Recommendations for chemoprophylaxis discontinuation are based on CD4 counts 
. However, there are limitations with using the peripheral blood CD4 count as the sole marker of immune response. It is estimated that only 2% of the total pool of CD4 T cells are present in blood at any time, and that even a small shift in the distribution of cells between the lymphoid tissues and blood cells may result in a large alteration in peripheral blood CD4 count 
. In many studies suggesting it is safe to discontinue PCP prophylaxis when CD4 count is >200 cells/µL, this rise in CD4 count was accompanied by a decline in VL 
. Furthermore, patients with similar CD4 counts frequently progress at different rates despite taking the same medications 
. These findings suggest that CD4 count alone is not the sole predictor of susceptibility to PCP.
Improvement in immune function associated with lower VL has been observed when assessing responses to vaccinations. For example, lower VLs have been associated with improved responses to hepatitis A, hepatitis B and varicella vaccinations independent of CD4 count 
. In a large survey of an HIV-infected cohort in the United States, a protective effect of PPV-23 vaccination was also shown against all-cause pneumonia, and this benefit was lost in patients with VL >100,000 copies/mL, irrespective of the CD4 count 
. Moreover, in a randomized, double-blind comparative trial of subunit and virosomal influenza vaccines for immunocompromised patients, suppression of VL with HAART was a more important predictor of response to influenza vaccination than the CD4 count 
In addition to discontinuation of prophylaxis after initiation of HAART, the COHERE study also analyzed the general incidence of PCP in persons who either never initiated prophylaxis or discontinued prophylaxis prior to initiating HAAART. They demonstrated an overall event rate of 0.12 events per 100 PYFU (95% CI 0.02–0.45) when the CD4 count is between 101–200 cells/µL and VL is suppressed in patients not receiving prophylaxis. This was similar to the event rate of 0.21 episodes per 100 PYFU (95% CI 0.08–0.43) in persons with CD4 counts between 101–200 cells/µL and suppressed VL receiving prophylaxis. The low overall incidence rate of PCP when CD4 count is 101–200 cells/µL may serve as supporting evidence that prophylaxis may not be required in patients with suppressed VL and CD4 count in the 101–200 cells/µL range.
Another important observation from the COHERE study was that the risk of PCP differed based on the CD4 strata in individuals who discontinued prophylaxis after initiating HAART. While it appeared safe to discontinue prophylaxis with CD4 count between 101–200 cells/µL, the upper confidence interval level of the incidence rate in persons who discontinued prophylaxis with CD4 count ≤100 cells/µL after initiating HAART was high (9.83 events per 100 PYFU), suggesting an unacceptably high risk of PCP in this one study.
The studies included in this review differed in several important aspects. Most notably, not all studies examined the same primary outcome (i.e. incidence of PCP during follow-up in patients off prophylaxis with CD4 counts <200 cells/µL and suppressed VL), thus we were unable to perform a meta-analysis. Furthermore, some studies examined discontinuation of PCP primary prophylaxis whereas others examined discontinuation of PCP secondary prophylaxis. Individuals receiving PCP prophylaxis for primary compared to secondary prophylaxis may differ in that having had PCP in the past increases one's risk for the development of subsequent episodes 
. Moreover, in addition to adults, the COHERE cohorts also include children and mother-infant pairs. However, the number of adults, children and mother-infant pairs who had CD4+ <200 cells/µL and suppressed VL on HAART is not reported.
Studies also differed significantly with respect to sample sizes and follow-up conditions. Sample size ranged from 19 subjects in the D'Egidio study to greater than 20, 000 in the COHERE study. In the COHERE study, one individual could provide several episodes of PCP to one CD4/VL stratum, while in the other studies individual follow-ups are examined. In the Soriano et al. study, follow-up included time with CD4 count >200 cells/µL, therefore likely underestimating the incidence of PCP and overestimating safety by prolonging follow-up. In the Cheng et al. study, follow-up time included persons with detectable VLs, thus likely overestimating the incidence. Follow-up durations whereby patients were off prophylaxis with CD4 counts <200 cells/µL and suppressed viremia differed among studies. Finally, studies were performed at different time points and current VL assays have increased sensitivity compared to those used decades ago. Whereby older assays could only measure VL to 400–500 copies/mL, newer assays can measure VL down to 40–50 copies/mL. This may be important due to deleterious effects associated with even low level plasma viremia, such as direct viral cytopathogenecity, persistent immune activation with apoptosis of uninfected cells, and decreased thymic output of naïve CD4 T cells 
Limiting the use of prophylaxis against PCP to situations where it has been shown to be most beneficial will minimize adverse drug reactions. The AIDS Clinical Trials Group (ACTG) 081 was a randomized clinical trial (RCT) of 842 HIV-infected individuals which compared the effectiveness of three alternative prophylactic regimens. There was no difference among treatment arms with regards to adverse events. Prevalence of patient self-reported fatigue was 67%, headache 61%, sleep disturbance 50% and dermatological problems 46%. Fever, sense of imbalance, paresthesias, diarrhea and cough all had self-reported prevalences greater than 30%, demonstrating a substantial burden of adverse effects 
. In an earlier meta-analysis by Ioannidis et al., which examined RCTs comparing efficacy and toxicity of different PCP prophylaxis regimens, trimethoprim-sulfamethoxazole was almost universally efficacious at preventing PCP when tolerated. The risk of discontinuing trimethoprim-sulfamethoxazole due to adverse reactions depended on dosing schedule and decreased by 43% if one double-strength tablet was given three times weekly instead of daily. For dapsone, among 100 patients given 100 mg orally once daily instead of twice weekly over a year, the authors determined that 7 fewer patients would develop PCP but 17 more would have significant toxic reactions 
. Of the 4 studies included in our systematic review, only Cheng et al. reported the incidence adverse effects due to PCP prophylaxis. They estimated the crude incidence of adverse effects for primary prophylaxis with trimethoprim-sulfamethoxazole to be 23.3%, of which 13.3% were due to skin rashes, 6.8% due to leukopenia, 2.9% due to gastrointestinal intolerance and 0.4% due to Stevens-Johnson syndrome. Of 47 patients (21.8%) who developed adverse events leading to discontinuation of secondary prophylaxis, adverse events included skin rashes in 17.6%, gastrointestinal intolerance in 2.3% and leukopenia in 1.9% 
. However, it should be noted that these rates do not apply exclusively to individuals with CD4 counts <200 cells/µL with suppressed VL on HAART.
An important limitation of our systematic review is the heterogeneity between the studies examined. Although this precluded us from performing a meta-analysis, we pooled results to include the totality of available, relevant evidence in our systematic review. Another limitation of our study is its retrospective nature in which we were dependent upon the availability and accuracy of data reported by others. Furthermore, none of the studies included were randomized controlled trials, eliminating the ability to control for confounding factors which may have influenced results.
The question of whether it is safe to discontinue prophylaxis with suppressed VL and CD4 count <200 cells/µL would be best addressed with an RCT. However, this is unlike unlikely to occur in the foreseeable future. As a meta-analysis was not possible due to the heterogeneity between studies, here we conducted a systematic review of the available evidence. Though studies are few in number, when examined together there is sufficient evidence to demonstrate that the risk of PCP infection in HIV-infected individuals on HAART with suppressed VL is sufficiently low to allow discontinuation of primary prophylaxis, despite a CD4 count between 101–200 cells/µL. Given the low overall risk of PCP with CD4 count between 101–200 cells/µL, combined with the drawbacks of prophylaxis for OIs, we suggest a revision of clinical guidelines for primary prophylaxis is merited to include consideration of the VL. However, as only one study examined the risk of PCP prophylaxis discontinuation at different CD4 strata, including ≤100 cells/µL, it may not be appropriate to apply this recommendation to individuals with CD4 counts of ≤100 cells/µL at this time.