整合酶抑製劑能降低死亡率嗎？
資料來源：www.thelancet.com/hiv Vol 9 June 2022/財團法人台灣紅絲帶基金會編譯
HIV 治療的最後一次重大進展是在 2007-08 年，第一個整合酶鏈轉移抑製劑 (integrase strand transfer inhibitor , INSTI) 拉替拉韋 (raltegravir) 獲得批准。 第二代 INSTIs，多替拉韋(dolutegravir) 和 bictegravir，具有更長的半衰期和更高的單片劑耐藥屏障，鞏固了這類抗反轉錄病毒藥物作為全球一線 HIV 治療的首選錨定藥物。
與非核苷逆轉錄酶抑製劑(non-nucleoside reverse transcriptase inhibitors) 和利托那韋增強的蛋白酶抑製劑(ritonavir-boosted protease inhibitors) 相比，INSTIs 與更快速和持續的病毒學抑制以及更高的 CD4 細胞計數增加有關。這些超過早期治療觀察到的死亡率特徵可以轉化為改善健康結果並降低死亡率。
INSTIs 在藥片數量、藥物動力學增強、耐藥性特徵和副作用方面的差異可能轉化為長期結果的差異。 直接比較未接受過抗反轉錄病毒治療 (ART) 的 HIV 感染者的各種 INSTI 的試驗未發現療效存在重大差異。然而，不同的毒性，如體重增加，可能會增加患糖尿病和心血管疾病的風險，這些疾病可能只有在接觸多年後才會顯現出來。
因此，需要觀察性研究來評估長期結果。迄今為止，此類研究並未顯示在長達 6 年的時間內與 INSTI 相關的發病率或死亡率降低。
在《刺胳針愛滋病毒》中，Adam Trickey 及其同事比較了 2013 年至 2018 年間在北美和歐洲開始抗反轉錄病毒治療的 62,500 名未接受過 ART 治療的 HIV 感染者的不同治療方案之間的全因死亡率。這是在 INSTI 使用普遍的時代迄今為止檢查首次接觸 ART與 HIV 相關死亡率的最大規模研究。
作者對按世代和日曆時間分層的六種錨定藥物進行了多重配對比較，並得出結論，幾乎沒有證據顯示大多數一線抗反轉錄病毒治療方案之間的死亡率存在差異。 然而，與所有其他 INSTI、非核苷類反轉錄酶抑製劑和蛋白酶抑製劑相比，基於 raltegravir 的方案與更高的死亡率相關。
鑑於先前發表的研究結果，這一發現既出乎意料又令人驚訝； 因此，它需要仔細審查。 raltegravir 死亡風險較高的可能解釋包括： 更高的病毒學失敗率和由此產生的跨類耐藥性；一些以前未知的臨床毒性僅與 raltegravir 相關； 暴露於 raltegravir 的人的特徵存在重要差異，這使他們與接受其他基於 INSTI 方案的人相比死亡風險增加。
在不同方案之間觀察到病毒學失敗的時間差異，多替拉韋 (dolutegravir) 的風險最低。 與多替拉韋 (dolutegravir)、艾替拉韋 (elvitegravir) 和利匹韋林 (rilpivirine) 相比，拉替拉韋 (raltegravir) 的風險更高，但與依非韋倫 (efavirenz) 或強化地瑞那韋 (boosted darunavir) 相比則沒有； 因此，較高的病毒學失敗率不太可能完全解釋相對於所有其他錨定藥物死亡率增加的風險。
在早期的臨床試驗中，有人擔心拉替拉韋 (raltegravir) 可能與癌症風險增加有關。 這種擔憂並沒有在隨後的研究中得到證實。自註冊以來的 15 年內，沒有出現其他特定於拉替拉韋 (raltegravir) 的長期臨床結果。不幸的是，本研究無法獲得特定原因死亡率。
作者承認結果可能會受到選擇性偏差或殘留干擾的影響。 例如，與多替拉韋相比，開始使用拉替拉韋的患者的 CD4 細胞計數較低（25% 的接受多替拉韋的人相較於 33% 的接受拉替拉韋的人每毫升有 100,000 個拷貝），並且更有可能出現較晚並有病史的愛滋病、結核病和在開始時非愛滋病定義之惡性腫瘤。值得注意的是，接受拉替拉韋 (raltegravir) 的人與開始使用地瑞那韋 (darunavir) 和依非韋倫(efavirenz) 的人有相似的結果。不能依靠回歸調整來完全消除干擾因素，並且當治療組之間存在重大差異時，對共變-結果關係建模的方式特別敏感。作者付出了相當大的努力來探索可能的偏差。 他們分析了一部分具有額外潛在干擾因素的世代； 擬合兩個不同時間段的模型，以反映開始抗反轉錄病毒治療的原因的潛在變化； 探索了表示缺失數據的不同方法； 為早到或晚到的人安裝單獨的模型； 並針對後續的潛在信息丟失進行了調整。 在每次敏感性分析中，raltegravir 與死亡率之間的關聯仍然存在。
像這樣的藥物警戒研究為新的藥物流行病學研究提出了假設。 這方面的一個教科書示例是 Cole 及其同事的目標試驗方法，其中接受 raltegravir 的人與接受 efavirenz 的人相比，4 年內罹患愛滋病或死亡的風險明顯更高，但在考慮到非隨機抽樣、治療的分配和流失退出等因素之後便消失了。 Cole 及其同事使用時間相關的逆概率權重來考慮包括後續損失可能的時間相關的干擾，和樣條函數(spline functions)以表示共變量關聯。他們還報告了每個方案的估計值，這些數據對 Trickey 及其同事分析的數據很有價值，其中超過 50% 的人在 2 年內停用了拉替拉韋 (raltegravir)。
儘管與 raltegravir 相關的研究結果很有趣並引發了問題，但實際意義充其量是適度的。隨著更簡單、更有效的第二代 INSTI 的出現，拉替拉韋 (raltegravir) 的使用已經減少。它不再被列為首選組合之一。對在懷孕期間使用多替拉韋 (dolutegravir) 的擔憂已經減輕，因此，儘管 2024 年專利到期，但即使在低收入環境中，未來也可能很少使用拉替拉韋 (raltegravir)。
因此，這項研究的重要信息是，儘管 INSTI 具有病毒學優勢，但現代抗反轉錄病毒藥物之間的死亡率差異似乎很小。 我們繼續需要對這些藥物進行長期研究，尤其是隨著 HIV 感染者年齡的增長。 Trickey 及其同事的研究顯示了使用觀察數據進行藥物警戒的能力和挑戰。對世代合作的支持仍然必須是優先事項，以便使用可靠的方法進行長期比較有效性研究。
MBK 報告了來自 ViiV Healthcare、AbbVie 和 Gilead Sciences 的研究支持，以及來自 ViiV Healthcare、AbbVie 和 Gilead Sciences 的諮詢費，這些都在提交的工作之外。 MBK 得到一級加拿大研究主席的支持。 JY 聲明沒有競爭利益。
Marina B Klein, Jim Young ; marina.klein@mcgill.ca
加拿大蒙特婁，麥吉爾大學健康中心，傳染病和慢性病毒疾病服務處，醫學系 (MBK)，流行病學、生物統計學和職業健康系 (JY)，
Can integrase inhibitors reduce mortality?
www.thelancet.com/hiv Vol 9 June 2022
The last major advance in HIV therapeutics came in 2007–08 with the approval of raltegravir, the first integrase strand transfer inhibitor (INSTI). Second generation INSTIs, dolutegravir and bictegravir, with longer half-lives and with a higher barrier to resistance in single tablet regimens have cemented this class of antiretrovirals as the preferred anchor drugs in first-line HIV therapy worldwide.
INSTIs are associated with more rapid and sustained virological suppression and higher CD4 cell count increases compared with non-nucleoside reverse transcriptase inhibitors and ritonavir-boosted protease inhibitors. These characteristics could translate into improved health outcomes and reduce mortality beyond that observed with earlier treatments.
Differences between INSTIs in pill count, pharmacokinetic boosting, resistance profiles, and side-effects could translate into differences in long-term outcomes. Trials directly comparing various INSTIs in people with HIV who are antiretroviral therapy (ART) naive have not revealed major differences in efficacy. However, differential toxicities, such as weight gain, could increase the risk of diabetes and cardiovascular disease, which might only manifest after many years of exposure.
Therefore, observational studies are needed to assess long-term outcomes. To date, such studies have not shown reduced morbidity or mortality associated with INSTIs over periods of up to 6 years.
In The Lancet HIV, Adam Trickey and colleagues compared all-cause mortality between different regimens in 62 500 ART-naive people with HIV starting antiretrovirals in North America and Europe between 2013 and 2018. This is the largest study to date to examine mortality associated with initial ART exposure in the era when INSTI use was common.
The authors make multiple pair-wise comparisons between six anchor drugs, stratified by cohort and calendar time, and conclude that there is little evidence that mortality rates differed between most first-line antiretroviral regimens. However, raltegravir-based regimens were associated with higher mortality compared with all other INSTIs, non-nucleoside reverse transcriptase inhibitors, and protease inhibitors.
Given the results of previously published studies, this finding is both unexpected and surprising; consequently, it needs careful scrutiny. Potential explanations for a higher risk of mortality with raltegravir include: higher rates of virological failure and resulting crossclass resistance; some previously unknown clinical toxicity associated only with raltegravir; and important differences in the characteristics of people exposed to raltegravir, which place them at an increased risk of death compared with people on other INSTI-based regimens.
Differences in time to virological failure were observed between regimens, with dolutegravir having the lowest risk. Raltegravir had a higher risk compared with dolutegravir, elvitegravir, and rilpivirine but not compared with efavirenz or boosted darunavir; therefore, a higher rate of virological failure is unlikely to completely explain the increased risk of mortality relative to all other anchor drugs.
In early clinical trials, concerns were raised that raltegravir might be associated with increased cancer risk. This concern was not borne out in subsequent studies. No other long-term clinical outcomes specific to raltegravir have emerged in the 15 years since its registration. Unfortunately, cause-specific mortality was not available for the present study.
The authors acknowledge that results could be influenced by selection bias or residual confounding. For example, compared with dolutegravir, those starting raltegravir had lower CD4 cell counts (25% of people who received dolutegravir compared with 33% of people who received raltegravir had 100000 copies per mL), and were more likely to present late and have a history of AIDS, tuberculosis, and non-AIDS defining malignancies at initiation. Of note, people who received raltegravir have similar results to those initiating darunavir and efavirenz. Regression adjustment cannot be relied upon to completely remove confounding and is particularly sensitive to the way covariate-outcome relationships are modelled when there are major differences between treatment groups. The authors made considerable efforts to explore possible biases. They analysed a subset of cohorts with additional potential confounders; fitted models in two different time periods to reflect potential changes in reasons for antiretroviral initiation; explored different methods for representing missing data; fitted separate models for people presenting early or late; and adjusted for potentially informative loss to follow-up. With each sensitivity analysis, the association between raltegravir and mortality persisted.
Pharmacovigilance research like this generates hypotheses for new pharmacoepidemiological study. A textbook example of this is the target trial approach by Cole and colleagues, in which an apparent higher 4-year risk of AIDS or death in people receiving raltegravir compared with those who receive efavirenz disappears after accounting for non-random sampling, treatment allocation, and dropouts. Cole and colleagues use time dependent inverse probability weights to allow for possible time dependent confounding in loss to follow-up and spline functions to represent covariate associations. They also report per-protocol estimates, valuable with data like those analysed by Trickey and colleagues9 where more than 50% of people switched off raltegravir within 2 years.
Although the findings associated with raltegravir are intriguing and raise questions, the practical implications are at best modest. Use of raltegravir has waned with the arrival of simpler and more effective second-generation INSTIs. It is no longer listed as one of the preferred combinations. Concerns have abated about the use of dolutegravir during pregnancy and consequently, although off patent in 2024, there is likely to be little use of raltegravir in future even in low-income settings.
Therefore, the important message from this study is that there appears to be very little difference in mortality between modern antiretrovirals despite the virological advantages of INSTIs. We continue to need long-term studies on these drugs, especially as populations with HIV get older. The study by Trickey and colleagues9 shows both the power and the challenges of using observational data for pharmacovigilance. Support for cohort collaborations must remain a priority to make possible long-term comparative effectiveness studies using robust methods.
MBK reports research support from ViiV Healthcare, AbbVie, and Gilead Sciences and consulting fees from ViiV Healthcare, AbbVie, and Gilead Sciences all outside the submitted work. MBK is supported by a Tier I Canada Research Chair. JY declares no competing interests.
Marina B Klein, Jim Young marina.klein@mcgill.ca
Division of Infectious Diseases and Chronic Viral Illness Service, Department of Medicine (MBK), Department of Epidemiology, Biostatistics and Occupational Health (JY), McGill University Health Centre, Montreal, QC H4A 3J1, Canada