(introduction...)

In this section, you will be presented with a description of the ARV regimens you wish to evaluate including their efficacies and dose patterns. The corresponding Excel spreadsheet is called “ARVs”.

The choice of ARVs, doses, timing of doses and estimated efficacy of these regimens are derived from clinical trials and are given by the model (blue cells). The trial-derived efficacy figures can then be adjusted to reflect imperfect adherence (relevant if you believe that adherence in your setting will be different from adherence in the trials); and late arrival for treatment (relevant for the three regimens with a prepartum component - ACTG 076; CDC-Thai; and PETRA-A). In addition, the user enters the cost of each dose to mother and to infant. Based on this cost data, the intended number of doses in the regimen, and the number of mothers who receive each component of the regimen, the model then calculates the drug costs per woman and annual drug costs for the expected number of women to be treated in the service area.

TABLE ARV-1: REGIMEN COMPONENTS

Background

The spreadsheet permits the analysis of five different ARV regimens that are based on clinical trials that were able to demonstrate clinical and statistical significance:

· ACTG 076 -- Zidovudine prepartum treatment starting at week 28 of gestation, plus intrapartum, and postpartum (mother and child) treatment.

· HIVNET 012 - One 200 mg NVP to mother at outset of labor; 2 mg/kg to infant once within 72 hours of birth.

· CDC-Thai -- 300 mg dose AZT twice daily from 36 weeks’ gestation and every 3 hours during labor through delivery..

· PETRA-A -- 300 mg AZT and 150 mg 3TC twice daily from 38 week’s gestation through labor and 300 mg AZT every three hours of labor through delivery; 300 mg AZT and 150 mg 3TC twice daily for a week postpartum; 5 mg/kg AZT and 2 mg/kg 3TC for infant every 12 hours for week postpartum

· PETRA-B -- Equivalent to PETRA-A without prepartum therapy.

C4 - G9. Table ARV-1: Regimen components. This table provides a summary of which component (prepartum, intrapartum and postpartum) is included in each regimen. C5 indicates that those who arrive too late for the full prepartum portion of ACTG 076 will receive a shorter prepartum course as a substitute. Since it may be hard to justify providing no treatment to those who arrive too late for the full treatment, we have set this to “yes”. D5 - G5 read “NA” to indicate that since the other three regimens do not include a long prepartum component, there is no substitute for a long prepartum component. No data entry is require for this table.

TABLE ARV-2: GESTATIONAL AGE AT START OF TREATMENT

Background

In the “Settings” sheet you entered information about the percentage of women who were available for VCT and treatment by gestational age. Table ARV-2 uses this information combined with an estimate of the latest prepartum treatment arrival time consistent with receiving the full efficacy of the prepartum portion of the regimen; and the latest arrival time consistent with receiving partial efficacy from the prepartum component. Unfortunately, there are no definitive data bearing on the relationship between length of prepartum treatment and efficacy. We therefore provided numbers based on reasonable guesses as follows:

Relationship between length of prepartum treatment and efficacy

Regimen	Treatment start time (gestational age)	Efficacy
ACTG076	34 weeks or earlier	67.5% (full efficacy)
ACTG076	> 34 weeks through 38 weeks	55% (reduced efficacy)
ACTG076	> 38 weeks	40% (reduced efficacy)
CDC-Thai	38 weeks or earlier	51% (full efficacy)
CDC-Thai	> 38 weeks	30% (reduced efficacy)
PETRA-A	38 weeks or earlier	50% (full efficacy)
PETRA-A	> 38 weeks	37% (reduced efficacy)

If you wish to explore the effect of a different set of efficacy adjustments, you may do so by altering the values in the table located in cells “ARVs” X99 - AE107. (You must first unlock the worksheet using the password as explained on page 6). Using the information provided in the table above) on arrival times and efficacy, combined with the data on the cumulative percent of women who arrive at various gestational ages, we calculated a weighted average efficacy adjusted for late arrival. This is shown in row 63 of “ARVs” and is discussed further in the section of the manual pertaining to “Table ARVs-5”.

Influence on cost-effectiveness: Medium. Since late arrival for prepartum treatment affects regimen efficacy, these estimates can have a significant impact on cost-effectiveness. This is particularly true if ACTG 076 is being evaluated since it has the longest prepartum treatment period, and relative to other regimens, the greatest portion of its benefit is presumed to occur during the prepartum period.

Expected effort of data collection: Medium. See explanation for “Setting” C27 -C34.

Adjusting for late arrival Enter for ACTG-076 only C13. Latest gestational age in weeks that a woman can arrive at clinic and still receive full efficacy for ACTG 076. The default is set at 34 weeks. C14. % of women who receive any prenatal care who arrive in time for ‘long’ course of ARV. This figure is derived from estimated cumulative arrival time from cells “Setting” C27 through C34. C16. Latest gestational age in weeks that a woman can arrive at clinic and still receive partial efficacy from the prepartum portion of ACTG 076. The default is set at 38 weeks. C17. % of women who receive any prenatal care who arrive in time for ‘short’ course of ARV. This figure is derived from estimated cumulative arrival time from cells “Setting” C27 through C34.

Adjusting for late arrival Enter for CDC-Thai and PETRA-A only C19. Latest gestational age in weeks that a woman can arrive at clinic and still receive full efficacy for CDC-Thai and PETRA-A. The default is set at 38 weeks. C20. % of women who receive any prenatal care who arrive in time for ‘short’ course of ARV. This figure is derived from estimated cumulative arrival time from cells “Setting” C27 through C34.

TABLE ARV-3: MOTHERS AND CHILDREN RECEIVING VARIOUS SERVICE COMPONENTS

Background

This table shows the number of women receiving each of the components of each ARV intervention. We assume no mother is denied treatment because she is not available for the full intended course; but rather all mothers who are tested and agree to ARV therapy receive some treatment. As shown in the table, all mothers receive intrapartum treatment even if they do not arrive in time for the prepartum component.

No Data entry is required.

C24 - G31 Table ARV-3. The figures in this table are derived from the estimates entered on the number of women completing VCT and the gestational age at start of treatment as shown in Table ARV-2.

TABLE ARV-4: DOSES, DRUGS AND DRUG COSTS.

Background

This table displays information needed to estimate the cost of each regimen. It requires data from the user on the cost per dose and, in the case of ACTG 076, the average number of weeks between the start of therapy and delivery. The CET then calculates the ARV costs for each intervention given their dosing schedules and amounts.

Influence on cost-effectiveness: Medium - High. The importance of these variables in determining cost-effectiveness varies a lot from regimen to regimen. As shown in the sensitivity analysis graph Figure 1c starting at cell “SAs” B20, the ARV-intensive interventions such as ACTG-076 and PETRA-A are more sensitive to drug costs than the less drug intensive regimens such as PETRA-B and especially HIVNET-012. If substitute feeding is also being evaluated, the drug cost inputs also assumes less importance because they represents a smaller portion of total program costs.

Expected effort of data collection: High. These data are both easy to obtain and important.

Estimating the cost of ARV doses to mother and to infant. The cost of ARVs to the public sector provider can vary depending upon agreements reached between the government and pharmaceutical companies or other third-party suppliers. The inputs required here are figures for the amount paid by the public sector provider for each dose of the relevant ARV agent. Since price information may be available in various forms (e.g., per bottle or per case), arriving at a per-dose price may require a calculation to be made outsides of the worksheet. If transportation and distribution costs are entailed that are additional to the price paid for the ARVs themselves, this too should be included in the per-dose cost estimate.

The default costs per dose currently used in the CET are shown below.

Default values of cost per dose of ARVs

Regimen	Mothers	Infants
ACTG-076	$0.78 per 300 mg dose AZT	$0.04/ml of Retrovir
HIVNET 012	$3.87 per 200mg NVP	$0.13 per 4 mg/kg NVP
CDC-Thai	$0.78 per 300 mg dose AZT	NA
PETRA-A	$2.22 per dose Combivir (combination of AZT/3TC)	$0.07/ml of Epivir (3TC); $0.04/ml of Retrovir (AZT)
PETRA-B	$2.22 per dose Combivir (combination of AZT/3TC)	$0.07/ml of Epivir (3TC); $0.04/ml of Retrovir (AZT)

Sources: Nevirapine - Johns Hopkins University Hospital Pharmacy, 1999; Combivir, Epivir. Retrovir - Personal communication, Joseph Saba, UNAIDS, 1999; AZT - Associated Press, Drug Company Cuts Cost of Aids Treatment, in New York Times. 1998: New York City. p. 87.

Cells ARV-C37 through ARV-I37. Cost per dose for mother. For each regimen, enter the cost per dose for mother for each ARV involved. For the Thai regimen, this is AZT (zidovudine) only, whereas the two PETRA regimens include both AZT and 3TC (lamivudine). Data sources: Project expenditure documents.

Cells ARV-C38 through ARV-I39; and ARV-D40. Cost per dose for infant. For each regimen, enter the cost per milliliter and number of milliliters of each postpartum dose to the infant. Since the HIVNET 012 regimen includes just one dose to the infant it is probably easier to enter this cost directly. Cell D40 is therefore a data entry (yellow) cell. Since the Thai regimen includes no postpartum treatment of infants, these cells read “NA”. As was the case with mothers, postpartum treatment of infants includes both AZT and 3TC for infants in the PETRA regimens. Data sources: Project expenditure documents.

Estimating the cost of the prepartum regimen. The number of doses and therefore the overall cost of prepartum treatment depends upon the average number of days of prepartum treatment and the number of doses per day. This period begins on the first day of treatment and ends on the day of delivery just prior to any change to a different intrapartum dosing pattern. The worksheet requires an estimate of the average number of days in this period for all women receiving each type of prepartum treatment. It then multiplies this number by the doses per day to calculate the total number of doses needed during the prepartum period. This product is in turn multiplied by the cost per dose in order to arrive at the overall cost for prepartum treatment. Because the possible start time for the ACTG 076 regimen can be so variable, the user is asked to enter figures for the actual number of weeks between the start of therapy and onset of labor. For each ARV agent used the prepartum treatment cost is simply: (Number of days of treatment) x (doses per day) x (cost per dose).

Cell ARV-C44. Average number of days for “long course” (only applies to ACTG-076). Enter figure for the average number of weeks of prepartum therapy experienced by mothers receiving ACTG 076 for women arriving in time for the long pre-partum course of treatment. Data sources: Project patient records.

Cell ARV-C49. Average weeks between therapy start and delivery for those receiving “short course” (only applies to ACTG-076). Enter figure for the average number of weeks of prepartum therapy experienced by mothers receiving ACTG 076 for women arriving too late for the full intended long prepartum course but in time for a shorter prepartum course of treatment. Data sources: Project patient records.

Low and high range of ARV drug costs. ARV prices tend to be fairly stable, and when they move, they usually move in a downward direction. Nevertheless, since they can be a major determinant of overall program costs we ask you to enter a percentage range above and below the base case estimate. Ideally the range would be based on knowledge of actual market conditions for the drugs or the state of negotiations with suppliers. However, in the absence of such knowledge, defining a range of 20% above and below the default prices will probably capture most of the potential short-term price variation. This will be used in the sensitivity analysis to show how cost-effectiveness varies with different estimates of drug costs.

C80 - G80. Enter the a figure for each regimen that represents the percent above and below the current drug cost estimate that you wish to examine in the sensitivity analyses. While it is possible to enter different figures for each regimen, the results of the sensitivity analyses will be easier to interpret if you use the same estimate across all five regimens. The default value is 20%.

C84 - Q85. Cost summary. The table in the blue cells “ARV” C84 through “ARV” Q85 provides summary information on the cost of ARVs for each of the regimens under consideration. The table includes both the cost of the entire cohort of women who complete the VCT sequence each year and the cost for each woman.

TABLE ARV-5: EFFECTIVENESS ESTIMATES

Background

ARV effectiveness refers to the percentage reduction in transmission in a treatment group compared with an untreated comparison group and is one of the most important determinant of overall cost-effectiveness. The model takes the relative percentage reduction in transmission as an input. Since effectiveness is sometimes expressed in absolute and sometimes in relative percentage terms it is necessary to understand the difference between the two and to be able to convert from one to another.

Relative efficacy: Relative percent is the more common way we think about applying percents. Multiplying the probability of transmission without treatment by the relative percent reduction would yield the efficacy. For example, if the intervention is 50% effective and the probability of transmission without treatment is 25%, the relative efficacy is 12.5%.

Absolute efficacy: The amount of reduction expressed directly as percentage points of transmission. For example, if an intervention has an absolute efficacy of 9%, the probability of transmission is reduced from 25% (to use the same background transmission rate as in the previous example) to 16%.

Influence on cost-effectiveness: High. ARV efficacy will have a large effect on program benefits and therefore on cost-effectiveness.

Expected effort of data collection: Low. Published efficacy estimates from the clinical trials are the accepted standard for these inputs. These estimates may be revised as the results from additional trials become available. We cannot foresee any circumstances under which it would be sensible to alter these estimates outside the context of the trials.

C91 - Q91. Relative efficacy. In the blue cells C91 through Q91 the relative reduction in HIV transmission for the ACTG 076, HIVNET 012, Thai, Petra-A and Petra-B regimens are shown as 67.5%, 47%, 51%, 50% and 37% respectively. These figures are based on the reported results of the respective clinical trials with the high and low estimates representing the limits of the 95% confidence interval around the point estimate. (See the ARV section of “Published studies and other resources” on page 76). These cells are locked to prevent casual changes in these values. Unless you have very good reasons to change them, such as new data from trials in your setting, we suggest that you not alter the default values given.

Background Transmission Rates and Efficacy Only the relative efficacy from the trials are being applied not the absolute values. A difference in background transmission rates between your setting and that of the clinical trials does not mean that the estimated efficacies from the treatment will be different in the two settings. For example, in absence of specific knowledge to the contrary, it is reasonable to suppose that a regimen which is 50% effective where transmission rates are 30% will also be 50% effective where transmission rates are only 20%.

As new trials are completed their results will be incorporated into subsequent versions of this model. Please visit the UNAIDS web site at http://www.UNAIDS/MTCT/*** for information about what additional regimens are being considered for inclusion.

TABLE ARV-6: ARV EFFECTIVENESS - ADJUSTMENTS FOR IMPERFECT ADHERENCE

Background

“Imperfect adherence” refers to the difference between the intended number and timing of doses in a regimen and the number and timing actually experienced by patients. The efficacy estimates incorporated into this model are based on the results of well-funded clinical trials. It is possible that patient adherence outside the trial setting will be lower than in the trials. The actual efficacy experienced in your client population may therefore also differ once adherence rates are taken into account. While lower adherence is undoubtedly associated with lower efficacy in general, the specific relation between the two have not been documented for the regimens under analysis. The default value of adjustments to reflect imperfect adherence is 0%. This means that adherence is expected to be the same as was attained in the trials. Depending on the setting, this may not be realistic. For example the infrastructure present in the industrialized countries where ACTG 076 was conducted may not be available in lower income countries. Adherence to this regimen may therefore be lower.

Treatment Adherence Rates

Country	Rates	Source
Northern Thailand	90% adherence	Thaineua et al., 1998
NTPHPT (Thailand)	99% of the women took at least 90% of the antenatal ZDV doses and 99% took at least 1 dose during labor	Weekly Epidemiological Record, 1998
UNAIDS PETRA trials (Uganda, Tanzania, South Africa)	7.7% of women did not take any of the prescribed medication	Marseille et al., 1998

Source: Adapted from John Stover, TFGI, 1999.

As shown in the table above, adherence can be variable even in the context of well funded clinical trials. The very sparse data that is available suggest that adherence may be more of a problem in sub-Saharan Africa than in Thailand. It is also possible that adherence will increase as people understand that they are taking a proven effective drug rather than a possible placebo or a drug which may or may not be effective. The adherence adjustment cells (ARVs-C98 - G98) can accept negative values. This would indicate that adherence in your project is better than that experienced in the trials and that this greater a adherence is associated with greater efficacy.

Unless good adherence data is available from your project we suggest you retain the default value of 0%. This makes for consistent comparisons with results generated by other users of this model. However, if you have reason to believe that adherence may be significantly above or below the levels experienced in the trials, cells C98 through G98 allow one to explore the consequences of different assumptions about adherence and its bearing on effectiveness.

Influence on cost-effectiveness: Potentially high. Adherence directly affects efficacy.

Expected effort of data collection: High. Since low adherence will render any intervention ineffective, most projects will find it worthwhile to monitor adherence independent of a cost-effectiveness analysis.

C98 - G98. Imperfect adherence. After reading the cautionary notes in the preceding paragraphs, enter changes in the adjustment for efficacy to reflect imperfect adherence if you believe they are warranted. Data sources: Project patient records.