GRÁFICO N° 4.10 Patrones a seguir
5. CONCLUSIONES Y RECOMENDACIONES
531 patients enrolled in I.Co.N.A were included in the analysis. They all started one PI and two nucleoside reverse transcriptase inhibitors (NRTIs). In particular, 66% (n=350) started HG-saquinavir and two NRTIs, and 34% (n=181) started ritonavir and two NRTIs. Median age in the two groups was 34 years (range: 18-98), and 35 (range:20- 65). Median CD4 count was 302 cells/pl (range: 1-1,178), and 183 (range:1-984). Median viral load was 4.80 logio copies/mL (range:2.76-6.72), and 4.96 (range: 2.80- 6.65). Finally, the percentages of patients infected through injecting drug use were 40.5%, and 44.2%, respectively.
For some patients the last available viral load was measured more than one year before the date of this analysis and these patients were defined as “lost to follow-up” at the date of the last viral load measurement. Other patients had discontinued ritonavir or HG-saquinavir before the end of virological follow-up and they did so for numerous reasons, including virological, immunological or clinical failure, drug-related toxicity (clinical signs and symptoms or abnormal levels of laboratory markers), or patient choice.
Table 4.4.1. Chronology of events In 531 patients enrolled in I.Co.N.A. and details of censoring strategy adopted Censoring strategy / approach
C hronology o f events N (%) DEF OT ITC TOX
Virological failure (VF) first event 258 (48.6) Failure at date
of VF Failure at date of VF Failure at date of VF Censored at date of VF
Stop due to patient’s choice 17(3.2) Failure at date
of stop Censored at date of stop Censored at date of last VL Censored at date of stop
Stop due to therapy failure 9(1 .7 ) Failure at date
of stop f^ailure at date O# stop Faiiure at of stop Censored at date of stop
Stop due to toxicity 32 (6.0) Failure at date
of stop Censored at date of stop Censored at date of last VL Failure at date of stop
Stop due to patient’s choice then Drop-out 4(0.8) Failure at date
of stop Censored at date of stop Failure at date of drop Censored at date of stop
Stop due to therapy failure then Drop-out 2 (0.4) Failure at date
of stop Failure at date of stop Failure at date of stop Censored at date of stop
Stop due to toxicity then Drop-out 16(3.0) Failure at date
of stop Censored at date of stop Failure at date of drop Failure at date of stop
Stop due to patient’s choice then VF 38 (7.2) Failure at date
of stop Censored at date of stop Failure at date of VF Censored at date of stop
Stop due to therapy failure then VF 9(1.7) Failure at date
of stop Failure at date of stop Failure at date of stop Censored at date of stop
Stop due to toxicity then VF 53(10.0) Failure at date
of stop Censored at date of stop Failure at date of VF FÎÉü'rë at date of stop
Total patients - Total Failure# 531 (100.0) 497 (93.6) 337 (63.5) 448 (84.4) 161 (30.3)
DEF=discontinuation equals failure; OT=on-treatment; ITC=lgnore treatment changes; TOX=Toxicity analysis
Stop = drug discontinuation; Drop-out = last viral load measured more than 12 months before the date of the analysis
VF = failure based on viral load data
When none of these events occurred, patients follow-up were censored at the date of the last viral load in all the approaches When drop-out was the first event, patients were defined as failure at the date of drop-out in all the approaches
Figure 4.4.1 I.Co.N.A - Kaplan-Meier estimates of the probability of therapy failure using the DEF event-defining censoring strategy
probability of therapy faiiure
O
9
HG-SQV RTV 87
6 5 iog-rank p =0.044
3
21
O
64
72
32
40
48
56
O 816
24
no. 350 at risk 181weeks from starting HAART 327 138 310 110 120 58 110 52 99 46 85 43 73 40 72 38 61 36
Of the 328 patients who had a therapy switch over follow-up, 314 (95.7%) had discontinued HG-saquinavir or ritonavir and only 14 (4.3%) had intensified the original regimen (i.e. had started a second PI or a NNRTI before stopping either HG-saquianvir or ritonavir). As mentioned above, changes in the NRTIs and intensifications have been ignored.
Table 4.4.1 describes the number of patients studied according to the chronology of events observed (therapy discontinuation for reasons other than toxicity, therapy discontinuation because of toxicity, and virological failure based directly on the viral load measurements). Losses to virological follow-up are considered as therapy failure in all of the four strategies and have been omitted from the table. For each of the chronology of events a detailed description of the censoring analysis adopted in the four different analyses (DEF, OT, ITC, and TOX) is also provided. In the following paragraphs, the results of using each of these strategies are reported separately.
4.4.1 Strateav 1 : effectiveness + tolerabilitv (DEF analvsis)
The objective of this part of the analysis was to compare HG-saquinavir (in combination with two NRTIs) and ritonavir (in combination with two NRTIs) with respect to their virological potency (or effectiveness) and, simultaneously, their propensity to be discontinued. Thus, therapy failure-defining “events” were virological failure, drug discontinuation (irrespective of the reason), or virological drop out (Table 4.4.1). Time to therapy failure is then the time to experience the first of these events. This definition of the endpoint is exhaustive as all outcomes other than successful sustained virological suppression are counted as failures.
For example, a patient whose last viral load had been measured one month before the date of the analysis, whose viral load had decreased below 500 copies/mL at week 16 without changing therapy, and who had discontinued ritonavir because of toxicity at week 18, would be defined as having an event at week 18 (event = stop due to toxicity in Table 4.4.1). This type of analysis can be defined as a “discontinuation equals failure” (DEF) analysis as a missing value for viral load due to therapy discontinuation determines an event or a “failure”. This often arises in clinical trials from pharmaceutical companies where patients' virological follow-up is stopped at the date of first therapy switch. In this sense, since all the patients originally receiving the HG- saquinavir-based or the ritonavir-based regimen are accounted for in the analysis, this follows the intention to treat principle. In section 4.4.3, another, less artificial, way of performing an intention to treat analysis will be presented.
but not easily tolerated, appears to be similar to that of a drug which is less potent but is associated with a lower rate of early therapy discontinuation.
Table 4.4.2 shows the distribution of the other failure-defining events in the DEF analysis in the group receiving HG-saquinavir and in the group receiving ritonavir. It is clear that an event such as virological drop out occurs in around 11-12% of the patients in clinical practice and this is equally distributed in the two arms (n=37 in HG saquinavir and n=22 in ritonavir-containing arm). All these patients were defined as therapy failures at the time of virological drop out. However, sequences involving therapy discontinuations because of toxicity were more frequent in patients who started ritonavir-containing regimens compared to those starting HG-saquinavir-containing regimens (highlighted in yellow). Similarly, as anticipated, a higher percentage of patients experiencing virological failure were observed in the group of patients starting HG-saquinavir (highlighted in green).
Figure 4.4.1 shows the Kaplan-Meier estimates of therapy failures when the DEF censoring strategy was adopted. Over the first 24 weeks, the rate of failure was clearly greater in the group of patients who had started ritonavir. This, however, was due to the fact that over the first 24 weeks of therapy ritonavir discontinuations (due to toxicity) are much more frequent than HG-saquinavir discontinuation and that discontinuations due to toxicity are a failure-defining event in this analysis. Indeed, at week 24, when patients who had still not suppressed their viral load below 500 copies/mL were defined as failure at this point in time, the Kaplan-Meier estimate of the probability of failure in the HG-saquinavir arm equals that of the ritonavir arm. This was probably due to the fact that the risk of failing to achieve a viral load below 500 copies/mL by week 24 is far greater in patients receiving HG-saquinavir compared to those receiving ritonavir arm. Also after week 24 there was no appreciable difference in rate of failure between the two groups suggesting that, when comparing patients who had achieved viral suppression, the rate of therapy failure (constituted by virological rebound, switching or stopping therapy) after week 24 would be similar in the two groups.
Table 4.4.2 Description of the failure-defining events in patients receiving HG-saquinavir and ritonavir (DEF analysis)
Chronology
HG-SQV RTV DEF Strategy
No event 18 (5.1) 16(8.8) Censored at last
VL
Virological failure first event 216(61.7) 42 (23.2) Failure at VF
Drop-out first event 37 (10.6) 22(12.2) Failiure at drop Stop due to patient's choice 13(3.7) 4 (2 2) Failure at stop Stop due to therapy failure 8 (2 3) 1 (0.6) Failure at stop Stop due to toxicity 9 (2.6) 23(12.7) Failure at stop Stop due to patient’s choice then Drop 3 (0.9) 1 (0.6) Failure at stop Stop due to therapy failure then Drop 2 (0.6) 0 (0.0) Failure at stop Stop due to toxicity then Drop 1 (0.3) 15(8.3) Failure at stop Stop due to patient’s choice then VF 20 (5.7) 18(7.2) Failure at stop Stop due to therapy failure then VF 8 (2 3) 1 (0.6) Failure at stop Stop due to toxicity then VF 15(4.3) M 38 (21.0) Failure at stop
Total 350 (100.0) 181 (100.0) 531 (100.0)
Censored 18 (5.1) 16(8.8) 34 (6.4)
Failures 332 (94.9) 165 (91.2) 497 (93.6)
The Kaplan-Meier estimate of the probability of failure at week 12 in the ritonavir arm (0.29, 95% Cl: 0.22-0.36) was significantly higher than that in the HG-saquinavir arm (0.08, 95% Cl: 0.05-0.11) which was explained by the higher rate of ritonavir- associated toxicity. However, there was no difference between the two groups at week 48: (0.76, 95% Cl: 0.70-0.82) in the ritonavir and (0.76, 95% Cl: 0.71-0.81) in the HG- saquinavir group, respectively. Several tests to compare the two curves over the whole period were carried out yielding conflicting results (log-rank test p=0.04, Wilcoxon p=0.0001, and likelihood ratio test p=0.86). This is due to the fact that these methods give different weights to events occurring at different time points during the observation period. Thus, the p-values for the log-rank and Wilcoxon tests, which give higher weights to events occurring earlier in time, were significant while the likelihood ratio test yielded a non-significant result. In conclusion, from this analysis, in which both discontinuations and virological failures were considered as therapy failure (i.e. HG- saquinavir and ritonavir were compared in terms of a combined measure of their virological potency and tolerability), in the longer term these two drugs appeared to be equivalent.
4.4.2 Strategy 2: effectiveness (OT analvsis)
In some cases it may be of interest to compare two drug regimens just in terms of their potency in suppressing viral replication. In order to do this, the previous analysis is
modified by eliminating the tolerability component from the outcome. Thus, in this analysis, follow-up times of patients who had discontinued therapy because of toxicity or patients’ decision before failure were no longer defined as failures but were right- censored. Virological failure (failing to suppress before week 24 or relapsing after initial suppression) and discontinuing because of virological failure as reported by clinicians were defined as failure (Table 4.4.1). Virological drop-outs were still defined as therapy failure. For example, if a patient's viral load had decreased below 500 copies/mL by week 24 and this patient had discontinued one of the drugs in the regimen because of toxicity at week 32, and the most recent viral load was measured one month before the date of the analysis, this patient’s follow-up would be censored at week 32 (Table 4.4.1). This kind of analysis is often called an “on-treatment” (OT) analysis as patients’ viral loads are considered for defining failure only if measured when the patients were still receiving the original treatment.
Table 4.4.3 Description of the failure-defining events in patients receiving HG-saquinavir and ritonavir (OT analysis)
Chronology
HG-SQV RTV OT
Strategy
No event 18(5.1) 16(8.8) Censored at last
VL
Virological failure #rst event 216(61.7) 42 (23.2) Failure at VF Drop-out first event 37 (10.6) 22(12.2) Failure at drop Stop due to patient’s choice 13(3.7) 4 (2 2) Censored at
stop
Stop due to therapy failure 8 (2.3) 1 (0.8) Failure at stop Stop due to toxicity 9 (2.6) 23 (12.7) Censored at
stop Stop due to patient’s choice then Drop 3 (0.9) 1 (0.6) Censored at
stop
Stop due to therapy failure then Drop 2 (0.6) 6(0.0) Failure at stop Stop due to toxicity then Drop 1 (0.3) 15(8.3) Cmsored at
stop Stop due to patient’s choice then VF 20 (5.7) 18(9.4) Censored at
stop
Stop due to therapy failure then VF 8 (& 3) 1(0-6) Failure at stop Stop due to toxicity then VF 15(4.3) 38 (21.0) Censored at
stop
Total 350 (100.0) 181 (100.0) 531 (100.0)
Censored 79 (22.6) 115(63.5) 194 (36.5)
Failures 271 (77.4) 66 (36.5) 337 (63.5)
Figure 4.4.2 I.Co.N.A - Kaplan-Meier estimates of the probability of therapy failure using the OT event-defining censoring strategy
probability of therapy failure
no. at risk 350 181 log-rank p = 0.0001
0.9
0.8
HG-SQV0.7
0 . 6
0.5
RTV0.4
0.3
0.2
0.0
0
8
16
24
32
4 0
4 8
56
64
72
weeks from starting HAART 332 151 315 129 124 81 115 75 105 72 94 71 82 68 80 66 71 62
Table 4.4.3 shows the distribution of failure-defining events in this modified analysis. Three hundred and thirty-seven patients (63.5%) were defined as failures in this analysis (as also shown at the bottom of Table 4.4.1). For the majority of the virological failure-defining events observed, the percentage of patients in the HG-saquinavir arm was higher than the percentage in the ritonavir arm (in green in Table 4.4.3).
However, in this OT analysis, discontinuations due to toxicity (in yellow) were no longer therapy failure-defining events and did not compensate in favour of HG-saquinavir as in the previous analysis. Overall, failing to achieve a viral load below 500 copies/mL by week 24 was more common than experiencing a viral rebound. As already shown, the risk of failure was much greater in the HG-saquinavir arm (61.7%, 216 out of 350 patients) than in the ritonavir arm (23.2%, 42 out of 181 patients) and was mainly due to an initial lack of suppression rather than to viral rebound. This is illustrated by Figure 4.4.2 that shows the Kaplan-Meier estimates of the probability of therapy failure in the two arms. The “jump” in the curve at week 24 (given by the proportion of patients failing to achieve a viral load below 500 copies/mL by this time) is significantly greater in the HG-saquinavir arm compared to the ritonavir arm. This is similar to that seen when using the DEF strategy because it is driven by the virological failures (Figure 4.4.1) and the result is probably explained by the difference in bio-availability of these
The probability of experiencing therapy failure after week 24 in the ritonavir group in this OT analysis was similar to that observed in the DEF analysis. However the cumulative probability of failure up to week 24 was much lower in the OT analysis because discontinuations were not counted as failures. This explains why, when taking into account the entire observation period, the difference in therapy failure between the two drugs was highly significant (log-rank p=0.0001, Wilcoxon p=0.0001, likelihood ratio p=0.0001).
Moreover, as in the DEF analysis, the rate of viral rebound following week 24 did not appear to differ significantly between the two groups of patients (note that for some patients the viral rebound may have occurred before week 24 with no difference in the rate of failure between the two arms, Figure 4.4.2). Overall, it is possible to conclude that, given the far greater rate of virological success at week 24 in the ritonavir arm, probably due to its better absorption, ritonavir in combination with two NRTIs appears to be superior to HG-saquinavir (also in combination with two NRTIs) in suppressing viral replication (log-rank p=0.0001).
4.4.3 Strategy 3: ignore treatment changes (ITC analysis) Rationale for this strategy and interpretation of resuits
If the aim of the analysis is to provide an estimate of the drug’s effectiveness under “real-life” conditions, the analysis should be conducted on an ITT basis. This kind of analysis is often known as a pragmatic anaiysis whose objective is to compare treatment strategies rather than a specific treatment. So, for example, as observed in the previous section, only a small percentage of patients starting a combination including ritonavir are likely to remain on this treatment by 24 weeks of therapy. This is generally due to drug-related toxicity or intolerance with potentially undesirable consequences for the patient’s future prognosis (e.g. the possible effect of toxicity on the response to second-line treatment, reduced number of options, etc.). As mentioned in chapter 2 one way of conducting the analysis on an ITT basis when using a time to event approach is to completely ignore treatment changes for the assessment of the outcome (ITC analysis). Treatment changes are, however, described and taken account of in the overall interpretation. If the ITC strategy is adopted, these patients would still be considered to belong to the group whose policy was the use of ritonavir in their first HAART regimen. One benefit of using the ITC strategy is that, since patients who stop or switch treatments are often those who are experiencing the worst response or the greatest adverse effects, by keeping these patients in their original treatment group, the results may be protected from a bias towards more favourable responses. Using this approach, the rate of therapy-failure in patients receiving either ritonavir and HG-saquinavir would then depend not only on the effectiveness of these drugs (in combination with two NRTIs) “per se” but also of the second and third line regimens consequent on the use of these drugs as a first HAART regimen. This is an important point as from this perspective this approach gives some insight on the potential longer-term consequences (i.e. overlapping toxicities, cross resistance, etc.) of using these drugs as first line treatment in terms of the future management of the patients. However, a caveat of this analysis is that the interpretation of the long-term response may not be applicable to current real-life responses of patients in the clinic. This is because, when the data have been collected, the common practice was to switch patients who discontinued ritonavir because of toxicity to another Pl-containing HAART while nowadays patients are most likely switched to a NNRTI-containing HAART.
Further, if the aim of the analysis is to compare the long term outcome of two consecutive drug regimens, the time to second virological failure (rather than the first as defined here) may be the best analytical approach as suggested by some
Results
Table 4.4.4 shows the chosen censoring strategy compatible with an ITC strategy in a time to virological failure analysis: therapy discontinuations due to toxicity or patients' decision are ignored (i.e. follow-up times are censored at the date of last viral load), patients who had a virological drop-out were treated as failures and virological failures or discontinuations due to virological failures were defined as therapy failures. As