Volume 11, Number 8—August 2005
Research
Optimizing Treatment of Antimicrobial-resistant Neisseria gonorrhoeae
Table 2
Variable description | Probabilities (%) |
|||
---|---|---|---|---|
Base | Range | Distribution* | Sources | |
Prevalence of gonorrhea in community among women | 1.0 | 0–15 | Triangular | 2 |
Prevalence of ciprofloxacin-resistant Neisseria gonorrhoeae | 0.1 | 0–20 | Triangular | 7 |
Prevalence of ceftriaxone-resistant N. gonorrhoeae | 0 | Assumed | ||
Treatment failure when strain is resistant to antimicrobial agent | 100 | Assumed† | ||
Treatment failure when strain is not resistant to antimicrobial agent | 0 | Assumed† | ||
Infected with gonorrhea and symptomatic | 30 | 20–50 | Triangular | 5,13,14 |
Infected with gonorrhea but without symptoms‡ | 70 | Residual‡ | Calculated | |
Not infected but with gonorrhea symptoms | 20 | 10–40 | Triangular | 5,13,15 |
Not infected and without gonorrhea symptoms‡ | 80 | Residual‡ | Calculated | |
Recalled patient returning to clinic | 40 | 20–80 | Triangular | 16,17 |
Sensitivity of nonculture-based tests | 95 | 85–100 | Triangular | 14,18,19 |
Specificity of nonculture-based tests | 97 | 95–99 | Triangular | 14,18,19 |
Sensitivity of culture-based tests | 93 | 85–95 | Triangular | 14,18,19 |
Specificity of culture-based tests | 97 | 95–97 | Triangular | 14,18,19 |
Concurrent HIV transmission§ | 0.066 | 0–0.5 | Triangular | 20 |
Develop pelvic inflammatory disease (PID) and sequelae, among untreated gonorrhea cases | 16 | 10–40 | Triangular | 5,13,14,21 |
Development of PID only (no sequelae)¶ | 70 | 70–72 | Uniform | 15,16,21,22 |
Developing sequelae of PID¶ | ||||
Infertility | 6 | 1–6 | Uniform | 15,16,21 |
Ectopic pregnancy | 8 | 5–9 | Uniform | 15,16,21 |
Chronic pelvic pain | 16 | 15–20 | Uniform | 15,16,21 |
Urethritis | 50 | 35–65 | Uniform | 15,16,21 |
Epididymitis | 2 | 1–5 | Uniform | 15,16,21 |
For strategy 1, % of culture-positive samples tested for antimicrobial susceptibility# | 80 | Assumed | ||
For strategy 3, % of culture-positive samples tested for antimicrobial resistance# | 20 | Assumed | ||
Female-to-male transmission of gonorrhea§ | 50 | 30–75 | Uniform | 5,13,23 |
Male-to-female transmission of gonorrhea§ | 50 | 30–75 | Uniform | 5,13,23 |
*The probability distributions used in the Monte Carlo sensitivity analysis. Uniform distributions were constructed with the minimum and maximum of the given ranges. Triangular distributions were constructed with the minimum and maximum of the given ranges and the base case as the "most likely" value.
†Assumes 0% drug failure if organism is not resistant.
‡The probability of being infected and without gonorrhea symptoms is the residual value after considering the probability of being infected and with gonorrhea symptoms. Likewise, the probability of being not infected and without gonorrhea symptoms is the residual value after considering the probability of not being infected and with gonorrhea symptoms.
§Describes probability of initially infected woman transmitting disease to male partner, who then has the probability of infecting another female partner (or reinfecting original female partner after she has been cured of initial infection). Further, with the initial female-to-male transmission, the probability of concurrent HIV transmission exists.
¶Rate of PID (only) and rates of PID-related sequelae are given as percentages of those that develop PID.
#See Table 1 for descriptions of strategy 1 (ST1) and strategy 3 (ST3). In ST1, culture-positive samples are tested for ciprofloxacin resistance. In ST3, culture-positive samples are tested for ceftriaxone resistance.
1In 2000, only 18% of gonorrhea tests performed by public health laboratories in the United States were culture-based tests.
2Monte Carlo simulation involves specifying a probability distribution of values for model inputs. A computer algorithm then runs the model for several iterations. During each iteration, the computer algorithm selects input values from the probability distributions, and calculates the output (e.g., cost per patient successfully treated). After the final run, the model provides results such as the mean, median, and 5th and 95th percentiles for each specified output.