The importance of evidence for “When to Start”: a response to Dr Myron Cohen

By: .

T-shirt from

Simon Collins, HIV i-Base

This article is prompted by approaches to the debate about when to start HIV treatment, which avoid both the complexity and lack of evidence to inform this key issue for HIV positive people, doctors and health workers, and public health policy.

While this is principally a response to an online interview with the respected researcher Dr Myron Cohen [1], many of the points are similar to other presentations and articles that call for earlier treatment.

In such examples, a radical public health approach to HIV care is presented as self evident, while neglecting to discuss the lack of important data or presence of contradictory evidence. This is a serious omission in an historical context of guideline recommendations that have been wrong on this question more often than they have been right.

The interview – which is far shorter than this response – included interesting ideas about other aspects of treatment, including that we have a wide range of effective drugs, that doctor experience is important to get the best results and that cure research and pipeline drugs may make today’s treatment unrecognisable in ten years time. This response is to emphasise that, like most things in HIV, even when headlines come from extreme statements, the real interest is in the detail.

Two summary action points set the tone for the interview but both have problems.

The first states:

“According to current recommendations, in the absence of contravening forces, when you detect HIV infection, therapy should be started immediately, independent of CD4 count.”

One problem is the implication that “all” guidelines provide this view – many others, produced on the basis of the same data, do not. Another is the urgency implied by using the word “immediately”. (I also have difficulty of picturing with any precision, an image of a “contravening force”.)

The second states:

“Point out that another benefit of early treatment is that, at least for heterosexual transmission, early treatment renders people no longer contagious”.

This seems more reasonable, but while the language of “contagion” is not ideal, protection is only dependent on the important caveats detailed several years ago in the Swiss statement.

The interview then starts by saying that both the IAS-USA and the DHHS US guidelines recommend “immediate” treatment irrespective of CD4 count, again emphasising urgency. Dr Cohen recognises: “This is a pretty big change, and it represents the accrued benefits, which are very, very strong”. So a huge change, with “very, very strong benefits”. Clearly something major has happened in the guidelines – which is true – but it is implied that this is due to incredibly certain benefits that must outweigh the risks. Because they are guidelines, the data must also be clear and incontrovertible.

We then learn that “earlier treatment is better” and that “this is pretty much written in stone”.

HIV is compared to hypertension, where people can experience kidney or eye damage while feeling fine, and still need treatment. This is presented as being so obvious that it doesn’t need discussing and the consequences will be scary for people not on treatment. The analogy with hypertension is not helpful, especially when we are focusing on clinical outcomes from the difference between starting treatment “immediately” compared to starting at CD4 350. The data for this remain extremely limited and inconsistent, and are the focus of the ongoing randomised START study.

Dr Cohen qualifies the statement by needing to start treatment immediately “to avoid another AIDS patient” which again sounds scary, but trying to connect these two things jumps several important steps. Symptomatic late stage HIV, or “AIDS patients” are predominantly due to late diagnosis. Just under half of all people diagnosed with HIV in the UK last year already had a CD4 count that was less than 350. This is linked to a 10-fold increased risk of death within a year compared to people diagnosed at higher CD4 counts. [2] The most effective way to reduce “AIDS patients” is to invest in programmes that reduce late diagnosis.

The context of guideline changes is then presented as a simple linear progression: “We started out at 200, then moved it up to 350, [some countries moved to 500] and the US has moved to start people immediately”.

For the first years of ART, guidelines originally recommended early treatment at high CD4 counts (500), even with AZT monotherapy. This was dropped to 350 (reluctantly by some) with awareness of the side effects and drug resistance and in some countries dropped further to 200. It is the improved safety and efficacy of more recent drugs that produced a risk:benefit balance weighted back towards earlier treatment. The current evidence does not show the same benefit from starting with a CD4 count higher than 500, and is perhaps not even there for starting above 350 (compared to waiting to 350).

Even with the best intentions, guidelines produced by experts, can be wrong. The limited evidence and lack of randomised data, restricts the ability to know the risks as well as the benefits. This is a lesson that should have been learned from the past. Also, even the best combinations, in best conditions (less complex, healthier patients, in a clinical trial), still fall short of 100% efficacy. Whether 70%, 80% or 90% of patients actually maintain viral suppression, this leaves 30%, 20%, or 10% likely to have complications including resistance, and on a population level these numbers are still way too high to be overlooked.

Dr Cohen stresses the importance of earlier diagnosis, but comes back to the hypertension analogy saying we have over-complicated HIV treatment, and stating that both observational and clinical studies support earlier treatment, and that the HPTN-052 study had a “fairly big impact on health and survival”.

HPTN-052 is a landmark study because it proved clearly that ART reduces the risk of heterosexual transmission and that the magnitude of protection exceeded that reported by condom studies. For this it was ground-breaking. [3] But although earlier treatment was not associated with harm, the clinical benefits were far less dramatic:

  • People in HTPN-052 were starting treatment at a CD4 count of 250 compared to 350-550. So any benefits could be driven by the low start threshold and not the benefit of treating above 350. It is already proven that starting at 350 is better than 250, and the study could just confirm this.
  • It is impossible to find a “fairly big impact on survival” because there was no statistical difference in the number of deaths between the two groups (out of nearly 900 people in each group there were 10 deaths in the early treatment group vs 13 in the group that waited until 250).
  • There was no difference in serious “non-AIDS” events such as major organ complications and some cancers that would be expected from reducing viral load and related immune inflammation/activation.
  • Higher events were driven by extrapulmonary TB from two study sites in India, and although pulmonary TB rates were also higher, this may be a finding that is specific to countries with high TB prevalence. Of interest, in high TB settings, earlier HIV treatment may reduce the risk of new TB infections.

In this context it is difficult to understand the “fairly big impact on health”.

The temptation to merge clinical and survival benefit does not reflect the evidence in the context of starting treatment above 350. Survival is a life and death decision. It is clearly important to doctors and patients. Anyone saying that “immediate” treatment for everyone upon diagnosis will reduce deaths, needs evidence to show this. Merging clinical benefits (less sickness – that is relatively minor and treatable) and survival benefits (less deaths) into a combined endpoint is not good enough unless the data support this. There was no significant difference in survival between the two arms in HPTN-052. It was balanced in each arm with a p-value of 0.5 showing no difference and no hint of a trend.

Combining survival and better health is common in studies. Better health is clearly important. But the results were that earlier treatment improved health by not getting some symptoms. Earlier treatment, at least in this study, did not save lives. Implying it gives a different weight to the story. The evidence does not support scaring HIV positive people with high CD4 counts into earlier treatment with the suggestion that their chance of dying will be reduced. None of the key studies included in Table 1 report a reduced risk of death from starting treatment at higher CD4 counts.

HIV treatment is not like treatment for hypertension. It is far more complex. The risk of resistance is permanent and reduces future options. There are more side effects. The extensive differences in progression rates driven by widely different immunological responses support an approach to treatment that is rightly emphasised in guidelines as demanding individualised patient management.

Although tiny numbers, in HPTN-052 the three cases of suicide in the early treatment group (compared to zero in the deferred arm) were all taking efavirenz. While the causes of death included missing data, broader roll-out of efavirenz-based treatment needs careful support and awareness of this as a potential outcome. Efavirenz is an important drug but it is not tolerated by at least 20% of people in the UK.

Dr Cohen acknowledges that side effects are a problem that needs to be effectively managed, but he hasn’t demonstrated that these risks are outweighed by treatment benefits from starting treatment on diagnosis compared to waiting until 350 or even 500, because the data does not exist. Not from randomised studies, nor from observational cohorts. Randomised data is dependent on the results from the currently ongoing START study, and cohort data has produced conflicting results, with only one database reporting a benefit at the highest CD4 counts (NA-ACCORD, using complex methodology) and others failing to see much or any difference.

Dr Cohen recognises that widely different HIV progression rates highlight a downside to universal treatment on diagnosis. While chronic infection results in a median CD4 count decline of 60 cells/mm3 annually, 25 percent of people progress very rapidly, dropping below 200 within two years. However, 25% of people maintain levels well above this for ten years and smaller percentages are able to maintain CD4 counts >500 for considerably longer, with a much smaller percentage of elite controllers also suppressing viral load without treatment.

The diversity of these responses does not argue for the urgency of “immediate” treatment for everyone. The most significant assault from HIV on the immune system probably occurs within the first weeks of infection. The extensive burst of viral load, commonly to levels higher than 1,000,000 copies/mL, is believed to reach all important parts of the body, persist in some compartments for many months, and decimate a significant bank of CD4 cells based in the gut. Despite this, most people then generate an immune response that fights back and holds HIV at bay for many years, even though gradually losing the fight. While ongoing viraemia during chronic infection is unlikely to be a good thing, and is an important and valid concern that supporting earlier treatment, quantifying the additional risk in context of broadly contained asymptomatic infection in someone with a high CD4 count, has still to be determined.

The relative contributions of primary (highly infectious) and chronic infection to the ongoing HIV transmissions varies from approximately 10%-60% depending on the population being studies and the modeling. This impact is likely to be highest for gay men because of the higher number of sexual partners, and is likely to shift oral sex from a low to high risk. In setting where new infections are driven by primary infection, a possibility in the UK, then earlier treatment is unlikely to have any major impact on a population level.

Dr Cohen notes that by treating everyone, there will no longer be elite controllers, because everyone will be controlled. The smaller number of people who don’t need treatment will be treated anyway, because this is better for everyone. Then, just in case elite controllers were feeling too confident, Dr Cohen uses a double-negative and summons doom saying that just because these people are doing fine – in rare cases maintaining undetectable viral load with treatment for more than 20 years – the lack of evidence doesn’t mean that it is not doing them “subtle” harm. Potential subtle harm is not good enough when trying to calmly assess the risk and benefit of individualised treatment.

Damage from ongoing viral replication off-treatment is quite possible, or even plausible. Some studies have shown biomarker differences between even elite controllers and HIV negative people. But this is currently a research hypothesis and we are talking about lifelong treatment, for possible “subtle” differences. Slow progressors or even elite controllers may not be in the clear, but there is no convincing data to support that they are in urgent danger that requires “immediate” treatment.

Finally, moving to the public health benefit of Treatment as Prevention (TasP), Dr Cohen states that in the context of heterosexual transmission “it becomes pretty clear that treatment renders them no longer contagious though this is modified to “nearly 100% protection when adherent” to recognise that some transmissions still occur. This is HTPN-052 data. Dr Cohen is the lead investigator for this study. The results are clear. But on a population level, being HIV positive does not necessarily mean that you transmit HIV. On an individual level, you may not even be having sex. You may be having exclusively safe sex, either by the choice of activity or careful and consistent condom use. Or your partner may also be HIV positive, and HIV transmission is only an issue in the context of drug resistance.

These are all big caveats when introducing a public health initiative to treat everyone, not least the premise that an HIV diagnosis equates to transmitting HIV, especially when 20-60% of new infections are likely to be come from people undiagnosed in acute infection and a significant percentage more from people undiagnosed in chronic infection.

In the context of informed choice, starting earlier treatment to reduce the risk to your sexual partner(s) can be incredibly important and liberating. Dramatically reducing the level of infectiousness can improve the quality of life for both HIV negative people and HIV positive people in ways that are difficult to describe or predict. TasP can reduce remaining anxiety for both partners and has given many people the peace of mind for improved sexual health when one partner is positive and the other is negative. This comes from knowing that even with continued condom use, if a condom breaks, that passing on HIV is highly unlikely.

But the temptation to merge treatment benefits (probably tangible but slight at CD4 counts over 350) and prevention benefits (dramatically reduced for anyone with sustained viral suppression) needs to be resisted. Some of the most important treatment guidelines, even when reviewing the additional benefit from TasP, are rooted in recommendations for treatment being primarily focused on clinical need. [15, 16, 17]

Dr Cohen acknowledges that there “may be adverse events (side effects) at higher CD4 counts” but says this will be balanced by the “clinical benefit […] of protecting organs”.

Scary language again – you do want to protect your organs, don’t you? This issue is important – the SMART study showed that people on treatment had fewer serious major organ complications (heart, liver, kidney) and reduced risk of some cancers, if they had a suppressed viral load on treatment, compared to people who interrupted treatment, even at CD4 counts over 350. These results from over five years ago showed treatment to be far safer than was originally realised, and have driven a whole field of research into the implications of untreated HIV. The results are real and important, but the absolute risks for all patients were generally low, and this is still part of a theory supporting earlier treatment rather than something that has been proven. However, these results were not seen by people using earlier treatment in HPTN-052, which may be why the clinical benefits in that study are highlighting the disadvantages of waiting until CD4 count reaches 250 rather than a benefit from starting at 350-550.

Global health is only briefly touched on in the interview. The context is important given that in most countries most patients have access to fewer choices. A significant proportion of people, perhaps 50% globally, are still using d4T. Most come to second-line treatment with far more extensive NRTI resistance due to lack or viral load and resistance monitoring, and many are still waiting for third-line options. Simplifying the benefits of earlier treatment irrespective of the setting is unhelpful, especially when the benefits become increasingly marginal at higher CD4 counts. Would the benefits of d4T-based treatment outweigh the risks on diagnosis, at 500, at 350 or at 200? When this was the option in Western countries, we set our guidelines at 200. When d4T is not a factor, then 350 has the strongest evidence. In the context of priority, in public health, in lives saved and health maintained, treatment above 500 for clinical benefit is not even on the radar in most countries.

To further emphasise the urgency of immediate treatment, Dr Cohen says that the concept of a person being “ready to start” treatment suggests that a doctor-patient discussion to arrive at this position might project “a false sense of security” that “all is well”. This central tenet of treatment guidelines – readiness to start – is one that activists have demanded and supported and wish to hold on to. Whether someone is starting treatment on diagnosis, using a CD4 threshold of 500, 350 or 200 – for their personal health or to reduce the risk to their partner – it should be an informed choice. Scaring people into the decision, whether for future health risks or on a public health agenda, will help no-one.

The advantage of earlier treatment to maintain a higher CD4 count is discussed. But the examples given are easy – starting at 440 compared to 220. The context of immediate treatment needs to aim higher, and with confidence. What about starting at 1100 compared to 900. Or with a CD4 count of 900 compared to 700. After 15 years experience with ART, the best cohort studies struggle to find consistent evidence for 700 compared to 500.

All we have is the plausibility of benefit and risks that are presently unquantified.

CD4 counts have a wide “normal” range that for most tests have a lower range of about 450, and 2.5% will have normal levels lower than this. Although the decision to start treatment is broadly taken based on observing over time an impact from HIV on this count, a threshold of 500 for starting treatment, makes it likely that some people will be recommended to start treatment without any immune damage relative to their pre-infection levels.

Similarly, the benefits of treatment may reach a cap after five or ten years. In this case, the argument that over a lifetime there is little difference between 35 or 40 years on treatment becomes seriously flawed. For people with no clinical urgency for treatment, this might become a difference of between 25 or 40 years, and many people might see this as significantly different.

As an aside, but clearly of  interest, a comprehensive review of the risks from the move to earlier treatment, was presented by Dr Cohen two years ago at the BHIVA conference and is still available as webcast. This was a pretty vociferous presentation against what he described then as “The Evangelical Test And Treat Movement”.  [4]

While that talk was an overly cautious and pessimistic review, it is difficult to see the volume of evidence since 2010 for Dr Cohen to have such a rapid reversal in his views (see Table 1).  Even the NA-ACCORD study – controversial in its methodology, but providing the strongest observational cohort data supporting earlier treatment at CD4 >500 – had been published 18 months earlier (in April 2009).  And what happened to concerns about drug levels in the genital tract, persistent shedding, lack of data, modeling problems, cost, the “inconvenient truth” about acute transmissions and even “Presidential support for the bandwagon”?

The plausibility of potential benefits of treatment on diagnosis has been argued since the days of AZT monotherapy. HPTN-052 demonstrated solid evidence for the impact treatment has on prevention but had far more tentative data on clinical benefits, especially for people in Western countries with higher CD4 counts. It didn’t provide the evidence needed to a shift to universal treatment on diagnosis.

Table 1 below includes references for the most significant studies and guidelines relating to when to start treatment. Most of the studies struggle to find evidence for benefits at CD4 counts over 500 and most of the guidelines, recognise the limited evidence on which their recommendations are made.

In summary, HIV positive people should have the option to start treatment at any CD4 count, especially if this choice is driven by wanting to reduce the risk of transmission to sexual partners. To be an informed choice, this needs to acknowledge that the evidence for personal health benefits has plausibility, but limited data, especially at high CD4 counts.

Currently, the evidence (and expert interpretation of the same evidence in different guidelines) still supports equipoise on the question of whether benefits outweigh the risks of earlier treatment at CD4 counts above 350. Results from the START study, expected in 2016, will provide the strongest real data to inform this question. [19]

Prejudging the results from START – in other words, guessing – helps no-one. Until then, a wide range of studies suggest a low absolute risk from starting earlier treatment if this is an individual’s choice and a low absolute risk from deferring until 350 if that is an individual’s choice. This is especially important to remember for people enrolled in the START study, who will ultimately help settle this key question.

Table 1. Evidence base for earlier treatment

Study or document Summary comment Refs

Randomised clinical trials (RCTs)

This type of study provides the most reliable evidence based on both risks and benefits of an intervention.

CIPRA HT-001 (Haiti) (2010)

Clear health and survival benefits were shown from starting at 200-350 vs waiting until less than 200. [5]

SMART (naive sub group)

Waiting until CD4 <250 has increased risk of serious events (compared to group with median CD4 440). [6]


Reduced risk of health events when starting at 350-550 compared to waiting until <250 (ie starting late). No difference in survival. [3]

Cohort studies

This type of study is used when RCT data is not available.

It usually involves many more people than an RCT but the results are less reliable.

This is because without randomisation it is impossible to adjust for the other things that make people who start treatment earlier different from people who start late.

ART-CC (2009) ART-CC showed combined health and survival benefit from starting at 350-450 compared to waiting until 250-350 but no significant difference in survival. No survival difference when starting at 450-550 compared to 350-450. [7]
NA-ACCORD (2009) Starting at 350-500 reduced the risk of death compared to waiting until <350. Survival benefit reported when starting >500 compared to <500 but very small numbers of deaths so absolute impact of ART is not possible to estimate. The approach to analysing data has also been questioned. [8]
HIV CAUSAL (2011) Estimated that starting at 350-500 has a reduced risk of an AIDS defining illness compared to <350. No difference seen in survival. [9]
CASCADE (2011) No significant difference in illness or death starting at 350-500 compared to <350. No benefit in either health or survival from starting at 500-800. [10]
COHERE (2012) Benefits in health outcomes including deaths were reduced as CD4 category increased. This included people starting at >500 but the paper notes that this is so small as to be of “little clinical relevance for most patients”. [11]

Concern about inflammation

SMART (2008)

ANRS (2009)

Growing awareness and concern with period of detectable viral load. Link to increased risk of serious “non-AIDS” events including heart, kidney, liver disease, neurological complications and some cancers.  See the 2013 DHHS guidelines below for an excellent summary and discussion but the references (mainly from 113-127 in the when to start section E) outline basis for concern. Only two are from 2011 and most are significantly earlier. [12, 13]
Key guidelines

The guidelines help explain the complexity of the limited evidence.

It is important to read the text and not just the table summary recommendations.

All three guidelines come to different conclusions based on the same evidence.

US DHHS (2012)

Evidence base of clinical benefits is reduced as CD4 count gets higher. Strongest recommendation is for starting <350. Then for 350-500. The benefits for starting above 500 are based on expert opinion worried about the potential implications of keeping a detectable viral load. Prevention is discussed but clinical benefits are the main focus of these guidelines.

IAS-USA (2012) Recommendation to offer treatment to all patients at any CD4 count. Strongly influenced by the potential impact on reducing transmission. However the strength of evidence was highest for starting at <500. There was only moderate support for starting at >500, and this was based on the lowest rating for evidence. [15]
BHIVA (2012) UK guidelines recommend starting at 350 or before CD4 count drops below 350. Based on lack of randomised data showing earlier benefit and conflicting results from observational studies. [16]

(2009, 2011)

WHO treatment guidelines use 350 for clinical benefit based on moderate evidence for clinical benefit. In guidelines for serodifferent couples, ART use for prevention is supported by strong evidence. [17, 18]

Simon Collins is a community representative on the START study and a member of the Community Advisory Board for the INSIGHT group.


  1. Smith M. Interview with Dr Myron Cohen: When to start therapy, a clinical context report. MedPage Today. (5 January 2013).
  2. Health Protection Agency (HPA). HIV in the United Kingdom: 2012 report.  (PDF file)
  3. Cohen MS et al. Prevention of HIV-1 infection with early antiretroviral therapy. N Engl J Med 2011; 365: 493–505.
  4. Cohen M. The HIV ‘test and treat’ movement: the details really matter, and now. BHIVA Autumn Conference 2010. Web cast online.
  5. Severe P et al. Early versus Standard Antiretroviral Therapy for HIV-Infected Adults in Haiti. N Engl J Med 2010;363:257-65.  (PDF)
  6. Emery S et al. Major clinical outcomes in antiretroviral therapy (ART)-naive participants and in those not receiving ART at baseline in the SMART study. J Infect Dis. Apr 15 2008;197(8):1133-1144.
  7. Sterne JA et al. When To Start Consortium. Timing of initiation of antiretroviral therapy in AIDS-free HIV-1-infected patients: a collaborative analysis of 18 HIV cohort studies. Lancet. 2009;373(9672):1352-1363.
  8. Kitahata MM et al. NA-ACCORD Investigators. Effect of early versus deferred antiretroviral therapy for HIV on survival. N Engl J Med. 30 April 2009;360(18):1815-1826.
  9. Cain LE et al. When to initiate combined antiretroviral therapy to reduce mortality and AIDS- defining illness in HIV-infected persons in developed countries: an observational study. Ann Intern Med. Apr 19 2011;154(8):509-515.
  10. Writing Committee for the CASCADE Collaboration. Timing of HAART Initiation and Clinical Outcomes in Human Immunodeficiency Virus Type 1 Seroconverters Arch Intern Med. 2011;171(17):1560-1569  (PDF)
  11. Opportunistic Infections Project Team of the Collaboration of Observational HIV Epidemiological Research in Europe (COHERE) in EuroCoord.  CD4 cell count and the risk of AIDS or death in HIV-Infected adults on combination antiretroviral therapy with a suppressed viral load: a longitudinal cohort study from COHERE.  PLoS Med. 2012;9(3):e1001194.
  12. Mugavero MJ et al. Viremia copy-years predicts mortality among treatment-naive HIV-infected patients initiating antiretroviral therapy. Clin Infect Dis. Nov 2011;53(9):927-935.  (PDF)
  13. Guiguet M, Boue F, Cadranel J, Lang JM, Rosenthal E, Costagliola D. Effect of immunodeficiency, HIV viral load, and antiretroviral therapy on the risk of individual malignancies (FHDH-ANRS CO4): a prospective cohort study. Lancet Oncol. Oct 7 2009;10(12):1152-1159. 
  14. US DHHS. Guidelines for the Use of Antiretroviral Agents in HIV-1-Infected Adults and Adolescents. March 2012.  (PDF)
  15. IAS-USA. Antiretroviral Treatment of Adult HIV Infection 2012 Recommendations of the International Antiviral Society–USA Panel. JAMA, July 25, 2012—Vol 308, No. 4.
  16. British HIV Association guidelines for the treatment of HIV-1-positive adults with antiretroviral therapy 2012. HIV Medicine (2012), 13 (Suppl. 2), 1–85. DOI: 10.1111/j.1468-1293.2012.01029_1.x.  (PDF)
  17. WHO. Rapid advice. Antiretroviral therapy for HIV infection in adults and adolescents. November 2009.
  18. WHO. Guidance on couples HIV testing and counsApril 2012.elling – including antiretroviral therapy for treatment and prevention in serodiscordant couples.
  19. Strategic Timing of Anti Retroviral Therapy (START). 


Your email address will not be published. Required fields are marked *