Section written by Harvey S. Bartnof, MD
For a more expanded version of this report, see Highlights about Treatment Interruptions in HIV/AIDS

Treatment Interruptions
There were 22 presentations at the 8th Annual Retrovirus Conference that addressed treatment interruptions in HIV/AIDS, including one oral session about the topic. Treatment interruption strategies should be considered highly experimental and risky due to the potential for developing drug resistance in some cellular or organ compartments, reseeding immune system reservoirs with newly rebounded HIV, acute HIV symptoms, and a decrease in the CD4 count that can lead to life-threatening AIDS opportunistic conditions. Currently, treatment interruptions are best performed in the research setting. No patient should stop treatment for HIV until he/she has discussed the issue with a knowledgeable HIV/AIDS physician.

Bruce Walker, MD of Harvard Medical School reviewed the potential benefits of treatment interruptions for HIV infection. They include: to improve immune responses to HIV; to decrease toxicities associated with HAART; and to "repopulate" with drug-sensitive HIV. (This third goal applies to situation "III" below, when there is drug-resistant, detectable virus.)

The reports about treatment interruptions can be divided into 3 general categories:

I. Treatment interruption(s) when HAART (highly active antiretroviral therapy) was started during primary or acute HIV infection ("PHI," before HIV antibody test is positive and usually with severe "flu" symptoms), and HAART has led to an undetectable HIV RNA viral load;

II. Treatment interruption(s) when HAART was started during chronic HIV infection (after antibody test is positive, usually without HIV-related symptoms), and HAART has led to an undetectable viral load; and

III. Treatment interruption when HAART is associated with a detectable viral load, sometimes called "treatment failure."

Treatment interruptions have several names and readers need to be careful about the specifics when attempting to understand these reports. The most common name is STI or Structured (also Strategic or Supervised, planned) Treatment Interruption(s). Another is SIT or Structured Intermittent Therapy, which refers to stopping and starting treatment at regular time intervals, without regard to viral loads or CD4 counts. Due to a similar strategy that is used to treat certain cancers, this approach also has been called "pulse" therapy. Another specific is whether the report represents observations after interrupting therapy once ("discontinued") or several sequential STIs separated by periods when treatment was restarted. Lastly, an important consideration is that almost all of these reports represent very small numbers of patients that make it difficult to apply the results to all patients.

STI After HAART Was Started During Primary HIV Infection (PHI)
The 4 presentations about this topic included only 49 patients: this limits broad conclusions and applicability to other patients. Some generalizations about these studies are:

(1) In 3 of the 4 reports, HIV RNA was "controlled" spontaneously in 38-46% after one or two STIs, although the lower limit of detection ("undetectable") varied. In the fourth study with 15 patients, only 20% achieved virologic control.

(2) Among those patients who underwent sequential STIs, there was a trend towards improved virologic control (lower level) when comparing a second to a first STI.

(3) When measured, there was a trend towards increased immune responses to HIV when comparing a second STI to a first one.

(4) There is a suggestion that starting HAART earlier, i.e., during PHI, may lead to a higher rate of virological control during subsequent STI(s) than when HAART is started within a few months after PHI.

Bruce Walker, MD of Harvard Medical School presented an update about 14 patients enrolled in their study. When first diagnosed with PHI, they had a very high mean viral load of 7 log (10 million) copies per milliliter. The mean duration of HAART before STI was approximately 18 months. The study protocol included restarting HAART if, during the STI, the viral load increased to greater than 50,000 copies per milliliter once or to greater than 5,000 copies per milliliter for three consecutive weeks.

The results were that, after one STI, 46% of the patients spontaneously "controlled" HIV with an undetectable viral load for a minimum duration of over 2 months, up to a maximum of approximately 15 months. Control was achieved after an initial increase in viral load that spontaneously decreased. This meant that the other half of patients did not control HIV during the first STI and had viral rebound that did not spontaneously decrease. However, this group of 6 patients had enhanced control of HIV after the second STI. This is shown by a 5-fold increase in the mean time they were able to be off HAART, when comparing the second STI (184 days) to the first STI (36 days). However, two of those patients did have late viral rebound during the second STI, occurring as late as approximately 10 months into the second STI.

Dr. Walker also reported that the magnitude and breadth of the immune response to HIV had increased significantly during the second STI, when compared to the first. Those responses were "CTL" (cytotoxic [cell-killing]) CD8 lymphocyte responses. These 14 patients will be followed for additional observations.

Regarding starting HAART during PHI, Bruce Walker, MD reviewed several lines of reasoning as to why early treatment with HAART may provide significant benefits for patients, with or without subsequent STIs. First, 8 research groups have shown that decreasing the HIV viral load in PHI is associated with improved T helper cell (lymphocyte) responses to HIV, when compared to no treatment at that time or to starting treatment later. Second is that there is an increased magnitude of immune responses (measured by a "stimulation index" of CD8 cells against HIV "gag" [core protein] in the laboratory) and breadth of CTL responses against HIV. Also, there is much less HIV viral diversity in PHI or early infection than in chronic infection. Early treatment would hinder HIV’s becoming more diverse (hindering increased "quasispecies"), and a less diverse virus would be easier to control by the immune system.

Dr. Walker presented 9 conclusions about treating HIV during PHI and subsequent STI(s).

(1) It is possible for HIV infection to be controlled by the immune system.

(2) Treating PHI can influence the immune response to HIV.

(3) After treating PHI, subsequent STI(s) can lead to enhanced immune control of HIV.

(4) At the present time, there has not yet been any evidence of clinical benefits to STI strategies (patients will need to be followed longer).

(5) The most important measurement of the immunological success of STI is likely to be the resulting "viral set point" (the stable viral load measurement that correlates with the future rate of CD4 cell loss and the rate of progression to AIDS.)

(6) The strategy of STIs after HAART started during PHI that leads to persistent viral rebound will likely provide insights into "the correlates of relative immune failure."

(7) Treatment termination (one STI without subsequently restarting treatment) after HAART was started during PHI "may be less effective than [sequential] STI[s]."

(8) STI after HAART is started for chronic infection as a potential strategy for boosting the immune system represents a "more formidable obstacle" than STI after HAART that is started during PHI: the main reason is HIV viral diversity.

(9) It will be important to understand HIV viral diversity in order to understand the immune system’s response to HIV.

STI After HAART Was Started During Chronic Infection
Most of the reports about STIs fall into this category. C. Fagard, MD presented an update of the largest STI trial, the SSITT (Swiss-Spanish Intermittent Treatment Trial) with 128 patients (31% women). In this study, patients must have had persistent viral undetectability (limit 50 copies per milliliter) for at least 6 months on HAART, without any prior anti-HIV therapy. All were required to be naïve to the NNRTI drug class. The study calls for 4 successive STIs, each 2 weeks long, and each followed by 2 months of restarted HAART. At the end of those 4 cycles (40 weeks), HAART will be discontinued for 12 weeks. Study endpoints at that time (1year into the study) will include the percentage of patients who maintain a viral load less than 3.7 log (5,000) copies per milliliter, the amplitudes of viral rebound during the 4 STIs, and CD4 count changes.

The pre-HAART median baseline CD4 count was 388 cells per microliter, with a median viral load of 4.5 log (31,622) copies per milliliter. Pre-STI, the median duration of HAART was 26 months, with a median 22 months of viral undetectability. HAART included one PI and 2 NRTI drugs for 77%; 8% were taking non-HAART regimens with only 2 NRTI drugs. The median CD4 count before the first STI was 740 cells per microliter.

The interim results for the 75% of patients who completed their 4th STI, 11% had no viral rebound, defined as a viral load less than 2 log (100) copies per milliliter. The median viral load during the 4th STI was only approximately 2.9 log (708) copies per milliliter. High rebound (greater than 5 log or 100,000 copies per milliliter) during the 4th STI occurred among 3%. The highest percentage of no rebound was 13% that occurred during each of the first 2 STIs. These various differences were not statistically significant. During the 8-week retreatment periods, 5-8% of patients during each period were not able to resuppress virus to a level less than 50 copies per milliliter. Another 4 patients had a CD4 count less than 400 cells per microliter during the 4th STI. These two negative outcomes represent a major risk for those who start an STI during chronic infection after successful HAART.

Completing the 52-week study to date includes 99 of participants (interim analysis). Among those, 21% had a viral load less than 5,000 copies per milliliter, the main study end-point ("responders"). However, 24% of those had a viral load less than that level before anti-HIV therapy was started. The "non-responders" at one year included many patients who were unable to resuppress virus to undetectability and a few with low CD4 counts, as mentioned above. Some were non-responders due to one or more high viral load measurements (results greater than 4.7 log [50,000] copies per milliliter). Half of non-responders were so designated due to a viral load greater than 5,000 copies at 52 weeks.

Dr. Fagard reported some newer analyses that compared STI responders with non-responders. Even though the median CD4 count before anti-HIV therapy was not significantly different, pre-therapy median HIV viral load was significantly lower among responders (4.2 log or 16,595 copies per milliliter) than among non-responders (4.6 log or 39,810) copies per milliliter). The odds for not responding to STIs were 6-fold greater for each log increase of baseline viral load ("odds ratio" of 5.9). Moreover, none of 32 patients with a pre-treatment viral load greater than 4.8 log (60,000) copies per milliliter were responders. Also, the density of cellular HIV DNA (incorporated among human genes in "chromosomes") in blood immune mononuclear cells was significantly lower at STI start among responders (43 copies per million or "PBMCs," Peripheral Blood "Mononuclear" Cells") than among non-responders (182 copies).

Dr. Fagard concluded that approximately "1 in 5 patients responded (stabilized viral load less than 5,000 copies per milliliter) after 12 weeks off drugs [after 4 consecutive STIs]." Also, a "viral load of greater than 60,000 copies per milliliter before ART (antiretroviral therapy) precluded response." In addition, "CTL [cytotoxic T lymphocyte], as measured by ‘spot-forming’ cells, increases during the 4 cycles of interruption and re-treatment but the correlation between CTL and response remains to be investigated." This largest study of STI is ongoing.

Characteristics of the CD4 cell decline after interrupting HAART for at least 20 weeks (one treatment interruption) were described in a poster by Pablo Tebas, MD of Washington University in St. Louis, Missouri. A total of 31 patients (55% women, 58% African American) had an undetectable viral load (limit 500 copies per milliliter) for at least 3 months while taking HAART. Patients stopped treatment due to personal preference (56%), with the other 43% stopping due to drug toxicity. Before treatment, the median nadir (lowest) CD4 count was 383 cells per microliter, with a viral load of 4.5 log (29,845) copies per milliliter. After a median 20 weeks of HAART, the median CD4 count increased to 635 cells per microliter, with an undetectable viral load (limit 500 copies per milliliter). The mean duration of follow-up after discontinuing HAART was 50 weeks.

The results showed that the mean "decay" (decline) was –16 CD4 cells per microliter per month. In general, there was an insignificantly faster decline of CD4 cells during the first 6 months after treatment interruption, when compared with months after that. Also, the rate of CD4 cell loss was significantly inversely associated with the magnitude of CD4 cell gain after HAART had been started previously. This meant that those who gained more cells on HAART had a faster rate of cell loss during the treatment interruption, while those with a minimal increase of cells on HAART had a slower rate of loss during interruption. The rate of CD4 cell loss was not related to the lowest CD4 count, baseline viral load, CD4 count before treatment interruption, type of HAART, or gender (sex).

Regarding HIV viral load changes, there were no significant differences when comparing the pre-HAART level to the "set-point" (stable level) achieved during the treatment interruption. For 55% of patients, the pre-HAART and post-interruption viral loads were within 0.5 log (3-fold) copies per milliliter and for 88%, were within 1 log (10-fold).

An update about short and long cycles of treatment interruptions was presented by Mark Dybul, MD of the NIAID (National Institute of Allergy and Infectious Disease). The main purpose of these two studies was to determine whether 1- or 2-week durations of treatment interruptions might lead to less toxicity with lower drug costs, yet without losing virologic suppression or CD4 count increases. Observations about an SIT (structured intermittent therapy) of 7 days off-7 days on included 10 patients. Entry requirements included an undetectable viral load (limit 500 copies per milliliter) for at least 6 months (less than 50 copies per milliliter twice at study entry) and a CD4 count greater than 300 cells per microliter while taking HAART. The regimen was ritonavir/indinavir (Norvir/Crixivan) plus d4T (stavudine, Zerit) and 3TC (lamivudine, Epivir). The study protocol called for all patients stopping HAART for 7 days, and then restarting the same HAART regimen for 7 days, in cycles, for a total of 24 months. "Failure" in this study was defined as a viral load greater than 500 copies per milliliter or a CD4 count decrease that was greater than a 25% decline from baseline on two measurements taken after the "off-drug" periods.

The interim results showed that 9 of 9 patients maintained virologic suppression after 3-11 months. However, one of these patients had a 10-day discontinuation period at the second cycle due to non-adherence and transiently rebounded to approximately 3.5 log (3,101) copies per milliliter. After restarting the study protocol, virus was resuppressed with 3 cycles of follow-up. Three other patients did have occasional, low level "blips" in detectable virus. The 10th patient was off study due to non-adherence. At week 12, this patient stopped treatment for 3 weeks and rebounded to 4.3 log (20,142) copies per milliliter. Thereafter, continuous HAART led to resuppression of viral load to less than 50 copies per milliliter.

Regarding lymphocyte counts, no patient had a "consistent decline" in their CD4 cell count during 3-11 months of observation (6-22 cycles). Similarly, there was no significant change in the CD8 ("suppressor-cytotoxic") cell counts. There was no drug resistance detected in 3 patients whose CD4 cells were cultured and tested after 3 and 6 months, including genotype and phenotype (Virco Antivirogram) testing. Blood levels of indinavir after the 7th day off treatment were zero to trace when 4 of the patients were tested at baseline and after 3 and 6 months into the study.

Dr. Dybul also measured the amount of HIV in reservoirs of blood immune mononuclear cells. In the limited sampling, there were no increases in proviral DNA or in infectious HIV particles in cells that were examined, as a result of 7 days cycles of being off HAART. The absence of detectable drug resistance to indinavir and the maintenance of viral suppression are very encouraging. Potential changes in blood lipids, insulin resistance, fat redistribution, and immune responses against HIV were not reported. This study is ongoing.

Dr. Dybul also authored a second poster about a different schedule of interrupted therapy in cycles. This regimen is called "long cycle" structured intermittent therapy of HAART with 4 weeks off then 8 weeks on therapy, in cycles, up to 22 months. The results would represent 30% less medication than continuous HAART. The enrollment criteria were the same as Dr. Dybul’s "7 days on-7 days off" regimen above. A total of 70 patients will be enrolled and randomized to intermittent therapy or to continuous therapy without treatment interruptions. The study endpoints will be viral load and CD4 counts in the two arms, in addition to immune responses against HIV. The interim analysis included 47 patients who have enrolled. The baseline (pre-HAART and pre-interruption) CD4 counts and viral loads were similar to the "7 days off-7 days on" report above.

The interim results were as follows. All patients in the "interruption" arm had viral rebound during the "off drug" periods, up to 6 cycles. Dr. Dybul described "four patterns" of viral load changes during the "off" period for those randomized to the treatment interruption arm. Those results included 15 patients who had completed 3-6 cycles of "4 weeks off--8 weeks on" therapy. Only 20% were able to maintain greater than a 0.5 log (3-fold) copies per milliliter reduction in viral load when comparing the first with subsequent interruptions (pattern 1). In contrast was13% that maintained greater than a 0.5 log increase in viral load when comparing the first to subsequent interruptions (pattern 2). Another 27% maintained similar viral load levels during each interruption, although each patient in this category (pattern 3) had less than 3.8 log (6,000) copies per milliliter during the off periods. The remaining 40% (pattern 4) had variable increases and decreases in greater than 0.5 log when comparing treatment interruptions. When analyzing the viral load changes for all 15 patients that compared the first with the third treatment interruption, the mean decrease was 0.3 log (2-fold) copies per milliliter, which approaches the variability of the lab test itself and might not represent any significance.

CD4 count changes were reported for ten "interruption" patients who had completed four cycles. During the first "off" period, the mean CD4 count decreased by 17%. However, the levels returned to baseline during the "on-HAART" periods. When measuring CD4 levels during subsequent interruptions, they decreased by 3% during the 2nd one, 8% during the 3rd one and 1% during the 4th one. At the same time points, patients in the other "continuous HAART" arm had CD4 count changes that were: 4% increase (1st time point), 2% increase (2nd), 6% decrease (3rd), and 13% increase (4th).

Drug resistance testing was reported for the 15 patients in the "interruption" arm who completed 2-5 cycles. No genotypic or phenotypic resistance (Virco Antivirogram, decreased susceptibility) was detected for 87% of patients. However, one patient had genotypic and phenotypic resistance to AZT (Retrovir) and 3TC (Epivir, both NRTI drugs) before entering the study and developed genotypic and phenotypic resistance to efavirenz (Sustiva, NNRTI drug) during the study. (Note that this patient’s regimen included indinavir [Crixivan, PI drug], AZT, ddI [didanosine, Videx], and efavirenz.) However, four other patients who were taking efavirenz did not develop any drug resistance when tested. Even though the newly detected drug resistance was found only in one patient, this suggests extreme caution against using drug interruption strategies for certain patients with pre-existing drug resistance—particularly for a regimen with an NNRTI drug that has a long "half-life." (During an STI, efavirenz or nevirapine [Viramune] would remain active in the body for hours to days longer than the other discontinued drugs, creating a period of "monotherapy;" this would increase the risk of developing resistance.)

Dr. Dybul concluded that for those "patients who cannot receive continuous therapy…long cycle (structured intermittent therapy) may be an efficacious means to decrease the amount of HAART medication by 30%." The complete analysis of risks and benefits are unknown, since the study is ongoing. It will also be interesting to find out whether the STI strategy of 30% less medication will have any beneficial effects in terms of lower rates and/or severity of adverse effects associated with anti-HIV medication, including increased blood cholesterol, triglycerides (fats), insulin resistance and diabetes (high blood sugar), vascular (artery) disease including "heart attacks," osteopenia/osteoporosis (loss of bone mineral density) and fat redistribution.

STI after a Detectable Viral Load with HAART or "Drug Failure"
There were four presentations that addressed treatment interruption after a detectable HIV RNA with HAART. This has sometimes been called "drug failure" with drug-resistant HIV, although it usually represents drug-related virologic failure (detectable viral load) and sometimes, "immunologic failure" (referring to the CD4 count), but not necessarily "clinical failure" (referring to progression to clinical disease or illness or even death). As mentioned, one reason for this strategy is the observation that interrupting HAART after virologic failure may lead to repopulation with drug-sensitive or "wild-type" (non-resistant, most common in the population) HIV that would respond better to a new course of therapy. Although, most wild-type HIV tested to date has a higher "replication capacity" (ability to grow or reproduce) as measured by one cycle of the PhenoSense HIV (phenotype drug resistance) test from Virologic and is associated with a higher viral load and lower CD4 count. Some researchers have classified certain drug-resistant virus strains as being less "fit," but the Grant report below is the second one to truly document this fact. Note that STI with "drug failure" usually occurs in a more advanced stage of HIV infection, often with previous AIDS illnesses and multi-drug resistant HIV. It is in this population that STIs have led to life-threatening low levels of CD4 counts and documented opportunistic conditions and death in some patients.

One of the more interesting presentations in this STI classification was a "meta-analysis" (combined studies) of 481 patients from 7 locations including New York City, San Francisco, London (U.K.), Calgary (Alberta, Canada), Milan (Italy) and Frankfurt (Germany). The lead author was Dr. C. Sabin of the Royal Free University in London, UK. The inclusion criteria into the retrospective (analyzed after events have occurred) meta-analysis was a viral load greater than 3.7 log (5,000) copies per milliliter after a treatment interruption of at least 2 months that occurred after anti-HIV treatment with at least 3 drugs for at least 3 months.

Before any anti-HIV treatment, the median nadir (lowest) CD4 count was 90 cells per microliter with a peak viral load of 5.5 log (301,995) copies per milliliter. The median baseline CD4 count immediately prior to the STI was 203 cells per microliter with a median viral load of 4.9 log (72,443) copies per milliliter. The median durations of total anti-HIV therapy, of total HAART, and of the current regimen before STI were 33 months, 17 months and just over 5 months, respectively. Ever achieving an undetectable viral load (limit not stated) had occurred among 20% for a median of 6 months. Patients had been exposed to median 6 prior anti-HIV drugs, including 92% with PI drug experience and 49% with NNRTI drug experience. Therefore, the 481 patients had advanced HIV infection with multiple drug experience.

The results were as follows. During the treatment interruption, the CD4 count decreased significantly by a median (normalized) 53 cells, to 117 cells per microliter. This placed many patients at risk for life-threatening opportunistic infections. The median (normalized) increase in viral load was 0.26 log (1.8-fold) copies per milliliter, to 5.3 log (181,970) copies per milliliter. Changes in the CD4 count during the treatment interruption were greatest among those with a higher CD4 count before interruption and a lower nadir (lowest level) CD4 count.

A total of 445 patients (93%) restarted an anti-HIV regimen with a median 3 drugs after the treatment interruption. Factors significantly associated with a CD4 cell response to the new regimen in two different statistical "multivariate models" included: higher CD4 count at the end of the treatment interruption (TI) period; lower viral load at the end of the TI period; and the total number of anti-HIV drugs started at the end of the TI period.

Steven Deeks, MD of the University of California at San Francisco authored a presentation about 22 patients and their response to "salvage therapy" after an STI that occurred during "drug failure." A "salvage" therapy was constructed using phenotype drug resistance results from samples obtained before the STI. The regimen for 86% of the 22 patients included 2 PI drugs (including low-dose ritonavir [Norvir]), at least 2 NRTI drugs and an NNRTI drug. In spite of these patients having moderately advanced HIV disease with significant drug resistance before the STI, the 24-week response to salvage therapy was good, particularly when a new drug class was able to be included as part of the salvage regimen. 11 of 22 patients had a salvage regimen that included one new drug class: NNRTI for 9 of them and T-20 (experimental fusion inhibitor drug) for the other 2 patients. These 11 patients had a median viral load reduction of approximately 3 log (1000-fold) copies per milliliter with 100% of achieving an undetectable level (limit 200 copies per milliliter) and a CD4 count increase of approximately 150 cells per microliter. Correlation with baseline resistance was not presented.

Viral Fitness (Reduced Capacity to Replicate)
Robert Grant, MD, PhD of the University of California at San Francisco presented true viral "fitness" information about HIV isolates from 12 patients in the Deeks study above. Dr. Grant compared the fitness of the dominant HIV "strain" before and during an STI. During the STI, this was "wild-type" (no mutations) HIV, while pre-STI, PI- and NRTI- drug resistant HIV dominated under the "selection pressure" of HAART. The rate of "wild-type" HIV overgrowth during the STI led to the determination that those strains were "more fit" by a factor of 3- to 36-fold. Those "in vivo" (in the living person) changes directly correlated with increased "replicative capacity" on one cycle of the PhenoSense HIV phenotype resistance test from ViroLogic. In turn, the fitness differences correlated with the increase in HIV viral load during the STI. Dr. Grant concluded that "PI-resistant HIV is markedly less fit than PI susceptible virus…[and that] may account for persistent partial viral suppression observed with PI-resistant viremia." Also, that during treatment with HAART, "wild type virus appears to persist, albeit at very low levels." These results are important, since it may be possible one day to obtain a "fitness" (replicative capacity) test on the dominant HIV strain during "drug failure" while on HAART to help determine whether it would be beneficial to continue the same regimen. This might not be necessary if the nadir (lowest) CD4 count and peak viral load levels were available and those numbers were worse than those while "failing" a current HAART regimen.

In a related presentation, Dr. A. Hance of CRIV Bichat-Claude Bernard in Paris, France showed that an STI after "drug failure" with HAART does not cause drug-resistant HIV to disappear. He found that after 13 patients underwent an STI for 3 months, drug-resistant HIV could still be detected, even though genotype resistance testing indicated "wild type" HIV without any mutations. Such "minority" species were present in 1-10% of the total plasma (blood liquid without cells) virus, as measured by "real-time PCR" (polymerase chain reaction) testing. he point is that drug-resistant HIV does not disappear during an STI, and may still be present as ‘minority species’ that could re-emerge and become dominant when specific HAART is restarted.