.26–0.85)	0	1.00	3/2774
Non-placebo RCTs	—	—	1.21 (0.54–2.67)	0	0.90	2/632
All RCTs	—	—	0.66 (0.41–1.05)	0	1.00	5/3406

Subgroup analyses did not reveal any significant differences regarding PARP inhibitor used, PARP inhibitor treatment duration, presence of BRCA1/2 mutations, previous systemic therapy, or treatment setting.

Retrospective analysis of WHO database

In VigiBase, 178 cases of MDS (n = 99) and AML (n = 79) related to PARP inhibitor therapy were identified. Patient characteristics are shown in Table 3. The majority of cases were in RCTs in ovarian cancer, and the latency period from first PARP inhibitor exposure to MDS or AML diagnosis was similar to that found in the safety meta-analysis. Progression of MDS to AML was reported in 14 patients (8%).

PARP inhibitor exposure data was available for 96 cases, with a median treatment duration of 9.8 months (IQR, 3.6–17.4). This was similar between patients who developed MDS (9.8 months; IQR, 3.6–17.4; n = 56) and AML (9.4 months; IQR, 3.4–19.6; n = 40). Median follow-up after MDS or AML diagnosis was 5.6 months (IQR, 3.2–9.5). Of the 104 cases with available outcome data, there were 47 (45%) deaths; 20 (37%) in patients who developed MDS and 27 (54%) in those with AML.

Table 3. Patient characteristics in RCTs extracted from VigiBase (adapted from Morice et al.¹)

AML, acute myeloid leukemia; IQR, interquartile range; MDS, myelodyplastic syndromes; PARP, poly(ADP-ribose) polymerase.
Characteristic	All cases (N = 178)	MDS (n = 99)	AML (n = 79)
Male, %	7	0	15
Median age at onset, years (IQR)	64 (58–69)	62 (56–71)	66 (60–69)
Indication, %
Breast cancer	5	6	3
Ovarian cancer	85	90	79
Pancreatic cancer	2	3	2
Prostate cancer	7	0	16
Vulval cancer	1	1	0
PARP inhibitor, %
Niraparib	18	18	18
Olaparib	75	76	73
Rucaparib	6	4	8
Talazoparib	1	1	0
Veliparib	1	1	1
Median latency period, months (IQR)	17.8 (8.4–29.2)	17.8 (8.6–27.9)	20.6 (8.4–29.7)

Conclusion

Across the 28 RCTs included in this meta-analysis, PARP inhibitor therapy significantly increased the risk of secondary MDS and AML compared with placebo treatment. All MDS and AML cases were reported in RCTs in ovarian cancers, which the authors suggested may be due to the longer follow-up in these studies.

From the WHO’s real-world pharmacovigilance database, it was found that MDS and AML occurred several months after PARP inhibitor exposure. Whilst this latency is shorter than that described following conventional chemotherapy,² there was a high proportion of deaths. As such, the authors recommended that further studies were needed to improve clinical understanding of these often lethal adverse events and highlighted the need for clinicians to be vigilant in monitoring delayed hematologic toxicities in patients treated with PARP inhibitors.

A number of limitations were identified by the authors, including the short follow-up in some RCTs, discrepancies between the different sources used, and under-reporting of AEs in the WHO database.