The Second DSMB Meeting (First Data Review): February 22, 2020
There were 231 patients consented and 215 randomized (144 in the remdesvir group and 71 in the placebo group). Fifty-four (25.1%) completed the 10 days treatment and none reached 14-day follow-up time yet. Ninety-six (44.7%) were within the 10-day treatment period; 53 (24.7%) did not start treatment yet. Twelve (5.6%) discontinued the treatment due to reasons including AE (n = 8) and other (n = 4).
In the closed session where unblinded data were examined by the DSMB, baseline characteristics and AE proportions were similar between the two groups. No unexpected adverse events were observed, judged by the medical experts of the DSMB. The stratified clinical 6-category scale results were examined with the bar charts and with the stratified WMW rank-sum test displayed on the dynamic “radar” screen. For example, Fig. 1 shows the bar chart for Day 10 results. (Additional bar charts were also requested by the DSMB for Days 3, 5, and 7, but not shown here.) Most (n = 118, or 83%) patients’ baseline ordinal category = 3. Four patients’ baseline category = 2, 20 patients’ baseline category = 4, and none were category = 5.
Figures 2, 3, 4, and 5 show the dynamic graphs on the movement of the stratified WMW rank-sum test (Z-value) as patient enrollments accumulated on each of the post-randomization Days 3, 5, 7, and 10, respectively. The upper boundary of the “promising zone” was set for conditional probability (CP) = 90% and the lower boundary was set for CP = 5%. The middle dash line represented CP = 50%. As shown, remdesivir did not have a quick effect on the clinical scale compared to the placebo (on Days 3, 5, 7), but was trending upward on Day 10. This fact provided a hopeful scenario for the DSMB to recommend continuing the trial and planning for the next DSMB meeting a week later. Note that, unlike Day 7, the data on the Days 3, 5, and 10 were unplanned but requested by and provided to the DSMB instantaneously with the efficient eDMC system. The supplementary data provided a useful “trend” information for the DSMB review. In addition, at this time, the COVID-19 epidemic in Wuhan was slowing down, and many other studies also started in the region in February, competing for the patient resources. Facing the decline of enrollment, DSMB recommended the investigators to consider enhancing their enrollment effort and to the trial sponsor to re-evaluate the planned time-line for study completion date (the original projection date was April 27, 2020).
DSMB also urged the investigators to verify the patient’s eligibility criteria, especially the baseline SPO2 and/or the PAO2/FIO2 levels for this trial of severe COVID-19.
Aside: On February 21, 2020, the US National Institute of Allergy and Infectious Diseases (NIAID) launched a multicenter, adaptive, randomized, blinded, controlled trial involving remdesivir for the treatment of COVID-19 in hospitalized adults , as the disease was spreading fast globally. The remdesivir regimen was the same as in the Chinese trials. DSMB noticed that the NIAID trial used a similar 8-category ordinal scale for the trial’s primary endpoint. However, the in-patient follow-up time was 15 days, shorter than the 28 days as in the Chinese trials. Sample size was n = 440, similar to the sample size n = 453 for the Chinese trial with severe cases.
The Third DSMB Meeting (Second Data Review): February 29, 2020
Prior to the meeting, the investigator team communicated that the patient enrollment to the trial of the mild–moderate cases had also been adversely impacted by the decline of COVID-19 epidemic, which was fortunate for the disease control, but sorry news for the trial per se. The focus of DSMB review remained on the first trial with severe cases.
There were 228 patients randomized (152 in the remdesivir group and 76 in the placebo group). Only 13 new patients were enrolled since the last DSMB review 7 days ago. However, more patients had completed the 10 days treatment and reached 14 days post-treatment follow-up. With carrying-over the status of discharged and death, the 6-category clinical scale showed 201 patients with data on Day 10, 177 patients on Day 14, and 97 patients on Day 21. Still, most (191/228, or 83.8%) were with baseline category = 3.
For safety, total deaths = 24 (10.53% mortality), evenly distributed between the two treatment groups. No concerned AEs were found, but DSMB noticed that more patients in the remdesivir group discontinued treatment than the placebo group (13/149, 8.7% versus 3/75, 4.0%). However, 20/149 (13.4%) of the remdesivir-treated patients had serious AEs, less than the placebo group (15/75, or 20.0%).
For efficacy, the total number of discharged from hospital was 55 (24.12%), slightly higher in the remdesivir group (40/152, or 26.32%) than in the placebo (15/76, or 19.74%). The 6-category clinical scale analysis added more enthusiasm for the DSMB as shown by the stratified distribution at Day 14 (bar chart, Fig. 6; other bar charts on earlier days were also examined by DSMB but omitted here), as well as by the dynamic “radar” graphs of the stratified WMW rank-sum tests at Days 7, 10, 14, and 21 (Figs. 7, 8, 9, 10, respectively; other graphs on earlier days were also examined by DSMB but omitted here). Based on the Z-value entering the favorable region on Day 14 (see Fig. 9), DSMB requested Kaplan–Meier plot (without p-value) be displayed for the TTCI endpoint at the next review meeting. Altogether, the tests seemed to be moving toward a favorable direction for remdesivir at this point in time, but the DSMB remained cautious on the yet to harvest data, especially since the Day 21 data were not adequate and lacked a trend in harmony with Day 14. The recommendation was to continue the trial, but DSMB urged the sponsor to re-evaluate the time-line for study completion date and the original planned interim analysis strategy on the primary endpoint, TTCI.
Aside: The protocol planned n = 453 severely ill patients to complete the 28 days study by April 28, 2020. With n = 228 randomized as of 2/28/2020, DSMB informed the sponsor that in the month of March, the investigator team must randomize 7 or 8 patients per day on average in order to complete the target enrollment at the end of March.
The Fourth DSMB Meeting (Third Data Review): March 8, 2020
There were 235 patients randomized (157 in the remdesivir group and 78 in the placebo group). Only 7 new patients were enrolled from the last DSMB review 8 days ago. Baseline characteristics were comparable between the two groups. Among the 235, 217 (92.3%) had nasal catheter or mask, 27 (11.5%) had high flow oxygen, 15 (6.4%) had noninvasive ventilation, and 2 (0.9%) had invasive ventilation for their respiratory support before randomization. The median day of nasal catheter or mask was 3 days and high flow oxygen support was 2 days. Four patients withdrew their informed consent.
The baseline SPO2 level was also similar between the two treatment groups. The in-room temperature, not requiring supplemental oxygen SPO2 range was 75 to 94% (n = 75, out of 235 patients; median = 92%). With supplemental oxygen, the range of SPO2 was 37 to 100% (n = 158, out of 235 patients; median = 93%). One patient (in the placebo group) had hypercapnia respiratory failure at baseline, whose SPO2 was 88%.
Of the safety data set (n = 230), 26/153 (17%) patients in the remdesivir group had SAE, compared to 17/77 (22.1%) in the placebo group; few SAEs were judged drug-related by the investigators (2 in each group). However, more patients discontinued treatment due to AE in the remdesivir group than the placebo group (17/153 or 11.1% versus 3/77 or 3.9%), as also noted in the previous DSMB review. The mortality rate was 27/230 (11.74%), slightly increased from the last review and was comparable between the two groups.
For efficacy, although only 7 new patients enrolled since last review, more patients had completed the 10 days treatment and reached Day 14 and Day 21 post-treatment follow-ups. Carrying over the status of discharged and death, the 6-category clinical scale cumulated 220 patients with data on Day 10, 218 patients on Day 14, 197 patients on Day 21, and 143 patients on Day 28. Still, mostly (192/230, or 83.5%) were with baseline category = 3. For this reason, DSMB firmly believed that the analysis of “clinical improvement” (as for the TTCI endpoint) would need to be supplemented by an analysis of “clinical no-change or worsening.” The DSMB’s analysis of the distribution of the 6-category scale using the stratified WMW rank-sum test was actually on-target.
The total number of discharged from hospital was 92 (40.0%), lower in the remdesivir group (58/153, or 37.91%) than in the placebo group (34/77, or 44.16%). This reversal from the previous data review worried the DSMB. Less optimistic than before, DSMB examined the 6-category clinical scale data, which were also disappointing in the following sense, as shown by the stratified distribution at Day 21 (bar chart, Fig. 11; other bar charts on earlier days were also examined by DSMB but omitted here), as well as by the dynamic “radar” graphs of the stratified WMW rank-sum tests at Days 10, 14, 21, and 28 (Figs. 12, 13, 14, 15, respectively; other graphs on earlier days were also examined by DSMB but omitted here). Rather than maintaining in the favorable region on the “radar” screen, Day 14 rank-sum test Z-value seemed moving downward. Moreover, rather than strengthening the Day 14 result, Day 21 and Day 28 tests seemed moving away from favorable region for remdesivir. For the first time, the eDMC system, at the request of the DSMB from the last meeting, displayed the TTCI’s Kaplan–Meier plot (without p-value); the median TTCI was about 23 days versus 24 days for the remdesivir and placebo groups, respectively. The 1-day difference was much lower than the 6-day difference expected in the trial protocol.
Learning from the sponsor that, due to the continuing decline of COVID-19 cases in Wuhan, investigators would not be able to enroll new patients in order to reach the protocol-planned number, DSMB recommended sponsor to consider expanding the study to other cities in China or to other countries. Otherwise, sponsor needs to revise the protocol: (a) change the planned sample size for the declining enrollment reason; (b) cancel the scheduled interim analysis of TTCI; (c) continue the trial till all randomized patients finish their 28 days follow-up; and (d) prepare the final DSMB review meeting at the end of March for the trial with severe cases.
Aside: On March 3, 2020, Gilead Science launched two clinical trials of remdesivir [11, 12]. For the moderate cases , the primary endpoint was time to discharge in 5 to 10 days. For the severe cases , the primary endpoint was proportion of participants with normalization of fever and oxygen saturation through Day 14. On March 11, 2020, the World Health Organization declared the rapidly spreading COVID-19 outbreak a pandemic.
The Fifth DSMB Meeting (Fourth And Final Data Review): March 29, 2020
DSMB received information from the sponsor that our last recommendation of revising the protocol of the first trial with severe cases was accepted. Hence, this would be the last DSMB review on the data of the severe cases. The second trial of mild–moderate patients was still too early for analyzing the efficacy data; only baseline and safety data were presented by the CRO. Our focus was still on the severe cases as follows:
There were 237 patients randomized (158 in the remdesivir group and 79 in the placebo group). Only 2 new patients were enrolled from the last DSMB review three weeks ago. Of the 237, 195 (82.3%) completed the study, 1 was still on-going, 35 (14.8%) died, and 6 (2.5%) withdrew for reasons other than death. The 1 on-going patient was to complete the Day 28 follow-up by 4/1/2020.
The intent-to-treat all-cause mortality rate was 24/158 (15.2%) in the remdesivir group and 11/79 (13.9%) in the placebo group, including 2 patients who died after randomization prior to receiving remdesivir and 3 patients died after Day 28 (1 in placebo and 2 in remdesivir group). These rates were lower than the corresponding rates (19.2% vs 25.0%) reported in the trial of lopinavir–ritonavir vs SOC by the same investigator team .
On the 6-category ordinal scale at baseline, the distributions (%) were (0, 0, 81.6, 17.7, 0, 0.6) for the remdesivir group and (0, 3.8, 83.5, 11.4, 1.3, 0) for the placebo group, in Categories 1 (discharged or met discharge criteria) to 6 (death). As seen, majority (> 80%) were Category 3 patients, who were hospitalized, required supplemental oxygen (but not NIV/ HFNC).
The live discharge rate continued to be lower for the remdesivir group (107/157, 68.2%) compared to 57/77 (74.0%) in the placebo group. (Three missing data were noted.)
Although only two new patients enrolled since last review, more patients had completed the 10 days treatment and reached Days 14, 21, and 28 post-treatment follow-ups. Carrying over the status of discharged and death, the 6-category clinical scale cumulated 228 patients with data on Day 10 (one patient was missing on Day 10), 229 on Day 14, 227 patients on Day 21, and 225 patients on Day 28. (The 6-category data excluded the 2 deaths occurred prior to receiving treatment and 3 deaths occurred after Day 28.)
Figure 16 displays the stratified distribution of the 6-category scale on Day 28. (Other distribution bar charts on earlier days were also examined by the DSMB but omitted here). Figures 17, 18, 19, and 20 display the stratified WMW tests on the “radar” screen for Days 10, 14, 21, and 28. Other earlier days were also examined by the DSMB but not shown here. Similar trends were observed compared to the previous data review, but confirmed with more subjects. Z-values of the stratified WMW test on Days 10, 14, 21, and 28 were all in the “promising” region, with the conditional power just touching 50% from above on Days 10 and 14, then lowering below 50% (dash line) on Days 21 and 28. The conditional power was calculated to project the chance that the final Z-value would be ≥ 1.96 (unadjusted for multiplicity), if the trial were to continue to the original final sample size of 435 patients, given the current estimate of treatment effect.
The DSMB also reviewed the Kaplan–Meier plot (without p-value) of the TTCI endpoint. The median TTCI was 22 days for the remdesivir group versus 24 days for the placebo groups. The 2-day difference was disappointingly much smaller than the 6-day difference projected in the protocol.
In an ordinary situation, we would recommend continuing the trial, perhaps even with an increase of the sample size. However, it was unfortunate that the sponsor could not feasibly continue enrolling patients outside of China, where the pandemic had begun. Thus, the DSMB agreed to complete the study with the current cohort of 237 randomized patients. DSMB recommended to let the last on-going patient finish the Day 28 follow-up before unblinding. DSMB also urged the investigator team to conduct the final analysis as soon as possible after unblinding to share the trial results with the international community.