Background Exercise programmes are a relatively inexpensive, low-risk option compared with other, more invasive therapies for treatment of leg pain on walking (intermittent claudication (IC)). This is the fourth update of a review first published in 1998. Objectives Our goal was to determine whether an exercise programme was effective in alleviating symptoms and increasing walking treadmill distances and walking times in people with intermittent claudication. Secondary objectives were to determine whether exercise was effective in preventing deterioration of underlying disease, reducing cardiovascular events, and improving quality of life. Search methods For this update, the Cochrane Vascular Information Specialist searched the Specialised Register (last searched 15 November 2016) and the Cochrane Central Register of Controlled Trials (CENTRAL; 2016, Issue 10) via the Cochrane Register of Studies Online, along with trials registries. Selection criteria Randomised controlled trials of an exercise regimen versus control or versus medical therapy for people with IC due to peripheral arterial disease (PAD). We included any exercise programme or regimen used for treatment of IC, such as walking, skipping, and running. Inclusion of trials was not affected by duration, frequency, or intensity of the exercise programme. Outcome measures collected included treadmill walking distance (time to onset of pain or pain-free walking distance and maximum walking time or maximum walking distance), ankle brachial index (ABI), quality of life, morbidity, or amputation; if none of these was reported, we did not include the trial in this review. Data collection and analysis For this update (2017), RAL and AH selected trials and extracted data independently. We assessed study quality by using the Cochrane 'Risk of bias' tool. We analysed continuous data by determining mean differences (MDs) and 95% confidence intervals (CIs), and dichotomous data by determining risk ratios (RRs) and 95% CIs. We pooled data using a fixed-effect model unless we identified significant heterogeneity, in which case we used a random-effects model. We used the GRADE approach to assess the overall quality of evidence supporting the outcomes assessed in this review.