We determine the optimal adaptive rate and power control strategies to maximize the total throughput in a multirate code-division multiple-access system. The total throughput of the system provides a meaningful baseline in the form of an upper bound to the throughput achievable with additional restrictions imposed on the system to guarantee fairness. Peak power and instantaneous bit energy-to-noise spectral density constraints are assumed at the transmitter with matched filter detection at the receiver. Our results apply to frequency selective fading in so far as the bit energy-to-equivalent noise power spectral density ratio definition can be used as the quality-of-service metric. The bit energy-to-equivalent noise power spectral density ratio metric coincides with the bit-error rate metric under the assumption that the processing gains and the number of users are high enough so that self-interference can be neglected. We first obtain results for the case where the rates available to each user are unrestricted, and we then consider the more practical scenario where each user has a finite discrete set of rates. An upper bound to the maximum average throughput is obtained and evaluated for Rayleigh fading. Suboptimal low-complexity schemes are considered to illustrate the performance tradeoffs between optimality and complexity. We also show that the optimum rate and power adaptation scheme with unconstrained rates is in fact just a rate adaptation scheme with fixed transmit powers, and it performs significantly better than a scheme that uses power adaptation alone.