scholarly journals Addressing the issue of bias in observational studies: Using instrumental variables and a quasi-randomization trial in an ESME research project

PLoS ONE ◽  
2021 ◽  
Vol 16 (9) ◽  
pp. e0255017
Author(s):  
Monia Ezzalfani ◽  
Raphaël Porcher ◽  
Alexia Savignoni ◽  
Suzette Delaloge ◽  
Thomas Filleron ◽  
...  

Purpose Observational studies using routinely collected data are faced with a number of potential shortcomings that can bias their results. Many methods rely on controlling for measured and unmeasured confounders. In this work, we investigate the use of instrumental variables (IV) and quasi-trial analysis to control for unmeasured confounders in the context of a study based on the retrospective Epidemiological Strategy and Medical Economics (ESME) database, which compared overall survival (OS) with paclitaxel plus bevacizumab or paclitaxel alone as first-line treatment in patients with HER2-negative metastatic breast cancer (MBC). Patients and methods Causal interpretations and estimates can be made from observation data using IV and quasi-trial analysis. Quasi-trial analysis has the same conceptual basis as IV, however, instead of using IV in the analysis, a “superficial” or “pseudo” randomized trial is used in a Cox model. For instance, in a multicenter trial, instead of using the treatment variable, quasi-trial analysis can consider the treatment preference in each center, which can be informative, and then comparisons of results between centers or clinicians can be informative. Results In the original analysis, the OS adjusted for major factors was significantly longer with paclitaxel and bevacizumab than with paclitaxel alone. Using the center-treatment preference as an instrument yielded to concordant results. For the quasi-trial analysis, a Cox model was used, adjusted on all factors initially used. The results consolidate those obtained with a conventional multivariate Cox model. Conclusion Unmeasured confounding is a major concern in observational studies, and IV or quasi-trial analysis can be helpful to complement analysis of studies of this nature.

Biostatistics ◽  
2017 ◽  
Vol 20 (1) ◽  
pp. 65-79 ◽  
Author(s):  
Torben Martinussen ◽  
Ditte Nørbo Sørensen ◽  
Stijn Vansteelandt

2002 ◽  
Vol 17 (4) ◽  
pp. 231-238 ◽  
Author(s):  
B. De La Lande ◽  
K. Hacene ◽  
J.-L. Floiras ◽  
N. Alatrakchi ◽  
M.-F. Pichon

Up to 80% of breast cancer patients developing metastases have high levels of CA 15.3. We studied the prognostic implications of CA 15.3 kinetics in 119 patients before and at first metastasis by univariate and multivariate statistics. At first metastasis, CA 15.3 was elevated in 82.4% of patients, with a lead time (median 162 days) in 42.0% of them. Kaplan-Meier analysis showed overall survival (median 1477 days) to be significantly related to estrogen receptor (ER) and progesterone receptor (PgR) status (p=0.0001) and tumor size (p=0.025). The interval between diagnosis and first abnormal CA 15.3 (p=0.0001), the CA 15.3 concentration (p=0.013), and the presence or absence of a lead time (p=0.001) also had prognostic value. ER and PgR status (p=0.0005 and p=0.0103, respectively), metastasis-free interval (p=0.0003), existence of a CA 15.3 lead time (p=0.0028), and days from diagnosis to first abnormal CA 15.3 (p=0.0055) entered in the Cox model. After first metastasis (median survival 573 days), ER and PgR status (p=0.0001 and p=0.0004, respectively), existence of a lead time for CA 15.3 (p=0.0138), and the concentration of first abnormal CA 15.3 (p=0.0145) had individual prognostic value. In the Cox model ER status (p=0.0001), nodal status (p=0.0191), existence of a lead time for CA 15.3 (p=0.0033), days from diagnosis to first abnormal CA 15.3 (p=0.0132), and concentration of first abnormal CA 15.3 (p=0.0320) were found to be independent prognostic variables. Compared to a matched historical control group that was not monitored by CA 15.3 assaying (n=140), the study group had a significantly longer survival after the first metastasis (p=0.0005). In conclusion, the kinetics of CA 15.3 before the first metastasis is of prognostic value. When associated with 18-fluorodeoxyglucose imaging, serial CA 15.3 assays may help to implement early treatment of metastases.


Blood ◽  
1998 ◽  
Vol 92 (8) ◽  
pp. 2712-2718 ◽  
Author(s):  
S. De Botton ◽  
H. Dombret ◽  
M. Sanz ◽  
J. San Miguel ◽  
D. Caillot ◽  
...  

All trans-retinoic acid (ATRA) syndrome is a life-threatening complication of uncertain pathogenesis that can occur during the treatment of acute promyelocytic leukemia (APL) by ATRA. Since its initial description, however, no large series of ATRA syndrome has been reported in detail. We analyzed cases of ATRA syndrome observed in an ongoing European trial of treatment of newly diagnosed APL. In this trial, patients 65 years of age or less with an initial white blood cell count (WBC) less than 5,000/μL were initially randomized between ATRA followed by chemotherapy (CT) (ATRA→CT group) or ATRA with CT started on day 3; patients with WBC greater than 5,000/μL received ATRA and CT from day 1; patients aged 66 to 75 received ATRA→CT. In patients with initial WBC less than 5,000/μL and allocated to ATRA→CT, CT was rapidly added if WBC was greater than 6,000, 10,000, 15,000/μL by days 5, 10, and 15 of ATRA treatment. A total of 64 (15%) of the 413 patients included in this trial experienced ATRA syndrome during induction treatment. Clinical signs developed after a median of 7 days (range, 0 to 35 days). In two of them, they were in fact present before the onset of ATRA. In 11 patients, they occurred upon recovery from the phase of aplasia due to the addition of CT. Respiratory distress (89% of the patients), fever (81%), pulmonary infiltrates (81%), weight gain (50%), pleural effusion (47%), renal failure (39%), pericardial effusion (19%), cardiac failure (17%), and hypotension (12%) were the main clinical signs, and 63 of the 64 patients had at least three of them. Thirteen patients required mechanical ventilation and two dialysis. A total of 60 patients received CT in addition to ATRA as per protocol or based on increasing WBC; 58 also received high dose dexamethasone (DXM); ATRA was stopped when clinical signs developed in 30 patients. A total of 55 patients (86%) who experienced ATRA syndrome achieved complete remission (CR), as compared with 94% of patients who had no ATRA syndrome (P= .07) and nine (14%) died of ATRA syndrome (5 cases), sepsis (2 cases), leukemic resistance (1 patient), and central nervous system (CNS) bleeding (1 patient). None of the patients who achieved CR and received ATRA for maintenance had ATRA syndrome recurrence. No significant predictive factors of ATRA syndrome, including pretreatment WBC, could be found. Kaplan Meier estimates of relapse, event-free survival (EFS), and survival at 2 years were 32% ± 10%, 63% ± 8%, and 68% ± 7% in patients who had ATRA syndrome as compared with 15% ± 3%, 77% ± 2%, and 80% ± 2% in patients who had no ATRA syndrome (P= .05, P = .003, and P = .03), respectively. In a stepwise Cox model that also included pretreatment prognostic variables, ATRA syndrome remained predictive for EFS and survival. In conclusion, in this multicenter trial where CT was rapidly added to ATRA in case of high or increasing WBC counts and DXM generally also used at the earliest clinical sign, the incidence of ATRA syndrome was 15%, but ATRA syndrome was responsible for death in only 1.2% of the total number of patients treated. However, occurrence of ATRA syndrome was associated with lower EFS and survival. © 1998 by The American Society of Hematology.


2007 ◽  
Vol 25 (18_suppl) ◽  
pp. 1052-1052 ◽  
Author(s):  
A. Chan ◽  
V. Ganju ◽  
D. Becquart ◽  
P. Conte ◽  
L. Petruzelka ◽  
...  

1052 Background: Chemotherapy (CT) plus H is the standard first-line treatment for HER2-positive MBC. H plus vinorelbine is an active and well-tolerated regimen in this setting. The all-oral combination of NVBo and X also appears active and well-tolerated in MBC. We report efficacy and safety results from the first 34 patients (pts) included in an international trial evaluating NVBoXH in HER2-positive MBC. Methods: In this multicenter trial, main eligibility criteria included: HER2-positive disease (IHC 3+ or FISH+), documented measurable MBC previously untreated by CT, relapse 6 months after completing neoadjuvant or adjuvant CT, Karnofsky PS = 70, age =18 years. Pts received 3-weekly cycles of NVBo 60 mg/m2 (cycle 1) escalating to 80 mg/m2 (from cycle 2) days 1 and 8; × 1,000 mg/m2 bid (750 if = 65 years) days 1–14; H 4 mg/kg day 1 as a loading dose then 2 mg/kg i.v. weekly starting on day 8. Treatment was continued until progression or unacceptable toxicity. Primary endpoint is overall response rate. Results: Baseline characteristics: median age 54 years (20% = 65); prior (neo)adjuvant CT 21 pts (62%); type of CT: anthracycline 52%, anthracycline + taxane 29%, CMF 14%, taxane 5%; visceral involvement 29 pts (85%), >2 metastatic sites 13 pts (38%). Treatment administered: median 8 cycles, median relative dose intensity: NVBo 77%, X 81%, H 95%; NVBo dose escalation to 80 mg/m2 in 91% of pts. Safety (n=34, G3/4 NCI CTC v2 adverse events): neutropenia 22 pts (65%), diarrhea 4 pts (12%), febrile neutropenia 3 pts (9%), vomiting 3 pts (9%), hand-foot syndrome 3 pts (9%), asthenia 3 pts (9%), infection without neutropenia 2 pts (6%), LVEF decline 2 pts (6%), stomatitis 1 pt (3%), nausea 1 pt (3%), constipation 1 pt (3%). Efficacy (n=31 evaluable pts): objective response rate (RECIST) 71% (95% CI [52–86]), CR 13%, PR 58%, SD 23%, PD 6%, disease control (CR+PR+ SD for =6 months) 84%. Progression-free survival, overall survival and duration of response data are not yet mature. Conclusions: This is the first trial, in pts with HER2-positive MBC, to show high efficacy with first-line NVBoXH therapy. This regimen can be safely administered in this pt population. No significant financial relationships to disclose.


2009 ◽  
Vol 27 (15_suppl) ◽  
pp. 1070-1070
Author(s):  
Y. Wang ◽  
D. Tai ◽  
L. Zhao ◽  
J. Gill ◽  
C. K. Obasaju

1070 Background: Population-based studies often attribute racial disparities in breast cancer outcomes to differences in access to treatment, quality of care, or other socioeconomic factors. In a controlled clinical trial setting, these systemic differences between races should be minimal. To evaluate the potential impact of race on outcomes in a controlled clinical setting, we retrospectively analyzed data from a phase III trial (B9E-MC-JHQG; NCT00006459 ) of patients (pts) with MBC. Methods: Analyses were performed on the JHQG trial database after 440 total pt deaths for both study arms. Demographics, safety, and efficacy were analyzed using safety population data from 3 racial groups: Caucasian (CA), Asian (AS), and Hispanic (HP). CA was the reference for all pair-wise comparisons. The logistic model was used to calculate odds ratios for tumor response and the Cox model was used to calculate hazard ratios for time-to-event parameters, adjusting for significant prognostic factors. Results: We report analyses of the gemcitabine (G) + paclitaxel (T) treatment arm. Demographics were balanced across the 3 groups with the exception that ER+/PR+ status was lower for AS compared to CA and HP; unknown ER/PR status was higher for AS. AS had significantly less neutropenia, fatigue, and nausea, but more anemia compared to CA and HP. Median number of treatment cycles completed was lower, but mean dose intensities for G and T were slightly higher, for AS. Response rate and progression-free survival were similar in the 3 groups. Overall survival (OS) and post-study chemotherapy (PSC) were significantly reduced for AS. Conclusions: Our analysis suggests that AS pts were better able to tolerate GT therapy compared to CA and HP pts. However, AS pts had the poorest OS outcome of the 3 racial groups, potentially due to reduced participation in PSC. [Table: see text] [Table: see text]


2017 ◽  
Vol 35 (15_suppl) ◽  
pp. 11523-11523
Author(s):  
Roseana Melo Borba ◽  
Tiago Cordeiro Felismino ◽  
Alexandre Andre B. A. Da Costa ◽  
Vladmir C. Lima ◽  
Marcelo Corassa ◽  
...  

11523 Background: Bev is a monoclonal antibody that binds to VEGFA that demonstrated improved progression free survival (PFS) in MBC clinical trials. VEGFR2, NOTCH1, Integrin a1b2 and ILK are angiogenesis-related proteins possibly related with Bev efficacy. The correlation of these proteins expression and Bev survival variables was evaluated. Methods: We retrospectively analyzed 1st line chemotherapy in two HER2 negative MBC cohorts. Pts were treated between May-07 and July-14. Cohort 1 (C1) was treated with paclitaxel and Cohort 2 (C2) with paclitaxel and Bev. Expression of biomarkers was determined by immunohistochemistry. Tumor samples were arranged on a tissue microarray. Survival curves were calculated by Kaplan-Meier method and log-rank test. Cox model was used in multivariate analysis. Results: C1 had 42 pts. Median age was 63y. Tumor subtypes were divided in luminal (92.9%) and triple negative (TN) (7.1%). Visceral metastasis (mets) were present in 71.4%. Median follow-up (mFUP) time was 32.1m. mPFS was 8.0m and mOS was 33.5m. C2 had 29 pts. Median age was 57y. Luminal 79.3%; TN 20.7%; Visceral mets 79.3%; mFUP 38m. mPFS was 10.5m and mOS was 47m. In C2, high VEGFR2ce was correlated with improved PFS (high VEGFR2 16.5m x low VEGFR2 6.8m, p = 0.025). Breast cancer subtype, metastasis pattern and VEGFR2 expression were included in the multivariate analysis for PFS. VEGFR2 remained as independent factor (HR 0.35; IC95% 0.14 – 0.85, p = 0.021). In C1, VEGFR2 was not correlated with improved PFS (high VEGFR2 8.6m x low VEGFR2 8.0m, p = 0.24). Other markers were not associated with PFS. Conclusions: High VEGFR2ce was associated with increased PFS in patients treated with Bev. In MBC VEGFR2 may have a role as a predictive tool on benefit of antiangiogenic therapy.


Sign in / Sign up

Export Citation Format

Share Document